Irrigation and rice production evidence in vietnam

PROBLEMSTATEMENT

Ricecanbegrowninawiderangeofweatherconditions,sothatmanycountrieshavegrownrice,e s p e c i a l l y i n A s i a w i t h about9 0 % riceproductiono f t h e world(Maclean,Hardy,& Hettel,2013).I nVietnam,riceisoneofthemaincrops,ratiooflandcultivatingricetototall a n d growingcerealsi supto86%

(VietnamGSO,2015).In2014,Vietnamwasrankedfifthi n theworldforriceproduction,afterCh ina,India,Indonesia,andBangladesh(FAOSTAT).From1 9 9 0 t o 2 0 1 4 , r i c e p r o d u c t i v i t y i n V i e t n a m increasedbyn e a r l y 1 0 0 %

(FAOSTAT).T h i s successiscontributedbymanyfactorssuchasapplicationofnewvarieties,tech nologyandextension ofirrigationsystem.

As of 2011, Vietnam boasted an extensive canal system spanning 254,180 km, which irrigated 85.5% of its rice cultivation land (Ha, Nguyen & Nguyen, 2015) However, this system faces significant operational challenges, including weak management, inadequate maintenance and dredging, and poor water quality (Tran, 2016) These issues raise critical questions about the effectiveness of irrigation and its impact on rice production in Vietnam While some studies have explored these concerns (Walle, 2003; Biltonen, Hussain & Tuan, 2003; Ut, Hossain & Janaiah, 2000), they primarily rely on data from as far back as 1991.

2000 Meanwhile,from2 0 0 0 t o nowadaystherehavebeenmanychangesi n Vietnami r r i g a t i o n systemandriceproduction.

Theroleo firrigationonagricultureandp o v e r t y alleviationhasb e e n research objectiveo fm a n y e c o n o m i s t s A largen u m b e r o f studiesw e r e conductedi n m a n y countries.A l m o s t researchresultsshowthatirrigationhelpstoincreasecrops’productivityandisasolutiontoreduc epoverty(Hussain&Hanjra,2004).Nonetheless,someofthemsuggestthatimpactofirrigationvar iesacrosssystemandevenacrosslocationsoncanalsystem(Biltonen,Hussain

&Tuan,2003;Jin,JansenandMuraoka,2012;Hussainetal.,2006;Hussain etal.,2004).More over,theseresultsaredifferentamongcountriesandregions.Hence,itisnecessarytoinvestiga tetheeffectofirrigationsystemonriceproductioninthecaseofVietnam.Inthist h e s i s , em ployeddataseti s VARHS2 0 1 4 , whichi s t h e latestsecondarydataandp r o v i d e s v e r y specifici nformationrelatedtoagriculturaloperationofhouseholdsinVietnamrural.

RESEARCHOBJECTIVES

Thisthesisfocusesoninvestigatingtheimpactofirrigationonannualriceproduction,thencomp areth e effecto f u s i n g waterf ro m canalf o r irrigationo n annualr i c e productionw i t h t h o s e o f o t h e r waters o u r c e s Finally,t h e e f f e c t s o f differentp o s i t i o n s o f communecanalsyste mwill beassessed.

RESEARCHQUESTIONS

(2) Howdoesi m p a c t o f i r r i g a t i o n o n ricep r o d u c t i o n v a r y acrosswatersource sandlocationsoncommune canalsystem?

SCOPEOFTHERESEARCH

The survey conducted in 12 provinces of Vietnam, including 5 in the Northern region, 3 in Central Coastal, 3 in Central Highlands, and 1 in the Mekong River Delta, encompasses data from 3,648 households It focuses on various aspects of rural development, particularly in agriculture, by gathering specific information on plot characteristics, types of plants, inputs, outputs, and irrigation methods Additionally, the survey includes details about household members and their respective communes, with a particular emphasis on rice farmers at the household level.

Douglasp r o d u c t i o n functioni s u s e d asl o g a r i t h m formf o r investigatingtheimpactofirrig ationonriceproduction.Toestimatethemodel,the clusterspecificfixedeffectatcommunele velwithclusteringstandarderrorisemployed.Inaddition,i n ordert o checkt h e robustness,O L S andFGLSestimationw i t h c l u s t e r - r o b u s t standarderrorareconducted.Althoughthethesiscancontrolunobservedvariablesaco mmunelevel,heteroskedasticityandcorrelationamongerrortermswithincommune,ithass o m e l i m i t a t i o n s , suchase n d o g e n e i t y o f i n p u t choice,p o t e n t i a l e n d o g e n e i t y o f irriga tionvariables, errormeasurementandomission ofvariables.

T HESIS ’ S STRUCTURE

Literaturereviewispresentedinthenextsection.Thethirdsectionisdataandmethodology.T h e f ourthsectionwillbeanoverviewofVietnam’scircumstanceinriceproduction,irrigationsysteman dimpactofirrigationonriceproduction.Finally,thefifthsectionismainfindings,policyimplicationa ndlimitations.

DEFINITIONOFIRRIGATION

“Irrigationisthecontrolledapplicationofwatertoarablelandsinordertosupplycropswitht h e wat errequirementsnot satisfiedbynaturalp r e c i p i t a t i o n ”(UnitedStates DepartmentofAgricultu re,1991).Accordingly,irrigatedl a n d i s l a n d whichi s s u p p l i e d waterfrommanysourcesbesides rainorsnow.Non- irrigatedlandiscalled“rainfedland”.Theterm“rainfedagriculture”isalsousedforfarmingapproa chonlydependsondirectrainfall.Themainroleo f irrigationtoprovideneededwaterfor plantsintheirgrowthstages.Furthermore,irrigationhasotherusessuchascontrollingweedandavoid ingsoilconsolidation.

Therearemanysourcesofwatercanbeusedforirrigating,theyincludesurfacewaterandground water.Forpublicirrigationsystem,canalsarebuilttoconnectfarmwithwatersource,f o r example,r iver,springandlake.Manydamsarealsobuilt to keepingwaterfordryseason.Intermsofprivateirrigation,thereisawidediversityofirrigationmetho dsthatfarmerscane m p l o y suchasflood,drip,center- pivot,andsprinkler.Methods’efficiencyarevarious,butchoosinganappropriateirrigationsys temalsor e l y o n cost,t e c h n i q u e , typeofp l a n t s andscale.

THEORETICALLITERATURE

Roleof irrigationinagriculture

Understanding the relationship between soil, water, and plants is essential for plant growth Water serves four primary functions: it is a major component of plant mass, acting as a solvent for transporting air, nutrients, and organic compounds; it participates as a reactant in vital biochemical processes like photosynthesis and hydrolysis; and it maintains turgidity, which is crucial for cell growth For instance, water constitutes 80-90% of herbaceous plants and about 50% of woody plants, highlighting its significance in overall plant health and development.

Plants extract water from the soil through their roots, primarily driven by the process of transpiration, which involves the release of water vapor from leaves and stems During the day, stomata in the leaves remain open to absorb carbon dioxide for photosynthesis, leading to increased water evaporation Approximately 95% of the water absorbed by plants is lost through transpiration, while only 5% is utilized for carbohydrate production The rate of transpiration is influenced by various factors, including weather conditions such as temperature, humidity, sunlight, wind speed, the type of plant, and the plant's growth stage.

Besidest h e functionofs t o r i n g nutrients,s o i l h o l d s waterf o r p l a n t s ( U n i t e d StatesDepart mentofAgriculture,1991).T h i s indicatesthatwater,s o i l andplantshavecloserelations.Water holdingcapacity ofsoilreliesonitschemicalandphysicalcharacteristics.T h e amountofwat erwhichsoilcanstoredetermineshowlongtheplantshaveenoughwatert o u s e u n t i l t h e n e x t irr igationo r rainfall.Ino t h e r words,characteristicso f s o i l determineirrigationfrequency.Inadditio n,m o s t o f wateri s evaporatedatt h e s o i l s u r f a c e , whichm a i n l y depends onweather,sothatweatheris anotherfactorofirrigationfrequency.

Plants require water for transpiration, which is supplied by the soil, but this process also results in water loss through evaporation The combined water demand for both transpiration and evaporation is referred to as evapotranspiration While some regions receive sufficient rainfall during the rainy season for plant growth, droughts can occur at other times, particularly in low rainfall areas where high-water-demand plants struggle to thrive and may even die if reliant solely on rainfall This highlights the necessity of providing adequate and timely water to plants to enhance crop yields Since the primary goal of agriculture is to produce profitable crops, irrigation serves as a vital solution to mitigate dependence on weather conditions, especially in arid regions.

Thedeterminantsofirrigationfrequency(toprovideenoughwaterforplant’sgrowth)canbesumma rizedasbelow:

Frequency and amount of rainfall

Intermsofeconomicperspective,irrigationpositivelyimpactsonannualcropyieldthroughincrea singoutputperseasonandthenumberofcroppingseasons(Hussain&Hanjra,2004).T h i s appea rstoberightforannualcropsbecauseitcanbegrownmorethanoneseasonperyear.Indryseason ofsomeareas,rainfallis too little tocultivate cropwithoutirrigation,landi n t h i s periodi s l e f t f a l l o w o r usedf o r growingo t h e r p l a n t s whichn e e d l e s s water.Incontrast,iflandisirrigatedsufficientwater,thenumberofcultivate dseasonscanbetwooreventhreeperyearforannualcrops.

Propertiesof riceplant and waterneedfor riceproduction

Ricep l a n t i s classifiedasgenusO r y z a whichi n c l u d e around2 3 species.Therea r e t w o sp eciesofOryzawhicharecommonlygrownforprovidingfood.TheyareOryzasavitaandOryzagl aberrima,whicharealsoknownasAsianriceandAfricanricerespectively.Atthepresenttime,

OryzasavitaismorepopularthanOryzaglaberrimasinceithashigherp r o d u c t i v i t y (Wopereis etal,2008).Asianricespeciescanbedividedintotwogroupindica andjaponica.Indicai scharacterizedbyh a v i n g non-sticky,l o n g andt h i n k grains,w h i l e japonicahassticky,shortandroundgrains.

The lifecycleofriceplantislikelytobedividedintothreebroadstageswhicharevegetativestage,reproduc tivestage,andmaturitystage(Wopereisetal,2008).Durationsofstagesaredifferent.Reproductiv estageandmaturitystagehaveafixeddurationwhichare30-

35daysand30daysrespectively.Meanwhilethelengthofvegetativestageisvariousamongvariet iesandday- length,sothatthelengthofrice’slifecycledependsondurationofvegetativestage.Commonly,rice’sl ifecyclecanlastfrom90to180daysor3-6months,soriceisannualcrop.

