a tale of two countries pot

NYY wrersstionas ro00 À' POLICY RESEARCH INSTITUTE IFPRI" ‹...«.x IEPRI Discussion Paper 00758 A Tale of Two Countries Spatial and Temporal Pattems of Rice Productivity in China and Bra

NYY wrersstionas ro00

À' POLICY RESEARCH INSTITUTE

IFPRI" ‹ «.x

IEPRI Discussion Paper 00758

A Tale of Two Countries Spatial and Temporal Pattems of Rice Productivity in China and Brazil

Liangzhi You

Environment and Production Technology Division

INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE

gnicltral esearch enters that reeve prinipal funding from governments, private foundations, and

International Agricultural Research (CGIAR),

FINANCIAL CONTRIBUTORS AND PARTNERS

‘ontnbutors and partners IFPRI raeflly acknowledges the gencrouswnrestctd funding rom

‘Austalia, Canada, Chia, Finland, Frane, German, India, leeland aly, Japan, Netherlands, Noeway Philippines, Sweden, Switzeriand, United Kingdom, United States, and World Bank

par utieaton 58B đoooslo papers wun De deere ste The eer ste te avaltechFPRIC

“robste at wi lơ oigpabeobetplbe hhntdp

IFPRI Daeusion Papers contin prtminay mater and esschresute, They have nl bee subject fra

‘ena trea managed ty FPR Pubeaona Review Cormies bl have been tevcwed by me sne

"Re and xtema fevewer They ae clted nero somata dacuesion an real Comment

Smuncitons ices csgratncsar

Contents

‘Acknowledgements

Abstract

1 Inraduetion

2, Data and Rice Production Systems

3 Spatial and Temporal Patterns of Rice Vield

4 Underlying Causes

5 Conclusions

Appendix A: Generalized Entropy Index of Spatial Yield Variability

Appendix B: Rice Production Systems in China and Brazil

List of Tables

1b Spatial variability of rice yield, Brazil (1975-2000),

2 Rice production by seed varieties i ei

Spatial change of rice yield in China, 1998-2000

Spatial change of rice yield in Brazil, 1998.00

Rice yield distribution in China, 1980-82

Rice yield distribution in China, 1998-00

Rice yield distribution in Brazil, 1975-77

Rice yield distribution in Brazil, 1998-00

Spatial variability of rice yields in China and Brazit

Spatial variability of rice yield in China, 1980-2000

Spatial variability of rice yield in Brazil, 1975-2000

Spatial variability in rainfall and upland rice yield in Brazil

‘Temporal variability in rainfall and upland rice yield in Brazil

Rice production systems in China

B2 Rice production ystems in Brazil

"

2

7

1b 15s l6

2 2

ACKNOWLEDGEMENTS,

‘We thank Xiaobo Zhang for sharing his SAS codes on calculating generalized earopy index and Eduardo Casclo-Magahaes for eitoal help We are thankful to seminar participants at 2007 UNU- WIDER Conference on Southern Engines of Global Growth: China, India, Brazil and South Atica at elsinkt for elpl discussions and comments on preliminary ess We would als thank the two anonymous reviewers andthe eter of IFPRI discussion pape series for valuable comments and suggestions onthe paper Any remaining errors ae solely our responsibility

ABSTRACT

‘This paper examines differences inthe spatial and temporal variations of rice veld in China and Brail

bocome increasingly homogeneous In contrast, ice vielde in Brail have diverged over ime, primi due to variations in upland ee yields Thee hypothetical explanations may account forthe efeent teshaviors of rice vel in Bra and China, namely: 1 erences in production systems (i.e iigated in Chia vs upland im Braz 2) changes in rainfall pattems and 3) bias in agncutural sear and

tstablishing that: 1) upland rice shows much more variation in isis compared to iigated ie; and 2)

‘hanging rainfall pater have primarily affected upland rice We also provide evidence of the bias

‘Keywords: rice productivity, sp

convergence, technology spillover, China, Brazil

4 INTRODUCTION

‘lee is widely produced and consumed in China and Bravia sa valued commodity in both

