• TABLE OF CONTENTS
HIDE
 Front Cover
 Half Title
 Title Page
 Table of Contents
 List of Tables
 List of Figures
 Foreword
 Acknowledgement
 Introduction
 Part I: Evidence on patterns of...
 Part II: Breeding and yield...
 Part III: Input management and...
 Part IV: Impacts of yield variability...
 References
 Contributors
 Index














Title: Variability in grain yields
CITATION THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00085378/00001
 Material Information
Title: Variability in grain yields implications for agricultural research and policy in developing countries
Physical Description: xix, 395 p. : ill. ; 24 cm.
Language: English
Creator: Anderson, Jock R., 1941-
Hazell, P. B. R
International Food Policy Research Institute
Publisher: Published for the International Food Policy Research Institute by the Johns Hopkins University Press,
Published for the International Food Policy Research Institute by the Johns Hopkins University Press
Place of Publication: Baltimore
Publication Date: c1989
Copyright Date: 1989
 Subjects
Subject: Wheat trade -- Seasonal variations -- Developing countries   ( lcsh )
Wheat trade -- Government policy -- Developing countries   ( lcsh )
Wheat -- Varieties -- Research -- Developing countries   ( lcsh )
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
 Notes
Bibliography: Includes bibliographical references (p. 357-380) and index.
Statement of Responsibility: edited by Jock R. Anderson and Peter B.R. Hazell.
 Record Information
Bibliographic ID: UF00085378
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: oclc - 18780860
lccn - 88032079
isbn - 0801837936 (alk. paper)

Table of Contents
    Front Cover
        Front Cover 1
        Front Cover 2
    Half Title
        Page i
        Page ii
    Title Page
        Page iii
        Page iv
    Table of Contents
        Page v
        Page vi
        Page vii
        Page viii
    List of Tables
        Page ix
        Page x
        Page xi
        Page xii
    List of Figures
        Page xiii
        Page xiv
        Page xv
        Page xvi
    Foreword
        Page xvii
        Page xviii
    Acknowledgement
        Page xix
        Page xx
    Introduction
        Page 1
        Page 2
        Page 3
        Page 4
        Page 5
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
    Part I: Evidence on patterns of changing yield variability
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
        Page 41
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
        Page 70
        Page 71
        Page 72
        Page 73
        Page 74
        Page 75
        Page 76
        Page 77
        Page 78
        Page 79
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
        Page 102
        Page 103
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
        Page 110
        Page 111
        Page 112
        Page 113
        Page 114
        Page 115
        Page 116
        Page 117
        Page 118
        Page 119
        Page 120
        Page 121
        Page 122
        Page 123
        Page 124
    Part II: Breeding and yield variability
        Page 125
        Page 126
        Page 127
        Page 128
        Page 129
        Page 130
        Page 131
        Page 132
        Page 133
        Page 134
        Page 135
        Page 136
        Page 137
        Page 138
        Page 139
        Page 140
        Page 141
        Page 142
        Page 143
        Page 144
        Page 145
        Page 146
        Page 147
        Page 148
        Page 149
        Page 150
        Page 151
        Page 152
        Page 153
        Page 154
        Page 155
        Page 156
        Page 157
        Page 158
        Page 159
        Page 160
        Page 161
        Page 162
        Page 163
        Page 164
        Page 165
        Page 166
        Page 167
        Page 168
        Page 169
        Page 170
        Page 171
        Page 172
        Page 173
        Page 174
        Page 175
        Page 176
        Page 177
        Page 178
        Page 179
        Page 180
        Page 181
        Page 182
        Page 183
        Page 184
        Page 185
        Page 186
        Page 187
        Page 188
        Page 189
        Page 190
        Page 191
        Page 192
        Page 193
        Page 194
        Page 195
        Page 196
        Page 197
        Page 198
        Page 199
        Page 200
        Page 201
        Page 202
        Page 203
        Page 204
        Page 205
        Page 206
        Page 207
        Page 208
        Page 209
        Page 210
        Page 211
        Page 212
        Page 213
        Page 214
        Page 215
        Page 216
        Page 217
        Page 218
        Page 219
        Page 220
    Part III: Input management and yield variability
        Page 221
        Page 222
        Page 223
        Page 224
        Page 225
        Page 226
        Page 227
        Page 228
        Page 229
        Page 230
        Page 231
        Page 232
        Page 233
        Page 234
        Page 235
        Page 236
        Page 237
        Page 238
        Page 239
        Page 240
        Page 241
        Page 242
        Page 243
        Page 244
        Page 245
        Page 246
        Page 247
        Page 248
        Page 249
        Page 250
        Page 251
        Page 252
        Page 253
        Page 254
        Page 255
        Page 256
        Page 257
        Page 258
        Page 259
        Page 260
        Page 261
        Page 262
        Page 263
        Page 264
        Page 265
        Page 266
        Page 267
        Page 268
        Page 269
        Page 270
        Page 271
        Page 272
        Page 273
        Page 274
        Page 275
        Page 276
        Page 277
        Page 278
        Page 279
        Page 280
        Page 281
        Page 282
        Page 283
        Page 284
    Part IV: Impacts of yield variability and implications for policy
        Page 285
        Page 286
        Page 287
        Page 288
        Page 289
        Page 290
        Page 291
        Page 292
        Page 293
        Page 294
        Page 295
        Page 296
        Page 297
        Page 298
        Page 299
        Page 300
        Page 301
        Page 302
        Page 303
        Page 304
        Page 305
        Page 306
        Page 307
        Page 308
        Page 309
        Page 310
        Page 311
        Page 312
        Page 313
        Page 314
        Page 315
        Page 316
        Page 317
        Page 318
        Page 319
        Page 320
        Page 321
        Page 322
        Page 323
        Page 324
        Page 325
        Page 326
        Page 327
        Page 328
        Page 329
        Page 330
        Page 331
        Page 332
        Page 333
        Page 334
        Page 335
        Page 336
        Page 337
        Page 338
        Page 339
        Page 340
        Page 341
        Page 342
        Page 343
        Page 344
        Page 345
        Page 346
        Page 347
        Page 348
        Page 349
        Page 350
        Page 351
        Page 352
        Page 353
        Page 354
        Page 355
        Page 356
    References
        Page 357
        Page 358
        Page 359
        Page 360
        Page 361
        Page 362
        Page 363
        Page 364
        Page 365
        Page 366
        Page 367
        Page 368
        Page 369
        Page 370
        Page 371
        Page 372
        Page 373
        Page 374
        Page 375
        Page 376
        Page 377
        Page 378
        Page 379
        Page 380
    Contributors
        Page 381
        Page 382
        Page 383
        Page 384
    Index
        Page 385
        Page 386
        Page 387
        Page 388
        Page 389
        Page 390
        Page 391
        Page 392
        Page 393
        Page 394
        Page 395
        Page 396
Full Text
Anderson
Hazell

Variability
in
Grain
Yields


Variability


Implications
for Agricultural
Research and
Policy in
Developing
Countries


,ran
rrall*


fields


edited by


Jock R. Anderson and
Peter B. R. Hazell


Johns
Hopkins
F)






.Als published in cuoperaiion with ih It nrnanonal Food PolacY Research Instistte

Food Subsidies in Developing Countries
Costs. Benefits, and Policy Options
EDITED BY PER PINSTRLIP--NDERSEN

Nineteen leading economists, nutritionists, and police) analysts present
key findings from the International Food Polic. Research Institute. syn-
thesizing practical experiences with subsidy programs in countries
throughout Asia. Africa. and Latin America The authors emphasize les-
sons applicable to decisionmaking on current and future food policy for
developing countries.


Agricultural Price Policy for Developing Countries
EDITED BY JOHN \\. MELLOR AND RAISUDDIN AHMED

Distinguished specialists recommend approaches for managing price fluc-
tuations and exchange rates, relating domestic to international prices.
and balancing the needs of producers and consumers in a coherent and
consistent strategy of economic development.
"[A] pragmatic blend of theory, administrative experience, and empirical
analysis of field practice."-Choice


Crop Insurance for Agricultural Development
Issues and Experience
EDITED BY PETER HAZELL, CARLOS POMAREDA, AND sLBERTO VALDES
Smith the assistance of Joan Straker Hazell

The authors present the economic theory behind crop insurance, test it
against empirical data from several countries, and put it into perspective
with other police> options
"A comprehensive analysis ... An important contribution to the under-
standing of crop insurance and its relationship with agricultural develop-
ment."-Carlos W. Cue% as, Journal of Economic Literature










The Johns Hopkins University Press
BALTIMORE AND LONDON


ISBN 0-801S-3"








Variability in Grain Yields








This publication is the outcome of collaboration between the International
Food Policy Research Institute and the Deutsche Stiftung fir Internationale
Entwicklung / Zentralstelle fiir Ernihrung und Landwirtschaft (German
Foundation for International Development / Food and Agriculture
Development Center).







Variability in Grain Yields

Implications for Agricultural Research and Policy
in Developing Countries


EDITED BY
JOCK R. ANDERSON AND PETER B. R. HAZELL























Published for the International Food Policy Research Institute
The Johns Hopkins University Press
Baltimore and London





























1989 The International Food Policy Research Institute. Copyright does not include the
separate content of chapter 14, which is a work of the U.S. government.
All rights reserved
Printed in the United States of America

The Johns Hopkins University Press, 701 West 40th Street, Baltimore, Maryland 21211
The Johns Hopkins Press Ltd., London

The paper used in this publication meets the minimum requirements of American
National Standard for Information Sciences-Permanence of Paper for Printed Library
Materials, ANSI Z39.48-1984.


LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA
Variability in grain yields: implications for agricultural research and policy in developing
countries/edited by Jock R. Anderson and Peter B. R. Hazell.
p. cm.
Bibliography: p.
Includes index.
ISBN 0-8018-3793-6 (alk. paper)
1. Wheat Trade-Developing countries-Seasonal variations. 2. Wheat
trade-Government policy-Developing countries. 3. Wheat-Developing
countries-Varieties-Research. I. Anderson, Jock R., 1941- II. Hazell,
P. B. R.
HD9049.W5D448 1989
338.1'7311'091724-dcl9 88-32079
CIP









Contents


List of Tables and Figures ix
Foreword xvii
Acknowledgments xix
1 Introduction 1
JOCK R. ANDERSON AND PETER B. R. HAZELL

PART I Evidence on Patterns of Changing Yield Variability
2 Changing Patterns of Variability in World Cereal Production 13
PETER B. R. HAZELL
3 Changing Patterns of Variability in Chinese Cereal Production 35
BRUCE STONE AND TONG ZHONG
4 An Analysis of Variability in Soviet Grain Production 60
JOHN R. TARRANT
5 Agricultural Planning Policy and Variability in Syrian Cereal
Production 78
HUNG NGUYEN
6 High-Yielding Varieties and Variability in Sorghum and Pearl
Millet Production in India 91
THOMAS S. WALKER
7 Variability in Wheat Yields in England: Analysis and Future
Prospects 100
ROGER B. AUSTIN AND MICHAEL H. ARNOLD
8 Variability in Wheat and Barley Production in Southeast
England 107
PAUL WEBSTER AND NIGEL T. WILLIAMS
9 Variability in Winter Wheat and Spring Barley Yields in
Bavaria 118
G. FISCHBECK






vi Contents


PART H Plant Breeding and Yield Variability
10 Plant Breeding and Yield Stability 127
MICHAEL H. ARNOLD AND ROGER B. AUSTIN
11 Modern Rice Varieties as a Possible Factor in Production
Variability 133
W. RONNIE COFFMAN AND T. R. HARGROVE
12 Possible Genetic Causes of Increased Variability in U.S. Maize
Yields 147
DONALD N. DUVICK
13 Yield Stability in Bread Wheat 157
W. H. PFEIFFER AND H. J. BRAUN
14 Genetic Improvement and the Variability in Wheat Yields in the
Great Plains 175
C. JAMES PETERSON, V. A. JOHNSON, J. W. SCHMIDT, AND
ROBERT F. MUMM
15 Yield Stability of CIMMYT Maize Germplasm in International and
On-Farm Trials 185
H. N. PHAM, S. R. WADDINGTON, AND J. CROSS
16 Variability in the Yield of Pearl Millet Varieties and Hybrids in
India and Pakistan 206
JOHN R. WITCOMBE

PART III Input Management and Yield Variability
17 Fertilizer and Crop Yield Variability: A Review 223
JAMES A. ROUMASSET, MARK W. ROSEGRANT,
UJJAYANT N. CHAKRAVORTY, AND JOCK R. ANDERSON
18 Irrigation and Crop Yield Variability: A Review 234
SUSHIL PANDEY
19 Pest-Resistant Varieties, Pesticides, and Crop Yield Variability: A
Review 242
GERALD A. CARLSON
20 Yield Stability and Modern Rice Technology 251
JOHN C. FLINN AND DENNIS P. GARRITY
21 Influence of Nitrogen Fertilizer and Fungicide on Yield and Yield
Variability in Wheat and Barley 265
H. HANUS AND P. SCHOOL
22 Influence of Technology and Weather on the Variability in U.S.
Maize and Wheat Yields 270
JAMES B. FRENCH AND J. C. HEADLEY






Contents vii


PART IV Impacts of Yield Variability and Implications for Policy
23 Can Yield Variability Be Offset by Improved Information? The
Case of Rice in India 287
VISHVA BINDLISH, RANDOLPH BARKER, AND TIMOTHY D. MOUNT
24 Are Modern Cultivars More Risky? A Question of Stochastic
Efficiency 301
JOCK R. ANDERSON, CARLY J. FINDLAY, AND G. H. WAN
25 Yield and Household Income Variability in India's Semi-Arid
Tropics 309
THOMAS S. WALKER
26 The Implications of Variability in Food Production for National
and Household Food Security 320
DAVID E. SAHN AND JOACHIM VON BRAUN
27 Synthesis and Needs in Agricultural Research and Policy 339
JOCK R. ANDERSON AND PETER B. R. HAZELL
References 357
Contributors 381
Index 385








Tables and Figures












Tables
2.1 Variability of world cereal production around linear trend:
1960/61-1982/83 15
2.2 Changes in the mean and variability of world cereal production:
1960/61-1970/71 to 1971/72-1982/83 16
2.3 Changes in the mean and variability of total cereal production by major
countries: 1960/61-1970/71 to 1971/72-1982/83 18
2.4 Mean yields and coefficients of variation by crop and country:
1960/61-1970/71 and 1971/72-1982/83 20
2.5 Summary of trends in coefficients of variation between periods reported
in Table 2.4 26
2.6 Components of change in production covariances 27
2.7 Disaggregation of the components of change in the average of world
cereal production: 1960/61-1970/71 to 1971/72-1982/83 28
2.8 Disaggregation of the components of change in the variance of world
cereal production: 1960/61-1970/71 to 1971/72-1982/83 30
3.1 Changes in the mean and variability of national foodcrop production,
area, and yield in the People's Republic of China 44
3.2 Analysis of the components of change in mean foodgrain production in
China, 1952-1957 to 1979-1983 46
3.3 Analysis of the components of change in the variance of total grain
production in China, 1952-1957 to 1979-1983 48
3.4 Analysis of the components of change in the variance of rice production
by province in China, 1952-1957 to 1979-1983 50
3.5 Analysis of the components of change in the variance of wheat
production by province in China, 1952-1957 to 1979-1983 52
3.6 Analysis of the components of change in the variance of other crop
production by province in China, 1952-1957 to 1979-1983 54






x Tables and Figures

3.7 Provinces exhibiting variance changes between 1952-1957 and
1979-1983 56
3.8 Wheat yield correlation coefficients among five major producing
provinces, 1979-1983 57
4.1 Spatial variability in cereal production within Soviet regions 66
4.2 Annual variability in Soviet cereal production through time 66
4.3 Cereal production trends in Soviet regions 69
5.1 The mean and coefficient of variation of areas, yield, and production
of cereals in Syria during the nonreform, land-reform, and
post-land-reform periods 80
5.2 Changes in the mean and variability of barley and wheat areas in
Syria 82
5.3 Changes in the mean and variability of barley and wheat yields in
Syria 83
5.4 Changes in the mean and variability of barley and wheat production in
Syria 84
5.5 Components of change in the average production of barley and wheat in
Syria: 1950-57 and 1962-64 to 1971-81 85
5.6 Disaggregation of the components of change in the variance of total
barley and wheat production in Syria: 1950-57 and 1962-64 to
1971-81 86
5.7 Yield correlations between provinces for barley and wheat in Syria 87
5.8 Yield correlations between barley and wheat in Syria 87
5.9 Percentage of wheat areas planted to high-yielding varieties in
Syria 88
6.1 Contribution of different sources to increased interregional covariance
in sorghum and pearl millet production in India, 1956/57-1967/68 to
1968/69-1979/80 93
6.2 Means and ranges of the data used to explain the increase in
interregional yield covariance in sorghum and pearl millet production
in India 97
6.3 Estimated regression coefficients of the determinants of changes in
interregional yield covariance in sorghum and pearl millet
production 98
7.1 Historical trends in English wheat yields 101
7.2 Interannual variation in English wheat yields 101
7.3 Interannual variation in wheat yields at individual sites in the United
Kingdom 104





Tables and Figures xi


7.4 Variation in wheat yields from field to field in the United
Kingdom 105
8.1 Changes in the mean and variability of wheat and barley production on
two groups of farms in Southeast England: 1964-1974 to
1975-1984 109
8.2 Components of change in wheat production variance on 16 farms in
Southeast England: 1964-1974 to 1975-1984 110
8.3 Interfarm yield correlations of wheat and barley significantly different
from zero on selected farms in Southeast England 112
9.1 Coefficients of regression and standard deviation for kernel yield of
winter wheat and spring barley in Bavaria in 10-year periods,
1950-1984 119
9.2 Changes in the mean and variability of winter wheat and spring barley
yields in Bavaria, 1950-1966, 1956-1975, and 1967-1984 123
9.3 Mean and variability of kernel yield of old cultivars (0) or landraces
and modern cultivars (M) of winter wheat and spring barley in different
crop management systems in Bavaria 124
10.1 Coefficients of variation of yield computed for hypothetical cases 130
10.2 Components of variance in 174 winter wheat trials during
1967-1978 131
13.1 Proportion of genotypes in international spring wheat yield nurseries
from 1964-1965 to 1978-1979 162
13.2 Correlations between parameters from one year to the next for a
common set of genotypes and environments 173
14.1 List of locations in the southern and northern regional performance
nurseries from which yield data were used for calculation of trends in
yield and variance in productivity 177
15.1 Grain yield and stability parameters of maize varieties in consecutive
cycles of selection tested in 1979-1984 191
15.2 Summary comparison of improved maize varieties with farmers' local
varieties in on-farm variety trials, where inputs and management are at
the farmer level 196
15.3 Grain yield and stability parameters for six variety X N fertilizer
combinations at seven sites on the Cayes Plain, Haiti, 1983 202
15.4 Yield, slope, and economic data for three packages of inputs at 418
sites in Ghana, 1982-1983 204
16.1 Summary of international pearl millet adaptation trials 207
16.2 Characteristics of the two highest yielding entries in the mean and
lowest yielding environments 214






xii Tables and Figures


16.3 Yield and stability characteristics of highest yielding entries averaged
over all environments and in the lowest yielding environment 215
16.4 Mean yields, S2 values, and regression coefficients for hybrids and
varieties 216
16.5 Sources of genetic variation relative to environmental variation
standardized to one hundred 217
16.6 Entries selected on basis of complex criteria of yield and stability 217
17.1 Elasticity of mean and variance of yield with respect to nitrogen
fertilizer and optimal nitrogen use with risk-neutral and risk-averse
decision preference 228
19.1 Sources of risk, utility formulation, and evidence on marginal risk
effects of pest control inputs 246
19.2 Export price ratios of pesticides for rice and wheat 249
20.1 First-period coefficients of variation and their changes in production,
area, and yield of rice in eight major rice-growing countries for periods
before and during modern variety rice adoption 252
20.2 Adaptability and stability of rice cultivars tested in the international
rice-testing program 255
20.3 Mean rice yields and yield distributions on farmers' fields in irrigated
and upland rainfed sites in the Philippines 261
22.1 Crop-reporting districts with statistically different variance measures
between low- and high-technology periods 277
22.2 Crop-reporting districts with no statistical difference in error variance
but with statistical difference in normalized weather variance between
low- and high-technology periods 277
22.3 Percentage of statistically significant crop-reporting district yield
correlations for maize in the United States 281
22.4 Percentage of statistically significant differences between crop-reporting
district yield correlations for periods 1931-1946 and 1967-1981 for
maize in the United States 282
23.1 Slope estimates for the structural model, the model of farmers'
expectations, and the model of the government's trend-based
expectations 292
23.2 Slope estimates for the model of the government's adaptive
expectations 296
23.3 Estimated variance of aggregate rice yield with and without the
high-yielding varieties 297






Tables and Figures xiii


23.4 Sums of district variances and interdistrict covariances with and
without the high-yielding varieties 298
24.1 Wheat yields over time with changing cultivars 302
25.1 Agroclimatic and technological features for three villages in India's
semi-arid tropics from 1975/76 to 1983/84 310
25.2 Descriptive information on the common crops sown in the study villages
in India from 1975/76 to 1983/84 312
25.3 Simulated risk benefits from perfect crop yield stabilization 314
26.1 World cereal production, trade, and price of wheat during the "food
crises," 1970-1976 323
26.2 Coefficient of variation for per capital cereal production, cereal
consumption, and total food consumption 324
26.3 Correlation coefficients for cereal production, cereal consumption,
total food consumption, and noncereal consumption, 1966-1980 327
26.4 Regressions of the coefficient of variation of daily calorie consumption
on gross national product and the coefficient of variation of food
production, and resulting elasticities 328