Rice farming systems can be categorized into five groups based on their water environment: irrigated lowland rice, rainfed lowland rice, deepwater rice, floating rice, and upland rice Lowland rice is distinct for its bunded and puddled fields In irrigated lowland rice systems, farmers actively manage water supply, maintaining a depth of five to ten centimeters, while rainfed lowland rice depends solely on rainfall Deepwater and floating rice systems are utilized in flood-prone areas, where flooding can persist for more than ten days and is largely uncontrollable In contrast, upland rice is grown in dryland conditions, relying exclusively on rainfall without puddling.

Irrigated and rainfed lowland rice are two primary rice cultivation systems, with irrigated lowland rice yielding the highest productivity (Maclean, Hardy, & Hettel, 2013) Over 90% of global rice is grown in these systems, covering approximately 93 million hectares for irrigated and 52 million hectares for rainfed lowland rice Together, they contribute to 75% and 19% of global rice production, respectively While the area dedicated to irrigated lowland rice is nearly double that of rainfed lowland rice, it produces rice yields that are almost four times higher.

Ricecultivationneedsalotofwater,especiallyforoutputmaximizingpurpose.Accordingly,t h e a mountofwaterisneededforriceplantstoproduceonekilogramofroughricevariesfrom500to1000litters,whichdependsonricefarmingsystemsandiscalculatedforonlytranspiration(Haef eleetal,2009).Thefigurewillbemuchlargeriftheevaporationisalsoaccountedfor,itis1432litte ronaverageforonekilogramofproducedroughrice(Bouman,

2009),andchangesf r o m 6 2 5 t o 1667 littersi n lowland ricefields(Zwart& Bastiaanssen,2004).Meanwhileitisjust900to1150littersformaizeonaverage(Falkenmark&Rockstrửm,2 0 0 4 ; C hapagain& Hoekstra,2 0 0 4 ) Inaddition,waterc a n b e l o s t throughseepingthroughbundsan dpercolatingdowntodeepersoillayers.Hence,intotalaricefieldcantakefrom800to5000litters(2500littersincommon)ofwatertoproduceonekilogramo f roughrice,andthisamountofwaterist wotothreetimeshigherthanthatofothermajorcereals(Bouman,2009).

Watersources

Water demand in rice production is significant, making the quality of irrigation water sources crucial Fields located near rivers, springs, lakes, and ponds can utilize water directly from these sources, while others rely on groundwater or canal water Since most fields cannot access water directly from rivers or lakes and groundwater is limited, canal systems are essential for providing irrigation Water levels in rivers and lakes can fluctuate significantly and unpredictably, while canal water supply tends to vary less across seasons compared to natural sources Additionally, canals may also serve drainage purposes in some areas The efficiency of canal systems largely depends on their quality and management, and disparities in water distribution among the head, middle, and tail ends of canals can impact crop productivity in these regions.

Analysismodel

Inagricultureproduction,inputsconsistofland,labor,seed,fertilizer,pesticide,machinery,an dothers.Therearemanyspecificformsofproductionfunctionsuchaslinearform,Cobb-

�=�� 1 � 2 � 3 … Leontiefp r o d u c t i o n f u n c t i o n i s employedi n t h e assumptiont h a t i n p u t s arecombinedi n a fixedratio.Accordingly,anincreaseinoneinputwithoutrespectiveincreaseinotherinputsw i l l notleadtochangeinoutput.However,thisisnotappropriateinagricultureproduction.T h e for mofLeontiefproductionfunctionisasbelow:

�=�(𝑉 1 ,…,𝑉 � ;� 1 ,…,� � ) WhereVii s t h ei th variableinput,Fji s t h ej th f i x e dinput.Inshortterm,profit offirmis:

∗i s p r i c eof thei th variableinput rii s n o r m a l i z e dpriceof the i thvariable input,whichequalsr ∗ /p

SupposethatthefirmwillchooseanoptimizedbundleofvariableinputsVtomaximizeits short-termprofitP w i t h a givenseto f p , r * a n d F.T h i s indicatest h a t profitfunctioni s a functiono f p , r * andF.Lett h a t o p t i m i z e d seto f variablei n p u t s b e V*,t h e n t h e profitfu nctioncanbedefinedas:

Supposingthatinordertoproduceanoutputy,thefirmwillchooseasetofinputsxwhichminimizesit scostc,giveninputs’price.Thisshowst h a t costfunctioni s a functiono f o u t p u t andinputs’price Thent h e c o s t functioncanb e writtenas:

The production function is widely utilized for its simplicity; however, it is affected by endogeneity in input choice, a challenge that can be addressed by employing profit or cost functions (Quisumbing, 1996) Nevertheless, profit and cost functions necessitate more detailed information, particularly regarding the prices of inputs and outputs, which can be a significant drawback due to the limited availability of comprehensive datasets In this context, the VARHS 2014 dataset lacks sufficient pricing information for all inputs, leading to the decision to solely utilize the production function in this thesis.

Cobb-Douglasproductionfunction

Douglasproduction functionwasfirstintroducedin1 9 2 8 byCobbandDouglas(1928).Theorigi nalCobb-Douglas production functionhasonlytwo inputswhicharecapitalandlabor,andhas assumption ofconstantreturnstoscale.Thefunction can bewrittenas:

�= 𝐴� 𝛼 � 1−𝛼 WhereYis output,Lis laborandK iscapital.

Douglasproductionfunctionarenotappropriatei n practice,manygeneralizationshavebeensug gestedwhicharecalled“Cobb-

Douglastypeo f function”.Twoofthemarerelaxing theassumption ofconstantreturnstosc aleandtheassumptionofthenumberofinputs(Debertin,2012).

As thenumberof inputs isconcerned,thefunctionisexpandedwith morethan two inputs:

4 𝛼 4 … Thisfunctioni s m o r e appropriatef o r a n a l y z i n g agricultureproductionbecauset h e r e a r e manyimportantinputs besidescapitaland labor,suchasseed,fertilizerand pesticide.Debertin(2012)suggestsanumberofpropertiesofCobb-Douglastypeoffunction:

(3) Outputequalszerowhenthereis at least one input is not used.

(4) Whenincreasingani n p u t Xi,outputw i l l increaseatt h e levelcorrespondingt o t h e parameteronthatinput� � Thevalueof� �can besmaller,equaltoorgreaterthan1.

(6) Ifall� �are smallerthan1,thereisusuallyapointwhereprofitisgloballymaximized with finite inputs X.

Douglasp r o d u c t i o n function,i t canb e transformedi n t o logarithmicformorTranslogformw hichcorrespondtofirstorderandsecondorderTaylor’sseriesexpansion.

Douglasp r o d u c t i o n functionsincei t accountsforrelationsamonginputsuse(Quisumbing,1996).Nonetheless,similarlytoCobb- Douglasfunction,i t candescribeonlyo n e p r o d u c t i o n stage(Debertin,2 0 1 2 ) Inaddition,o n e ofproblemsofestimatingTranslogformis possibilityofcollinearitybecauseofthelargen u m b e r ofvariables,especiallywhenthereareman ytypes of input(Pavelescu,2011).

Otherproductionfunctions

Douglasp r o d u c t i o n function,Debertin(2012)s h o w s s o m e o t h e r productionfunctionsareal sousedinagriculturaleconomics.TheyareSpillmanfunction,transcendentalproductionfunction,C obb-

Douglasw i t h variablei n p u t elasticity,deJ a n v r y Modification,andpolynomialforms.Spillma nfunctionwassuggestedin1923-1924,butithasbeenrarelyusedsinceCobb-

Douglasfunctionareconstant,t h i s functioncannotrepresentthreestageso f productionwhicharei ncreasinginaverageproduct,decreasinginaverageproductandnegativemarginalreturns.P olynomialformsalsohaveelasticityofinputwhichchangeswithvalueofthatinput,butiti s nots imilartoCobb-Douglasfunction.Thegeneraldisadvantageofthese productionfunctions,in comparisonto Cobb-

EMPIRICALLITERATURE

Therearemanypapersinvestigatingtheeffectofirrigationonpoverty.Areviewofempiricalstudiesco nductedbyHussainandHanjra(2004)showsthattherelationshipbetweenirrigationandpover tyareverystrong.Irrigationimpactsonpovertythroughbothdirectandindirectpaths.Indirectpat h,poorfarmerscangetbenefitfromirrigationthroughhighercropproductivity,higherf a r m i n g pr ofit,choosinghigherv a l u e crops,andparticipatingi n t h e market.Inlongterm,benefitofirrigati onwillspreadthroughoutthepoorpeoplewhodonothave land.

Research indicates that irrigation significantly impacts agricultural growth and poverty alleviation Bhattarai and Narayanamoorthy (2004) analyzed macro data from 14 Indian states between 1970 and 1994, finding that irrigation positively affects Total Factor Productivity (TFP) and plays a crucial role in reducing poverty They emphasize that the efficiency of irrigation systems can significantly influence the outcomes of poverty reduction programs Additionally, Butzer, Mundlak, and Larson (2002) examined agricultural growth factors in Indonesia, the Philippines, and Thailand, revealing that irrigation contributes between 10% and 16% to agricultural growth However, Walle and Gunewardena (2001) caution that relying solely on macro data may introduce bias by overlooking regional and household diversity, suggesting that microdata might provide a more accurate estimation of irrigation's benefits.

Int e r m s o f u s i n g microdata,manyp a p e r s a s s e s s t h e effecto f irrigationo n likelihoodo f bei ngpoor,andhousehold’sincomeorexpenditure.Particularly,Hussainetal.

(2006)estimateaLogitmodelwithdependentvariableisdummyvariablewhichequal1ifhouseh oldispoor.ThestudyemploysprimarydatainJava,Indonesiainyear2000-

2001andshowsthathouseholdswhoaccesstoirrigationsystemarelesslikelytobepoorthanothe rs.However,theeffectisvariousacrossirrigationsystemsandlocationsonthesystem.Opeyemiand Babatunde(2014)attempttoevaluatetheimpactofKampeirrigationdaminNigeria.Theycollectd atafrom140householdswhicharedividedintotwostrataastreatmentandcontrolgroups.ALogit modelisestimatedwithregressandisdummyvariableo f beingpoorbasingonFoster,Greerand Thorbeckepovertyindex.Theoutcomeindicatest h a t theprojecthelpstoreducepovertyinth eregionitisconducted.TheresearchresultsofBergandRuben(2006)inEthiopiashowthatirrigat edlandhaveapositiveandmuchhighermarginaleffectonhousehold’s expenditure thanth o s eofrainfedland.Similarly,inChina, H u a n g etal.