‘many fanners Over the past ew decades, these countries have invested significant efforts toward

improving ree prodetvty and increasing production Thee efforts hae largely paid off in terms of roducien amlsields tothe pont tat China and Brarl together hive accounted for roughly one thitd of

important and islusatal players in the word's ice market,

Increases in rice productity have hoon the major source of production growth in both Beil and (China, The development and eventual adoption of highesielding vaites(HYVs) during the Green Revolution played an important and significant ole inthis productivity improvement (Fan tal 2008, Sanint, 2008) Rice yields increased 25 an 15 percent pe year for China and Bra respectively botween 1970 and 2000 This rapid growth in producti allowed China nd Brat meet the sowing

however, were not uniforms dstbuted aross rieegrowing areas Ih Fel, sgnfieant ariation ean be iserved across diferent rie ecologies, agroscologieal zones, demographic pressures and policy

land suitable for nce production suzyes that China and Brazil ned to further increase ice producti if they hope te continuc meeting the inreasing demand for food The search for new sources of productivity sromth canbe aided by improving our understanding of the spatio-emporal evaluation of rice yield (Wood, You and Zhang 2001)

Technology spillovers account fora significant sare of agricultural productivity growth, and some studies suggest that research and development (R&D) silovers might account for half or more of the total pradtisiy growth (Alston, 2002), Given the generally easy acces lo aicltral echnologes, technology Tatzcomers may readily "eat yp simply by adopting existing technologies superior thee

‘own (Wood, You and Zhang, 2004), This should be the case tn parteular fr countries like China and Braviwhere agricultural extension services ar relatively song and effective Ihe adoption of new and beer technologies is indced a simple process in China and Lal given the widespread

‘dissemination of such technologies (though extension services) andthe effects of spillovers, then we

developing counties converged! to levels found in developed countries fom 1961 to 1999 fr most ofthe

‘ight erops included in he suy (bares, eoto, mae, mille, sorghum, sosbean and wea), Using hybrid ie in India san example, Zhang, Fan and Cai (2002) showed that erly suecessfol HYV

‘Mopters had a age effect on seighboring farmers, which vanslated inc higher technological adptio by father farmers This suggests that technological spillover i the centrpcal force fr productivity

production i subjet to substantial spatial heterogeneity in tems of soi terrain and climate, which can

‘pede technologieal transfer and adoption This isthe centrifugal free fr eropvield convergence

Caribbean didnot converge between 1975 an 1998, Given the variability of yields aeross production systems, raps and reions, aswell the lack of consensus fom peviows stds, the khu of rạp iElf onvergenee oer time and space remains largely an empirical question

‘Although large body of hterature deals with eehnology adoption and wansfer, most oF these

‘Boras and Serato-Domingo, 2007), Using a panel dataset of ie viedsin China and Bra the resent paper ills dis anal ial gap by examining spatial parts of rice yield variation and variability on a

the spatio-temporal changes for rice yields 2) Tests for yield convergence in the two counties ae applied: the results suggest convergcnee for China but not fr Brazil 3) Given that viclds converged for China bat not for Bra, we ws the Shorock inequality: decomposition method and Geograph

Information System (GIS) ool to analy the underiying eases ofthe diferenees observed betwen the

“hice hypotheses ar offered to explain the differences in rice yield convergence inthe Wo stuied counts

1 Differences in rice production systems: The majority of ice in China is iigated, whereas

‘hatin Bra i produced in a combination of igated and upland ecologies We hypothesize that these differences in rodtion systems contbute othe ysl! divergence in Braz

patter have changed over the past fee decades due to eimats change Ineeasing afl arias has exacerbated vied divergence in rainfed areas, thereby alfecting rainfed ie production hich relies on consistent rainfall during the growing cao

Agricultural RED bias toward iigated aoa: Intational snl domestic investments in

sicultaral R&D over the past few decades have been hea biased low ards gated

production systems, Ths bias benefits iigated ice mote than anf vce, We eliove tha the divergence in yoks in Braal is derived primarily fom the variability im upland rice

veld

‘The remainder ofthis paper i organized as follows We fst deseribe the panel dataset and ice production s\stoms in Brazil and China Next, We analy2etempoval and spatial yield vais in China tnd Bra, The final section investigates the underiving causes fr the differences in rice productivity convergence betweon these vo gounirics, We conclude witha summary and some policy implications

2, DATA AND RICE PRODUCTION SYSTEMS

‘We compiled time-series dala of rie production statistics (production, af and yield) atthe county level

‘ountes in China and 3.800 municipalities i Bra shich corresponds to 98 poreent ofall Chinese ountgs and BÉ fall Bravlan muneipliies Two GIS boundary les for Chinese counts and Bra

‘municipalities wor linked wo the corresponding slatstieal data In ation, we cleulated average

‘ainfall during the riee-growing season for all cunts in China fro 980 to 2000 an fr all

sveraging the rainfall valves ofall pixels within the counts municipalies, Anaual aiafall measures

the countiesimunicipalites in China and Brazil

Daring the std perio rice was grown via tre different production systems in China and Braviiigate lowland, rainfed lowland, and upland, The uilized production system impacts rez

ptformanee, and fundamental difeenees in plant chascersies and physiology make particular wes oF| Fice mote or ess suited to diffrent production systems For example, the modem semi-avae,high-