Figures
2.1 World cereal production, 1960/61-1982/83 14
3.1a Coefficients of variation for sown area of foodgrains in China,
1954-1982 36
3.1b Coefficients of variation for yields of foodgrains in China,
1954-1982 37
3.1c Coefficients of variation for production of foodgrains in China,
1954-1982 37
3.2 Fertilizer nutrient application per sown hectare in China, 1983 42
4.1 Cereal production in the Soviet Union and net cereal imports 61
4.2 Spatial variability in Soviet cereal production 64
4.3 Annual variability in Soviet cereal production 67
4.4 Diverging trends in Soviet regional cereal production 68
4.5 Spatial contrasts in Soviet production 70
4.6 Climatic factors in variability of Soviet cereal production, 1965 72
4.7 Climatic factors in the variability of Soviet cereal production,
1975 73
4.8 Regional compensation effects in the Soviet Union 74






xiv Tables and Figures


6.1 Adoption of pearl millet hybrids in Bhir (Maharashtra) and South
Arcot (Tamil Nadu), 1966-1980 95
7.1 U.K. wheat yields, 1948-1984 103
8.1 Number of sample farms showing change in variability of wheat and
barley yields in Southeast England, 1964-1974 to 1975-1984 111
8.2 Nitrogen application per hectare, England and Wales,
1974-1984 114
8.3 Fungicide application on cereals, England and Wales,
1970-1984 115
8.4 Fungicide use and the change to winter-sown barley, England and
Wales, 1969-1984 116
9.1 Changes in yield increase and yield variation of winter wheat and spring
barley in Bavaria, 1950-1984 120
9.2 Mean yield of winter wheat and spring barley in Bavaria,
1960-1978 122
11.1 Twenty landraces, or traditional rice varieties, from eight nations in
the genetic ancestry of IR64, released in May 1985 136
11.2 Maternal derivation of IR varieties 138
11.3 Comparison of breeding objectives, 1984 and 1975 141
12.1 Annual average grain yield of U.S. maize, 1930-1985 148
13.1 Response of four hypothetical varieties to higher productivity
levels 160
13.2 Distribution of mean yields, coefficients of regression, and yield
stability parameters for four groups of genotypes across International
Spring Wheat Yield Nurseries 1 through 15 163
13.3 Yield of Veery "S" in the 73 environments of the 15th International
Spring Wheat Yield Nursery 165
13.4 Relative yield performance of the respective highest yielding genotype
(variety) in the Asian and tropical highlands regions 167
13.5 Relative grain yield of the CIMMYT variety Veery "S," the best locally
developed variety (LDV), and the longtime check variety Siete Cerros
in eight different environmental groupings for the 15th International
Spring Wheat Yield Nursery 169
13.6 Input efficiency of old and new varieties under differing production
conditions 171
13.7 Nitrogen response curves of Veery "S" and two old cultivars 172
14.1 Mean squares associated with the interaction of genotype and
environment effects on yield of cultivars in the Southern Regional
Performance Nursery, 1958-1984 179





Tables and Figures xv


14.2 Mean squares associated with the interaction of genotype and
environment effects on yield of cultivars in the Northern Regional
Performance Nursery, 1958-1984 180
14.3 Coefficients (b values) from regression of check cultivar yields on
nursery mean yields at locations in the Southern Regional Performance
Nursery, 1959-1984 181
14.4 Coefficients (b values) from regression of check cultivar yields on
nursery mean yields at locations in the Northern Regional Performance
Nursery, 1959-1984 182
15.1 Steps in population improvement and experimental variety
development, evaluation, and use 188
15.2 Linear regressions of variety yield on the mean grain yield at a site for
three improved maize varieties and the local variety in eastern
Paraguay, 1984 197
15.3 Linear regressions of variety yield on the mean grain yield at a site for
seven improved maize varieties and the local variety at 18 sites in
Ghana, 1983 198
15.4 Linear regressions of variety yield on the mean grain yield at a site for
V455 and the local variety at 14 sites in Guerrero, Mexico, 1984 and
1985 200
15.5 Linear regressions of variety/input treatment yields on the mean grain
yield at a site for four variety/input treatments at 26 sites in Veracruz
state, Mexico, 1973-1980 201
15.6 Linear regressions of input "package" yield on the mean grain yield at
a site for four input packages at 25 sites in Thailand during 1984 203
16.1 to 16.5 Relationship of regression coefficients, yield, and s2 values,
International Pearl Millet Adaptation Trials, 1979-1983 208
16.6 Mean yield and standard deviation of the entries in the International
Pearl Millet Adaptation Trial, 1983 218
16.7 Cumulative probability and predicted yield in the lowest yielding
environment in the International Pearl Millet Adaptation Trials, 1979,
1980, 1981, and 1983 219
21.1 Yields and standard deviations of winter barley with increasing
nitrogen levels with and without fungicide treatments based on averages
over the different application systems 267
21.2 Yields and standard deviations of winter wheat with increasing nitrogen
levels with and without fungicide treatments based on averages over
the different application systems 268
24.1 Cumulative distribution functions for two cultivars and two crop
histories 303





xvi Tables and Figures

24.2 Example of cumulative distribution functions with second-period wheat
yield that is first stochastic dominant over first-period yield, as found
in seven local government areas 305
24.3 Example of cumulative distribution functions with second-period wheat
yield that is first stochastic dominant over first-period yield, as found
in 14 local government areas 306
24.4 Example of cumulative distribution functions with second-period wheat
yield that is second stochastic dominant over first-period yield, as found
in 24 local government areas 307
26.1 Production, price formation, and consumption: tracing market-level
effects of instability for consumption 322








Foreword


Variability in foodgrain yields and production has entered into the food
policy agenda in the wake of the green revolution, but debate and decision
making have been stifled for lack of a systematically gathered body of co-
gent evidence.
Research by the International Food Policy Research Institute (IFPRI)
on countries and crops shows that, in most cases, increases in yield vari-
ability and, more important, a loss in offsetting patterns of variation (in-
creased correlations) in crop yields between regions are the predominant
sources of the increase in production variability. There has been a ten-
dency by some researchers to attribute this increased yield variability to
improved seed- and fertilizer-based technologies. Some researchers have
also argued that plant breeders should focus less on maximizing average
yields and more on reducing yield sensitivity to environmental stress. Such
recommendations may prove costly for future growth in foodgrain produc-
tion, and they cannot be warranted before more thorough and quantitative
studies of the sources of increased variability have been undertaken.
In view of the importance of this issue to national breeding programs
and to the international agricultural research centers, the Deutsche Stif-
tung fur Internationale Entwicklung (DSE) and IFPRI convened an inter-
disciplinary workshop for an intensive four-day discussion of a broad
range of issues associated with increasing yield variability. There were
about 60 participants, including biologists, social scientists, and policy-
makers, with particularly strong representation from the centers of the
Consultative Group on International Agricultural Research (CGIAR).
Workshop participants discussed the relationship between changes in
yield variability and yield correlations and such causal factors as changes
in agricultural technology, weather, irrigation, input availability, and re-
lated variables. They also discussed the consequences of increasing yield
variability, including its effect on different types of farmers and on poor
urban and rural consumers. Participants were asked to make specific rec-





xviii Foreword

commendations for agricultural research policy in the fields of plant breed-
ing, farming systems, and management of irrigation, fertilizers, and pesti-
cides, and to address the need for changes in national and international
agricultural policies. (Summary Proceedings of a Workshop on Cereal
Yield Variability, edited by Peter B. R. Hazell, was published by IFPRI
and DSE in 1986.)
Selected papers from the workshop plus papers commissioned to fill
gaps in coverage of issues constitute this volume. It is our hope that this
collection of papers will stimulate debate and further research on the im-
portant topic of yield variability and that it will lead to improved policies
and agricultural research priorities for coping with yield risks in the future.

JOHN W. MELLOR
ERHARD KRUSKEN








Acknowledgments


The publication of this book completes a collaborative effort extending
over more than four years. We are most grateful to the many people who
planned and participated in the 1985 Feldafing Workshop and to those
who have assisted subsequently in preparing papers to provide a more
complete coverage of the subject.
The International Food Policy Research Institute and the Deutsche
Stiftung fir Internationale Entwicklung generously provided financial and
institutional support. Lloyd T. Evans, through his exceptional synthesiz-
ing skill exercised at the workshop, provided important parts of the per-
spectives described in the concluding chapter. Robert E. Evenson reviewed
the entire manuscript and contributed many helpful suggestions. Eliza-
beth A. Anderson helped to edit and rework the papers into the more cohe-
sive forms herein.
JOCK R. ANDERSON
PETER B. R. HAZELL








1 Introduction

JOCK R. ANDERSON AND PETER B. R. HAZELL










Many countries have achieved impressive rates of growth in national food-
grain production in recent years. Much of this growth can be attributed to
new technologies, especially improved varieties, and the increased use of
irrigation, fertilizers, and pesticides. These increases in production have
provided a lifeline for many developing countries and prevented the mass
starvation predicted by some observers in the mid-1960s.
As agricultural output has grown, however, so has its variability, and
this presents other problems and concerns that need to be addressed by the
agricultural research and policymaking community. Prominent among
these concerns are:

a. increased income risk, which may make new technologies less attrac-
tive to farmers and hence slow agricultural growth in developing
countries;
b. increased instability in national and world food supplies, which may
act to destabilize domestic prices, national income, and the food con-
sumption of the poor, especially in poor agrarian countries;
c. increased variability in domestic production, which may add to the
difficulties and cost of price support and stabilization schemes in
many industrial countries.

There has been a tendency by some researchers to attribute this in-
creased variability to the improved crop varieties underlying the new tech-
nologies (e.g., Mehra 1981, Barker et al. 1981, and Griffin 1988). Some
researchers have also argued that plant breeders should focus less on maxi-
mizing average yields and more on reducing yield sensitivity to environ-
mental stress. Such recommendations may prove costly for future growth
in foodgrain production, and they cannot be warranted before more thor-
ough and quantitative studies have been undertaken of the sources of in-
creased variability or of alternative ways of redressing the problem.





2 Variability in Grain Yields


In view of the importance of these issues to policymakers, to national
breeding programs, and to the international agricultural research centers,
this book attempts to bring together a significant body of empirical evi-
dence on production variability, drawing on work in several disciplines,
including plant breeding, agronomy, and economics. Chapters review
available evidence on patterns of variability in cereal production and how
these patterns have changed in recent years. The biological, climatic, and
economic factors underlying these changes are discussed and implications
sought for both agricultural research and policy.
To make the task more manageable, the scope of the material has
been limited in two ways: First, there is a focus on yield risks. These are the
most important source of increase in the variability of world food produc-
tion (see ch. 2). Yield risks also lead to much of the variability in domestic
prices in many countries, and they are often the predominant source of risk
in rainfed agriculture. Second, attention is confined to cereals. These are
the predominant foodcrops for most of the world's poor. They are also the
crops that are most directly linked to recent advances in high-yielding
varieties.


Structure of the Book
The book has four parts. The first examines patterns of yield variabil-
ity for the world and for selected countries and how these patterns have
changed in recent decades. While the chapters in this section are largely
descriptive, they do suggest some important hypotheses about causal fac-
tors behind changing patterns of variability. Part II explores the relation-
ship between plant breeding and yield variability. It includes empirical
evaluations of the stability of many modern cereal varieties in comparison
with their more traditional counterparts. Part III reviews the relationships
between cultural practices, particularly the use of key inputs, and yield
variability. The final part discusses the conditions under which increasing
yield variability may be a problem for farmers, poor consumers, and gov-
ernments; the consequences for these interest groups; and reviews the im-
plications for agricultural policy and agricultural research priorities.

Part I. Evidence on Patterns of Changing Yield Variability
In chapter 2, Hazell introduces a variance decomposition procedure
and examines the evidence for changing patterns of variability in cereal
production for the world (excluding China because of data difficulties) and
for the 34 most important cereal-producing countries. While the results
vary by crop and country, it seems clear that there has been a general pat-
tern of increase in the variability of total cereal production since the early
1970s. Hazell finds that this increase is predominantly due to increases in






Introduction 3


the variances and covariances of yield. This finding justifies the focus of
this book on yield variability in that it is the major variable of interest in
comprehending production variability. Perhaps the most important con-
tribution of chapter 2 is the identification of yield covariances, and partic-
ularly increases in yield correlations between countries and crops, as a ma-
jor source of increase in the variability of world cereal production. It is,
therefore, as important to understand the causal factors underlying these
increases in correlation as it is to understand those underlying changes in
yield variances within crops and countries.
Additional insights into changing patterns of yield variability can be
obtained by examining individual countries in more depth. This has the
added advantage of allowing the analysis to be carried out at a more disag-
gregated regional level. Chapters 3 and 4 are devoted to analysis of chang-
ing patterns of variability in cereal production in the People's Republic of
China and the Soviet Union. These countries were chosen because of their
great importance in world cereal production, and because results for the
other two major cereal-producing countries, that is, India and the United
States, have been presented elsewhere (Hazell 1982, 1984). A similar anal-
ysis is undertaken for Syria in chapter 5. Syria is one of the riskiest cereal-
producing countries in the world and is also one where public policy may
be a most important factor underlying recent changes in variability.
In chapter 6, Walker reports an analysis of changing patterns of vari-
ability in sorghum and pearl millet production in the semi-arid tropical
areas of India where these grains are major staple foods. His analysis at a
district level provides additional evidence for the importance of increasing
interregional yield correlations. Chapters 7 and 8 are devoted to changing
patterns of variability in cereal production in the United Kingdom. Austin
and Arnold (chapter 7) examine data on wheat yields for several centuries,
whereas Webster and Williams (chapter 8) analyze variability in wheat and
barley yields at the individual farm level over 21 recent years. Finally, in
chapter 9, Fischbeck examines yield data for winter wheat and spring bar-
ley in Bavaria since 1950. During this period average yields more than dou-
bled, and there is an excellent data base for examining the associated
changes in variability.
In several of these chapters the authors use time-series data and vari-
ance decomposition methods to attempt to identify changing patterns in
cereal yield variability. While the authors are able to identify the important
components of change in yield variability and suggest some important hy-
potheses about why these changes may be occurring, they are unable to
identify cause and effect relationships. This is an important deficiency be-
cause, without such an understanding, it is difficult to establish what kinds
of interventions will be most appropriate and effective. The task of identi-
fying causal relationships is taken up in parts II and III. However, before





4 Variability in Grain Yields


overviewing these, it is useful to introduce a conceptual scheme of sources
of crop yield variability and in this way indicate how the chapters in these
parts can be related.

A Conceptual Scheme of Sources of Crop Yield Variability
The variability observed in cereal yield results from the interaction of
many factors-some emerging from the physical environment, such as
those related to climate; some from the economic and political environ-
ment, such as prices and access to inputs; some from the intervention of
farm decision makers themselves, such as choice of levels of factors of pro-
duction (e.g., fertilizers, pesticides) and of other aspects of technique
(e.g., varieties, mechanization). While any simple conceptualization
surely cannot do justice to the complexity of such matters, an attempt is
made here to sketch a framework into which much of the material in this
book can be integrated. The treatment follows that of Byerlee and Ander-
son (1969) and Anderson, Dillon and Hardaker (1977, p. 174).
For brevity, symbols are introduced for prices (P), per hectare mea-
sure of yield (Y), fertilizer (F), irrigation (I), variety or cultivar (V), soil
nutrients (S), rainfall (R), other stochastic climate variables (C), mechani-
zation (M), and other controlled inputs such as pesticides (Z). In an indus-
trial economy, some factors can reasonably be regarded as totally control-
lable (V, F, M, Z, and possibly I) and not, in themselves, as sources of risk
or unpredictable variability. Over time, however, especially in response to
changing P, decision makers may choose different levels of such controlla-
ble inputs so that consequent yields will vary and will show up as time-
series variability that, in fact, is not risk per se to the extent that it is pre-
dictable and controlled.
The defined variables can be combined in a functional relationship in
which the productive factors can be ordered from the most to the least
predictable/controllable. This speculative exercise is attempted for the
context of many developing countries where access to fertilizer is often un-
predictable from season to season, desired quantities frequently cannot be
acquired, and irrigation water supplies are highly unreliable, either from
storage or via discontinuous supplies of electricity or fuel for pumps:
Y =f(V, M, F, Z, II S, P, C, R). (1.1)
The first five variables are technological variables more or less under
the control of farmers, and the next four variables are environmental and
essentially uncontrollable from the farmer's point of view. Part II of the
book deals primarily with the first variable, V, while part III deals primar-
ily with F, Z, and I.
There are important interactions among most of the explanatory vari-
ables so that much important complexity is assumed away if any of them





Introduction 5


are examined singly, as is done at several points in this volume. The diffi-
culty is that attempts to grapple with several or all factors simultaneously
falter on impracticalities and cognitive dissonance, if not contemporary
analytical impossibility. Either exceptionally rich data sets that may admit
detailed econometric investigation using methods that have been sug-
gested by Just and Pope (1978) and Antle (1983c), or exceptionally detailed
bioeconomic simulation models (Dent and Anderson 1971, Dent and
Blackie 1979), would be required for such an ambitious task.
It is plausible that multifactorial investigation of some of the factors
may provide greater insight than presently is readily available to some of
the interactive effects of several variables. There is mounting evidence on
the pervasive interactions between V (Simmonds 1962, 1981) and other
factors such as fertilizer (F) (Roumasset et al., ch. 17), pesticides (Z)
(Carlson, ch. 19; Hanus and Schoop, ch. 21), irrigation (I) (Pandey, ch.
18), soil nutrients (S) (Hanus and Schoop, ch. 21), stochastic climatic vari-
ables (C) (Coffman and Hargrove, ch. 11; Pfeiffer and Braun, ch. 13), and
rainfall (R) (e.g., Thompson, 1975; Parry and Carter 1985; Walker, ch. 6;
Duvick, ch. 12; Pham et al., ch. 15; Witcombe, ch. 16; French and
Headley, ch. 22). In the considerable relevant literature, and especially in
part II, such effects are described as genotype-environment (GE)
interactions.

Part II. Plant Breeding and Yield Variability
While plant breeders are primarily interested in maximizing yield re-
sponsiveness to environment, they have for some years also been concerned
with the stability of the genotypes that they select. In chapter 10, Arnold
and Austin provide an overview of the methods that plant breeders use in
selecting stable and adaptable varieties. They also discuss the limitations
of commonly used methods and make some suggestions for improvements
in the future.
There are numerous characteristics of a plant that determine its sta-
bility properties. These aspects are examined in detail by Coffman and
Hargrove in chapter 11 for the case of rice. Studies of this kind are useful
in identifying those agronomic features of plants that need to be empha-
sized by breeders if stability is to be enhanced. The fact that some charac-
teristics that are favorable for yield responsiveness are also unfavorable for
stability suggests that there may be real trade-offs between breeding for
higher average yield and breeding for stability.
One issue raised by Coffman and Hargrove is that the widespread
adoption of a few rice varieties throughout much of Asia has narrowed the
genetic diversity of cultivated varieties. Most popular semidwarf rice vari-
eties have similar cytoplasm, and virtually all have the same dwarfing
genes. The common ancestry of many modern rice varieties does not neces-





6 Variability in Grain Yields


sarily imply that they increase production variability, but their common
susceptibility to the same kinds of pest and weather stresses may mean that
yields will tend to be more covariate across regions. Duvick addresses this
same issue in chapter 12 for maize yields in the U.S. Corn Belt. He con-
cludes though that observed increases in interregional maize yields in the
United States are probably due more to changing weather patterns and
more homogenous cultural practices than they are to narrowing of the ge-
netic base.
Experimental data from varietal trials in numerous (mainly develop-
ing) countries are analyzed in chapters 13 through 16 to test the stability of
modern cereal varieties. These chapters illustrate the common methods
used by breeders for screening for stability and the superiority of modern
varieties of wheat (chapters 13 and 14), maize (chapter 15), and pearl
millet (chapter 16). The results demonstrate that high-yielding varieties
typically have at least the same yield stability/adaptability as local vari-
eties, but on a higher yield plateau. They are also more responsive to favor-
able environments and higher levels of inputs. There is also evidence that
the modern varieties yield at least as well as traditional varieties in poor
environments, although the range of environments tested may not be suffi-
ciently wide to encompass the farming conditions of some of the most mar-
ginal farmers in the world.

Part III. Input Management and Yield Variability
The management of other factors of production can be at least as im-
portant as genotype in determining the variability of yields in farmers'
fields, particularly when the potentially large interactions between geno-
type and environment, or in this case the farmer-manipulated environ-
ment, are allowed for. Chapters 17 through 19 provide reviews of the litera-
ture on the relationship between yield variability and the use of fertilizer
(chapter 17), irrigation (chapter 18), and pesticides (chapter 19). These
reviews suggest that the proper management of these inputs can have a
stabilizing effect on yield, and particularly so for modern varieties.
The remaining chapters in part III address some of the joint effects of
input management and shed some light on some of the interaction effects.
In chapter 20, Flinn and Garrity discuss the relationships between nitro-
gen use, pesticide use, and agronomic practice in contributing to yield sta-
bility for rice in East Asia. Hanus and Schoop describe the relationship
between fertilizer and fungicide use on barley in West Germany in chapter
21. In chapter 22, French and Headley use regression analysis to analyze
causal factors behind changing yield variability and interregional yield cor-
relations for maize in the United States. They are able to separate the ef-
fects of weather variables and technology factors and in this way challenge
some of Hazell's (1984a) conclusions about change in correlation effects.