(2005)suggestt h a t irrigationh e l p st o i m p r o v e household’stotali n c o m e andreducepoverty.

Tot h e besto f myknowledge,i n V i e t n a m , r e c e n t l y there i s n o p a p e r relatedt o irrigation.B eforet h a t , s o m e studiesu s e datacollectedi n 1 9 9 2 -

1 9 9 3 and2 0 0 0 t o assessbenefito f irrigationtoagricultureproductionandpovertyreduction inVietnam.Indeed,Walle(2003)usesVietnaml i v i n g standards u r v e y 1 9 9 2 -

Research indicates that both irrigated and rainfed lands significantly enhance crop income, with irrigated land yielding twice the benefits of rainfed land Additionally, farmers with higher human capital can gain more from the expansion of irrigation A study by Ut, Hossain, and Janai in 2000 analyzed data from various farming systems in Vietnam, revealing that modern rice varieties cultivated in irrigated conditions outperform those grown in rainfed areas in terms of yield and profit Furthermore, Biltonen, Hussain, and Tuan (2003) examined agricultural production in two irrigation systems in Vietnam and found that improved irrigation performance correlates with increased crop yield, although the impact of irrigation varies across different locations.

Studiesrelatingto agricultureproductivity showt h a t accessingto irrigationimprovescropyiel d,thebenefitconsistsoflandintensityandcroppingintensity.Accordingly,Ahmadetal.

A study conducted in 2002 employed a stochastic frontier production function to analyze the impact of irrigation on wheat productivity in Pakistan, utilizing secondary data from 2,228 farmers The findings revealed that irrigation significantly enhances wheat productivity, particularly when water is sourced from both canal and tube wells, leading to optimal efficiency In a separate evaluation of the Zimbabwe European Union micro-project program, Nhundu, Gwata, and Mushunje (2010) demonstrated that farms with access to irrigation exhibit higher crop productivity Additionally, research by Jin, Jansen, and Muraoka (2012) in India, using long panel data at the plot level, confirmed that irrigation positively affects land productivity, with variations depending on whether the irrigation systems are public or private Their results indicated that a combination of both public and private irrigation yields the highest benefits, primarily through increased cropping intensity Dhawan and Datta (1992) also supported the strong correlation between irrigation and cropping intensity, utilizing cross-sectional data to measure the dependent variable as the ratio of gross to net sown area.

HYPOTHESISTESTING

(2)Assessdifferencesinimpactofwatersources forirrigationonriceproductionH 0:differentwatersourcesforirrigationdo not havedifferentimpacton riceproduction.H 1:differentwatersourcesforirrigationhavedifferentimpacton riceproduction.

(3) Exploredifferencesi n impacto f locationso n canals y s t e m f o r i r r i g a t i o n o n rice production

Age Gender Education Climate Infrastructure

Land Labor Capital Seed Fertilizer

D ATASOURCE

The survey conducted in 12 provinces of Vietnam, including 5 in the Northern region, 3 in Central Coastal, 3 in Central Highlands, and 1 in the Mekong River Delta, involved selected communes for data collection It encompasses information from 3,648 households, focusing on various aspects of rural development, particularly in agriculture The data includes specific characteristics of rice cultivation, such as plot details, types of plants, inputs, outputs, and irrigation practices Additionally, information on household members and communes was gathered The study specifically analyzes data from 2,490 rice-growing households, representing over 68% of the total sample size.

MODELSPECIFICATION

Buildingmodel

Asmentionedabove,profitfunctionandcostfunctioncannotbeemployedbecauseofdatal i mitat ionininputs’price,sothatthisthesisusesproductionfunctiontoaccomplishallthreeresearchobject ives.Becauseo f t h e s i m p l i c i t y inestimating,a C o b b -

Model 1: For investigatingtheimpact ofirrigationonannualriceproduction ln(�)=� 0+ � 1ln (���𝑖���𝑖��) �+ � 2 ln(������) �+ � 3ln (������_�ℎ����

Model 2: For assessthe impact ofdifferentwatersourceson annualriceproduction ln(�)=� 0 +� 1 ln(𝐶����) � +� 2 ln(𝑅𝑖���) � +� 3 ln(����) � +� 4 l n (�� �� �ℎ _

Model 3 : For e x p l o r e t h e impacto f differentlocationso f canalsystemo n annualriceprodu ction ln(�)=� 0 +� 1 ln(����) � +� 2 ln( 𝑖� ����) � +� 3 ln(���) � +� 4 l n (𝑅𝑖���)

Constructingvariables

Regressandisriceproductionwhichismeasuredbytotalkilogramofroughriceinlogarithmformpro ducedbyhouseholds in 12 monthsbefore interviewedpoint of time.

Independentvariablesconsistofirrigation,inputs,characteristicsoffarmers,communefactors,and otherfactors.

Irrigation is a significant variable in assessing rice production Various methods can evaluate the impact of irrigation, including dummy variables and the proportion and area of irrigated land While dummy variables are suitable for plot-level data and homogeneous irrigation conditions, the ratio of irrigated land to total cultivated land is rarely utilized Therefore, this thesis will adopt the logarithm of irrigated land measured in square meters (m²) as the irrigation variable for analysis.

Inaddition,inordertoassesstheimpactofwatersources,irrigatedlandisdividedintolands i z e irr igatedbycanal,riverandspring,lakeandpond,andothersources.Sincetherearefewfarmusingwater fromwellforirrigation,itisclassifiedasothersources.Similarly,landsizeirrigatedbyc a n a l i s d i v i d e d i n t o head,m i d d l e andendb a s i n g o n l o c a t i o n o f plotso n commune canalsystem.These variablesarecalculated inlogarithm.

Intermofexpectedsign,manymentionedstudiesshowthatirrigationimprovescropproductivity,c ropprofit,andhousehold’sincome,soexpectedsignofvariableirrigatedlands i z e ispositive.Si milarly,expectedsignso f variables, irrigatedl an d sizedividedbywatersourcesandlocation oncanalsystem,arepositive.

Inputsaretheoryvariablesofproductionfunction.Inagricultureproduction,inputsinclud el a n d , labor,capital,f e r t i l i z e r andseed(Bindlish& Evenson,1 9 9 3 ; B i n d l i s h etal.,1 9 9

The study analyzes rice cultivation by measuring land area dedicated to rice production in square meters It distinguishes between household labor, quantified by the total working days of family members, and hired labor, represented by the total cost of hiring labor in thousands of dong Additionally, the total cost of hiring assets, machinery, equipment, and cattle, along with the total value of productive assets, serves as proxies for capital variables Fertilizer and seed costs are also included as inputs for rice production All input variables are expressed in logarithmic form, reflecting their expected positive impact on rice production.

Asm e n t i o n e d a b o v e , o n e o fp ro bl em s o f Cobb-

Douglasproductionf u n c t i o n i s t h a t o u t p u t w i l l bezeroifanyinputisnotused.Forlogarith mform,theproblemnowisexpressedthatl o g ofzerovaluecannotbedetermined.Hence,tosolv ethisproblem,logofzerovalueissetaszero.Thissolution isappliedforallvariables inlogarithmformin the model.

Numerous studies examine the impact of farmers' characteristics, such as age, education, and gender, on agricultural productivity, incorporating these factors as control variables However, the effects remain controversial A review by Quisumbing (1996) of eight studies on gender differences in agricultural productivity reveals that the efficiency of female farmers is generally comparable to that of male farmers, with only one study indicating lower efficiency for females, albeit with weak significance Additionally, three out of the eight studies demonstrate that female education significantly enhances agricultural productivity, while the remaining studies show a positive but insignificant effect It is also anticipated that an increase in farmers' age may lead to a decrease in productivity.

27 productivitybecauseo f lowers t r e n g t h , b u t higheragemaygivem o r e experiencewhichimpr ovesagricultureproduction.Particularly,oneineightpapersindicatessignificantlynegativeimpact ofhead’sageontotalvalueofcropproduction,andanothershowss i g n i f i c a n t l y positive impactof ageonvalueofproduction.

This thesis analyzes the age, education, and gender of household members responsible for rice production as control variables Since households may manage multiple rice plots with different managers, average measures of age and years of schooling for these plot managers are utilized, both expressed in logarithmic form Gender is represented using two dummy variables: one indicating whether all rice plot managers in a household are male, and another indicating the presence of both male and female managers While the expected sign for formal education is positive, the anticipated signs for age and the two gender-related dummy variables remain uncertain.

Communal factors play a significant role in influencing agricultural productivity, encompassing both infrastructure and climate Key infrastructure elements include the availability and quality of roads, irrigation systems, input markets, and information systems Three specific variables serve as proxies for infrastructure: the distance from the commune center to the main road, the distance from the commune center to the district center, and the percentage of concrete roads within the commune, with the first two measured in logarithmic form While there is no direct variable to represent rainfall, these infrastructure variables will be excluded from the models due to the application of cluster-specific fixed effects estimation methods, which will be elaborated on later.

Othercontrol variablesinclude la nd qu al i ty (Huangetal.,2005; J i n , Jansen andMu r a o k a , 2012),intra- farmfragmentationofland(Huangetal.,2005;Markussenetal.,2013),visitingagriculturee x t e n s i o n ( A h m a d etal.,2 0 0 2 ;Moock,1 9 7 6 ;Saito,M e k o n n e n & Spurling,1994),andhitb ydisaster(Huangetal.,2005).Accordingly,landqualityismeasuredasratioo f ricecultivatedl a n d

28, whichhashigher quality thana v e r a g e quality ofl an d invillage, t o totalricecultivatedla nd.Landintra-farmfragmentationisthefragmentationoflandowned

8 Areaoflandirrigatedbyothersources(log) Positive byafarm,itismeasuredasthenumberofplotsgrowingrice(inlogarithmform).Visiting agr icultureextension isaddedtomodelas dummyvariablewhichequals1if anyhousehold’smembervisitsextensionagentatleastoncein12monthsbeforeinterviewedpo intoftime.Finally,hitbydisaster ismeasuredasratioofricecultivatedland hitbydisasterin 12monthst o t o t a l ricecultivatedland.

Expectedsignsofvariablesratiooflandquality,visitagricultureextensionandratiooflandh i t by disasterarepositive,positiveandnegativerespectively.Intermofnumberofplots,itsexpectedsigni s n o t clear.Ifl a n d i s dividedi n t o m a n y p l o t s , d i f f i c u l t y i n m o v i n g amongp l o t s oflab orandmeanswillincrease.However,highintra- farmfragmentationmayleadtolowerriskandgivechancetotestnewvarietiesortechnology.