‘elding rites dosoloped during the Green Revolution forthe gated and favorable rainfed iow land som could not be grown in upland systems In China, irigated rice was the primary ie production

‘sem, accounting for over 93 percent of tol area sown orice, Ranfed lowland ice and upland rice sccounied fr 5 perent and 2 percent of the remaining ara, respectively Upland rice was tpiclly found

Im provinces that have mountainous regions, such asin Yunnan Guizhou, Guan, and Janes Ranfed losand rice was mainly planted in walr-limited areas, seh a those found in te provinces of Hebe Henan, Shangdong Shai nd Lisonng (ste Figure B fora map on rie production systems in China)

Jn Bra about one-third ofthe area planted with rice wat igaled, The remaining two-thirds were

predomisanly culated under upland systems, with only a small poreentage grown unde ante

inmigated A fow other stats such as Tocantins, S30 Paulo, and Mato Grosso do Sul produced limited

‘mounts of irigated rice Rainfed lowland sce was grown in only tree sates Sergip Minas Geras and Rio de Janeiro,

Since clatvely litle ofrie are in China and Bal was rainfed Howland, we would herein focus

an inigated and upland ce

‘nicer 20)

3, SPATIAL AND TEMPORAL PATTERNS OF RICE YIELD Figure 1a Spatial change of ree yield in China, 1980-2000

Figure 2a, Spatial change of rice yield in Brazil, 1975-77

Figures | and 2 show the spatial changes in rice yal over (he pas vo decades in China and Brazil providing snapshots of spatial yield variation a the sar and end years ofthe examined period Tivo specific alles emerge ffom these maps Fis, thee is significant spatial varauon of ice vilds in

the evant spatial variation sn veld performanes, For insanes, ie yields in the Noreen Chia Pin and Xinging province averaged abot 3 toma n 2000, while those in Northeast China were considerably higher, aeraging over? toma Likewise, in Beal, highly provetive states such s Santa Catarina and

performed considerably poorer with yields averaging 15 wha,

largest veld gains occurred in the Norheas egien andthe proines of Xinjiang In Bra the areas with largest yeldncreses included tte such as Koraima, Mato Grosso, nd Minas Gea, vhoreas Sana

Inner Mongolia andthe Sichuan provinces, Snir, comparison of Figures 2a) and () provides

‘evidence thatthe ries area expanded into the Brian sans, or “certados.” Mos of the nondfce-

Rondénia, Mato Grosso, and Bahia Indeed, upland rice cultivation has plea crucial role in bringing the Brarilan savannas under cultivation asthe low Rrully and acidic soso the region has ime the

‘ulation of the erops (Pineiro, Casto and Guimarses, 3010)

Figure 3a, Rice yeld distribution in China, 1980-82

A more quantitative sense of ie vild changes ma’ be gained fiom Figures 3 and 4, which show the viol istibution a the county (or China) and municipality (or Braz levels These histograms oF

‘eld disebution are plots of the harvested area within cath yield class, and represent about 2.300,

‘ountes in China and 3.800 muniipaies im Broil We ean se that he yield distbution in China

rice veis boh increased and converped during this period However the ease is rather differen in Bravil On aerage, Bravia rice siell also increased rom I-46 gna in 1970s 0298 ona i the late 1990s (compare Figures 4a) and (b), However, the ie viel in Bravil fr this period shaw 9 bimodal distribution, reflecting the to distinct ie production ssiems used in this county the st lustering of ice area inthe range of 0.6 10 26 toma presumably represents ce grown under the upland system, hile hatin the 46 1 62 tonha (34 104.6 tha in Figure (a) range mos ks represents 'mmgwed rice The bimodal distribution implies that yield growth has not been uniform across the two

dhferences in yield pattems between China and Bra, we used the decomposable generalized entropy” (GE) cass of inequality measres developed by Shorrock (1980, 1983), The GE mex which measires the overall spatial variability of yes, ean also be decomposed into sarople groups inorder to asess the

‘ontibation of individual groups otal savabilty and the sabi within and between eroaps

1 2000, eveaing a muck highor spatial variability in Brazilian yields compared to Chinese vals This apparent diferenee in the els of variability i confirmed by th ests ofthe GE analysis The GE

Peaks in T984 and 1988 In contrast the GE index fr Brazil iereass by 4.5% per yoar fom 1975

101993 and gradually decreases tereater These sll confi our nding that rice yields converged in Cina but not Brasil fram T9801 2000

10