Introduction 7


Part IV. Impacts of Yield Variability and Implications for Policy
Parts I, II, and III are largely concerned with whether yield variability
and yield correlations are increasing and, if so, why. In the remaining part
the question of whether increasing yield variability is important for deci-
sion makers, particularly for farmers and poor consumers, is addressed.
Measured from the farmers' point of view, high or increasing levels of
yield variability need not be a problem. There are several reasons for this.
First, if farmers were able successfully to anticipate yield outcomes each
year, and thereby make appropriate adjustments to their resource alloca-
tion decisions in order to maximize returns each year, yield variability
would not necessarily involve any economic cost. It is only when yield
changes cannot be perfectly anticipated at the time of making resource
allocative decisions that the possibility of resource misallocation and eco-
nomic loss arises. This begs the question of how accurately farmers and
others can forecast yield variations from year to year. This subject is taken
up in chapter 23 by Bindlish, Barker, and Mount. Using district-level data
from India, they explicitly model the expectations behavior of rice farmers
and then calculate production risks over time as the difference between
actual yield outcome and expected yield each year.
Second, from a decision-making point of view, measures of yield dis-
persion in themselves do not indicate much about riskiness except under
rather extreme probability distributional assumptions, such as normality.
For more general risk analysis it is necessary to employ more general con-
cepts of stochastic efficiency, preferably using the economic returns from
the crop rather than simply yield. Such an analysis is developed and illus-
trated in chapter 24 by Anderson, Findlay, and Wan for an Australian
wheat farm.
Third, yield variability may not be important for farmers because
yield risks are only some of many faced by farmers. For example, even for a
single crop, yield variations could be compensated by price variations, pro-
viding the latter move in the opposite direction. Further, total household
income usually comprises the returns from several crops and livestock ac-
tivities, most of which are not perfectly correlated, as well as nonfarm
sources of income, such as off-farm earnings. In chapter 25, Walker exam-
ines conditions for Indian farmers in the semi-arid tropics under which
reduced yield risk for a major cereal crop might have a stabilizing effect on
total family income.
Significant yield risk could have an important bearing on farmers' de-
cision to adopt new technology. To the extent that they also destabilize
national food supplies and prices, yield risks may also be detrimental to
poor consumers. High production years for major cereals should, in princi-
ple, be good for poor consumers. They should gain from more plentiful





8 Variability in Grain Yields


food supplies, from lower prices, and perhaps from increased agricultural
employment. The opposite might be expected in low production years.
However, since consumers typically purchase a number of different food-
stuffs, shortages or high prices for one may be offset by substituting other
foods of which supplies are more plentiful or prices are lower. There is a
surprising lack of evidence on the relationship between the variability of
individual food supplies and the variability of the income and nutritional
intake of the poor. Sahn and von Braun (chapter 26) muster much of the
evidence available on this issue.
The book concludes with a synthesis of the issues in chapter 27, in
which the editors attempt to summarize the main findings and draw impli-
cations for changes in agricultural research priorities and for agricultural
policy. A particularly important consideration is the cost-effectiveness of
different approaches. Should, for example, plant breeders be encouraged
to give greater emphasis to crop stability rather than increased yield re-
sponsiveness if this would lead to a significant loss in the growth of yield
over time? It may be more cost-effective to pursue alternative approaches
to the problem, such as improved agronomic practices, or policy interven-
tions that mitigate the consequences of yield variability.

Measures of Yield Variability
In interdisciplinary studies of this kind, difficulties can easily arise
when different concepts and approaches are encountered in addressing a
common problem. This problem is particularly acute when confronted by
the very different concepts of yield variability held by plant breeders and
economists. These differences do not invalidate the approach of either dis-
cipline, but rather they reflect differences in the clientele that breeders and
economists seek to assist, and differences in the sources of variability that
exist in yield data measured at different levels of aggregation.
Plant breeders are primarily concerned with developing higher yield-
ing varieties that also provide acceptable levels of risk to farmers. As such,
they tend to focus on reducing "downside" yield risks and on selecting va-
rieties that will perform well for farmers under diverse conditions. Their
analyses are based on yield data from experimental plots or farmer fields,
and, because most stability tests on specific genotypes are typically carried
out for only two or three years, there is a strong presumption that stability
across different locations (or more correctly, "adaptability") is a good
proxy for "stability" over time at specific locations. The evidence for this is
mixed; whereas supportive evidence is offered in chapters 16 and 20, con-
trary evidence is to be found in Watson and Anderson (1977), and Evenson
et al. (1979).
Economists tend to be more concerned with national food problems





Introduction 9


and, at this level of analysis, both high and low yields can present prob-
lems. Low national yields may result in food shortages, high food prices
(especially for the poor), and balance of payments problems, whereas high
yields may result in unacceptably low prices for farmers and excessive food
stocks, perhaps publicly owned. Both upside and downside risks can,
therefore, seem significant to economists. In addition, they often work
with regional or national yield data that embody more diverse sources in
their variability. The latter include covariance relations between more mi-
crolevel units of observation (e.g., fields or farms) which are usually ig-
nored by plant breeders (ch. 2). A major limitation of using aggregate yield
data is that it is much more difficult to isolate the impact of improved tech-
nologies (especially improved varieties) on yield variability.
There are usually differences in the types of yield distributions ob-
served at the farm and national levels. For example, experimental plot and
farm yields are often skewed (Day 1965, Anderson 1973), whereas national
yields tend to be more symmetric or even normal. This is to be expected
since the latter are a weighted sum of many individual farm yields, and
many of these are only weakly correlated.
When yield distributions are symmetric, measures of variability that
focus only on downside risks often give results similar to those based on
general measures of dispersion. The variance and the coefficient of varia-
tion (cv) of yields are then satisfactory measures of variability for a wide
range of purposes. If yield distributions are skewed, however, other mea-
sures of variability, such as the semi-variance or, indeed, part or all of the
cumulative distribution function itself may be more relevant.
Several measures of variability are used in this volume, reflecting the
purpose and preferences of individual authors. All authors choose to rep-
resent central tendency by the arithmetic mean (m). Dispersion is usually
measured by the variance (var) or the standard deviation (sd), in most
cases after trends have been removed. Procedures for decomposing vari-
ances of aggregate production into their component parts are introduced
in chapter 2. Many authors work with the standardized (or dimensionless)
coefficient of variation (cv) defined as cv = sd/m.
The sampling distribution for the cv is complex but tests are made in
this volume by appeal to normality in the parent population and estimation
of its variance (Kendall and Stewart 1977) as
var(cv) = c2(1 + 2c2)/(2n),
where c is the coefficient of variation in the parent population and n is the
sample size.
Presuming that some attempt at testing for statistical significance is
better than none, given the potentially controversial interpretation of
changing variability, the editors have intruded significance tests (usually





10 Variability in Grain Yields

"two tailed" at the conventional, albeit arbitrary, 5 percent level) for many
of the variances and cvs reported by several authors. Tests for changes in
variances involve standard two-tailed F ratio tests. Tests for cvs involve a
comparison of a cv for one period, cvl based on n observations, with that
for a second, cv2 and n2. The ad hoc procedure is adopted of assuming
(a) that the "parent" population cv is approximated by cvl, and (b) that
the estimates of the cvs are statistically independent, so that a standard
error of the difference between cvs is given by
D = c{[(1 + 2c2)/2](1/ni + 1/n2)}05,
and the approximately standard normal test statistic by
z = (cv2 cvl)/D.






PART I

Evidence on Patterns of
Changing Yield Variability








2 Changing Patterns of Variability
in World Cereal Production

PETER B. R. HAZELL










Total cereal production for the world (excluding China for data reasons)
grew at an average yearly rate of 2.7 percent between 1960/61 and
1982/83. The average yield during this period grew by 2.0 percent per
year, and the total gross cropped area allocated to cereals by 0.7 percent
per year.
As figure 2.1 shows, this growth in aggregate production has been
accompanied by a seemingly widening band of variability with annual
downswings of more than 50 Mt. (The unit here is for million metric tons,
106t; t is used throughout this volume to designate metric tons.) An en-
couraging feature is that successive trough low points in production are
monotonically increasing over time.
The calculation of 10-year moving averages for the mean and, after
linear detrending, the standard deviation and coefficient of variation (cv)
of production resulted in the data shown in table 2.1.
Absolute variability around trend has increased substantially and sig-
nificantly. In fact, it has increased faster than average production as
shown by the increase in the cv which, although seemingly substantial, is
not significant in a statistical sense (comparing successive decades). Note
that cv measured in this way peaked in the decade ending in the early
1980s, with some recent gain in stability after the turbulent 1970s.
An increasing cv means that the probability of a major shortfall, say
5 percent, below trend changes in a similar manner (table 2.1).1 This is
moderated by a change (again not statistically significant) from negative to

1. Let detrended production in year t be denoted by Q, = Q + e, where Q is the period
mean and e, is the deviation from the mean that year. Then the probability of a shortfall of 5
percent or more below trend is derived fromPr{Q + e, < 0.95 Q} = Pr{e,/o, < -0.05 Q/
a,} where o, is the standard deviation of e,. Assuming e, is approximately normally distrib-
uted, the desired probability can be obtained from tables of the cumulative standard normal
distribution.






14 Evidence on Patterns of Changing Yield Variability


FIGURE 2.1 World cereal production, 1960/61-1982/83

Million metric tons
1,300 r-


1,200





1,100





1,000





900





800


1960/61


1965/66


1970/71


1975/76


1 1
1980/81 1982/83


positive skewness (table 2.1). This shift implies that there is now less risk of
extreme catastrophe at the global level, but more frequent falling below
trend production.


Changes in World Cereal Production

To examine these changes more fully it is instructive to compare vari-
ability for important producing countries in two periods. U.S. Department
of Agriculture (USDA) data by country and crop (area and yield) were as-





Variability in World Cereal Production 15


TABLE 2.1 Variability of world cereal production around linear trend:
1960/61-1982/83

Average Standard
Production Deviation Probability of a
(million (million Coefficient 5 Percent Shortfall
Period metric tons) metric tons) of Variation Skewness below Trend

Decade
Beginning
1960/61 819 24.3 0.030 -0.31 4.65
1961/62 837 20.7 0.025 -0.39 2.17
1962/63 867 22.4 0.026 -0.64 2.62
1963/64 890 24.1 0.027 -0.21 3.22
1964/65 923 26.8 0.030 -0.32 4.18
1965/66 946 31.2 0.034 -0.32 6.55
1966/67 972 32.5 0.034 -0.39 6.68
1967/68 1,001 34.3 0.035 -0.47 7.21
1968/69 1,026 34.4 0.035 -0.27 6.81
1969/70 1,057 40.0 0.037 0.05 9.01
1970/71 1,081 40.0 0.037 0.34 8.85
1971/72 1,108 40.1 0.036 0.37 8.38
1972/73 1,132 39.5 0.035 0.46 7.64
1973/74 1,159 38.5 0.033 0.37 6.68

Note: Does not include China.


sembled for years 1960/61 to 1982/83 and split into two periods: 1960/61
to 1970/71 and 1971/72 to 1982/83. This split

a. corresponds to speculated changes in yield variability that are possibly
associated with the green revolution, usually regarded as occurring
around 1970 in many developing countries;
b. corresponds broadly with the dramatic increase in price variability in
the early 1970s;
c. more pragmatically, gives roughly equal sample sizes.

Data for each of the 34 more important cereal-producing countries
were processed individually, and all other countries (excluding China)
were pooled into a "rest of the world" category. The area and yield of each
crop in each country were detrended separately, and production derived as
the product of detrended area and detrended yield. Quadratic trends were
used together with a generalized least squares estimating technique (Ha-
zell 1984).
Table 2.2 summarizes the results by crop at the global level. Average
production of total cereals increased by 37 percent between periods, and
the cv increased by 21 percent-from 0.028 to 0.034. This increase is not
significant at the 5 percent significance level.
















TABLE 2.2 Changes in the mean and variability of world cereal production: 1960/61-1970/71 to 1971/72-1982/83

Average Production Coefficient of Variation
of Production F Ratios
First Second
Period Period Change First Second Change Area
Crop (million metric tons) (percent) Period Period (percent) Production Sown Yield

Wheat 253 353 39.3 0.054 0.048 -11.5 1.52 0.34 1.64
Maize 210 317 51.0 0.033 0.044 34.0 4.08* 1.65 4.17*
Rice 120 155 29.2 0.039 0.038 -4.3 1.52 2.45 0.88
Barley 95 150 58.5 0.048 0.075 55.9 6.18* 3.13 3.28
Millets 20 21 8.2 0.079 0.077 -3.2 1.10 2.17 0.67
Sorghums 40 53 32.7 0.052 0.057 9.2 2.10 0.78 2.32
Oats 49 48 -2.9 0.113 0.054 -52.6 0.21* 0.07* 4.42*
Other cereals 41 35 -14.9 0.046 0.093 103.7* 2.94 0.36 3.61*
Total cereals 829 1,134 36.7 0.028 0.034 21.2 2.75 2.18 2.73

Note: Does not include China.
*An asterisk denotes that the cv/variance in the second period is significantly different from that of the first period at the 5 percent level (two-tailed
test).






Variability in World Cereal Production 17


Table 2.2 also shows important differences by crop. The cv for wheat
and rice has diminished, notwithstanding these crops being the flag bear-
ers of the green revolution. But the cv for coarse grains (maize, barley, and
sorghum) has increased. The F ratios indicate the importance of increased
yield variability rather than of variability in sown area.
Table 2.3 shows results for total cereals by major countries. The cv of
production has increased in 14 and diminished in 20 countries. The results
by crop within countries are also diverse but often feature high absolute
levels of variability (Hazell 1985a). There is little observable relationship
between a country's performance in increasing cereal production and the
changes in production variability. The correlation across countries be-
tween the percentage change in average production and the change in the
coefficient of variation is -0.15, which is not significantly different from
zero at the 5 percent level.
Table 2.4 summarizes the changes in the mean and cv of yields of
major cereals by country. In the following counts of trends in pairs of cvs in
table 2.4, "total cereals" is regarded as another "crop," but the "rest of
the world" is not regarded as another country. Inspection of the pairs re-
veals no clear pattern within crops, there being about equal numbers of
increases and decreases in cv for each (except wheat and maize, if signifi-
cance is ignored), or within countries (except the United States where all
crops are more variable and Pakistan where all are less variable)-table
2.5. The situation seems similarly unclear with respect to such variables as
size of country, the importance of a country as a producer of the particular
cereal, and the rate of technological progress.


A Decomposition of the Components of Change
Method of Analysis
To analyze the components of change in the mean and variance of
world cereal production, a variance decomposition procedure is used (Ha-
zell 1982). Let Q denote production, A the area sown, and Y yields. Also,
letting subscripts i and j denote crops, and h and k denote countries, total
cereal production for the world is Q = EhEjAhjYhj. Average production is
E(Q) = E E(AhjYhj), (2.1)
h j
and the variance of production is


V(Q) = E E E cov(AhiYhi, AkjYkj).
h k i j


(2.2)

















TABLE 2.3 Changes in the mean and variability of total cereal production by major countries: 1960/61-1970/71 to 1971/72-1982/83

Average Production Coefficient of Variation
of Production F Ratios
First Second
Period Period Change First Second Change Area
Country (million metric tons) (percent) Period Period (percent) Production Sown Yield


United States
U.S.S.R.
India
France
Canada
Brazil
Argentina
Germany, F.R.
Indonesia
Turkey
Australia
Romania
United Kingdom
Italy
Mexico
Yugoslavia
Spain


182.0
138.4
74.8
27.5
30.0
16.5
17.2
16.0
13.5
12.9
12.6
11.6
12.4
14.2
10.5
11.4
9.3


265.0
181.0
104.0
41.0
40.0
26.1
23.8
22.2
20.3
18.4
17.4
17.4
16.8
16.7
15.6
15.0
13.7


0.068
0.122
0.077
0.060
0.170
0.051
0.118
0.091
0.061
0.071
0.195
0.109
0.087
0.034
0.070
0.100
0.081


0.066
0.143
0.054
0.092
0.107
0.089
0.140
0.060
0.052
0.097
0.232
0.099
0.083
0.057
0.111
0.052
0.137


-2.8
17.3
-29.2
52.9
-37.6
70.9*
19.0
-34.7
-15.4
37.5
18.5
-9.2
-4.5
65.1*
57.9*
-48.1
71.3*


1.97
2.35
0.97
5.26*
0.69
7.25*
2.72
0.82
1.62
3.80*
2.66
1.83
1.66
3.72*
5.58*
0.47
6.37*


1.24 8.23*
1.28 1.69
0.65 0.92
1.58 4.30*
0.22* 0.44
4.30* 2.47
1.04 2.12
3.24 0.59
0.74 2.89
3.98* 3.45
1.65 1.67
0.80 2.17
0.33 1.77
5.50* 0.66
3.99* 3.40
0.74 0.57
0.68 7.73*








Thailand
Pakistan
Poland
Bangladesh
Hungary
South Africa
Japan
Czechoslovakia
Nigeria
Bulgaria
Vietnam
German D.R.
Philippines
Iran
Burma
South Korea
Egypt
Rest of world
(excluding China)
Total world


54.9
71.9
56.9
22.0
65.0
60.0
-21.8
56.5
9.0
41.9
21.9
55.2
63.1
31.3
32.6
18.3
22.8


98.5 117.7


(excluding China) 829.2 1133.9


0.078
0.102
0.092
0.072
0.101
0.204
0.060
0.117
0.117
0.103
0.089
0.113
0.055
0.083
0.099
0.060
0.050


0.084
0.032
0.093
0.050
0.061
0.197
0.093
0.075
0.051
0.075
0.056
0.064
0.054
0.092
0.077
0.108
0.027


19.6 0.032 0.028

36.7 0.028 0.034


7.4
-69.2
1.0
-30.2
-40.0
-3.3
54.9
-35.7
-56.7
-27.3
-37.8
-43.3
-1.5
11.4
-22.3
80.4*
-46.1

-12.2

21.2


2.76
0.28
2.52
0.72
0.98
2.40
1.45
1.01
0.22*
1.05
0.58
0.78
2.56
2.15
1.06
4.62*
0.44

1.10

2.75


3.00 2.01
0.44 0.27*
0.12* 4.00*
0.20* 1.05
0.35 1.39
2.63 1.99
4.27* 1.58
0.07* 1.62
0.16* 0.14*
3.55* 0.72
1.26 0.41
1.18 0.65
6.87* 0.77
1.00 3.88*
0.45 1.77
0.96 7.76*
0.23* 0.37

0.47 0.75

2.18 2.73


*An asterisk denotes that the cv/variance in the second period is significantly different from that of the first period at the 5 percent level (two-tailed
test).