Dependentvariable:household’stotalamountofroughricein12 months(kg)

2 Householdmembers'workingdaysto producerice(log) Positive

3 Cost of hiringlabor(log) Positive

6 Cost ofrentingasset,machinery,equipment,meansandcattle(log) Positive

Average ageofplotmanagers(log) Positive/negative

Ratiooflandwith better qualitythanaveragein thevillage Positive

ESTIMATIONSTRATEGY

Cluster-specificrandomeffectmodel

Supposet h a t therei s a sampleo f N observationsw i t h C clusters.Thena cluster- specificeffectsmodeliswrittenas:

� �� =�′ �� �+� � +� �� where:the script jc denotethe j th observationin thec th cluster,j =1, 2,…,Nc, c=1, 2,…, C.

� ��i s aK×1vectorofregressors.� �i s clusterspecific effectwhichchangesacrossc l u s t e r s , anditisassumedthat� �~ [0,𝜎 2 ].�i sassumedtohavez eromeanandconstantvariance.UnderaCSREmodel,� �i s assumedtodistributerandomlyandi n d e p e n d e n t l y w i t h regressors.� �+ � ��is alsos u p p o s e d t o b e correlatedw i t h i n cluster,𝐶�� �( �+ � ��, � �+

IfthetruemodelisCSREmodel,OLSestimationcangiveconsistentestimatorsbutitisnotefficient (CameronandTrivedi,2005).Hence,thedefaultvarianceofOLSestimatorsneedst o b e adjusted.

T h e i s s u e canb e s o l v e d byu s i n g cluster- robustvariancematrix(CRVM)estimation(White,1984;Liang&Zeger,1986).CRVMofOLSe stimatorcanbeestimatedas:

Source:CameronandTrivedi,2005 where:XcisaNc×Kmatrix,�̂ � =� � −� � �̂

GeneralizedLeastSquare(GLS)estimation.GLScoefficientsandvariancesareestimatedbyt h e f ormula:

FeasibleGLSs i n c eΩ mayb e mis-specified.C l u s t e r - r o b u s t v a r i a n c e m a t r i x o f FGLSi s calculatedas:

Inordertotestforclustereffect,BreuschandPaganLM(LagrangeMultiplier)testcanbeemplo yed,t h i s t e s t i s a t w o - s i d e d t e s t u s i n g residualso f OLSregressionw i t h o u t c l u s t e r - specificeffect(Breusch&Pagan,1980).Nonetheless,Moulton(1987)suggestanone- sidedLMtestthatismorepowerfulthanBreuschandPaganLMintermofcluster-specificeffect.

Thenullhypothesiso f t hi s t e s t i s H0:𝜎 𝛼= 0,w it h H1:𝜎 𝛼> 0.IfH0isrejected,CSREis moreappropriatethanOLSwithoutcluster-specificeffect.Thistestiscalculatedas:

Where�̂i sresidualsofOLSregressionwithoutcluster-specificeffect.LMstatisticfollows standardnormaldistribution.

Cluster-specificfixedeffectmodel

UnderCSFE,bothregressorsandtheintercept� �need tobeestimated.Inthesituationo f fewcl usters,� �can beestimatedbyusingclusterdummyvariables.However,whent h e r e aremanyclusters,within-clusterestimationshouldbeapplied.Awithin- clusterestimationiswrittenas:

Thism o d e l c a n b e e s t i m a t e d byOLSprocedure,w h i c h givesconsistent�coefficients. Nonetheless,Cameron&Miller(2015)claimthatthereisapartofcorrelationamongerror termsinthesameclusterthatcannotbesolvedbywithin- clusterestimation.Therefore,itisn e c e s s a r y toappliedCRVMestimation.Fortunately,CR VMestimationcanalsosolveheteroskedasticityproblem.

TochoosingbetweenCSFEandOLSwithoutcluster-specific effect,asimpleFtestcanbe usedw i t h n u l l hypothesisi s H0:� 1= � 2= ⋯=� � IfH0i srejected,CSFEregressionshould beusedinsteadofOLSwithout cluster-specificeffect.

ChoosingbetweenCSRE and CSFE

CSFE estimation provides consistent coefficients when there is potential correlation between unobservable cluster-invariant variables and independent variables However, it has drawbacks, such as reducing variable variation and being unable to determine coefficients for cluster-invariant variables Therefore, it is essential to conduct a test to verify the validity of OLS with CRVM estimation or Feasible GLS If these procedures are found to be invalid, CSFE should be utilized To assess the assumption of no correlation between the independent variables and the unobservable variables, a Hausman test can be performed, which examines the differences between the regressors of CSRE and CSFE regressions.

Where:�̂ is vector ofestimatedcoefficientsof CSFE regression,�̂ isvectorofestimated coefficientsofCSREregression.Sincecluster-invariantvariablesaredroppedinCSFE regression,thesevariablesarenot contained in theformulaofHausmantest.Thistestfollows

𝜒2d i s t r i b u t i o n Ifthenullhypothesisis rejected,CSREregressionisnot appropriate,t h e n CS FEregressionneedstobeused.

However,thistestis validinthe situationthatCSREregressionisfullefficient.Ifthereis heteroskedasticityin� �o r � ��, orcorrelationamong� ��i n thesame cluster,anothert e s t isrequired.HausmantestcanbereplacedbyaWaldtestfortheauxiliaryO LSregressionwith cluster- robustvariance(Wooldridge,2010).T h e auxil iaryOLSregressioniswrittenas:

Onlyclustervariantvariablesareincluded.ArobustWaldtestisconductedwithnull hypothesisis�=0.IfH0isrejected,CSFEregressionshouldbeused.

India Indonesia Bangladesh Viet Nam China

V IETNAMRICEPRODUCTIONANDPUBLICIRRIGATIONSYSTEM

Intermofriceproduction,withnearly45milliontonnesofproducedricein2014,Vietnamwasr ankedfifthi n t h e world,afterC h i n a , India,Indonesia,andBangladesh(FAOSTAT).However,for rice productivity,VietnamwasjustafterChina(Figure4.1).

Figure4.1:Riceproductivity in 2014 ofsome countries(kg/ha)

From1 9 9 0 t o 2 0 1 4 , Vietnamr i c e p r o d u c t i o n s u b s t a n t i a l l y increasedf r o m about2 0 t o 4 5 m i l l i o n tonnes (Figure4.2) This increase canbepa rt ly explainedbyincreasein harvestedareafrom1 9 9 0 t o 2 0 0 0 (Figure4 3 ) However,i n t h e stageo f 2 00 0-

2 0 1 4 , Vietnamfocusedonimprovinglandproductivityratherthanincreasingricelandsize.T heincreaseinriceproductivityofVietnamhasbeencontributed bymanyfactors.Oneofthe misdevelopmentandapplication ofmodernvarietieswhicharetolerantoflocalclimate,pe st s, andgivehigheryielda n d value.Applicationofnewt e c h n o l o g y ino p e r a t i o n alsohelpst o redu cecostandi m p r o v e productivity.Moreover,e x t e n s i o n andimprovemento f irrigationsystemb ythegovernmenthasplayedaveryimportantr o l e inriceproductionbyreducingcostofirrigat ion,providingsufficientwaterforfields anddrainingin floodseason.

M ill io n To n n e s M ill io n h a 19 90 19 91 19 90 19 91

Aspublic irrigationsystemisconcerned,Ha,NguyenandNguyen(2015) showthatin thestage of2006-

2010investmentof Vietnamgovernmentinirrigationsystemis59%oftotalinvestmentinagri culturesector.Ratioofricelandsizeirrigatedincreasedfrom63%in2001t o 85.5%in2011.Until2011,Vietnamhas254180kmofcanalservingforirrigation,while

RedRiver DeltaandMekong Deltaaccountforover50%of th is figure.Ratioofconcretecanal t o totalkilometerso f canali n Vietnam2 0 1 1 i s 21.5%.U n t i l 2013,Vietnamp u b l i c irrigationc anirrigate upto 94% oftotalricecultivatedland.

Nonetheless,therearealsosomedisadvantagesofVietnamirrigationsystemwhichleadtol o w e f f i c i e n c y in irrigation.Tran(2016)suggests some pointsas:

(1) Weaknessinmanagementresultsinpoorperformanceofirrigationsystem.Particularly,wate ri s usedwastefully,financials u p p o r t i n g m a i n l y comefroms t a t e b u d g e t Lackingo f main tenanceanddredgingleads todegradationand efficiencyreducingofirrigationsystem.

(2) AlthoughVietnamhasover254000kmofcanal,only21.5%ofthemisconcreted.Thisfigure isevenlower forinterior canalsystemwith just 16%km ofcanalisconcreted(2013).

R ICEPRODUCTIONANDIRRIGATIONOFFARMERSIN VARHS2014

TotalareacultivatingriceinVARHS2014is1065.73ha.Ratioofareathatgrows1seasono f ricep e r yeari s 4 5 % , and5 3 % f o r 2 s e a s o n s o f ricep e r year.Meanwhileareat h a t cultivates

Only 1.73% of the sample consists of land growing rice in three seasons, while nearly 80% of rice cultivation is irrigated The ratio of irrigated land indicates a potential link between irrigation and the number of rice-growing seasons, with nearly 100% of land growing rice in two or three seasons being irrigated In contrast, only 55.5% of land growing rice in one season is irrigated Additionally, the cultivation of other crops alongside rice is minimal, with only 8% of rice land used for other crops, primarily occurring in areas with one rice season.

Table4.1:Areacultivatingrice,ratioofirrigationandratioofpolyculturebynumbero f seas ongrowingrice

Total 1seasongr owing rice 2seasongr owing rice 3seasongr owing rice

Percentageofl a n d g row in g other crops(%)

Intermsofirrigationsources,mostofhouseholdsusewaterfromcanalandriverorspringf o r irrigation.Accordingly,ratiooflandirrigatedbycanalandriverorspringis58.4%and1 8 % res pectively.Whilethisratiooflandirrigatedbylakeorpondandothersourcesisjust1 7 % and0.9

%respectively(Table 4.2).Forlocationon canalsystem,a largepartoflandirrigatedbycanalisi nthemiddleofcommunecanalsystem.Indeed,ratiooflandirrigatedbycanalwhichisinthehead, middleandtailofcommunecanalsystemis28.6%,61.5%and1 0 % respectively(Table4.2).