20 Evidence on Patterns of Changing Yield Variability


TABLE 2.4 Mean yields and coefficients of variation by crop and country:
1960/61-1970/71 and 1971/72-1982/83

Wheat Maize Rice

Mean Mean Mean
Country (t/ha) cv (t/ha) cv (t/ha) cv


United States
First period
Second period
U.S.S.R.
First period
Second period
India
First period
Second period
Canada
First period
Second period
France
First period
Second period
Indonesia
First period
Second period
Brazil
First period
Second period
Argentina
First period
Second period
Mexico
First period
Second period
Turkey
First period
Second period
Australia
First period
Second period
Thailand
First period
Second period
Germany, F.R.
First period
Second period
Bangladesh
First period
Second period


1.80 0.048 4.44
2.16 0.067 5.96

1.14 0.162 2.45
1.52 0.137 3.00

0.92 0.106 1.02
1.42 0.059 1.05

1.50 0.178 4.81
1.84 0.097 5.33

3.11 0.079 3.84
4.52 0.086 5.00


0.066 3.32 0.031
0.096 3.68 0.064*

0.109 1.81 0.041
0.113 2.54 0.038

0.084 1.01 0.080
0.078 1.19 0.079

0.078 -
0.077 -

0.193 2.61 0.117
0.105 2.21 0.191*


- 0.96 0.050 1.43 0.025
- 1.25 0.042 1.95 0.028


0.77 0.168 1.33
0.86 0.252 1.56

1.34 0.185 1.99
1.55 0.087 2.89

2.23 0.074 1.01
3.32 0.105 1.17

0.96 0.074 1.42
1.39 0.103 2.03

1.23 0.158 2.16
1.24 0.216 2.82


0.047 1.02 0.069
0.084* 0.96 0.052

0.090 2.38 0.050
0.149* 2.26 0.074

0.082 1.60 0.066
0.113 1.97 0.075

0.086 2.44 0.097
0.067 2.81 0.065

0.099 4.80 0.102
0.087 4.27 0.109


- 2.11 0.110 1.09 0.072
- 2.05 0.174 1.18 0.067


3.59 0.091 4.01
4.70 0.057 5.58

0.73 0.152 -
1.49 0.086 -


0.077 -
0.088 -

- 1.12 0.042
- 1.23 0.040






Variability in World Cereal Production 21


Barley Millet Sorghum Total Cereals

Mean Mean Mean Mean
(t/ha) cv (t/ha) cv (t/ha) cv (t/ha) cv


I


2.06 0.050 3.01 0.065 2.91
2.51 0.086* 3.45 0.120* 3.81

1.30 0.147 0.80 0.189 1.20
1.54 0.177 0.79 0.341* 1.52

3.89 0.089 0.43 0.137 0.50 0.083 0.78
1.07 0.090 0.51 0.108 0.63 0.104 1.02

1.82 0.121 1.69
2.32 0.072 2.14

2.94 0.086 1.32 2.82 0.131 3.04
3.78 0.085 2.08 4.14 0.107 4.28

- 1.30
- 1.79

0.86 0.134 1.20
1.14 0.190 2.14 0.108 1.30

1.15 0.165 1.13 0.094 1.95 0.148 1.56
1.32 0.128 1.19 0.109 2.79 0.124 2.09

0.86 0.058 2.31 0.097 1.19
1.29 0.197* 2.60 0.155* 1.50

1.22 0.090 1.28 0.059 1.06
1.71 0.098 1.39 0.127* 1.50

1.15 0.156 0.97 0.170 1.58 0.192 1.20
1.19 0.198 0.94 0.161 1.92 0.124 1.27

- 1.53 1.17
- 1.41 0.373 1.29

3.24 0.106 3.23
4.13 0.042* 4.19

0.56 0.094 0.73 0.060 1.11
0.61 0.058 0.77 0.071 1.24


0.040
0.088*

0.132
0.134

0.062
0.046

0.141
0.073

0.056
0.082

0.021
0.025

0.046
0.067

0.086
0.094

0.055
0.081

0.069
0.091

0.152
0.185

0.064
0.081

0.088
0.052

0.043
0.039






22 Evidence on Patterns of Changing Yield Variability


TABLE 2.4 Continued

Wheat Maize Rice

Mean Mean Mean
Country (t/ha) cv (t/ha) cv (t/ha) cv


Poland
First period
Second period
Romania
First period
Second period
United Kingdom
First period
Second period
Italy
First period
Second period
Pakistan
First period
Second period
South Africa
First period
Second period
Yugoslavia
First period
Second period
Burma
First period
Second period
Japan
First period
Second period
Vietnam
First period
Second period
Hungary
First period
Second period
Spain
First period
Second period
Philippines
First period
Second period
Nigeria
First period
Second period
Czechoslovakia
First period
Second period


2.10 0.051
2.88 0.091*

1.54 0.169
2.49 0.105

3.95 0.078
4.96 0.075

2.09 0.065
2.58 0.051

0.90 0.098
1.39 0.041

0.65 0.176
1.01 0.119

2.04 0.118
3.10 0.091

0.49 0.249
0.76 0.083*

2.47 0.196
2.84 0.105


2.50 0.206
3.75 0.186


1.96 0.091 1.78
3.08 0.098 1.53


3.66 0.085 3.44
6.19 0.028* 3.62

1.06 0.082 1.08
1.23 0.031* 1.57


0.156
0.177


0.098
0.127

0.119
0.039*


1.38 0.241
2.09 0.219


2.62 0.111 2.39
3.87 0.058 2.62

0.49 0.325 1.03
0.83 0.100* 1.32

2.62 0.063 3.69
2.75 0.057 4.12


- 1.10 1.26
- 1.14 0.065 1.33


2.09 0.125
3.76 0.106

1.11 0.106
1.57 0.150


2.89 0.084 1.20
4.76 0.067 1.25

2.68 0.189 3.95
4.26 0.062* 4.22


- 0.70 0.043 0.88
- 0.88 0.035 1.22

- 0.90 0.148 1.20
- 0.85 0.035* 1.26


2.62 0.077
3.87 0.091


0.107
0.114

0.066
0.075

0.040
0.057

0.080
0.056


0.043
0.038

0.078
0.048

0.085
0.040


2.91 0.146
3.91 0.192






Variability in World Cereal Production 23


Barley Millet Sorghum Total Cereals

Mean Mean Mean Mean
(t/ha) cv (t/ha) cv (t/ha) cv (t/ha) cv


2.09 0.085
2.81 0.074

1.81 0.100
2.73 0.103

3.52 0.059
4.10 0.061

1.44 0.067
2.54 0.071

0.61 0.076
0.69 0.036

0.57 0.248
0.89 0.228

1.54 0.107
2.13 0.092


-- -


2.69 0.177
2.99 0.084


- -

1.99 0.110
3.10 0.110

1.48 0.102
1.88 0.168*


2.53 0.121
3.52 0.076


0.96 0.064 1.97
0.91 0.146* 2.68

- 1.74
- 1.41 2.79

- 3.55
- 4.39

- 3.39 2.33
- 4.37 0.096 3.25

0.45 0.059 0.52 0.070 0.87
0.49 0.048 0.59 0.054 1.31

- 0.89 0.229 1.08
- 1.86 0.269 1.66

- 2.59 0.081 2.23
- 2.28 0.105 3.34

0.71 0.349 1.00
0.53 0.475 1.28

1.63 0.091 3.42
0.95 4.01

- 1.26
- 1.32

- 2.31
- 4.04

- 2.30 0.222 1.29
- 4.39 0.114 1.87

- 0.81
- 1.05

0.61 0.175 0.75 0.087 0.71
0.60 0.059* 0.62 0.043 0.66

- 2.45
- 3.60


0.062
0.092

0.087
0.080

0.061
0.066

0.039
0.023

0.086
0.030*

0.198
0.182

0.094
0.047

0.064
0.066

0.054
0.058

0.082
0.050

0.072
0.048

0.069
0.132*

0.059
0.040

0.093
0.038*

0.084
0.073


j






24 Evidence on Patterns of Changing Yield Variability


TABLE 2.4 Continued

Wheat Maize Rice

Mean Mean Mean
Country (t/ha) cv (t/ha) cv (t/ha) cv


German D.R.
First period
Second period
Iran
First period
Second period
Bulgaria
First period
Second period
South Korea
First period
Second period
Egypt
First period
Second period
Rest of world
(excluding China)
First period
Second period
Total world
(excluding China)
First period
Second period


3.39 0.095 2.03 -
4.19 0.060 3.19 -

0.84 0.055 1.09 0.090 2.22
0.99 0.054 1.33 0.146* 2.46

2.25 0.152 3.04 0.135 2.16
3.67 0.064 4.05 0.111 2.39

2.11 0.080 0.97 0.121 3.03
2.53 0.156* 2.87 0.175 3.95

2.58 0.102 3.14 0.140 3.46
3.30 0.043 3.77 0.035* 3.64


1.09 0.063 1.15 0.042 1.22
1.32 0.048 1.34 0.034 1.47


1.35 0.050 2.27 0.030 1.25
1.78 0.049 3.02 0.046 1.46


Note: The first period is from 1960/61 to 1970/71; the second period is from 1971/72 to
1982/83.
*An asterisk denotes that the cv in the second period is significantly different from that of the
first period at the 5 percent level (two-tailed test).



The variance can be expanded as


V(Q) = EE V(AhjY,) + E E cov(Ahihi,, AhjYhj)
h j h i*j j


(2.3)


(sum of individual
crop variances within
countries)


(sum of intercrop covariances
within countries)


+ E cov(AhjYh, AjYkj) + E EE cov(Ahighi, AkjYkj).
j h*k k h*k k i*j j


(sum of intercountry
covariances within crops)


(sum of covariances between
different crops in different
countries).


0.099
0.092

0.151
0.148

0.080
0.126

0.058
0.038


0.032
0.010


0.033
0.026






Variability in World Cereal Production 25


Barley Millet Sorghum Total Cereals

Mean Mean Mean Mean
(t/ha) cv (t/ha) cv (t/ha) cv (t/ha) cv


3.05 0.150 3.03 0.105
3.88 0.078 3.92 0.065

0.70 0.044 0.87 0.036
0.74 0.131* 1.01 0.061*

2.23 0.121 0.95 2.37 0.103
3.19 0.075 1.00 3.60 0.057

1.84 0.142 0.81 0.121 2.43 0.056
2.32 0.158 1.18 0.132 3.42 0.110*

2.40 0.206 3.76 0.018 3.07 0.045
2.80 0.053* 3.89 0.046* 3.60 0.023


1.44 0.045 0.60 0.044 0.75 0.034 1.20 0.032
1.78 0.068 0.56 0.062 0.82 0.032 1.36 0.025


1.63 0.043 0.53 0.073 1.01 0.040 1.45 0.026
1.97 0.064 0.56 0.058 1.27 0.046 1.85 0.034


Each of the component terms can be expanded as follows:

E(AhjYhj) = AhjYh, + cov(AhjYhj),


(2.4)


and, following Bohrnstedt and Goldberger (1969),

cov(AhiYhi, AkjYkj) = AhiA cov(Yhi, Ykj) + AhiYk, cov(Yhi, Ak) (2.5)
+ YhiAkj Cov(Ahi, Ykj) + YhiYkj cov(Ahi, Akj)

cov(Ahi, Yhi)cov(Akj, Ykj) + R,

where A and Y denote mean area and yield, and R is a residual term con-
sisting of higher order cross moments.





26 Evidence on Patterns of Changing Yield Variability


TABLE 2.5 Summary of trends in coefficients of variation between periods reported
in Table 2.4
Increases Decreases

Number Number Number Number
Crop Measured Significanta Measured Significanta
Wheat 9 2 20 1
Maize 11 3 19 6
Rice 12 2 14 1
Barley 13 4 13 2
Millet 6 3 4 1
Sorghum 6 3 7 0
Total cereals 17 4 17 2
aA significant change at the 5 percent level using the procedure described in chapter 1.


The decomposition analysis partitions the changes in V(Q) and E(Q)
between the first and second periods into constituent parts. This involves
decomposing the changes in each of the terms in equations (2.1) and (2.3)
with the aid of equations (2.4) and (2.5), and then summing the changes in
different components over countries and crops. For a full exposition of this
method, see Hazell (1982).
Using equation (2.4), but dropping crop and country subscripts for
simplicity, average production in the second period is

E(Q2) = A2Y2 + cov(A2, Y2). (2.6)
Each variable in the second period can be expressed as its counterpart
in the first plus the change in the variable between the two. Equation (2.6),
therefore, can be written as

E(Q2) = (A, + AA)(Y1 + AY) + cov(Al, Y1) + A cov(A, Y). (2.7)
The change in average production is then obtained from

AE(Q) = E(Q2) E(Ql) = AiAY + YAA + AAAY + A cov(A, Y).
(2.8)
There are four sources of change in AE(Q). Two parts, AIAY and
YIAA, arise from changes in the mean yield and the mean area. These
"pure" effects arise even in the absence of other sources of change. The
term AAAY is an "interaction" effect, and A cov(A, Y) arises from
changes in the covariability of areas and yields.
The change in the variance of production can be decomposed in an
analogous way. Using equation (2.5), the change in each of the production
variance and covariance terms can be decomposed as in table 2.6. The first






Variability in World Cereal Production 27


TABLE 2.6 Components of change in production covariances


Source of Change

Change in mean yields


Change in mean areas

Change in yield variances
and covariances
Change in area variances
and covariances
Change in area-yield
covariances


Interaction between
changes in mean yields
and mean areas
Interaction between
changes in mean areas
and yield variances
Interaction between
changes in mean yields
and area variances
Interactions between
changes in mean areas
and yields and changes
in area-yield
covariances

Change in residual


Components of Change

,;AliA Cov(Yi, Aj) + AiAYCov(Ai, Ylj) + [YiAYj
+ YI, A Y + AYiA ,] Cov(Ai, Ali)

li,,AACov(Ali, Ylj) + YjAAiCov(Yi, A,) + [A ,,AAj
+ AjAAi + AAiAAj] Cov(Y,;, Yi)

A,,AACov(Y;, Yj)

,Yi YyA Cov(A,, Aj)

A,i YFACov(Y, ,A) + YiAISACov(A;, Yj) [Cov(Ali,
Y,,) + ACov(Ai, Y,)]ACov(Aj, Yj) Cov(AIy, YIj)
ACov(A,, Yi)


AAA YCov(Yi,, Al) + A YAAjCov(A,l, Yj)


[AiAAj, + AAAA AA + AA A] ACov(Y,, Yi)


[FliA ~ + FyjAY, + AyA~ ] ACov(Ai, A)




[YIFAA + AiAY + AAiAYj] ACov(Y,, Aj) + [YFi
AAj + AiA Y + A YAAJ] ACov(Ai, Yj)
ACov(Ai Y,, Aj Yj) sum of the other components


five sources of change are "pure" effects, the next four are interaction ef-
fects that occur because of simultaneous changes in all the constituent
parts, and the final term is a higher order term that is typically small, is of
little importance, and has no ready interpretation.

Components of Change in World Cereal Production
Table 2.7 shows the results from decomposing the changes in average
cereal production for the world. Increases in mean yields account for about
70 percent of the increase in total cereal production, and area expansion
accounts for about 20 percent. Yield improvements are even more impor-
tant in expanding the production of wheat. They are also more important
than area expansion in increasing the production of maize, rice, and mil-















TABLE 2.7 Disaggregation of the components of change in the average of world cereal production:
1960/61-1970/71 to 1971/72-1982/83 (percent)

Components Other Total
of Change Wheat Maize Rice Barley Millet Sorghum Oats Cereals Cereals

Change in mean
yields 80.93 64.21 60.62 39.52 63.64 45.63 -528.08 -179.99 72.40
Change in mean
areas 14.94 28.61 33.64 49.11 44.76 44.42 534.84 220.53 22.36
Change in area-yield
covariances 0.19 0.09 -0.02 0.45 2.96 0.20 15.21 -1.08 0.14
Change in interaction
term 3.95 7.08 5.77 10.93 -11.36 9.76 78.05 60.54 5.10

Contribution of
crop to change in
mean production of
total cereals 32.65 35.18 11.50 18.28 0.55 4.34 -0.47 -2.03 100.00

Note: Does not include China.





Variability in World Cereal Production 29


lets. Area increases, however, are more important in expanding barley
production.
Table 2.8 shows the results from the decomposition of the change in
the variance of world cereal production. The rows in these tables corre-
spond to the four groups of production variances and covariances deline-
ated in equation (2.3). The columns correspond to the 10 sources of change
defined (table 2.6) for a production variance and covariance, though the
four types of interaction terms have been added together. All entries in the
table are expressed as a percentage of the change in the variance of total
cereal production, hence both the rows and the columns sum to 100
percent.
The row sums in table 2.8 show that 35 percent of the increase in the
variance of world cereal production is attributable to increases in the pro-
duction variances of individual crops within countries. Wheat, maize, and
barley account for nearly all of this 35 percent increase. The remaining 65
percent of the increase in the variance of world cereal production is due to
increases in production covariances. Of these, the important ones are be-
tween different crops within and between countries. Changes in intercoun-
try covariances within crops turn out to be a relatively minor component of
the total variance increase (only 5 percent).
The column sums in table 2.8 show that 96 percent of the increase in
the variance of world cereal production is directly attributable to changes
in the variances and covariances of crop yields. Changes in yield variances
within countries account for one-quarter of this increase, and most of this
is attributable to increased yield variances for wheat and maize.
For most of the individual crops, increased yield variances account for
the lion's share of the contribution to the variance of total cereal produc-
tion. For example, when summed over countries, the increased production
variances for wheat account for 7.61 percent of the increase in the variance
of total cereal production. Of this, 100(5.27/7.61) = 69 percent is due to
increased yield variances. Similarly, the yield variance shares for other
crops are: maize 124 percent, rice 36 percent, millets 57 percent, sorghum
77 percent, and total cereals 77 percent.
Changes in yield covariances seem to be more important than changes
in yield variances for the variability of world cereal production. However,
part of the increase in the yield covariances is itself a direct consequence of
increased yield variances. Part of it may also be due to changing correla-
tions between crops and countries. To separate these effects, it is useful to
pursue the decomposition one step further.
Using the same kind of decomposition procedure, the change in a
yield covariance between two periods can be decomposed into three terms
(Hazell 1982). These are, respectively, the changes in the yield variances









W TABLE 2.8 Disaggregation of the components of change in the variance of world cereal production:
1960/61-1970/71 to 1971/72-1982/83 (percent)

Source of Change

Change Change
in Yield in Area
Change Change Variances Variances Change in Change in Change
Variance in Mean in Mean and and Area-Yield Interaction in Row
Component Yields Areas Covariances Covariances Covariances Terms Residual Sums

Crop variances
Wheat 2.06 -2.38 5.27 -0.57 3.57 -0.49 0.15 7.61
Maize 6.67 1.94 17.16 -6.15 -5.01 -1.54 0.73 13.80
Rice 0.11 0.25 0.45 0.12 0.16 0.13 0.05 1.26
Barley 0.43 2.30 1.87 0.86 1.37 4.67 0.96 12.46
Millet 0.01 -0.01 0.04 0.01 0.06 -0.02 0.00 0.07
Sorghum 0.19 0.07 0.57 -0.23 0.12 0.07 -0.05 0.74
Oats 0.83 0.27 0.11 -1.25 -0.54 -1.06 -0.19 -1.85
Other 0.14 -0.15 0.93 -0.14 0.29 -0.77 0.06 0.36

Sum crop variances
within countries 10.44 2.28 26.40 -7.36 0.01 0.99 1.70 34.45
Intercrop covariances
within countries 0.97 4.48 36.68 -0.94 -9.38 1.89 1.65 35.35
Intercountry covariances
within crops 0.09 1.61 11.49 -3.61 -4.40 -0.98 0.49 4.70
Covariances between
different crops in
different countries 2.75 0.85 21.36 19.13 -28.51 6.43 3.55 25.50
Column sums 14.24 9.22 95.93 7.22 -42.28 8.33 7.40 100.00

Note: Does not include China.





Variability in World Cereal Production 31


alone, autonomous changes in the yield correlations, and the interaction
between these two terms.
Application of this decomposition procedure shows that, for the world
as defined, only 4 percent of the 69.5 percent increase in the variance of
total cereal production arising from changes in yield covariances is directly
attributable to changes in yield variances. Some 56 percent of the increase
is attributable to changes in yield correlations alone, and the remaining 40
percent is due to interaction effects. Of the correlation increases, the pre-
dominant ones are between the yields of the same or different crops in dif-
ferent countries. Increases in the intercrop yield covariances within coun-
tries are nearly all attributable to increased yield variances.
Other results in table 2.8 show that changes in area-yield covariances
had an important stabilizing effect on world cereal production; they re-
duced the variance of total cereal production by 42 percent. A further de-
composition of these covariance terms showed that virtually all of this re-
duction can be attributed to a decline in area-yield correlations. Of the
various area-yield correlations, the most important declines are between
the crop yields in one country and the sown areas of the same or different
crops in other countries.

Discussion
Three major components of the change in the variability of world ce-
real production since the 1960s have been identified. These are increased
yield variances; an increase in correlations between the yields of different
crops and countries; and a decline in area-yield correlations, particularly
between the crop yields in one country and the sown areas of the same or
different crops in other countries. Why have these changes occurred? Ad-
ditional insights are to be found elsewhere in this book, but a number of
hypotheses can be offered here.
Some researchers (e.g., Mehra 1981) have argued that the high-yield-
ing varieties (HYVs) associated with the green revolution are more risky,
hence their introduction since the late 1960s is an important source of in-
creasing variability in farm and national yields. Several authors in this vol-
ume suggest, however, that, while high-yielding varieties (HYVs) typically
lead to higher yield variances, their cvs are not generally larger than those
of alternative varieties when grown under trial conditions; in many cases,
HYVs have smaller cvs.
HYVs are selected to be more responsive to good growing condi-
tions-indeed, this is largely how yield increases are obtained. This im-
plies, however, that yields will fluctuate more if the use of inputs varies.
Input variation may have increased with increases in price variability since
the early 1970s, and with the difficulties of reliably supplying inputs in
developing countries to meet the growing demands of the green revolution





32 Evidence on Patterns of Changing Yield Variability


(Jain, Dagg, and Taylor 1986). Input variation thus leads to behaviorally
induced variability in yields, and the phenomenon is perhaps an inevitable
consequence of the modernization of agriculture.
Weather patterns may have become more variable since the 1960s,
although there is little evidence for this (Carter and Parry 1986). If any-
thing, weather in some areas may have become more stable, at least in the
U.S. Corn Belt (French and Headley, ch. 22). National yields may also
have become more variable because increases in the areas cropped have
pushed some cereals into more marginal land (e.g., barley in Syria,
Nguyen, ch. 5). Other sources of increased variability include changes in
policy and land reform (e.g., Tarrant, ch. 4; Nguyen, ch. 5).
Yields have become more correlated across regions within some coun-
tries (e.g., Hazell 1984; Anderson et al. 1988; Walker, ch. 6) and this may
have contributed to increasing the variability of national yields. Possible
causes of this phenomenon, as well as of the increase in yield correlations
between countries, are still speculative but include:
a. the potentially narrowing genetic base-or is it the "too few" vari-
eties problem? Whatever, there does seem to be more susceptibil-
ity to the same weather and pest stresses (Coffman and Hargrove,
ch. 11);
b. varieties that are screened for stability across sites are likely to be
more highly correlated across sites too;
c. more homogeneous cultural practices (Duvick, ch. 12);
d. yield variability induced by input variations is also likely to be
more covariate (e.g., fertilizer application is adjusted similarly by
farmers facing the same price movements). This problem is com-
pounded by the green revolution, which has resulted in more
farmers becoming more dependent on fertilizers, and other mod-
ern inputs.
e. irregularities in input supplies are likely to have covariate effects
on yields, for example, electricity blackouts in India worsened just
when more farmers had become dependent on electric pumps for
irrigation (Hazell 1982);
f. an increase in irrigated area. Although irrigation may be effective
in reducing yield variability within fields, it may, by reducing some
dispersed climatic influences on yields, lead to more synchronized
patterns of variability across locations (Pandey, ch. 18);
g. more covariate patterns of rainfall and climatic variation (Walker,
ch. 6; French and Headley, ch. 22).
The tendency for correlations between crop yields to increase may well
be a price-related phenomenon. This is suggested by examining the change






Variability in World Cereal Production 33


in correlations among world prices since the early 1970s after detrending
(Hazell 1985a).