Table 4.2 highlights the impact of irrigation on rice productivity, showing that irrigated land supports a higher number of growing seasons and greater productivity compared to rainfed land Specifically, rainfed areas yield an average of 1.12 rice seasons per year, whereas irrigated land achieves 1.79 seasons annually Additionally, the productivity per season for irrigated rice is 5.06 tons per hectare, which is double that of rainfed rice at 2.47 tons per hectare Overall, the annual productivity of rice on irrigated land reaches 9.31 tons per hectare, more than three times the 2.65 tons per hectare found in non-irrigated areas.

Rainfed Irrigated Irrigated Irrigated Irrigated Irrigated Locationoncommune bycanal byriver orsprin g bylake or pond byotherso urces canalsystem

Onaverage,r i c e grownunderl a n d irrigatedbycanali s m o r e p r o d u c t i v e t h a n o t h e r w a t e r s ources.Particularly,t h e n u m b e r o friceseasonperyearun de r l a n d irrigatedbycanalandriverors pringis1.85and1.61respectively(Figure 4.2).Rice productivityperseasonoflandirrigatedbycanali s 1 7 % highert h a n t h a t o f l a n d irrigatedbyr i v e r ofspring.Similarly,annualr i c e p r o d u c t i v i t y o f l a n d u s i n g canalf o r irrigationi s 3 5 % h i g h e r t h a n t h a t o f l a n d u s i n g waterfromr i v e r o r spring.T a b l e 4 2 s h o w s t h a t l a n d i rrigatedbyw a t e r froml a k e , p o n d andothersourceshashighernumberofriceseasonandprod uctivitythanthatoflandirrigatedbyriver orspring.

Finally,t h e averagefiguresi n Figure4 2 i n d i c a t e t h a t ricep r o d u c t i v i t y ishighestath e a d e ndsofcommunecanalsystemandlowerformiddleandendlocations.Indeed,numberofricesea sonperyearoflandinthehead,middleandtailofcommunecanalsystemis2.03,

1.91and1.83respectively.Riceproductivityperseasonofland onheadends,middleandtailendsofcommunecanalsystemis6.06,5.20and4.81(ton/ha/season)respectively.Si milarly,

42 landath e a d endscanalhashighestannualp r o d u c t i v i t y ofrice(12.32t o n / h a ) , l a n d i n t h e m i d d l e hasthesecondhighestproductivity(10.07ton/ ha),andriceproductivityoflandintailendsofcommune canalislowest(8.91ton/ha).

E FFECTOFIRRIGATIONONRICEPRODUCTIONINRURAL V IETNAM

IntermsofchoosingthemostappropriatemodelamongOLSwithoutcluster- specificeffects,C S R E andCSFE,threesets of tests areused Theyare Ftest forcluster- specificfixedeffects,two-sidedandone- sidedLMt e s t f o r clustere f f e c t s , andfinally,officialHausmant e s t andr o b u s t W a l d t e s t o f auxiliaryOLSregressionf o r correlationbetweenclusterunobservedvariablesandregressors.R esultsofthesetestsaresummarizedasbelow:

Cluster- specificfi xe d e ffects(CSFE)

Source:p e r f o r m e d b y author.T h e resultso f t h e s e t e s t s a r e c o n s i s t e n t f o r a l l threet h e s i s o b j e c t i v e m o d e l s D e t a i l resultsoftestsareshowedinAppendices

Thesetests suggestthat CSFEisthe mostappropriatemodel Particularly,resultsofFtest

2 showt h a t therearedifferencesa m o n g� � ,s o t h a t CSFEs h o u l d b e usedinsteado f OLSwith outcluster-specificeffects.Sincetwo-sidedandone-sidedLMtestindicatesthat𝜎 𝛼> 0,

CSREismoreappropriatethanOLSwithoutcluster- specificeffects.AlthoughofficialHausmantestshowssignificantcorrelationbetweenunoservab lecluster-specificeffectsand regressors,t h e t e s t i s i n v a l i d becauset h a t variancematrixi s n o t p o s i t i v e definite.Hence choosingbetweenCSREandCSFEreliesonly onrobustWaldtestfor auxiliaryOLSregression,andthetestsuggeststhatCSREmodelis invalid.

4.4.Sincetherearesomeintervieweeswhodon’tknowlocationoftheirplotsoncommunecanal system,these households willbe droppedoutof sample.Hencemodel 3hassmaller n u m b e r ofobservationsthanmodel1andmodel2.VIFresultoftestingformulticollinearitys h o w s t h a t t h e r e is nohighmulticollinearityinthemodel.Meanwhile,modifiedWaldtestforf i x e d effectmodelsugg estsheteroskedasticityinerrorterms,sothatCRVMestimatorshouldb e used.

Inbrief,signsofallvariablesaresimilartoexpectation,andregressionoutcomesareconsistentacros smodels,butsomevariablesarenotsignificant.Irrigationhasapositiveandsignificantimpacto n r i c e production.However,o n l y u s i n g w a t e r f r o m canalandriverors p r i n g cangivesignifica ntbenefittoriceproduction.Thetestshowsthatthereisnodifferenceintheimpactofcanalirrigation andriver- springirrigationonriceyield.Inaddition,allthreelocationsoncommunecanalsystemwhicharehead ,middleandtailhaveap o s i t i v e andsignificanteffectonriceproduction.Nonetheless,theimpa ctoftaillocationislowerthanthat ofmiddlelocation,whichissignificantat 10%level.

Intermofrobustnesscheck,OLSandFGLSwithCRVMmodelsareregressedtocomparew i t h CSFEregression.BesidesvariablesinCSFEregression,thesetwomodelsalsocontrolcommunein frastructurevariablesanddummyvariablesofprovinces.Thespecificoutcomeso f t h e s e regressi onsareexpressedi n A p p e n d i c e s Regressionresultsi n d i c a t e t h a t t h e r e i s highconsistenc ybetween thesetwo modelsandCSFEregression.

Inmodel1,coefficientofvariableAreaofirrigatedland(log)ispositiveandsignificantat1 % le vel.Valueofthecoefficientis0.04whichistheelasticityofriceproductiontoareaofirrigatedl a n d Itmeanst h a t riceproductionw i l l increaseby0 0 4 % i f irrigatedl a n d s i z e increasesby1%,giveno t h e r factors.Itindicatesthat irrigationhasa significantly positiveimpactonricepro duction.

In Model 2, irrigation is categorized into four types based on water sources: canal, river or spring, lake or pond, and other sources The coefficients for land irrigated by canal and land irrigated by river or spring are both positive and statistically significant at the 1% level, with values of 0.027 and 0.021, respectively This indicates that a 1% increase in the area irrigated by canals or rivers/springs will lead to a corresponding increase in rice yield of 0.027% and 0.021%, holding other factors constant Conversely, the coefficients for the last two variables—irrigation from lakes, ponds, and other sources—are positive but not significant, suggesting that using these water sources for irrigation is less efficient and does not enhance rice production compared to rainfed systems.

Althoughu s i n g waterf r o m b o t h canalandrivero r s p r i n g f o r irrigations i g n i f i c a n t l y a ndp o s i t i v e l y affectrice production, canal’sbenefit m a y behigherthan t ha t o f rive rorspring s i n c e coefficiento f canalvariablei s larger.However,t h e F t e s t s h o w s t h a t the rei s n o significantdifferencebetweenimpactsofirrigationbycanalandbyriverorspring.Speci ficresultof thetest ispresentinAppendices.

The area of land irrigated by canals is categorized into head, middle, and tail locations within a communal canal system The coefficients for these locations are positive and significant at the 1% level, with values of 0.019, 0.022, and 0.016, respectively This implies that a 1% increase in the area irrigated by canals in the head, middle, and tail locations will result in a corresponding increase in rice production of 0.019%, 0.022%, and 0.016%, respectively Overall, the use of canal water has a significantly positive effect on rice production, although the magnitude of this effect varies by location Notably, the head and middle locations appear to have a greater impact on rice production compared to the tail location F-test results further indicate that the impact of canal irrigation on rice production varies across locations, with the middle location showing a higher effect than the tail location at the 10% significance level, while no significant differences exist between the head and middle locations or between the head and tail locations.

The coefficients for all inputs in rice production models are positive and consistent, with significant contributions from cultivated land size, household labor, hired labor, fertilizer, and seeds Notably, the coefficient for cultivated land size is approximately 0.48, indicating that a 1% increase in land size results in a 0.48% increase in annual rice production, assuming other factors remain constant This highlights the critical role of land in rice production in rural Vietnam, suggesting that if economies of scale are maintained, land could account for nearly half of the rice production output.

Both household labor and hired labor significantly impact rice production, with coefficients of 0.166 and 0.011, respectively, at a 1% significance level This indicates that a 1% increase in household labor leads to a 0.166% rise in rice production, while a similar increase in hired labor results in a 0.011% increase The larger coefficient for household labor suggests greater efficiency in utilizing household members for labor However, to draw more accurate conclusions, further tests and additional regressions are necessary, which are beyond the scope of this thesis.

Coefficientsofvariablescostoffertilizerandcostofseedarepositiveandsignificantat1%level,t h i s indicatesimportantr o l e o f fertilizerandseedi n ricep r o d u c t i o n Particularly,coefficient softhesetwovariablesare0.035and0.12, whichmeanst h a t if expenditure forfertilizeran dexpenditureforseedincreaseby1%,riceproductionwillincreaseby0.035%and0.12

%respectively,ceterisparibus.Unlikefertilizerandseed,althoughestimatedparameterso f costo f rentingassetandcattlef o r producericeandh o u s e h o l d ’ s productiveassetsarepositive,theyare notsignificant.Thismaybebecausethatonlyusingmachinefortillageaffectsriceproduction,whil ethesevariablesareasumofmanyitemsthatservesomedifferentprocessinproducingrice.Anot herreasonisthatthesevariablesmaybenotgoodp r o x i e s forcapital.

Althoughpositive,t h e i m p a c t o f a v e r a g e a g e ofp l o t managerso n riceproductioni s n o t s ignificant.T h i s mayb e becauset h a t o l d managerst e n d t o h i r e l a b o r r a t h e r t h a n d i r e c t l y workingonfield.Inaddition,agevariabledoesnotonlycontainimpactofhealth,butitalsoincludeso therfactorssuchasriskattitude andexperience.

Formaleducationofplotmanagersmeasuredbyaverageschoolingyearshasapositiveandsignif icant( a t 1 % level)impacto n ricep r o d u c t i o n T h e coefficiento f formaleducationi s about0.04 whichmeansthatiftheaverageschoolingyearsofplots’managergoesupby1%,ricep r o d u c t i o n w i l l increaseby0.04%,o t h e r t h i n g s equal.Thisf i n d i n g i s consistentw i t h J a m i s o n andLau (1982)forthecaseofKoreaandThailand.

Asgendero f p l o t managersareconcerned,e s t i m a t e d parameterso f t w o d u m m y variabless h o w thathouseholdwhichhaveonlymalemanagersofplotshaslessriceproductionthano t h e r householdswhichhavef e m a l e o r b o t h m a l e andf e m a l e m a n a g e r s , ceterisp a r i b u s However,thedifferenceisweaklysignificantat 10%level.

Thecoefficientofratiooflandwith goodquality ispositive butinsignificant.Thismayares ultofweakproxyforlandquality.Sincetherearenotgoodmeasuresoflandquality,thisvariablei s u s e d H o w e v e r , i t measurest h e r e l a t i v e q u a l i t y o f l a n d i n villageinsteadofco mparingw i t h q u a l i t y o f alll a n d i n sample.Quantitativevariablesm a y bem o r e appropriate.

In 1993, it was found that the ratio of land affected by disasters has a negative and significant impact on rice production, with a decrease of approximately 0.12% for every 1% of land hit by a disaster, ceteris paribus Additionally, households that engage with agricultural extension services at least once a year experience a notable benefit, yielding 5.5% higher rice production compared to those who do not participate This positive correlation is significant at the 1% level and aligns with the findings of Saito, Mekonnen, and Spurling (1994).

Dependentvariable:loghousehold'sriceproduction Model1 bet a t-statistic Model2 bet a t-statistic Model3 bet a t-statistic 1

Source:calculatedbyauthor.*,**and***denotesignificanceat10%,5%and1%respectively.Within-clusterestimatorisemployedwithOLSprocedureand cluster-robustvariancematrixatcommunelevel

This thesis examines the relationship between irrigation and rice production in Vietnam, focusing on the impact of different water sources and the variability of canal irrigation effects across various locations Utilizing the VARHS 2014 dataset, which includes data on rice production, household input use, plot manager characteristics, irrigation quality, and community infrastructure, the study employs a cross-sectional data model with cluster-specific fixed effects and robust standard errors The findings align with previous research by Hussain and Hanjra (2004), highlighting key insights into the effects of irrigation on rice production This section will summarize the main findings, along with policy implications and limitations of the study.

M AINFINDINGS

Theth es is hast h r e e m a i n findings.Firstly,r i c e grownunderirrigatedl a n d producesm o r e ann ualyieldthanricegrownunderrainfedland,givenotherconstant.Thisfindingisconsistentwithma nypreviousstudieswhichshowthatirrigationplaysanimportantroleinagriculturep r o d u c t i v i t y andp o v e r t y alleviat ion (Bhattarai& Narayanamoorthy,2 0 0 4 ; Butzer,Mundlak&Larson,2 002;Walle&Gunewardena,2001;Huangetal.,2005;Walle,2003).

Secondly,differentwatersourcesusedforirrigationhavedifferenteffectonriceproduction.U s i n g waterforirrigationfromonlycanalandriverorspringgivessignificantbenefittoriceproductioni n comparingt o cultivatingu n d e r rainfedl a n d Althoughcanalirrigationhashighercoefficientt h a n t h a t o f rivero r s p r i n g irrigation,t h e d i f f e r e n c e i s n o t significant.M a n y previo uspapersalsosuggestvariationacrossirrigationsystems(Ahmadetal.,2002;Bhattarai&Naraya namoorthy,2004)

Thirdly,t h e r e i s i n e q u a l i t y i n waterd i s t r i b u t i o n o f communecanals y s t e m P a r t i c u l a r l y benefito f u s i n g canalsystemf o r i r r i g a t i o n o n ricep r o d u c t i o n i s sign ificantf o r allt h r e e locationswhicharehead,middleandtail.Howeverthehouseholdshavela ndattailendsofcommunecanalsystemtake lessbenefitfromirrigationthanthoseatmiddlecanal.Similarly, somepreviousstudiesshowsignificantimpacto f locationo n benefitofcanalsystemi n agriculturep roduction and povertyreduction(Hussain,2006;Hussain&Hanjra,2004).

P OLICYIMPLICATION

Theimportantroleofirrigationonriceproductionsuggeststhatgovernmentshouldextend i rrigationsystemsince there isa partofriceland whichisnot irrigated.Inaddition,since differe ntwatersourceshavedifferentimpactsonricep r o d u c t i o n , s o q u a l i t y ofi r r i g a t i o n system significantlyaffectsriceproductivity.Hence,thegovernmentshouldalsoincreasetheq u a l i t y o firrigationsystem.Theyincludeconcreting,maintaininganddredgingcanal.Researchresultsa lsoindicateinequalityinwaterdistribution ofcommunecanalsystem,sot h a t t h e governm entn e e d s t o i m p r o v e waterregulationandirrigationsystemt o p r o v i d e sufficientwater tolandattail endscanal.

L IMITATION

The primary limitation of this thesis is the drawback of the production function, which has been widely utilized across various sectors since its inception However, following the cautionary notes from Marschak and Andrews (1944) regarding the endogeneity of input choices, economists began to adopt richer models The issue with the production function lies in farmers' simultaneous selection of input quantities and output levels Although profit or cost functions could address this concern, they necessitate input price data, which is unavailable in the VARHS dataset Consequently, this thesis cannot utilize these functions.

Douglasproductionfunctioncanrepresentonlyo n e stageofproduction.Someotherproduction functionssuchastranscendentalproductionfunction,Cobb-

Douglasw i t h variablei n p u t elasticity,andpolynomialforms,canexpressthreestageso f p r o d u c t i o n However,disadvantageo f t h e s e functionsi s a largen u m b e r o f parametersneedt o b e e s t i m a t e d T h i s cancausemulticollinearityp r o b l e m Moreover,t h e logarithmformofCobb-Douglasproductionfunctiondoesnotallowthedependencyamongi n p u t s Atranslogfor mcansolvetheproblem,butitalsorequiresalotofparameterstobeestimated.

Thethirdlimitationispotentialendogeneityofirrigationvariables.Unobservedfactorsthatc ontroldecisionofirrigationmay affectriceproduction Thesefactorsmay containannual rainfallanditsdistributionduringayear,abilityandpersonalityoffarmers.Althoughusingcl uster- specificf i x e d effectcancontrolunobservablecommunevariables,manycharacteristicsoffarmers cannotbecontrolled.Anotherproblemisthathouseholdscanbuyl a n d whichisirrigatedbywa tersourceswithhighquality.Nevertheless,thisproblemmayben o t serioussincethemarketofland cultivatedriceappearsnot to be developed inVietnam.

Fourthly,us i n g dataathouseholdlevelf o r pr od uc ti on functionmaycausepr ob le m Particular ly,irrigationisdummyvariableatplotlevelbutinputsaremeasuredathouseholdlevel.Thereis possibilitythattheamountofinputsuseisdifferentbetweenirrigatedplotandrainfedplot.Hence,spe cificdataatplotlevelmayleadtobetterregression.Moreover,paneldatamayp r o d u c e betteresti mationt h a n c r o s s - s e c t i o n d a t a s i n c e i t givem o r e v a r i a n c e ofi n p u t variables.

The price level differences among provinces are not accounted for, leading to measurement errors in input costs, as these variables are collected at nominal prices Additionally, some variables, such as household labor and water sources for irrigation, are inadequately measured Household labor is recorded as total working days of household members, which can be inaccurate due to varying definitions of a working day and potential memory lapses among interviewees Furthermore, the quality of a working day may differ between males, females, and children Regarding irrigation, the questionnaire only inquires about the main water source used, neglecting the fact that farmers may utilize multiple sources simultaneously, which could provide greater benefits than relying on a single source.

Lastbutn o t least,therea r e s o m e omittedvariableswhichcans i g n i f i c a n t l y affectricepr oduction,suchasirrigationapproachappliedbyfarmers,u s i n g o f n e w varietiesandt h e amountof waterusingforirrigation.Givenasourceofwater,differentirrigationapproachesm a y givediffer entimpactsonproductivity.Similarly,modernvarietiesmayproducehigherriceyield.Therei s p o s s i b i l i t y t h a t farmersapplydifferentvarietiesf o r irrigatedandn o n - irrigatedland.Whiletheamountofwaterusingforirrigationdirectlyaffectsriceproduction.

(2002).WheatProductivity,Efficiency,andSustainability:A StochasticProductionFront ierAnalysis[with C o m m e n t s ] ThePakistandevelopmentreview,643-663.

Asayehegn,K.(2012).IrrigationversusRain- fedAgriculture:DrivingforHouseholds’IncomeDisparity,A

Ayele,G K (2011).T h e impacto f selecteds m a l l - s c a l e i r r i g a t i o n schemeso n householdincomea n d t h e likelihoodo f povertyi n t h e LakeT a n a b a s i n ofEthiopia(Doctoraldissertation,CornellUniversity).

( 1 9 9 2 ) Frontierproductionf u n c t i o n s andtechnicalefficiency:a s u r v e y o f empiric alapplicationsinagriculturaleconomics.A g r i c u l t u r a l economics,7(3),185-2 0 8

(2004).Impacto f irrigationo n agriculturalgrowthandpovertyalleviation:macrolevelana lysisin India.WaterPolicy ResearchH i g h l i g h t IWMI-

Biltonen,E ; Hussain,I.,Tuan,D D ( 2 0 0 3 ) P r o - p o o r InterventionStrategiesi n IrrigatedAgricultureinAsia.VietnamCountryPaper.Col ombo,SriLanka:InternationalW a t e r ManagementInstitute.

(1993).EvaluationoftheperformanceofT&Vextensionin Kenya(Vol.23).Washington, DC: WorldBank.

Bindlish,V.,Evenson,R.E., Gbetibouo,M.,& Mundial,B.(1993).E v a l u a t i o n ofT& V-

BasedExtensionin Burkina Faso.Washington,DC: WorldBank.

Breusch,T.,&Pagan,A.(1980).TheLagrangeMultiplierTestand itsApplicationstoModelSpecificationi n Econometrics.T h e Reviewo f EconomicStudie s,47(1),239-253.Retrievedfromhttp://www.jstor.org/stable/2297111

( 2 0 0 2 ) Determinantso f AgriculturalGrowthi n Indonesia,thePhilippines,andThailand.

Cameron,A.C.,&Miller,D.L.(2015).Apractitioner’sguidetocluster -robustinference.

(2004).W a t e r footprintso f nations.Valueo f waterresearchreportseriesNo.16.Delft(Net herlands):UNESCO-IHE.76 p

FAOSTAT(n.d.).RetrievedNovember21,2016,fromhttp://faostat.fao.org

(2014).ProductionEconomics:A DualApproacht o T h e o r y andApplications: TheTheoryofProduction(Vol.1).Elsevier.