Maize Wheat

Wheat 1961-71 0.30
1974-81 0.89
Rice 1961-71 -0.62 -0.13
1974-81 0.78 0.82
The dramatically increased correlations presumably reflect the pro-
gressive development of international markets in terms of numbers of trad-
ers, growth in net transfers from the industrial to the developing world, the
increasing role of China in foodgrain arbitrage, and perhaps the greater
substitutability of grains as greater quantities of foodgrains become feed-
grains, to mention just a few possibilities. Whatever the true explanation,
to the extent that producers of cereals are price responsive in yields (Guise
1969, Houck and Gallagher 1976), more correlated prices for grains will
predispose more correlated crop yields.


Conclusions
World cereal production (excluding China) grew at an average yearly
rate of 2.7 percent between 1960/61 and 1982/83, largely as a result of
improved yields. This growth has been accompanied by a more than pro-
portional increase in the standard deviation of production. The coefficient
of variation of production around trend was 0.028 during the period 1960/
61 to 1970/71. It increased to 0.034 over 1971/72 to 1982/83.
Increases in yield variances and a simultaneous loss in offsetting pat-
terns of variation in yields between crops and countries are the overwhelm-
ing sources of the increase in production variability. Although more re-
search is required before firm conclusions can be drawn about the cause of
these changes, the increased use of improved varieties and fertilizer-inten-
sive technologies since the 1960s may have been an important factor. How-
ever, this is less likely to be because of any higher sensitivity of new technol-
ogies to environmental stress, than because these technologies use
purchased inputs and hence lead to more variable and synchronized pat-
terns of input use across crops and regions in response to changing prices.
This effect has been amplified by the sharp increase in the variability of
world cereal prices since the early 1970s, and particularly by the increase in
price correlations between crops.
Continued high levels of variability in world cereal prices seem likely.
The United States is unlikely to return to its stockpiling policies of earlier





34 Evidence on Patterns of Changing Yield Variability

years, and cereal imports by the Soviet Union remain unpredictable.
World prices will also be affected by the levels of production variability
now established. These factors, together with a continuing trend towards
more input-intensive technologies, suggest that world cereal production is
also likely to remain quite variable in the years ahead.








3 Changing Patterns of Variability in
Chinese Cereal Production

BRUCE STONE AND TONG ZHONG









In 1984 the People's Republic of China produced 20 percent of the world's
cereal and 42 percent of the developing countries' cereal output. It was the
world's largest producer of both wheat and rice; the second largest pro-
ducer of maize, sorghum, and millet; and a major producer of other cere-
als (State Statistical Bureau [SSB] 1985b, p. II; He Gang et al. 1984, pp.
85-90; FAO 1985, pp. 107-22). China is also becoming an important,
though variable, trading nation in world cereal markets. The People's Re-
public was a net exporter of more than 2 million metric tons (Mt) of cereals
in 1985 (General Administration of Customs of the PRC 1986, pp. 20, 28),
following net imports averaging more than 10 million metric tons during
the 1977-84 period. While the major trend shifts in China's cereal produc-
tion and international trade performance have been more or less predicta-
ble (e.g., Eckstein 1966, Stone 1980, Stone 1985, 1986b), changes in typi-
cal year-to-year variations are much less so. Both are inevitably important
for the rest of the world because the absolute magnitudes are so large. The
more fundamental of these changes, production variability, is the focus of
this paper.
Figure 3.1 plots coefficients of variation (cvs) of production, area, and
yield for wheat, rice, maize, and total foodgrains for consecutive five-year
periods against the midpoints of these periods. Two general conclusions
can be drawn about these trends in cvs: (a) the coefficients for area have, if
anything, been falling during the 28-year period analyzed, although
slightly rising cvs during the past decade cannot be ruled out; (b) there is
little evidence of single 28-year trends for the production and yield cvs,
although the late 1950s and early 1960s were considerably less stable than
recent periods for all crops, as well as for total foodgrains.
In seeking to understand the sources of variability in Chinese cereal
production, it may be most useful to generate hypotheses regarding the
relatively anomalous periods of higher cvs for particular crops, and then
look at the structure of variability changes to the extent that the quality of






36 Evidence on Patterns of Changing Yield Variability


FIGURE 3.1a Coefficients of variation for sown area of foodgrains in China,
1954-1982


Coefficients of variation
0.16 r


Total foodgrains
- ---- Maize
---- Rice
............ Wheat


0.00 1
1954


1959 1964 1969 1974


1979 1982


NOTE: Each data point represents a coefficient of variation, expressed as a proportion, of a
five-year data period, plotted against the midpoint of that data period. Paddy rice, wheat,
and corn comprised a 53-72 percent rising share of "total foodgrain" sown area during the
1952-82 period and a 63-83 percent rising share of production of "total foodgrains," which
also include sorghum, millet, root and tuber crops (at one-fifth fresh weight), soybeans, and
pulses.



disaggregated Chinese data may justify attempting a variance decomposi-
tion exercise.


Hypotheses Regarding Abnormally High Coefficients of Variation for
Particular Crops and Periods

The outstanding subperiod to explain is undoubtedly the late 1950s
and early 1960s. At the time, Chinese reports blamed the poor and erratic
performance on consecutive years of catastrophic weather. But the period
of greater sown area variability appears to predate those for production
and yield, and Western observers have emphasized the dislocating policies
of the Great Leap Forward period. Recent Chinese analysts have also em-
phasized policy-induced disaster (Kueh 1984, p. 80), as have Lardy (1983)
and Walker (1984).
During this period, the State Statistical Bureau was disbanded and
reporting units came under considerable pressure to report policy suc-







Variability in Chinese Cereal Production 37


FIGURE 3.1b Coefficients of variation for yields of foodgrains in China, 1954-1982

Coefficients of variation
0.25
Total foodgrains
S----- Maize
0.20 \ ---- Rice
\ .": ............. Wheat


0.00 L
1954


1959 1964 1969 1974


1979 1982


FIGURE 3.1c Coefficients of variation
1954-1982

Coefficients of variation


for production of foodgrains in China,


- Total foodgrains
----- Maize
-- Rice
............ Wheat


Year





38 Evidence on Patterns of Changing Yield Variability


cesses. One of the destabilizing results of this development was that, with
grossly exaggerated reports of grain production growth in 1958, the area
sown with nongrain crops in 1959 was increased at the expense of grain. By
the time the true grain production level was understood in April 1959, it
was too late (Li 1962, Chao 1970, Walker 1984). To make matters worse,
1959 yields were low, due to poor weather. Catastrophic weather in 1960
and the political and administrative chaos precipitated by famine had dev-
astating effects on subsequent production.
In contrast to the period from the late 1950s to the early 1960s for
which major increases in cv were common to production, area, and yield
for each crop, major cv increases since the mid-1960s were confined to pro-
duction and yield. The declining trend in the cvs for areas sown to grain, as
well as for each grain crop individually, may be ascribed to increasing ad-
ministrative control over sown area in China.
The area cvs tended to increase after the mid-1970s, with increasing
decentralization of production decisions to individual farm families. Yet
the relative weakness of this increase is a testament to the inertia present in
areas operating through agricultural administration and perhaps em-
bedded in farmers' own decision-making processes. Though the decontrol
is genuine in principle, substantial sown area planning efforts continue in
many regions.
Significant cv increases in production and area since the mid-1960s
are confined to particular crops, unlike the initial period of instability. The
marked increase in maize yield cvs during the 1960s and early 1970s is
difficult to investigate due to lack of data. The pattern is not easily explain-
able by weather or rapid expansion into risk-prone farm areas. Most Chi-
nese maize is subject to damage from frost, drought, heat, and waterlog-
ging, and planting is concentrated in particularly risky regions (e.g., North
and Northeast China). But wheat has also been grown in very risk-prone
areas, and irrigation of wheat did not accelerate until the 1970s, when its
rising cv path paradoxically crossed the declining path for maize. Given
the scanty data available, the following hypothesis may explain maize
yields and yield cvs.
Maize yield cvs increased rapidly in the 1950s, due to rapid growth in
sown area concentrated in the Northeast provinces of Liaoning, Jilin, and
Heilongjiang, and the adjacent north China provinces of Hebei and Shan-
dong. Maize areas in these regions are highly risk-prone (Stone et al.
1985), and weather patterns are highly correlated, further accentuating
the increase in cvs.
The anomaly then becomes not the higher maize cvs of the late 1960s,
but the sharp decline in cvs during the 1962-64 period to values below
those of the 1950s. There is some statistical evidence that maize area was
cut back sharply in 1962 (Zhongguo Kexueyuan 1980) in response to a





Variability in Chinese Cereal Production 39


maize disease epidemic in 1961 (Wiens 1978, p. 677). While maize area
recovered somewhat in 1963, it remained quite constant during 1963-65,
well below the aggregate levels of the middle and late 1950s (Zhongguo
Guojia Tongjiju 1984b). Weather, according to Kueh's (1984, 1986) analy-
ses, remained relatively favorable throughout this period and featured sim-
ilar and low aggregate estimates of gross crop damage. Subsequently, as
maize area recovered in the Northeast, yield variability increased. The year
1966, in particular, was reported as one of especially widespread maize
disease (Wiens 1978), while throughout the late 1960s the increased
weather variability and highly correlated weather patterns of the expansion
region would contribute to higher cvs.
Disease-resistant hybrids covered 40 percent of maize area by 1973
(Wiens 1978, p. 677) and 60 percent by 1978 (Stone 1980, p. 158). With
their success, maize area surpassed the 1950s peak years for the first time
in 1974 and exceeded wheat area in a few years at the turn of the decade.
Although irrigation of maize is increasing, most maize remains rainfed
and the drop in cvs cannot be primarily ascribed to irrigation. It may be
associated with the lower disease susceptibility of the single-cross hybrids
and rapid expansion into regions for which weather patterns are not corre-
lated with those of the Northeast and the northeastern North China Plain.
Rapid dissemination of semidwarf wheat varieties and rapid area ex-
pansion in highly correlated areas may also explain the increase in wheat
cvs in the mid-1970s and early 1980s. The principal growth region in the
1960s was North China, characterized by highly correlated internal
weather patterns and increased irrigation of winter wheat. In fact, irriga-
tion provision was often a requisite condition for winter wheat introduction
in the North.
More irrigated wheat, however, did not necessarily provide greater
stability against drought, the predominant yield risk in the region. Irriga-
tion facilities may have proved unreliable, in some years providing much
higher yields and in others no change from the "unirrigated" state. This
could be particularly destabilizing in the aggregate if unreliabilities and
weather patterns were highly correlated. For surface irrigation there is
considerable evidence that this is so. When winter wheat in North China
requires irrigation, no rain usually means no river flow (Stone 1983 and
Chinese papers in the same collection). The Yellow River remains the
major exception, but its water has been difficult to use due to its high silt
content. The years 1976, 1977, and 1978 all had poor weather with the
principal disaster areas being within North China, and the principal disas-
ter being drought.
Since a large portion of North China irrigation development has been
tubewell construction during the late 1960s and 1970s, correlated yield
failures may be due to a major administrative failure to deliver key requi-





40 Evidence on Patterns of Changing Yield Variability


site inputs. This could indeed have happened during the chaotic period
associated with the Gang of Four successional struggle. For example, there
is some evidence that agricultural users were among those most heavily
penalized during the 1978 electrical shortage within the North and North-
east power grids (Stone 1980), but the availability of detailed information
is low.
Why have foodgrain production cvs not fallen during the past four
decades? If the data midpoints in figure 3.1 corresponding to the Great
Leap Forward (1958-59) policies and the catastrophic weather (1960-61)
are taken as an exceptional period of instability, then study must focus on
the periods represented by the midpoints of 1954-56 and 1964-83, which,
in turn, provide no clear evidence of long-term decline in cvs. Yet China's
reduction of weather-related agricultural risk for thousands of localities is
well known. Irrigated area increased from 16 to 45 million hectares (ha)
(around 45 percent of cultivated area) and its reliability vastly improved.
Flood control efforts have achieved considerable success. Modern varieties
combining improved maximum yields with locality-specific weather and
disease tolerances have been developed, along with considerable breeding
for earlier maturing varieties. The resulting aggregate multiple cropping
index increased from 1.2 to 1.5, thereby reducing the annual aggregate
production risk associated with very bad weather in any one season.
Consequently "high and stable yield" areas now comprise one-third of
cultivated area, around one-half of sown area, and account for not less
than three-quarters of national grain production. Why then have produc-
tion cvs not declined since the 1950s? The succeeding sections will try to
answer this question and to clarify the issues relating to the increases in
wheat production cvs since the early 1970s.

The Variance Decomposition Exercise:
The Data, Implications of Selected Data Periods, and Detrending
To better examine structural change in production variance across
different periods for China, the variance decomposition method of chapter
2 has been employed. To identify the important components of the change
in aggregate production variance, however, the selection of data period has
been severely constrained by the availability and quality of provincial and
crop data.
From original source material compiled in provincial statistical col-
lections, including Chen (1967), Social Science Research Council (1969),
CIA (1969), and Wiens (1980), an analysis of additional disaggregated
provincial data by Walker (1984) and IFPRI has culminated in disaggre-
gated series for 1952-57 of quality sufficient to attempt an initial variance
decomposition study. Data reported by PRC provincial authorities prior to





Variability in Chinese Cereal Production 41


1952 and for 1958-61 are unreliable, while full reliability and complete-
ness of data even for 1962-78 present problems (Wiens 1980, Stone 1983,
1984a).
The data for the study are, therefore, constrained yet were the only
disaggregated data available of sufficient quality and continuity for a vari-
ance decomposition during the People's Republic period, 1952-57 and
1979-83. Even within these temporal limitations, data for a number of
crops and provinces had to be excluded due to insufficiently independent
estimates for consecutive years or other problems of reliability. The final
data cover only wheat (for 21 of 26 provinces), rice (for 18 provinces) and
"other foodcrops" as a group (consisting of all other cereals, plus soy-
beans, other miscellaneous leguminous crops, and sweet and white pota-
toes). It was possible to include, however, almost all the important wheat-
and rice-producing provinces.
While chosen solely on the basis of data availability, continuity, and
reliability, the two periods are fully appropriate for a comparison of pre-
and post-green revolution patterns. Seed selection and strain improvement
along with organic fertilization had been traditional in Chinese agricul-
ture. However, manufactured fertilizer application was not widespread in
China until the late 1960s (Dalrymple 1978, p. 41; Stone 1980, pp. 122,
155-71; Stone 1986a,b).
Semidwarf rice varieties were first developed in Chinese research insti-
tutions during 1956-59 (Wiens 1978, p. 676; Dalrymple 1978, p. 78;
IRRI/CAAS 1980, p. 9) and high-yielding hybrid rice, a fully Chinese
technical innovation, was first grown by farmers in 1975 (Wiens 1978,
p. 679; IRRI/CAAS 1980, p. 10). Semidwarf wheat varieties from Chi-
nese-Mexican crosses were initially planted on farmers' fields in 1973
(Dalrymple 1978, p. 41; Stavis 1978, pp. 638, 645; Wiens 1978, p. 677),
although imported fertilizer-responsive varieties were used on farms as
early as 1957, and Chinese research institutions and breeding stations were
releasing their own varieties by 1960 (Stone et al. 1985). Double-cross hy-
brid maize was released around 1958-59 and occupied a large share of
maize area in the mid-1960s but proved susceptible to disease over large
areas in 1961 and 1966 (Wiens 1978, p. 677). New single-cross maize hy-
brids were developed and released in the late 1960s and recovered to 40
percent of maize area by 1973 (Wiens 1978, p. 677) and 60 per cent by 1978
(Stone 1980, p. 158).
By 1977, high-yielding dwarf rice varieties were sown on around 80
percent of China's rice area (Wiens 1978, p. 677) and in 1985 the propor-
tion was around 95 percent (Dalrymple 1986a). On-farm use of hybrid rice
expanded quickly from its introduction in 1975 to 20.4 percent of rice area
in 1983 (Stone 1984b, table 1). If semidwarf wheat can be construed to
include all varieties under 105 cm at maturity, semidwarfs exceed 70 per-





42 Evidence on Patterns of Changing Yield Variability


FIGURE 3.2 Fertilizer nutrient application per sown hectare in China, 1983


30-65 140-165
65-90 165-190
90-115 190-215
115-140


SOURCE: Stone (1986a, 460).


cent of Chinese wheat area. If the standard is 100 cm, the proportion is
around 55 percent. If a standard of 85 cm is employed, the proportion is
much smaller, but grew rapidly from 10 percent in 1980 to more than 30
percent in 1983 (Stone et al. 1985, Dalrymple 1986b).
Application of manufactured fertilizers averaged 0.6 kg/ha in 1952
and 2.4 kg/ha in 1957. From 1979 through 1983, average application grew
from 73 to 115 kg/ha to exceed average rates in most other nations (Stone
1986a; Zhongguo Guojia Tongjiju 1984b, p. 137; FAO 1984, pp. 44-55).
Such fertilizer is the principal purchased input in most Chinese provinces
(figure 3.2). All in all it is clear that the period 1952-57 predates the green
revolution in China and that, by 1979-83, the technical transformation of
Chinese agriculture was already considerably advanced. The second pe-
riod also falls within the peculiar years of higher wheat production cvs,
while the first period considerably precedes them.
The data period ends prior to the formation throughout most of China





Variability in Chinese Cereal Production 43


of the people's communes and the radical policy initiatives of the Great
Leap Forward (1958-59), as well as the unusually severe natural disasters
which caused or seriously accentuated the famine of 1960-61. However,
this first period spans the formation of agricultural cooperatives (1952-
56), the establishment of a state grain market (1953), and the imposition of
fixed quotas of grain to be delivered from each unit of farmland (1955)
(Chao 1970, pp. 44-68; Stone 1980, pp. 147-8). It also spans a period of
some instability in purchase prices for cereals, although the cash costs for
inputs were consistently minor throughout the 1950s (Stone 1980, 1984c).
The second period postdates major rural organizational reforms, a
three-tiered structure for grain prices including free-market cereals prices,
and significant procurement pricing changes after the Cultural Revolution
(Stone 1980, pp. 147-53). And while farm price and market structures re-
mained reasonably constant during 1979-83, rapid growth in yields be-
cause of greater fertilizer availability allowed average and especially mar-
ginal prices that many farmers faced, as well as the degree with which they
interacted with private versus public markets, to change considerably dur-
ing the period. Thus while 1952-57 and 1979-83 (or even 1979-84) are
periods of greater internal consistency than surrounding periods, such
continuity is imperfect in both cases.
Detrending required to complete the variability comparisons and the
variance decomposition exercise posed a problem, because of the lengthy
period separating the two internally contiguous data sets. Due to major
structural changes in Chinese agriculture during the 1960s and 1970s
(Stone 1980), the factors governing trend increase in yields during the two
data periods were thoroughly dissimilar. Yet the periods are so short that
separate linear trends could jeopardize the results through loss of the extra
degree of freedom. A single quadratic trend was, therefore, estimated for
total foodcrops and for each crop (production, area, and yield) for the en-
tire 1952-83 period.