Gujarati,D.N.,&PorterD C.(2009).Basiceconometrics.TataMcGraw-HillEducation.

2 0 1 5 : T h ủ y lợi h ó a đ ư ợ c ư u tiênt ậ p t r u n g đầut ư pháth u y rõh i ệ u quả.Retrievedf ormhttp://www.tongcucthuyloi.gov.vn/

Haefele,S.M.,Siopongco, J.D.L.C.,Boling,A.A.,Bouman,B.A.M ,&Tuong,T.P.

Harris,D.N (2006).Watermanagementi n p u b l i c i r r i g a t i o n schemesi n Vietnam(No.113170).Canberra,Australia:A u s t r a l i a n C e n t r e f o r InternationalAgriculturalResearch.

Huang,Q , Dawe,D , Rozelle,S , Huang,J , & Wang,J.

Hussain,I.,Mudasser,M.,Hanjra,M.A.,Amrasinghe,U.,&Molden,D.

(2004).Improvingwheatproductivity inP a k i s t a n : econometr ic analysisusing pa neldata f r o m Chajint h e upperIndusBasin.WaterInternational,29(2),189-200.

(2006).Irrigation,productivity andp o v e r t y linkagesinirrigationsystemsinJava,I ndonesia.WaterResourcesManagement,20(3), 313-336.

Jamison, D T., &Lau,L.J.(1982).Farmereducation and farmefficiency.PublishedfortheW o r l d Bank[by]Johns Hopkins UniversityPress.

(2012,August).Theimpactofirrigationonagriculturalp r o d u c t i v i t y : EvidencefromI ndia.InInternationalAssociationo f A g r i c u l t u r a l Economists(IAEE)T r i e n n i a l Conference,F o z d o Iguacu,A u g u s t ( p p 18-24).

Kramer,P J., &Boyer,J S.(1995).Waterrelations ofplantsand soils.Academic press.

(2003).Estimating productionfunctionsusinginputsto controlf o r unobservables.TheReviewofEconomicStudies,70(2),317-341.

(1980).Farmereducationandfarmefficiency:Asurvey.Economicdevelopmentand cultural change,29(1),37-76.

Markussen,T., Tarp,F.,Do , H.T.,& Nguyen,D A.T (2013).Inter-andintra- farm landfragmentationinVietNam.CIEMWorkingPaper.Hanoi:CentralInstitutefor

(1976).Theefficiencyofwomenasfarmmanagers:Kenya.AmericanJournalo f Agricult uralEconomics,58(5),831-835.

Newey,W K , & W e s t , K D (1986).A simple,p o s i t i v e semi- definite,heteroskedasticityandautocorrelationconsistentcovariancematrix.

( 2 0 1 0 ) Impactso f ZimbabweEuropeanU n i o n micro-projectprogramme(Zim/ EuM P P ) i n f u n d i n g smallholderi r r i g a t i o n projectso n f o o d s e c u r i t y andi n c o m e levels:A cases t u d y o f Mopanei r r i g a t i o n schemei n Zvishavane,Midlandsprovin ce,Zimbabwe.AfricanJournalofAgriculturalResearch,5(14), 1759-1771.

(2014).Irrigation,i n c o m e andp o v e r t y alleviation:anassessments t u d y o f k a m p e i rrigationdami n ejiba(KogiState,Nigeria).ВестникОрловскогогосударственногоагра рногоуниверситета,51(6).

Quisumbing,A R ( 1 9 9 6 ) M a l e - f e m a l e differencesi n agriculturalproductivity:Methodologicalissuesa n d empiricalevi dence.WorldDevelopment,2 4(10),1 5 7 9 - 1 5 9 5

(2016).Thựctrạngvàgiảiphápxâydựnghoànchỉnhhệthốngthủylợinộiđồngđápứ n g p h á t triểnnềnn ô n g nghiệpđ a d ạ n g vàh i ệ n đại.Retrievedfromhttp://www.pim.vn/Web/ Content.aspx?distid09.

Irrigation.Retrievedfromhttp://www.wcc.nrcs.usda.gov

VanDenBerg,M , & Ruben,R (2006).S m a l l - s c a l e irrigationandi n c o m e d i s t r i b u t i o n i n E t h i o p i a JournalofDevelopmentSt udies,42(5),868-880.

(2015).StatisticalYearbookofVietnam2014.Hanoi:GeneralStatisticsOffice(GSO)Vietn am.

White,H (1980).A heteroskedasticity- consistentcovariancematrixestimatoranda directt e s t f o r heteroskedasticity.Econom etrica:J o u r n a l o f t h e EconometricS oc ie ty,8 1 7 - 8 3 8

Wopereis,M.C.S.,Defoer,T.,Idinoba,P.,Diack,S.,&Dugué,M.J.

(2008).Participatorylearningandactionresearch(PLAR)forintegratedricemanageme nt(IRM)ininlandvalleysofsub-

SaharanAfrica:technicalmanual.WARDATrainingSeries.Cotonou,Benin:AfricaRic eCentre, 128.

(2004).Reviewo f measuredc r o p waterp r o d u c t i v i t y valuesforirrigatedwheat,rice,c ottonandmaize.A g r i c u l t u r a l WaterManagement,69(2),115-133.

1.1 Cluster-specificfixedeffectmodel:within-clusterestimatorwithcluster- robustvariancematrix.

Model 1: investigating impact ofirrigationon annualriceproduction

Fixed-effects(within)regression Numberofobs =

R-sq:within=0.7488 Obspergroup:min= 2 between=0.8326 avg= 6.9 overall=0.7630 max= 50

(Std.Err.adjustedfor341clustersincommune) ln_production Coef.

The analysis reveals significant relationships between various agricultural factors and productivity Specifically, the natural logarithm of irrigated land size (ln_land_size_irrigated) shows a strong positive correlation with productivity (t = 5.89, p < 0.001) Similarly, ln_land_size (t = 10.69, p < 0.001) and ln_working_day (t = 5.63, p < 0.001) also demonstrate significant positive impacts Other notable factors include ln_fertilizer (t = 3.34, p = 0.001) and ln_seed (t = 5.56, p < 0.001), both contributing positively to productivity Conversely, the land hit by disaster ratio (land_hit_disaster_ratio) has a significant negative effect (t = -4.03, p < 0.001) Additionally, the average education level of managers (ln_manager_avg_edu) is positively associated with productivity (t = 2.81, p = 0.005), while the gender of the rice manager shows a marginal effect (rice_manager_male, t = -1.88, p = 0.061) Overall, these findings underscore the importance of land management and resource allocation in enhancing agricultural productivity.

Fixed-effects(within)regression Numberofobs = 2365

R-sq:within= 0.7451 Obspergroup:min= 2 between=0.8251 avg= 6.9 overall=0.7558 max= 50

(Std.Err.adjustedfor341clustersincommune) ln_production Coef.

The analysis reveals significant factors influencing agricultural productivity, with ln_land_size showing a strong positive correlation (β = 0.485, p < 0.001) and ln_working_day also significantly impacting output (β = 0.168, p < 0.001) Additionally, ln_labor (β = 0.012, p < 0.001) and ln_fertilizer (β = 0.038, p < 0.001) contribute positively to productivity Conversely, the land_hit_disaster_ratio negatively affects productivity (β = -0.126, p < 0.001) The education level of managers (ln_manager_avg_edu) is positively correlated with productivity (β = 0.039, p = 0.003), while the gender of rice managers shows a marginally significant negative effect (rice_manager_male, β = -0.036, p = 0.058) Furthermore, agricultural extension visits (visit_agri_extension) positively influence productivity (β = 0.054, p = 0.008), highlighting the importance of support services in enhancing agricultural outcomes.

_cons 1.50658 2574451 5.85 0.000 1.000194 2.012965 sigma_u 39540812 sigma_e 35663154 rho 55142516 (fraction ofvariancedue to u_i)

(1) ln_land_size_canal- ln_land_size_irrigated_river= 0

Model 3: exploring impact ofdifferentlocationsofcanalsystem on annualriceproduction

Fixed-effects(within)regression Numberofobs = 2159

R-sq:within= 0.7336 Obspergroup:min= 2 between=0.8086 avg= 6.6 overall=0.7460 max= 49 corr(u_i,Xb)= 0.2549

(Std.Err.adjustedfor328clustersincommune) ln_production Coef.

The analysis reveals significant relationships between various factors and agricultural productivity Notably, land size, whether for heads, middles, or tails, shows a strong positive correlation with productivity, with coefficients of 0.0193, 0.0220, and 0.0159 respectively, all statistically significant at p < 0.001 Additionally, the overall land size (ln_land_size) has a substantial impact with a coefficient of 0.4798 Working days and fertilizer also contribute positively, with coefficients of 0.1728 and 0.0384 respectively, both significant at p < 0.001 Labor and seed usage further enhance productivity, with coefficients of 0.0103 and 0.1121, significant at p < 0.01 and p < 0.001 respectively Conversely, the presence of male rice managers negatively impacts productivity, with a coefficient of -0.0477, significant at p < 0.05 The number of cultivated plots and agricultural extension visits also positively influence productivity, with coefficients of 0.0634 and 0.0525, significant at p < 0.05 However, the land disaster ratio has a detrimental effect, with a coefficient of -0.1200, significant at p < 0.001 Overall, the findings underscore the importance of land size, management, and external support in enhancing agricultural productivity.

Ft e s t checksw h e t h e r t h e r e i s differencei n impacto f locationso n canalsystemo n riceprod uction.