Changes in Means and Variability
Table 3.1 summarizes the changes in Chinese foodgrain production
between the two periods in the included provinces. Average foodgrain pro-
duction increased from 166 to 317 million metric tons, or by 90 percent.
This accords quite well with the increase for all provinces between the two
periods: from an average of 173 (1952-57) to 344 million metric tons
(1979-83) or by 98 percent (Stone 1984a, p. 605; Zhongguo Guojia Tong-
jiju 1984c, p. 145). Paddy rice production for the group of included prov-
inces increased from 67 to 131 million metric tons, or 96 percent between
periods (from 76 to 152 Mt, or 99 percent for all provinces). Wheat pro-









TABLE 3.1 Changes in the mean and variability of national foodcrop production, area, and
yield in the People's Republic of China

Means Coefficient of Variation

Change Change No. of
Crop I II (percent) I II (percent) F Ratio Provinces

Production (Mt)
Total grains 166 317 90.5 0.022 0.085* 286.4 56.35* 8/11/21
Rice 67 131 95.6 0.050 0.063 26.0 6.12 6/7/18
Wheat 21 63 199.3 0.062 0.137* 121.0 44.02* 7/12/21
Other crops 79 123 56.6 0.022 0.063* 186.4 19.51* 4/6/21
Area (million ha)
Total grains 123 109 -11.6 0.007 0.031 342.9 16.01* 2/5/21
Rice 27 30 9.7 0.050 0.005* -90.0 0.01* 2/2/18
Wheat 26 28 7.8 0.024 0.020 -16.7 0.79 3/3/21
Other crops 71 52 -26.6 0.011 0.041* 272.7 8.13 1/4/21
Yield (t/ha)
Total grains 1.35 2.91 115.6 0.020 0.056* 180.0 35.17* 6/15/21
Rice 2.52 4.49 78.2 0.026 0.061* 134.6 17.14* 5/5/18
Wheat 0.82 2.27 176.8 0.050 0.120* 140.0 45.35* 11/17/21
Other crops 1.11 2.37 113.7 0.025 0.037 48.0 10.46* 3/7/21

Note: I = First Period: 1952-57, II = Second Period: 1979-83. Asterisks on the F ratios indicate statistically
significant homogeneity of variance tests and asterisks on the second period cvs denote that the cv in the second
period is significantly different from that of the first, both at least at the 5 percent level. Although these are de-
scribed as "national" results, they are based on sums of data for only the provinces included in the study (see text).
The three numbers in the final column on the right indicate the number of provinces for which the F tests for
homogeneity of variance over the two periods indicated apparently significant differences between the two periods
at the 1 and 5 percent levels, and the total number of included provinces, respectively.





Variability in Chinese Cereal Production 45


duction for included provinces increased 199 percent, from an average of
21 to 63 million metric tons (from 22 to 65 Mt, or by 200 percent, for all
provinces).
Among remaining cereal crops, the outstanding contributor to food-
grain production growth was maize, which increased throughout China
from a mean of 19 million metric tons in the first period to 62 million met-
ric tons in the second period, or by 223 percent (Zhongguo Guojia Tongjiju
1984c, p. 145). Maize represented a rapidly increasing proportion of
"other foodcrops" in China (23 percent in the first period and 49 percent
in the second period). Together, rice, wheat, and maize increased their
share of total Chinese food crop production (Chinese definition) from 68 to
81 percent. Production of "other foodcrops" (including maize) among in-
cluded provinces increased from 79 to 123 million metric tons.
All of the increase in national foodgrain production is attributable to
changes in average yields, which increased 116 percent to 2.9 tons per hect-
are among included provinces, while area declined by 12 percent. But this
increase in average total grain yields was partly due to a shift in composi-
tion of grain sown area, with wheat and rice area among included prov-
inces rising by 8 and 10 percent, respectively, while those of other food-
crops declined by 27 percent. Among other foodcrops, however, area sown
with maize, the most widely planted "other" crop, increased throughout
China by 36 percent, implying a very sharp decline in area sown with
millet, sorghum, barley, soybeans, potato crops, and other minor grains
and bean crops (Zhongguo Guojia Tongjiju 1984b, p. 138). Among these
latter crop categories, only white potatoes increased in area (Stone 1984a,
p. 628).
Although part of the total grain production growth was attributable to
changes in crop composition, it was dominated by increasing yields.
Among included provinces, average yields increased from 0.8 to 2.3 tons
per hectare (177 percent) for wheat, from 2.5 to 4.5 tons per hectare for
paddy (78 percent) and from 1.1 to 2.4 tons per hectare (114 percent) for
other foodcrops. Among other foodcrops, final average yields and their
rates of growth differed considerably, although all registered gains. For all
of China the leaders in growth between the two periods were maize, sor-
ghum, and potato crops, for which yields increased by 139, 129, and 84
percent to second period averages of 3.2, 2.6, and (at 1/4 weight) 3.5 tons
per hectare, respectively. In contrast, millet and soybean yields increased
by 48 and 42 percent to 1.6 and 1.1 tons per hectare. Yields for barley and
other minor grains and bean crops also grew slowly from very low base
levels (Stone 1984a, p. 608, 1984b; Zhongguo Guojia Tongjiju 1984b,
p. 47).
The coefficient of variation of total grain production increased 286
percent among included provinces, which was statistically significant at





46 Evidence on Patterns of Changing Yield Variability


TABLE 3.2 Analysis of the components of change in mean foodgrain production
in China, 1952-1957 to 1979-1983 (percent)

Interaction
"Pure" "Pure" Changes in between Mean
Changes Changes Area-Yield Yields and
Crop in Yields in Area Covariances Mean Areas
Rice 81 10.3 0.08 9
Wheat 95 6.5 -0.05 -1
Other crops 205 -45.0 0.53 -61
Total foodcrops 122 -7.1 0.18 -15



the 5 percent level or better. This significance result was echoed individu-
ally for wheat and "other crops" (for which the cvs increased by 121 and
186 percent, respectively) but not for rice. The cv for total grain area grew
343 percent, with the increase again significant but from a very small cv to
another small cv. Wheat and rice actually registered nominal declines in
area variability. The cv for yield increased by 135 percent for rice, 140 per-
cent for wheat, 48 percent for "other crops" and 180 percent for total food-
grains, each (except for "other crops") proving to be a statistically signifi-
cant increase.


Results from the Mean and Variance Decomposition Procedures
Applying Hazell's decomposition method to the change in average
foodgrain production between the 1952-57 and 1979-83 periods, changes
in yields are confirmed to be the most important component (table 3.2).
For total foodcrops, changes in yields accounted for 122 percent of the
mean production change between periods, while loss of area reduced the
production impact of increasing yields by 7 percent, and interaction be-
tween mean yields and mean areas, by 15 percent. This last figure indi-
cates that, where area increased (decreased), average yields decreased (in-
creased). Reduction of foodcrop-sown area primarily involves low-yielding
lands, despite well-publicized administrative complaints that considerable
areas of high-yielding land have been lost to irrigation structures, roads,
and buildings. Hence, where area increased (especially in the major recla-
mation provinces of Heilongjiang and Inner Mongolia) average yield
tended to decline due to low productivity on the newly reclaimed land.
The area-yield interaction effect was very weakly negative for wheat (1
percent) and actually positive for rice (9 percent). In provinces where rice
yields were rising, a particular effort was probably made to increase rice
area, an influence related to the primacy of the fine-grains procurement






Variability in Chinese Cereal Production 47


issue in central policy and planning (Stone 1984c, Walker 1984). Wheat's
weakly negative area-yield interactive effect could be considered a balance
between a similar (positively correlated) effort for wheat and the (nega-
tively correlated) expansion of wheat onto relatively unsuitable lands.
"Pure" yield changes accounted for 81 percent of the change in the mean
aggregate rice production level between periods (and 95 percent for
wheat), while ("pure") increased rice area accounted for 10 percent of the
aggregate rice production level change (and less than 7 percent for wheat).
Table 3.3 itemizes the contribution of various components of the in-
crease in the variance of total foodgrain production between the two pe-
riods. Increases in interprovincial covariances account for 91.4 percent of
the production variance increase for total foodcrops (table 3.3), 87.8 per-
cent for rice (table 3.4), 84.6 percent for wheat (table 3.5) and 72.6 percent
for "other foodcrops" (table 3.6). This is consistent with results obtained
for India (Hazell 1982, 1984), the United States (Hazell 1984), the
U.S.S.R. (Nguyen 1985, table 4), Syria (Nguyen, ch. 5), and Australia
(Anderson et al. 1988), though the effect appears strongest for China.
Unlike the United States and India, but like the Soviet Union, the
Chinese cross covariance contribution (between different crops in different
provinces) exceeds the within-crop covariance effect among provinces.
This may represent a characteristic feature of central planning where
state-supplied inputs are allocated for specific crops in provinces where
authorities aspire to purchase marketable surpluses. The state may also
contribute to this peculiar pattern of interprovincial correlation through
fixed quotas for specific grains in each province to facilitate exports,
through pricing policies which have provided much higher marginal re-
turns to grain deliveries above assigned quotas, and by allocating fertil-
izers in exchange for desired commodities (Stone 1980, 1984b, 1986b).
The changes in provincial crop production variances account for only
6.1 percent of the change in the variance of total foodgrain production (1.5
percent from rice, 2.4 percent from wheat, and 2.2 percent from other
crops). This is a particularly small proportion compared with that found in
other country studies. Also striking is the especially small contribution of
intercrop covariances within provinces (2.5 percent), perhaps because in-
tercrop covariances were already high in the first period.
The interaction effect between mean area levels and yield variances
and covariances, and the interaction effect between mean yield levels and
area variances and covariances are offsetting and particularly important
for rice (see table 3.4). The positive value for the first of these interaction
terms is not surprising. As area increases (onto less suitable and more
poorly serviced lands) the variability of yields tends to increase. This inter-
action effect is particularly strong but negative for "other crops" (table






48 Evidence on Patterns of Changing Yield Variability


TABLE 3.3 Analysis of the components of change in the variance of total grain
production in China, 1952-1957 to 1979-1983 (percent)


Changes Changes Changes
in Yield in Area in
Changes Changes Variances Variances Area-
Crop in Mean in Mean and and Yield
Variance Yields Areas Covariances Covariances Covariances

Rice 1.4 0.2 1.6 -0.8 0.2
Wheat 0.5 -0.0 1.9 -0.0 0.0
Other 0.8 -0.4 3.5 -0.1 0.5

Total variances
within provinces 2.6 -0.2 7.0 -0.9 0.8
Intercrop
covariances within
provinces 0.6 -0.0 1.1 -0.2 0.4
Interprovince
covariances within
crops 5.8 0.3 23.8 -1.8 1.9
Covariances between
different crops
in different
provinces 5.9 -0.4 33.8 -1.1 9.8
Column sums 14.9 -0.3 65.7 -4.1 12.9




3.6). This may indicate that land going out of "other crop" production was
relatively good land which left the remaining production of "other crops"
in a more unstable state on average.
The interaction term between mean yields and area variances and co-
variances is negative for the aggregated provinces in each crop category
and may also reflect central planning effects. The impact is especially
strong for rice. In provinces exhibiting rising yields, successful efforts were
made to stabilize area to help guarantee procurement despite relatively low
prices for rice (Stone 1984b). Conversely, in areas where effective control
was achieved, state authorities focused current inputs and infrastructural
investments, thereby raising yields. In any event, several items in the de-
composition suggest relative success in crop area stabilization between the
1950s and 1979-84.
The most important result of the variance decomposition is the partic-
ular dominance of the interregional covariances, especially among yields.
This is very likely the reason that the cvs for yield and production of total
foodgrains, as well as of the individual grains, do not appear to have de-





Variability in Chinese Cereal Production 49


Change in Interaction Terms between
Total
Mean Mean Areas Mean Yields Mean Yields, Contribution
Yields and Yield and Area Mean Areas, to Change
and Variances Variances and Changes in Variance
Mean and and Area-Yield in of Production
Areas Covariances Covariances Covariances Residual in China

0.0 0.1 -1.4 0.2 0.1 1.5
-0.0 0.1 -0.1 0.0 0.0 2.4
0.2 -1.8 -1.0 0.3 0.1 2.2


0.2 -1.6 -2.5 0.5 0.3 6.1


-0.0 0.3 -0.0 0.4 0.0 2.5


0.1 0.5 -4.3 2.9 0.8 30.0



-0.3 1.7 -2.2 12.2 2.0 61.4
-0.0 0.9 -9.0 15.9 3.0 100.0




dined during more than three decades (figure 3.1). The most plausible
explanation for this increase in interprovincial covariances is that Chinese
farmers are now much more responsive to central policy and to national
market influences in general.
Issues relating to the genetic base for high-yielding crop varieties con-
stitute an input hypothesis related to the central policy theme. It has been
suggested for the United States and for the U.S.S.R. (Hazell 1984, Nguyen
1985) that many regionally adapted varieties of a crop were replaced by a
very few higher yielding varieties sown broadly across regions, and that this
development may have contributed to increased interregional correlations
in yield and hence to increased aggregate variance. Comparing the 1960s
in China (a period of particular national production variability but poor
provincial data) with the 1950s, the proportions of wheat, rice, and maize
area planted with just a few closely related varieties increased dramati-
cally. Since the 1960s, however, more regionally adapted high-yielding va-
rieties have proliferated, somewhat reducing the area sown with any single
variety. Difficulties associated with central authorities promoting overly





50 Evidence on Patterns of Changing Yield Variability


TABLE 3.4 Analysis of the components of change in the variance of rice production
by province in China, 1952-1957 to 1979-1983 (percent)



Changes Changes Changes
in Yield in Area in
Changes Changes Variances Variances Area-
in Mean in Mean and and Yield
Province Yields Areas Covariances Covariances Covariances

Hebei, Beijing
Tianjin 270.7 0.7 -47.4 -126.8 173.9
Nei Monggol -101.7 -4.6 19.1 164.1 -70.3
Liaoning -333.3 -90.9 15.6 36.8 37.5
Jilin -212.6 -33.1 9.5 67.0 31.4
Heilongjiang 14.8 3.8 40.4 38.3 -29.7
Jiangsu and
Shanghai 4.6 5.3 48.0 -5.3 10.7
Anhui 529.9 1.3 318.5 -265.4 6.4
Jiangxi 20.1 11.2 71.7 -8.8 -1.5
Shandong 6.1 8.6 0.3 19.4 -8.5
Henan 8.9 -0.1 43.7 2.7 21.5
Hubei 45.8 38.2 64.9 -14.2 -37.8
Hunan 26.7 22.2 17.2 -19.2 38.6
Guangxi 156.0 4.6 78.1 -61.5 37.4
Guangdong 3.8 -0.2 114.4 -0.8 9.6
Sichuan -1,443.6 18.1 -878.0 1,040.1 -275.4
Guizhou 17.0 -0.8 145.8 -20.6 -5.7
Yunnan 0.2 0.1 29.7 9.5 35.7
Shaanxi 0.9 0.7 86.5 -0.1 -11.8
Interprovince
covariances 41.8 3.0 83.6 -18.3 9.3
All provinces 47.7 4.1 86.9 -22.3 10.1



rapid and intensive adoption of specific high-yielding varieties have been
documented for maize in the 1960s (Wiens 1978, p. 677; Stone 1980, pp.
157-8), and for hybrid rice in the late 1970s and early 1980s (Stone 1984c).
But it may be more valuable to focus attention on wheat, for which the
yield and production cvs increased most notably between 1952-57 and
1975-84.
During the late 1970s and early 1980s, nine varieties accounted for
14.5 million hectares (one-half of national wheat sown area) in a few of the
major wheat-growing zones; and a number of these varieties have closely
related genealogies (Stone et al. 1985). But although the numbers of vari-
eties were considerably fewer than during the 1950s, they had been bred
with greater experience and attention to disaster risks than those of the
1960s.





Variability in Chinese Cereal Production 51


Change in Interaction Terms between
Total
Mean Mean Areas Mean Yields Mean Yields, Contribution
Yields and Yield and Area Mean Areas, to Change
and Variances Variances and Changes in Variance
Mean and and Area-Yield in of Production
Areas Covariances Covariances Covariances Residual in China


-99.7 -80.8 -396.8 384.9 21.4 0.0
-4.0 -6.3 1,213.8 -5.6 -14.4 0.0
-65.5 43.8 263.1 175.4 17.6 -0.1
-13.5 10.3 181.1 56.8 3.1 -0.1
-2.2 16.6 19.0 -13.1 12.2 0.0

-3.3 35.7 12.5 15.3 1.4 3.5
-0.2 4.2 -558.5 4.9 58.9 0.5
-0.2 29.2 --20.7 -1.7 0.7 0.4
2.5 17.2 186.5 -135.2 3.2 0.0
0.5 -11.3 11.5 20.8 1.7 0.2
7.3 54.4 -22.8 -45.0 9.2 2.0
-5.1 10.3 -43.4 49.6 3.1 1.7
-6.7 29.8 -192.1 51.9 2.5 0.5
0.2 -31.7 -2.9 7.8 -0.2 2.9
-22.4 216.0 1,608.9 -106.1 -57.7 -0.4
-0.1 -18.1 -14.8 -1.3 -1.4 0.2
-0.0 4.6 6.6 14.3 -0.8 0.7
0.1 27.8 0.0 -5.0 0.8 0.0

0.2 16.1 -41.4 3.6 2.6 87.8
0.2 15.1 -48.0 4.1 3.0 100.0



Examining table 3.7, it is immediately clear that a much greater num-
ber of provincial yield (and even production) variances tested as nonho-
mogenous between the periods in the case of wheat than for rice and for
"other crops." Of those testing nonhomogenous, only two recorded de-
clines in variance (neither significant at the 1 percent level): for yields, only
Anhui and Jiangsu; for production, only Anhui and Yunnan. Contrasting
sharply with the results for rice and for "other crops," all of the largest five
wheat producers exhibited significant increases in production variability.
Among the next nine provinces of intermediate importance for wheat
(each with 1.2 to 7.3 percent of national production), production variabil-
ity increases were not significant at the 5 percent level only for Xinjiang,
Shanxi and Anhui. Some 81.5 percent of Xinjiang's farmland and virtually
all Xinjiang wheat are irrigated, by far the highest proportion in China






52 Evidence on Patterns of Changing Yield Variability


TABLE 3.5 Analysis of the components of change in the variance of wheat
production by province in China, 1952-1957 to 1979-1983 (percent)



Changes Changes Changes
in Yield in Area in
Changes Changes Variances Variances Area-
in Mean in Mean and and Yield
Province Yields Areas Covariances Covariances Covariances


Hebei, Beijing
Tianjin
Shanxi
Nei Monggol
Liaoning
Jilin
Heilongjiang
Jiangsu and
Shanghai
Anhui
Jiangxi
Shandong
Henan
Hubei
Hunan
Guangxi
Guangdong
Sichuan
Guizhou
Yunnan
Shaanxi
Gansu
Xinjiang
Interprovince
covariances
All provinces


6.0
41.5
2.7
-2,843.2
5.5
11.0

30.8
63.0
10.8
2.8
78.2
22.6
107.3
-727.2
101.6
0.1
2.5
8.1
-1.5
9.0
-1.2


0.8
-9.8
17.7
-4.7
9.5
5.5

-1.5
-12.0
-4.2
-0.8
-3.4
1.5
-1.8
-31.6
-4.0
1.4
262.4
22.6
0.2
1.7
283.9


49.5
161.5
34.7
-342.8
-6.7
14.7

60.9
186.6
156.0
94.5
142.3
48.7
74.2
-41.0
276.7
28.0
-101.7
70.5
75.2
74.3
84.3


-0.7
-5.1
-0.0
152.0
16.8
4.4

5.9
-2.4
-4.3
0.1
-5.2
-4.0
-16.2
167.6
39.8
0.2
49.2
122.8
0.8
-1.5
139.6


4.8 0.5 68.1 -0.2
7.0 0.4 69.8 -0.2


3.1
-0.5
-12.5
-29.2
-2.6
-1.5

-23.9
1.3
-12.9
8.0
-2.9
7.6
11.9
-51.7
-185.9
0.1
118.6
-161.3
12.0
1.3
-276.9

2.7
2.4


(Stone et al. 1985). The performance of Shanxi and Anhui may be related
to the unusually high cv in the first period for both provinces (25 percent),
coupled with relatively large wheat area reductions (21 and 26 percent),
augmented by irrigation expansion.
But from table 3.5 it is clear that it is not the individual provincial
changes in production variance, but the increased interprovincial covari-
ances that account for most (85 percent) of the increased aggregate wheat
production variance.







Variability in Chinese Cereal Production 53


Change in Interaction Terms between
Total
Mean Mean Areas Mean Yields Mean Yields, Contribution
Yields and Yield and Area Mean Areas, to Change
and Variances Variances and Changes in Variance
Mean and and Area-Yield in of Production
Areas Covariances Covariances Covariances Residual in China


-0.4
-3.0
1.1
-52.5
0.1
-0.6

1.4
-3.0
-0.6
0.0
-2.7
-2.4
0.3
-63.5
0.1
-0.8
-43.7
-0.1
-0.0
0.5
-5.6


34.9
-61.5
75.2
312.8
-8.7
65.2

-19.3
-84.5
-37.5
-22.6
-41.4
31.9
-16.2
39.3
-70.0
71.5
-422.6
253.8
1.1
16.1
274.1


-4.9
-26.0
-0.1
2,915.6
85.7
15.3

106.8
-49.6
-5.5
0.7
-59.9
-19.9
-65.7
774.2
142.2
0.9
29.3
145.2
1.9
-3.7
90.2


8.3
-0.5
-19.3
-9.6
-7.1
-5.8

-61.5
3.1
-4.1
14.3
-5.9
16.2
11.8
25.9
-145.2
0.3
211.1
-339.8
10.3
1.4
-453.0


0.5 10.0 -3.5
0.3 8.9 -3.3


3.5
3.5
0.4
1.7
7.4
-8.3

0.4
-2.4
2.4
3.1
0.9
-2.2
-5.5
8.0
-55.4
-1.7
-5.1
-21.8
0.1
1.1
-35.6

-0.4
-0.2


84.6
100.0


Hypotheses Explaining Increased Interprovincial Correlations and Yield
Variability

In table 3.8, detrended pairwise yield correlations for the 1979-83 pe-
riod have been calculated for only those provinces contributing more than
1 percent to the increase in the variance of national wheat production.
These provinces also happen to be the five largest wheat producers and five
provinces within which wheat production increased most between the two






54 Evidence on Patterns of Changing Yield Variability


TABLE 3.6 Analysis of the components of change in the variance of other crop
production by province in China, 1952-1957 to 1979-1983 (percent)



Changes Changes Changes
in Yield in Area in
Changes Changes Variances Variances Area-
in Mean in Mean and and Yield
Province Yields Areas Covariances Covariances Covariances


-43.0
54.2
-36.8
-0.2
0.2
6.6

36,841.2
13.1
-167.5
142.2
1.4
-101.0
-6.1
-255.1
8.1
-164.1
537.7
68.7
23.0
14.6
59.0


85.2
-5.3
68.1
-1.9
--1.1
-1.4

366.0
-53.6
76.2
202.2
-3.4
66.2
-40.1
36.9
-6.2
9.8
-0.6
17.7
-1.5
-96.0
83.7


Hebei, Beijing
Tianjin
Shanxi
Nei Monggol
Liaoning
Jilin
Heilongjiang
Jiangsu and
Shanghai
Anhui
Jiangxi
Shandong
Henan
Hubei
Hunan
Guangxi
Guangdong
Sichuan
Guizhou
Yunnan
Shaanxi
Gansu
Xinjiang
Interprovince
covariances
All provinces


71.4
172.0
15.3
122.5
122.4
102.1

16,885.2
102.7
170.8
-247.4
185.0
-4.2
-83.6
-88.7
3.6
99.8
227.1
31.6
103.8
186.5
502.7

98.0
115.9


18.0
-15.0
10.8
1.2
0.1
-5.0

-4,178.4
10.7
99.8
14.3
3.2
34.7
34.8
115.2
35.0
69.3
-205.8
-101.1
-16.3
14.6
769.9


-40.4
9.8
7.0
16.8
11.4
4.7

2,758.7
50.0
-107.4
-108.1
4.4
7.8
16.6
-11.4
21.2
-65.3
33.9
117.3
35.2
97.8
-1,215.4


3.1 9.8
0.4 13.6


periods. Although there was a decrease in wheat yield cv between the two
periods for Jiangsu (and increases for each of the other provinces), the ac-
tual yields among all these provinces are highly correlated in the 1979-83
period. This includes some of the most distant pairs of provinces such as
Jiangsu and Sichuan, and Hebei and Sichuan, for which reliability of the
correlation appears quite high. In fact, the least correlated and least reli-
able correlation is for the adjacent provinces of Jiangsu and Shandong.
Although some of these provinces have highly correlated weather patterns,
this relationship does not include all of those with highly correlated yields
during the period. Two important conclusions may be drawn from exami-


11.5 -1.7
18.6 -6.3






Variability in Chinese Cereal Production 55


Change in Interaction Terms between
Total
Mean Mean Areas Mean Yields Mean Yields, Contribution
Yields and Yield and Area Mean Areas, to Change
and Variances Variances and Changes in Variance
Mean and and Area-Yield in of Production
Areas Covariances Covariances Covariances Residual in China


-11.0
3.1
4.0
0.2
0.3
0.0

6,253.7
-1.7
-42.3
-96.8
1.3
7.8
8.2
-0.8
-0.4
-6.2
0.0
-9.1
4.0
1.0
-44.4


-27.5
-59.5
-6.0
-62.6
-48.2
-4.7

-12,274.5
-73.4
-120.3
164.5
-116.2
2.9
43.5
33.3
-2.3
-14.0
-3.1
10.6
-21.5
-123.8
460.4


86.2
-70.2
16.7
6.0
0.4
-5.5

-48,564.7
33.1
258.2
84.3
14.6
79.9
111.4
255.5
44.7
226.4
-531.1
-126.3
-56.8
9.9
601.3


-35.8
9.1
1.7
12.2
8.6
2.0

2,346.7
4.0
-3.1
-56.0
1.9
0.2
6.9
-4.8
-2.2
-59.7
29.8
85.9
31.1
-24.4
-1,014.3


-3.0
1.9
19.4
5.8
5.8
1.0

-333.7
15.1
-64.3
0.7
7.8
5.7
8.4
19.9
-1.3
4.0
12.0
4.6
-0.8
19.8
-103.0


-1.1
0.6
-0.4
4.9
10.7
3.5

0.0
1.7
-0.1
-0.7
8.2
-0.6
0.3
-0.1
0.4
-0.9
0.0
0.1
0.8
0.2
0.0


0.5 -45.1
2.6 -55.4


11.7
-3.7


2.7 9.4 72.6
5.8 8.6 100.0


nation of tables 3.5 and 3.8: (a) Those provinces that are individually con-
tributing the most to increased wheat production variance are not only
China's largest wheat producers, but exhibit highly correlated wheat yields
in the second period, and so are apt to be major contributors to the in-
creased interprovincial yield correlations that are the predominant factor
associated with increased wheat production variability; (b) While corre-
lated weather patterns and the similar response of similar varieties may
explain some of the yield correlation among these major wheat-growing
provinces in the 1979-83 period, these factors certainly do not explain all
of the yield correlation, nor is there convincing evidence to suggest that









TABLE 3.7 Provinces exhibiting variance changes between 1952-1957 and 1979-1983

Production Area Yield

Total Total Total
Province Grains Rice Wheat Other Grains Rice Wheat Other Grains Rice Wheat Other

National ** ** ** ** ** ** ** **
Hebei, Beijing,
Tianjin ** **
Shanxi n.a. n.a. n.a. *
Nei Monggol *
Liaoning ** ** ** ** ** ** **
Jilin ** ** ** ** ** **
Heilongjiang ** **
Jiangsu, Shanghai ** ** ** ** *
Anhui *
Jiangxi **
Shandong ** ** ** ** **
Henan ** ** ** ** ** ** **
Hubei ** **
Hunan *
Guangxi *
Guangdong ** ** ** ** ** *
Sichuan ** **
Guizhou ** ** **
Yunnan ** ** ** ** ** ** **
Shanxi ** ** ** *
Gansu n.a. n.a. n.a. **
Xinjiang ** n.a. n.a. n.a.

Note: **, indicate statistically significant F ratios (one-tailed tests) at 1 and 5 percent confidence levels, respec-
tively. n.a. = not available.






Variability in Chinese Cereal Production 57


TABLE 3.8 Wheat yield correlation coefficients among five major producing
provinces, 1979-1983

Province Jiangsu Shandong Henan Sichuan

Hebei (plus Beijing and Tianjin) 0.71 0.96 0.92 0.80
0.11 0.00 0.00 0.05
Jiangsu (plus Shanghai) 0.48 0.65 0.79
0.33 0.16 0.06
Shandong 0.88 0.66
0.02 0.15
Henan 0.61
0.19

Note: The upper number represents the correlation; the lower number the percent level of
"significance."



they would explain most of the increased correlation between the two
periods.
What, then, could be the cause of increased wheat yield correlation
among major wheat growing provinces, to the extent that it is not due to
weather and similarity among the varieties sown over broad areas? Among
the input-related hypotheses, the high correlation among distant provinces
such as Sichuan and Hebei would argue against linked irrigation or re-
gional power grid allocation being the dominant cause, although these
again cannot be ruled out within the North China provinces. Little is
known about difficulties with the seed distribution system, but some docu-
mentation related to manufactured fertilizer use is available.
The particularly rapid rate of growth of fertilizer use has elsewhere
been established as a principal contributory factor to increased foodgrain
yields since the 1960s and especially since the mid-1970s. Among food-
grains, wheat has been a high priority crop upon which yield-increasing
attention has been focused (Stone 1984b, 1986a). Wheat yields grew most
rapidly in these major wheat producing provinces but failed to reach the
1979 level in 1980 and 1981, resuming rapid growth in 1982 and 1984.
While 1980 was a poor weather year for China's main wheat producers,
1981 was not (Kueh 1984). When the state budget collapsed between 1980
and 1981 under the weight of unmanageable food subsidies and overex-
pansion of capital construction (Stone 1984b, 1985), one of several adjust-
ment measures (also aimed at alleviating foreign exchange difficulties) was
curtailment of imports, including fertilizers. Thus, imports of fertilizers,
though growing rapidly over the period, fell in 1981.
This decline began in late 1980, in time to affect the winter wheat
crop. Between October 1980 and January 1981, monthly procurement and
sales of manufactured fertilizer (both imported and domestically pro-





58 Evidence on Patterns of Changing Yield Variability


duced) by the central marketing organization were consistently 4 to 19 per-
cent below corresponding data for the previous year (Zhongguo Guojia
Tongjiju 1984a, pp. 312-13). Imports constituted 14-19 percent of total
Chinese application during the period, but considerably more for these in-
tensive user provinces (figure 3.2), and applications of imported fertilizers
were concentrated on cotton and wheat. The variability in wheat yields of
the major producing provinces and the high correlations among their
yields since 1979 may indeed be at least partly related to irregularities in
the institutional behavior affecting fertilizer supply and purchase.
Whether such difficulties were also contributory to the high cvs for the
remainder of the 1974-83 period (1974-78) is difficult to verify, but 1974-
76 was a foreign exchange conservation period when imports were reduced
somewhat, while 1977 and 1978 were the worst weather years in wheat-
growing regions since 1960-61.


Conclusions
Although the coefficient of variation for sown area of wheat, rice,
maize, "other crops," and total foodgrains showed some trend decline over
the entire three and one-half decade period due to increasing administra-
tive planning and control over sown area, the coefficients of variation for
yield and production showed considerable variability and no long-term
dominant trend, despite the measured increase between the 1952-57 and
1979-83 terminal periods. Between these two periods there were three
main subperiods of higher variability: the late 1950s and early 1960s for all
crops, 1966-73 for maize, and 1974-82 for wheat. Policy and administra-
tion seem to have been important causes of variability. Weather was highly
contributory in the first period and partially so in others. But aggressive
state-conducted area expansion efforts for wheat and maize in risk-prone
areas for which weather patterns were highly correlated must be consid-
ered as both weather and policy related. The proliferation of disease-sus-
ceptible maize hybrids may also have been contributory, but any major
problem appears to have been resolved before the mid-1970s. For wheat,
there may have been a similar phenomenon related to the rapid prolifera-
tion of semidwarf wheat varieties beginning around 1973, but increasing
concentration of production in areas with correlated patterns of weather
and irrigation water availability and variability associated with centralized
control of inputs, especially fertilizers, appear more important.
All in all, the lack of trend decline in production and yield cvs (despite
area stabilization, multiple cropping increases, and substantial risk-re-
ducing capital construction efforts and technology adoption) is predomi-
nantly related to increased yield correlations among provinces. Between
the two terminal periods for which variance decomposition exercises were





Variability in Chinese Cereal Production 59

conducted, the proportions of increased foodgrain production variance ex-
plained by increase in interprovincial covariances were high: 88 percent for
rice, 85 percent for wheat, 73 percent for other food crops, and 91 percent
for all food crops taken together. Hypotheses for explaining this phenome-
non include increased provincial responsiveness to central policy, in-
creased market involvement, and especially the interaction of centralized
policy with supply of the agricultural inputs now critical to yield levels,
particularly manufactured fertilizers.







4 An Analysis of Variability in
Soviet Grain Production

JOHN R. TARRANT









One of the most dramatic changes in the grain markets since 1970 has
been the rise in importance of the U.S.S.R. as one of the world's largest
importers (Tarrant 1984b). Apart from a brief period in the mid-1960s, the
U.S.S.R. was a net exporter of grain until 1972 when a fundamental policy
switch led to rapidly rising imports. This policy change was linked to other
policies to increase consumer expenditure. Incomes, especially of urban
Russians, had been rising (Laird 1982), but Soviet production of consumer
goods had not been able to meet the rising demand. One important area
where production had not kept pace with demand was in meat and other
livestock products. The policy reaction to poor grain harvests had been to
use livestock herds as a grain buffer by slaughtering in times of feed short-
age and building up herds when grain was available. Following the intro-
duction of the Five Year Plan in 1971 (Barbakov 1972) livestock numbers
were planned to increase at a much faster rate than had previously proved
possible. If livestock herds were not to be slaughtered in times of poor har-
vest, the only alternative was substantial grain imports. Protecting and
building up livestock numbers meant that, although the grain harvest in
1972 was not far below trend, imports were necessary, and they have con-
tinued to rise (figure 4.1).
Total grain imports reached about 55 million metric tons (Mt) in
1984/85. This is estimated to be at or even above the port handling capac-
ity of the U.S.S.R. without serious disruption to other forms of trade (Tar-
rant 1981). Although this capacity is expected to increase, the future level
of Soviet imports is far from clear. One might expect them to remain at or
close to the recent maximum until production of domestic grain (and other

The author gratefully acknowledges generous financial assistance for the research leading to
this paper from the Economic and Social Science Research Council of the U.K. and from the
International Food Policy Research Institute, Washington, D.C. Comments from the editors
and others have been most helpful, but the responsibility for the opinions expressed remains
with the author.





Variability in Soviet Grain Production 61


FIGURE 4.1 Cereal production in the Soviet Union and net cereal imports


Net imports
(million metric tons)


Production
(million metric tons)


-101I I I I I I OI100
1955 58 61 64 67 70 73 76 79 82 85

SOURCES: FAO trade yearbooks and compiled from various U.S.S.R. sources by the Foreign
Agriculture Service of USDA.


fodder crops) increases sufficiently to ensure that large livestock herds can
be fed from domestic resources. If, however, domestic feed production
does increase, there will be a falloff in import demand. Current world
prices for grains are low despite the record level of Soviet demand. The
price effects of removing a substantial part of this demand would be con-





62 Evidence on Patterns of Changing Yield Variability


siderable. It is, therefore, vital to understand the factors in Soviet agricul-
ture that might bring about such a change.
Soviet grain production is characterized by great variability-a fact
that has been noted by many authors (Laird 1982, Kogan 1983). Variabil-
ity is also increasing in both space and time. The five years until 1985/86
were characterized by poor harvests and this has encouraged the buildup
of imports. If the pattern of increasing variability continues, the market
might well expect substantial falls in Soviet import demand associated
with those years when harvests are good. The magnitude of the variability
is much greater than imports (figure 4.1), hence a relatively small upturn
in Soviet production, especially if it continued for more than one year,
could more or less eliminate the need for imports.
This level of production variability is closely and linearly associated
with variability in yields and has little to do with the area planted, which
has been falling marginally since the expansion in the 1950s into the
"newlands" of Kazakhstan and western Siberia (Nguyen 1985, Tarrant
1984a).' This chapter will concentrate on production rather than yield be-
cause there are generally more data on production, and it is grain produc-
tion which determines food consumption and trade. Aggregated cereal
production will be used due to the availability of data, rather than the pro-
duction of separate cereal crops. The study will be mostly confined to the
Russian Soviet Federated Socialist Republic (RSFSR), Belorussia, Molda-
via, Ukraine, and Kazakhstan. These regions account for approximately
90 percent of Soviet cereal production.
There are many problems with using such Soviet data. The time series
are short and there are gaps in the data runs-gaps which are not consis-
tent throughout the U.S.S.R. Very little agricultural information is avail-
able after 1975 at an aggregate or local level. The data which are published
have to be interpreted with caution. The crop area recorded is that planted
not that harvested. In severe conditions the crop planted may be ploughed
in and another planted-this is especially true in those areas where winter
wheat may not survive the winter dormancy period and may be replaced by
spring wheat. Yields have been recorded in many different ways in the
U.S.S.R. In the early years biological yields were used (Hedlund 1984).
Samples were taken of the growing crop and yield estimates were made on
this basis. These produced considerable overestimates of true yields, espe-
cially in years of poor harvesting weather conditions. Currently the Soviets
use "bunker yields"-that is, yields taken into the combines (Severin

1. A linear regression of production (Q), measured in millions of metric tons, against
yield (Y), measured in metric tons per hectare, gives
Q = -2.85 + 126.5 Y
with a coefficient of determination (R2) of 0.977.





Variability in Soviet Grain Production 63


1984). Excessive moisture and often a severe shortage of vehicles to deliver
the grain to elevators mean that the use of bunker yields probably overesti-
mates yields and production by at least 10 percent over methods used in
the United States and Europe (Hedlund 1984). Despite the difficulties, So-
viet data do show a high degree of internal consistency, and it seems un-
likely that the variations in overall production figures are seriously
unreliable.
The production variability in the U.S.S.R. has both spatial and tem-
poral elements interlinked. A marked feature is a two-year cycle where
good harvests are almost invariably followed by poorer ones (Kogan
1981a). Good production in one year tends to deplete soil nutrients and
moisture so that production is reduced the following year. This situation is
aggravated where production is close to the dry margin for grain cultiva-
tion and where supplies of fertilizer are not reliable. Both these examples
relate to the majority of Soviet grain lands and thus a marked two-year
cycle is not unexpected. One of the most striking features over the past five
years has been that the two-year cycle has been replaced by a run of poor
harvests.
Another notable feature of Soviet production is that it is spread over
at least three contrasting production regions. Most cereals are produced in
the chernozem soil belt which can be divided into the dry east and the
moist west. Increasingly important is the nonchernozem production area
to the northwest of the country. There is little doubt that the bulk of the
annual variability in production, superimposed on the two-year cycle,
results from climatic conditions (Tarrant 1984a, Meshcherskaya 1983,
Kogan 1981a,b) and that climatic conditions will often not be the same
over these different production regions. A poor harvest in one area may be
compensated by a good one elsewhere. This will be referred to here as the
regional compensation effect (RCE). Once established, such regional con-
trasts will tend to be perpetuated in most years by the two-year production
cycle. Exceptionally poor (or good) national harvests will be produced
when all the major producing regions (and different cereal crops) move
synchronously and there are no regional compensation effects.


Spatial Variability
If increased production is accompanied by no other significant
changes, absolute spatial variability of production will rise through time-
with fluctuations about a trend line (figure 4.2a)-as it will be directly de-
pendent on production levels. Relative spatial variability will, therefore, be
constant through time. There is an extensive literature in ecology concern-
ing the relationship between mean population size and variability (Taylor,
Woiwod and Perry 1978, Anderson, Turner and Taylor 1979, Taylor and





FIGURE 4.2 Spatial variability in Soviet cereal production


Spatial variability / production


Spatial variability


EL

I



Spatial variability

soURCES: Narodnoye Gospodarsvo Ukrainskoi: Ukraine 1957, 1969, 1971, 1972, 1983.
Narodnoye Khozyaistvo Ukrainskoi: Ukraine all other years. Narodnoye Khozyaistvo Ka-
zakhstanskoi: Kazakhstan all years. RSFSR v. Tsyfrakh: RSFSR 1966. Narodnoye Kho-
zyaistvo RSFSR: RSFSR all other years. Narodnoye Khozyaistvo Byelorusskoi: Belorussia all
years.
NOTES: (a) The expected relationships between spatial variability, production, and time.
(b) The spatial variability of cereal production within regions of the U.S.S.R. The standard
deviation of cereal production in all oblasts of each region is plotted for each year for which
there are data. The data for Kazakhstan, although following similar patterns, have been
removed for clarity. (c) The spatial variability of cereal production related to average produc-
tion levels within the U.S.S.R. For each year's data, the mean and standard deviation of






Variability in Soviet Grain Production 65


Woiwod 1980). Linear relationships have been established for a wide vari-
ety of species of plants and animals.
Evidence from the U.S.S.R. fits this general model well. The first test
applied was to take the concept of the mean correlation field, as used by
hydrologists to describe spatial patterns of precipitation within river catch-
ments (Ward 1967, Hendrick and Comer 1970). When applied to patterns
of Soviet grain production the mean correlation field is the correlation be-
tween the production of each oblast (political subdivision of a republic in
the U.S.S.R.) and all others over a period of time. Dividing the data set
into two time periods (table 4.1) shows that, in three regions, oblasts are
decreasing similarly over time. Spatial variability increases over time in
response to increased production (figure 4.2b). The rate of increase is a
reflection of the inherent variability within the regions analyzed (figure
4.2c). The RSFSR includes a large area in the central U.S.S.R., stretching
from the far north to the Caucasus. Belorussia, on the other hand, is a
small region on the border of Poland, which has more uniform conditions.
In all cases the increase in variability is linear with production and (figure
4.2d) confirms that variability is increasing in proportion with overall pro-
duction. Variability between regions is examined below.


Annual Variability
As with spatial variability, if stochastic processes alone were operat-
ing, one would expect year-to-year variability in production to increase in
proportion to increases in total production (figure 4.3a). Indexing the pro-
duction to 1960 (figure 4.3b) facilitates a comparison of both the relative
growth rates of production in the regions and the year-to-year variability.
During the period 1960 to 1975, production data are available for the 104
oblasts in these four regions. The mean production of each oblast over this
period is plotted with the mean deviation from its production trend. Al-
though there is a considerable scatter (figure 4.3c), 74 percent of the vari-
ance in the mean deviation from trend can be explained by changes in
mean production.
A comparison of absolute and relative annual variability through time
is made difficult by the short and broken data runs. An attempt is made in




production of all oblasts are plotted. The regular increase in spatial variability is clear as is
the different rate of increase depending on the inherent variability within the regions. (d)
Spatial variability of cereal production relative to production through time. Although abso-
lute spatial variability is clearly increasing (see b), it is increasing at no faster a rate than is
production. In the RSFSR, the Ukraine, and Belorussia relative spatial variability may be
falling slightly.