(1) ln_land_size_head-ln_land_size_tail=0

(1) ln_land_size_middle-ln_land_size_tail=0

(1) -ln_land_size_head+ln_land_size_middle=0

Dependentvariable:loghousehold'sriceproductionM o d e l 1 Model2 Model3 beta t-statistic beta t-statistic beta t-statistic

14 Costofrentingasset,machinery,equipment,meansandcattle(log) 0.008 (2.25)** 0.008 (2.1)** 0.007 (1.76)*

Source:calculatedby author.Othercontrolvariables includecommune’sinfrastructure(logofdistancetomainroad, logofdistancetodistrictcenter,andpercent ofconcretecommuneroads)anddummyvariablesofprovinces.*Significantat10%;**significantat5%;***significantat1%.Standarderrorisclusteredatc o m m u n e level

Model1 Model2 Model3 beta t-statistic beta t-statistic beta t-statistic

14 Costofrentingasset,machinery,equipment,meansandcattle(log) 0.006 (1.75)* 0.006 (1.66)* 0.005 (1.37)

Source:calculatedby author.Othercontrolvariablesincludecommune’sinfrastructure(logofdistancetomainroad, logofdistancetodistrictcenter,andpercent ofconcretecommuneroads)anddummyvariablesofprovinces.*Significantat10%;**significantat5%;***significantat1%.Randomeffectsandstandarderrora r e clusteredatcom munelevel

BreuschandPaganLagrangianmultipliertestforrandomeffects ln_production[commune,t]= Xb+ u[commune]+ e[commune,t]

Var sd=sqrt(Var) ln_prod~n 1.009284 1.004631 e 1251438 3537568 u 0540137 2324084

Test: Var(u)=0 chibar2(01) = 1801.81 Prob>chibar2= 0.0000

BreuschandPaganLagrangianmultipliertestforrandomeffects ln_production[commune,t]=Xb+u[commune]+e[commune,t]

Var sd=sqrt(Var) ln_prod~n 1.009284 1.004631 e 1271861 3566315 u 0537701 2318838

Test: Var(u)=0 chibar2(01) = 1613.52Prob>chibar2 = 0.0000

BreuschandPaganLagrangianmultipliertestforrandomeffects ln_production[commune,t]=Xb+u[commune]+e[commune,t]

Var sd=sqrt(Var) ln_prod~n 9727354 9862735 e 1298976 3604131 u 0546363 233744

Test: Var(u)= 0 chibar2(01) = 776.54 Prob>chibar2 = 0.0000

2.2.Testforcluster effect:choosingbetweenCSFEand OLSwithoutcluster effect

Fixed-effects(within)regression Numberofobs = 2365

R-sq:within=0.7488 Obspergroup:min= 2 between=0.8326 avg= 6.9 overall=0.7630 max= 50

The analysis reveals significant factors influencing ln_production, with an F-statistic of 374.18 and a Prob>F value of 0.0000, indicating a strong model fit Key coefficients include ln_land_size (0.4783) and ln_working_day (0.1663), both significant at p < 0.0001 Additionally, ln_fertilizer (0.0354) and ln_seed (0.1203) also show strong positive impacts on production, while ln_manager_avg_edu (0.0371) significantly contributes as well Interestingly, the variable land_hit_disaster_ratio has a negative coefficient of -0.1290, highlighting its detrimental effect on production Other variables, such as ln_rent_equi_cattle and rice_manager_male, show less significance, with p-values of 0.270 and 0.052, respectively Overall, the results underscore the importance of land size, working days, and education in agricultural production, while also indicating the risks posed by disasters.

3894805.35375675.54795428 (fraction ofvariancedue to u_i)Ftestthatallu_i=0: F(340,2008)= 5.63 Prob>F=0.0000

Fixed-effects(within)regression Numberofobs = 2365

R-sq:within= 0.7451 Obspergroup:min= 2 between=0.8251 avg= 6.9 overall=0.7558 max= 50

The regression analysis reveals significant relationships between various factors and ln_production, with an overall F-statistic of 308.50 and a p-value of 0.0000, indicating strong model significance Key predictors include ln_land_size (coef 0.4854, p < 0.0001), ln_working_day (coef 0.1682, p < 0.0001), and ln_fertilizer (coef 0.0381, p < 0.0001), all showing positive contributions to production ln_seed also has a notable impact (coef 0.1187, p < 0.0001) Conversely, the variable land_hit_disaster_ratio negatively affects production (coef -0.1262, p < 0.0001) Other significant factors include ln_manager_avg_edu (coef 0.0389, p < 0.001) and the number of cultivated plots (coef 0.0600, p < 0.002) The model suggests that agricultural practices and management education play crucial roles in enhancing production outcomes.

_cons 1.50658 1294971 11.63 0.000 1.252617 1.760543 sigma_u 39540812 sigma_e 35663154 rho 55142516 (fraction ofvariancedue to u_i)Ftestthatallu_i=0: F(340,2005)= 5.58 Prob>F=0.0000

Fixed-effects(within)regression Numberofobs = 2159

R-sq:within=0.7336 Obspergroup:min= 2 between=0.8086 avg= 6.6 overall=0.7460 max= 49

The analysis reveals a significant relationship between various factors and ln_production, with an F-statistic of 237.30 and a p-value of 0.0000 Key variables include ln_land_size, which has a strong positive coefficient of 0.4798, indicating its critical role in production Additionally, ln_working_day (0.1728), ln_fertilizer (0.0384), and ln_seed (0.1121) also show significant positive impacts In contrast, the variable land_hit_disaster_ratio exhibits a negative effect of -0.1200, highlighting the detrimental impact of disasters on production Other notable factors include ln_manager_avg_edu (0.0492) and ln_cultivated_plot_num (0.0634), both of which positively influence production outcomes The results underscore the importance of land size, working days, and education in enhancing agricultural productivity, while also emphasizing the risks posed by disasters.

40990813.36041309.5639883 (fraction ofvariancedue to u_i)Ftestthatallu_i=0: F(327,1810)= 5.10 Prob>F=0.0000

The analysis reveals various coefficients related to agricultural factors, with notable findings including a positive correlation between land size and agricultural productivity (ln_land_size = 0.4783) and a significant negative impact from land degradation (land_hit_d = -0.1290) Additionally, the variables ln_fertility and ln_seed show minor positive contributions to productivity, while labor-related factors exhibit minimal effects The results indicate consistency under the null hypothesis (Ho) and inconsistency under the alternative hypothesis (Ha), highlighting the efficiency of the xtreg model in analyzing these relationships.

Test: Ho: difference in coefficientsnot systematic chi2(16)=(b-B)'[(V_b-V_B)^(-1)](b-B)

Prob>chi2= 0.0000(V_b-V_B is not positive definite)

The analysis reveals significant insights into various agricultural factors The variable ln_land_size shows a substantial positive impact, with a coefficient of 0.4854, suggesting that larger land sizes contribute positively to agricultural productivity Additionally, ln_working_y and ln_seed also exhibit positive coefficients of 0.1682 and 0.1187, respectively, indicating their importance in enhancing crop yields In contrast, ln_land_hit_d demonstrates a negative effect with a coefficient of -0.1262, highlighting potential challenges in land management The results further indicate that variables such as ln_fertilizer and ln_manager_u have varying degrees of influence, underscoring the complexity of agricultural productivity determinants Overall, this analysis underscores the critical role of land size, labor, and management practices in optimizing agricultural outputs.

The analysis reveals significant relationships among various agricultural factors The variable ln_land_size shows a strong positive correlation with a coefficient of 0.4798, indicating its importance in agricultural productivity Additionally, ln_working_y and ln_seed also exhibit positive effects, with coefficients of 0.1728 and 0.1121, respectively Conversely, ln_land_s~ad and ln_l~w_other present negative correlations, suggesting potential challenges in these areas The results indicate that ln_fertili~r and ln_product~s have minor negative impacts, while ln_manager~u shows a slight positive influence The analysis underscores the complexity of agricultural dynamics, with ln_cultiva~m and land_hit_d~o exhibiting notable variances, highlighting the need for targeted agricultural policies.

Testofoveridentifyingrestrictions:fixedvsrandomeffectsCross- sectiontime-seriesmodel:xtregre robustcluster(commune)Sargan- Hansenstatistic180.083 Chi-sq(16) P-value=0.0000

Testofoveridentifyingrestrictions:fixedvsrandomeffectsCross- sectiontime-seriesmodel:xtregre robustcluster(commune)Sargan- Hansenstatistic184.385 Chi-sq(19)P-value=0.0000

Testofoveridentifyingrestrictions:fixedvsrandomeffectsCross- sectiontime-seriesmodel:xtregre robustcluster(commune)Sargan-Hansenstatistic220.140 Chi-sq(21) P-value=0.0000

The analysis of variance inflation factors (VIF) reveals critical insights into agricultural factors affecting rice management The variable ln_land_size shows a VIF of 3.04, indicating potential multicollinearity, while ln_working_day and ln_seed have VIFs of 2.42 and 2.06, respectively Other noteworthy variables include ln_fertilizer at 1.50 and the gender ratio of rice managers at 1.36 The VIF values for ln_cultivated_plot_num and ln_land_size_irrigated are both 1.33, suggesting they are closely related to other variables Additionally, ln_labor and ln_manager_avg_edu show VIFs of 1.28 and 1.27, respectively, highlighting the importance of labor and education in rice management The analysis also indicates that variables like ln_rent_equi_cattle and land_hit_disaster_ratio have VIFs of 1.23 and 1.14, underscoring their relevance Lastly, the ln_manager_avg_age and ln_productive_assets show VIFs of 1.11 and 1.09, while the impact of agricultural extension visits and land quality ratios are reflected in VIFs of 1.09 and 1.02, respectively.

The analysis of variance inflation factors (VIF) reveals key insights into agricultural factors influencing rice management Notably, the variable ln_land_size has a VIF of 3.10, indicating a strong correlation with other variables Other significant factors include ln_working_day (VIF 2.42) and ln_seed (VIF 2.07), which also demonstrate substantial interdependencies Variables such as ln_fertilizer (VIF 1.46) and ln_cultivated_plot_num (VIF 1.38) further contribute to understanding agricultural practices Additionally, the gender of rice managers (rice_manager_male_female VIF 1.37) and their average education level (ln_manager_avg_edu VIF 1.29) play a crucial role in rice production dynamics The analysis also highlights the importance of land quality and disaster impact, as indicated by land_hit_disaster_ratio (VIF 1.15) and land_good_quality_ratio (VIF 1.03) Overall, these findings underscore the intricate relationships among various agricultural variables, essential for optimizing rice management strategies.

The analysis of variable inflation factors (VIF) reveals key insights into agricultural productivity Notably, the variable for land size (ln_land_size) has a VIF of 3.04, indicating a strong correlation with other factors Following closely are ln_working_day at 2.38 and ln_seed at 1.98, which also show significant relationships Other important variables include ln_land_size_middle (1.61) and ln_cultivated_plot_num (1.48), highlighting their relevance Additionally, the VIF values for ln_fertilizer (1.40) and gender dynamics in management (rice_manager_male_female at 1.36) further emphasize the complexity of agricultural management Variables such as ln_manager_avg_edu (1.35) and ln_labor (1.28) contribute to understanding the impact of education and labor on productivity The analysis also considers the effects of land quality and disaster ratios, with land_hit_disaster_ratio at 1.17 and land_good_quality_ratio at 1.04, underscoring the importance of environmental factors Overall, these VIF values provide a comprehensive overview of the interrelated factors influencing agricultural outcomes.