66 Evidence on Patterns of Changing Yield Variability


TABLE 4.1 Spatial variability in cereal production within Soviet regions

Time Perioda

Region Whole Early Late

RSFSR 0.38 0.16 0.18
Belorussia 0.93 0.95 0.83
Ukraine 0.66 0.56 0.49
Kazakhstan 0.54 0.52 0.41

Sources: See figure 4.2.
Note: The table indicates the mean correlation of every oblast with every other over all years,
an early period, and a late period. The magnitude of the correlation coefficients reflects the
degree of internal diversity within the region. Data limitations require the use of different
time periods in each region.
a E l....


1940, 1953, 1959, 1960-65
1960, 1965-81
1940, 1960, 1965-69
1940, 1950, 1953, 1955, 1958-61, 1965


1966-73
1972-79
1970-75
1966-71, 1975, 1976, 1980, 1981


TABLE 4.2 Annual variability in Soviet cereal production through time

E(y-f)/(n-1). (E(y-y)y)/(n- i)b

Region Earlye Middle Late, Earlyc Middle, Late,

RSFSR 189.8 262.3 241.0 1.01 1.01 1.13
Belorussia 29.7 44.9 57.0 1.09 1.06 1.05
Ukraine 200.8 319.8 315.2 1.15 1.08 1.06
Kazakhstan 161.5 170.7 184.9 1.28 0.91 1.19

Sources: See figure 4.2.
Note: The three time periods are not continuous but have been arranged with overlaps to give
an equal number of years in each time category.
"Average deviation from trend X 103 metric tons for all oblasts in the region.
bAverage relative deviation from trend for all oblasts in region.
Early Middle Late
RSFSR 1940-1965 1962-1971 1966-1975
Belorussia 1960-1971 1968-1975 1972-1979
Ukraine 1940-1968 1966-1972 1969-1975
Kazakhstan 1940-1965 1958-1971 1966-1981



table 4.2. Each run of data for each region is broken into three time pe-
riods. These periods are not the same for each region because of differ-
ences in the regional data availability. The periods overlap to ensure that
there are at least seven years in each period. The selection of the time divi-
sions is made simply to ensure equal numbers of years in each case. Abso-
lute variability increases in each region, although showing a slight drop in


RSFSR
Belorussia
Ukraine
Kazakhstan







Variability in Soviet Grain Production 67


FIGURE 4.3 Annual variability in Soviet cereal production


(a)







Time


Production index 1960 = 100
400
360 (b)
320 Belo
280
240
200
160
120 Ukraine
80 RSFSR -,
40 Ka
40 Kazak i--

1940 48 56 64
Year


Production Time

(y P)/n 1 x 10 metric tons
2,000 -.
1,800 (c) o
russia 1,600 0 0
1,400 //
1,200

1,000
SJ\Ukraine 80 8
800 o
600 o~~
400 000 0 o
zakhstan 00
200 0

72 80 0 1,200 2,400 3,600 4,800 6,000
Mean production x 103 metric tons


SOURCES: See figure 4.2.
NOTES: (a) The expected relationships between annual variability, production, and time. (b)
U.S.S.R. regional cereal production (indexed to 1960), 1940-80. (c) Annual variability in
cereal production related to mean production. Mean production for each oblast compared
with its mean deviation from its trend. For each oblast the production trend through time is
established using available data. The average deviation from this trend (ignoring the sign) is
plotted with mean production for the same time period. The trends and mean production
levels are established for 1960, 65, 66, 67, 68, 70, 71, 75.




the RSFSR and Ukraine between the middle and late period. As expected
these increases are greatest where the production trend is steepest (figure
4.3b). Relative mean annual variability is shown by the second part of ta-
ble 4.2 where the difference between the actual and predicted (trend) val-
ues of production are divided by the mean production for that period. Rel-
ative variability shows much smaller changes and indeed slight declines in
three of the four regions. The statistical reliability of this analysis is not
high because of the data problems and limited numbers of observations,






68 Evidence on Patterns of Changing Yield Variability

FIGURE 4.4 Diverging trends in Soviet regional cereal production

Cereal yield
Metric tons per hectare
4 -
(a)


23


2 rE\ >


01 1 I 11 1
1955 57 59 61 63 65 67 69 71 73 75 77 79


SOURCES: (a) Compiled from various Soviet sources by the Foreign Agriculture Service of the
United States Department of Agriculture. (b) See figure 4.2.
NOTES: (a) Trends in cereal yields for a selection of Soviet economic regions. (b) The slopes
(3 coefficients) of the regression lines of cereal production through time for each oblast. The
trends were established for the eight years as in figure 4.3 for which there are data for all
oblasts. Over the majority of the country the slopes are not significantly different from zero
(partly as a result of the small number of degrees of freedom). The contrast between the
generally rapidly growing areas of the west and the stagnating east is striking.





Variability in Soviet Grain Production 69


but the results do fit closely to the relationships hypothesized in figure
4.3a.
Increases in spatial and temporal variability in the main Soviet grain
production regions are no larger, and in some cases slightly smaller, than
would be expected as a result of the increased levels of production. The
situation and estimates for the future are complicated, however, by the
effects of production conditions and rates of change in different produc-
tion regions. It is necessary to look at variability between regions.


Variability between Regions

Changes through time in the production of different regions result
from a combination of diverging trends in production and variability about
those trends. Figure 4.4a shows differing trends of cereal yields for some
economic regions in the U.S.S.R. Clearly, the rates of growth in cereal
yields are very different and this will be the major determinant of produc-
tion. The further east the slower the rates of growth. Figure 4.4b shows the
slope values for the trend lines of production of each of the 104 oblasts in
the regions under study. Over a large part of the country the slopes of lin-
ear trends are not significantly different from zero. In much of the western
producing areas the rates in growth of production are considerable-al-
though even here there are a number of oblasts that show little or no
growth. These data, however, show a marked difference in the trends of
production between the east and the west. Further evidence of this is pro-
vided by table 4.3
The establishment of trends in production data is complicated by the
selection of appropriate start and finish dates. This becomes a major prob-
lem when the data run is short and when it includes some extreme years.
Seven years of data (eliminating 1975 which was such a universally poor
year that it would unrealistically reduce the production trends for such a


TABLE 4.3 Cereal production trends in Soviet regions
Annual Annual Increment
Increment as Percent of 1960
Region R 2 (X 103 metric tons) Production
RSFSR 0.58 3250 4.6
Belorussia 0.69 235 10.6
Ukraine 0.72 1323 6.1
Kazakhstan 0.08 438 2.3

Source: See figure 4.3.
Note: Trends calculated for seven of the eight years for which data are available for all oblasts
in all regions.





70 Evidence on Patterns of Changing Yield Variability

FIGURE 4.5 Spatial contrasts in Soviet production


* >2 standard deviations below mean 0 >2 standard deviations above mean


* I to 2 standard deviations below mean

* 0 to 1 standard deviations below mean


0 1 to 2 standard deviations above mean

o 0 to 1 standard deviations above mean


SOURCES: See figure 4.2.

NOTES: Oblast deviations from trend for selected years. Deviations were measured even if the
trend established for that oblast was not significantly different from zero. In such circum-
stances the deviations are essentially the same as deviations from the mean production over
this time period.



short data run) show very different annual increments of production in the
four regions, ranging from about 10 percent of 1960 production in the fast-
growing western area of Belorussia to just over 2 percent per year in the
slower growing Kazakhstan. As annual variability is related to overall pro-
duction levels, so the effects of diverging trends will be accentuated by di-
verging variability about these trends.


Regional Compensation Effects

Overall production variability in the U.S.S.R. will be greatly reduced
if, in one year, poor production in one area is compensated for by good





Variability in Soviet Grain Production 71


production in another. An examination of the spatial pattern of each ob-
last's deviation from its trend shows that in 1966, for example (figure 4.5),
conditions in the nonchernozem area of the northwest of the country were
poor, and that production was well below trend-for many oblasts two or
even three standard deviations below. At the same time, however, much of
the rest of the country had production well above trend, especially in Ka-
zakhstan and the southern RSFSR. In 1975 almost all oblasts were below
trend-a fact which produced the disastrous overall production level for
the U.S.S.R. that year (figure 4.1).
The main causes of these regional contrasts are climatic. There have
been many studies of the interaction of climate and crop production in the
U.S.S.R. (Desai 1981, Kogan 1981a,b, Berentsen 1982, Kogan 1983,
Meshcherskaya 1983, Tarrant 1984a). Most have been based on some
form of multiple regression analysis where a number of weather variables
are regarded as the independent variables producing variations in the de-
pendent variable, either production or yield. Such models have shown a
high degree of statistical explanation of the annual variability in cereal
production. Unfortunately, it is difficult to test these models outside the
range of the data used to calibrate them. Soviet sources of agricultural
data have been very limited recently and such models have had to be tested
against Soviet production estimates derived from U.S. Department of Ag-
riculture (USDA) and Food and Agriculture Organization (FAO) sources.
These estimates are themselves based, at least partly, on the weather con-
ditions during the growing season in the U.S.S.R., thus ensuring a reason-
able fit of the regression models.
It is well known that plant growth, and therefore final production of
grain, will be radically affected by weather conditions at critical times dur-
ing the growing season (Mostek and Walsh 1981). This produces two com-
plications. The first is that very detailed climatic analysis becomes neces-
sary-using daily weather records if possible. This is difficult for a large
area with many weather stations, and daily weather data for many Soviet
stations have only recently become available in the West. The other com-
plication is that the critical time or times will not be the same from one year
to the next and from one area to another, which makes a time series analy-
sis very challenging. Using very generalized data, this chapter presents just
one illustration of the role of climatic variability.
Figures 4.6 and 4.7 show, for those weather stations reporting data
over the period from 1951 to 1984, deviation of the mean spring and sum-
mer temperatures and the total spring and summer precipitation in 1965
and 1975 from their 35-year means. As most of the south and east of the
Soviet grain growing area is in water deficit (Kelly et al. 1983), the combi-
nation of cool wet conditions would be expected to produce relatively good
production, and conversely, hot dry conditions to produce poor produc-






72 Evidence on Patterns of Changing Yield Variability


FIGURE 4.6 Climatic factors in variability of Soviet cereal production, 1965


0 >2 standard deviations below mean

* I to 2 standard deviations below mean

* 0 to 1 standard deviations below mean


0 >2 standard deviations above mean

0 1 to 2 standard deviations above mean

o 0 to I standard deviations above mean


SOURCEs: World weather records, U.S. Department of Commerce and Monthly Climatic
Data for the World, NOAA, United States.

NOTES: (1) Precipitation in spring; (2) Temperature in spring; (3) Precipitation in summer;
(4) Temperature in summer. Summer is defined as the months of June, July, and August and
Spring as March, April, and May. For most of the weather stations the mean weather vari-
ables were established over the period 1951-83. Not all stations reported data for all months
in all years, thus there is not an equal number of stations on all maps. In all cases weather
variables showed no significant trends so are expressed as deviations from the mean.





tion. In 1965, when production in the west was above trend and production
in the east below trend (figure 4.5) the spring and summer were generally
cooler than average in the west, and hotter than average in the east. Pre-
cipitation distribution, although more spatially variable in the summer,
generally reinforces this picture. In 1975 the very poor year for cereal pro-
duction resulted from a very hot and dry spring and summer over most
grain-growing areas (figure 4.7).





Variability in Soviet Grain Production 73


FIGURE 4.7 Climatic factors in the variability of Soviet cereal production, 1975


SOURCES AND DETAILS: See figure 4.6.
NOTES: (1) Precipitation in spring; (2) Temperature in spring; (3) Precipitation in summer;
(4) Temperature in summer.


The Role of Regional Compensation Effects

In years when regional compensation effects do not work, there will be
marked "blips," either upward or downward, from trend. These blips will
tend to grow in size as production rises and variability increases. As there
is a marked two-year cycle in Soviet production, once regional compensa-
tion is established, it will tend to continue in most years even if not specifi-
cally reinforced by contrasting climatic conditions. Comparing yields in
Belorussia and Kazakhstan, for example, shows compensating yield
changes in most years (figure 4.8a). Compensation is not confined to those
years when production in one region goes up and in another, down. Winter
wheat is mainly confined to the western parts of the cereal belt, while
spring wheat is grown everywhere but concentrated in the east. The annual
changes in production of wheat always lie somewhere between annual
movements of winter and spring wheat (figure 4.8b). Years when winter
and spring wheat production change in the same direction by a similar
percentage are rare. In 12 of the years illustrated, winter wheat had the
highest annual increase in production, and in 13 years, spring wheat pro-






74 Evidence on Patterns of Changing Yield Variability

FIGURE 4.8 Regional compensation effects in the Soviet Union
Cereal yield
Metric tons per hectare
3
(a) Belorussia I


2 i





Kazakhstan
1955 58 61 64 67 70 73 76 79 82
1955 58 61 64 67 70 73 76 79 82


% change in production from previous year
I I i I I I I I I I I I I I I I I I i I I I I


W = Winter wheat
S = Spring wheat


it w_
S S WS W W

So W S W

1 s. I Ii


(b)


* Total wheat


64 66 68 70 72


74 76 78 80 82 84


SOURCES: a. see figure 4.4a; b. see figure 4.1.

NOTES: (a) Cereal yields in Belorussia and Kazakhstan. Marked regional compensation can
be seen, especially in 1965, 1971, and 1972. Even in the recent years with poor harvests the
compensation has continued. (b) Percentage change in winter and spring wheat production
over the previous year. The percentage change in total wheat production is usually between
the two extremes of the winter and spring wheat (only the wheat crop with the fastest rate of
growth is labeled). The compensation effect is not confined to years when one wheat crop
increases and the other decreases.


1958 60 62





Variability in Soviet Grain Production 75


duction grew fastest. As winter and spring wheat have different planting
seasons, it is possible to induce these compensation effects by additional
planting of the spring wheat crop if the winter wheat fails during the winter
or early spring. The compensation effect between the two wheats is not
entirely human induced, however, and the movements are often in the
other direction.
Evidence has been presented (Nguyen 1985) that the regional com-
pensation effects have reduced as intercrop and interregion yield correla-
tions have increased. It is suggested that this is partly because of the intro-
duction of common technologies throughout the Soviet Union and a policy
of "sharing the shortage" of necessary inputs. This conflicts with Kogan's
view (1983) of the very unequal allocation of inputs to agriculture. Figure
4.8b shows that the annual percentage change in winter and spring wheat
production may have been unusually small recently (although representing
a large volume of production) but, within that reduced variability, com-
pensation between winter and spring wheat appears to be of continued
importance.
Of greater significance is that regional compensation effects are
eroded by strongly diverging trends in the cereal production in the western
and eastern regions. As faster growing regions have increasing annual vari-
ability, and regions that are not growing have little change in their annual
variability, total production soon fluctuates almost as greatly as produc-
tion in the region with the steepest production trend, even with a complete
coincidence between upturns in one region and downturns in the other. If
production levels in regions with contrasting growing conditions are grow-
ing in parallel then the damping effect will remain constant. There is every
reason to expect, therefore, that unless the trend of production in Kazakh-
stan and western Siberia improves, the production differences between
east and west will increase and the annual variability in total Soviet pro-
duction will grow faster than would be expected from the growth in total
Soviet production alone.


Policy Implications
If livestock herds are not to be adjusted to the size of the cereal and
other feedcrop harvest, annual variability becomes a determinant of the
size of the import requirement, the size of the necessary buffer stock, or
both of these. Marginal production conditions over much of the grain-
growing area of the Soviet Union ensures that there is great annual fluctua-
tion in yields and harvests-fluctuations that are probably relatively
greater than in any other major cereal-producing country. As production
increases, so will this annual variability, thereby increasing the uncertain-
ties of the size of the Soviet import demand. These uncertainties will be





76 Evidence on Patterns of Changing Yield Variability


accentuated by the diminishing importance of regional compensation ef-
fects if current trends continue. As imports are already close to port-han-
dling capacity and there are obvious practical and financial limits to the
size of security stocks, future shocks to the world cereal market are likely
to follow a sharply reduced Soviet import need following a coincident up-
ward movement in harvests in all Soviet grain-producing regions.
The Soviet Union obviously has a need to increase overall production
and thereby to increase meat production. Meat consumption is still very
low (Markish 1982, Hedlund 1984), and poor cereal harvests led to a
downturn in cattle numbers and milk production in the late 1970s and
early 1980s in spite of high levels of grain imports, and in sharp contrast to
the demands of economic plans (USDA 1983c, 1984a, Tarrant 1984b). Al-
though grain production may have recovered in 1985 (USDA 1984a), it will
still be below the trend established from 1960 to 1980. Equally important,
however, is that production increases should be achieved with the mini-
mum increase in the annual variability that is already too great to be easily
absorbed through the use of buffer stocks. Policies that are directed to
stability are at least as essential as those directed at increased production.
There are two regions where there is potential for rapid increases in pro-
duction. The first of these is the nonblack earth region of the northwest-
where increased production can be associated with land drainage and soil
improvements (Rostankowski 1980). Unfortunately, much of this region
has already shown very rapid improvements (figure 4.4) and a policy to
accelerate the production trend even further, although it would certainly
increase overall production, would also accelerate the end of effective re-
gional compensation effects and sharply increase the annual variability of
total Soviet production. The second region where there remains potential
to increase production is in the dry east. Here new irrigation investment
and more effective use of existing facilities would provide the increase
(Micklin 1978). This increased production may bring with it increased
variability. Provided that the variability moves, at least in most years, in a
less extreme manner than in the west, it will represent a positive benefit.
As production increases, so upturns in eastern harvests can compensate
more fully for poor western harvests and vice versa. This provides the
strongest case for the Soviet plans to divert rivers southward to the Aral
Sea and lower Volga, and other less grandiose irrigation investments
(Kelly, et al. 1983, Micklin 1983, Pryde 1983).

Conclusions
Absolute spatial and annual variability in Soviet grain production will
continue to rise from already high levels. Absolute variability is higher
than can be economically and practically catered to by buffer stocks, and





Variability in Soviet Grain Production 77


imports are possibly near their maximum. Additional downward variabil-
ity in production will have to be accommodated largely through further
reduced consumption. The Polish food riots in the 1970s showed the Soviet
government how dangerous such policies can be. If production varies
sharply above trend, western exporting countries can expect an abrupt re-
duction in imports (USDA 1984a). Soviet import demands can be expected
to be much more variable in the future with a great potential for market
disruption (Kogan 1983).
Provided concerns over environmental damage can be overcome,
there is likely to be continued pressure for river diversion schemes. Other
policy and research efforts in agriculture are apt to be continued with in-
creased urgency. These include the development of new techniques of fod-
der production as well as land improvement projects in the northwest. It is
unlikely that further newlands remain to be developed-unless there is
widespread increase in irrigation, which might come from the river diver-
sion schemes.
There is no evidence that increased uniformity in the application of
agricultural technology has led to significantly reduced spatial variation in
production and, therefore, to increased annual variations. Access to agri-
cultural technology is far from uniform and, together with the range of
climatic conditions throughout the country, that ensures continued spatial
variability.
Future changes in relative annual variability will be dependent on
changes in relative spatial variability, itself dependent on changes in (a)
scatter about trends in the different production regions, (b) the divergence
of trends in different production regions, and (c) the compensation effects
of deviations from trends within different production regions. There is no
evidence for increased relative scatter about trends, but the trends them-
selves are sharply diverging. Therefore, although the compensation re-
mains at present, its effectiveness is rapidly being reduced.
Diverging production trends will increase annual total harvest fluctu-
ations relatively and absolutely, unless adoption of technology throughout
the country can reduce the divergence of production trends, and thereby
reduce spatial variability, maintain the effectiveness of regional compensa-
tion effects, and reduce annual variability. Of various possible scenarios,
the most probable is that diverging trends in production between east and
west will reduce regional compensation effects and produce greater than
expected increases in annual variability.








5 Agricultural Planning Policy and Variability in
Syrian Cereal Production

HUNG NGUYEN









Few countries experience such an extraordinarily high degree of variability
in national cereal production as Syria. In 1986, for example, national bar-
ley production was only 15 percent of the previous year's production
(Bakour 1984, p. 8). Such fluctuations are a long-standing phenomenon
and originate largely from Syria's highly variable rainfall. There is evi-
dence that the variability of national cereal production has increased in
recent years, and this may be related to change in technology and agricul-
tural policy.
This chapter examines recent changes in the patterns of variability in
Syrian cereal production. Variance decomposition is used to identify the
importance of various contributing factors to the increase in variability.

Background
Because of Syria's dry climate only about 3.4 million hectares are un-
der cultivation while 2.5 million hectares of cultivable land are left fallow
(Mukhitdinov 1974, p. 61). In areas receiving less than 250 millimeters
(mm) of rainfall annually (which is the case for 60 percent of cultivable
land), a two-year crop rotation system is adopted-the first year under
grain, the second fallow. In areas receiving 400 to 600 millimeters annually
(about 10 percent of cultivable land), a three-year crop rotation system is
the common practice-the first year under wheat, the second under bar-
ley, and the third fallow. Rainfall occurs primarily during the winter, and
the planting of winter crops depends crucially on the level of rainfall in
November. El-Sherbini (1979, p. 145) has shown that the correlation be-
tween early and total rainfall is highly significant for every province of
Syria. There has been a large increase in the area of irrigated wheat since
the early 1970s and in the adoption of high-yielding Mexican wheat vari-
eties. Barley remains an essentially rainfed crop. These two crops provide
an interesting contrast for the study of production variability. Oats, maize,




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs