Impact of zero tillage adoption on household welfare in Pakistan

Transcription

1 Journal of Agrcultural Technology 013 Vol. 9(7): Journal of Agrcultural Avalable Technology onlne 013, Vol. 9(7): ISSN Impact of zero tllage adopton on household welfare n Pakstan Akhter Al * and Olaf Erensten Agrcultural Economs, CIMMYT, Pakstan and Ethopa Akhter Al and Olaf Erensten (013) Impact of zero tllage adopton on household welfare n Pakstan. Journal of Agrcultural Technology 9(7): The current study was carred out to estmate the mpact of zero tllage technology adopton on household welfare n Pakstan. For the study cross sectonal data set was collected from 34 households n the rce-wheat area of Pakstan Punjab. The data was collected through stratfed random samplng technque from 3 man dstrcts of rce-wheat area namely Gujranwala, Shekhupura and Hafzabad. The emprcal analyss was carred out by employng the propensty score matchng approach to correct for potental sample selecton based ness that may arse due to systematc dfferences between adopters and non adopters. The emprcal result ndcates that adopton of zero tllage technology has postve and sgnfcant mpact on wheat yeld and household ncome whle non sgnfcant mpact on rce yeld. Most mportantly the adopton of zero tllage technology can help to reduce poverty among rural households n the range of 8-10 percent. Keywords: Adopton, Impact Evaluaton, Zero Tllage, Propensty score matchng, Punjab, Pakstan. Introducton Rce-Wheat Croppng Zone n Pakstan In South Asa, rce-wheat croppng systems cover 13.5 mllon hectares and are source of ncome and food to many mllons of people (Gupta et al., 003; Tmsna and Connor, 001). The rce-wheat system s prmarly rrgated wth 85 percent concentrated n the Indo-Gangetc Plans (IGP) (Tmsna and Connor, 001). In the face of ncreasng competton for water from ndustral, domestc and envronmental sectors, concerns are beng rased about the productvty of water used n agrculture (Kjne et al., 003). Increasng water scarcty s also seen as a major contrbutor to stagnatng productvty n the rce-wheat croppng systems of the IGP (Byerlee et al., 003; Kumar et al., 00). In the face of unrelable canal water supples, many farmers have * Correspondng author: Akhter Al; e-mal akhter.al@cgar.org 1715

2 ncreased ther relance on prvate tube wells, placng tremendous pressure on groundwater supples (Abrol 1999; Ahmad et al. 007; Quresh et al. 003). Negatve envronmental effects related to rrgaton are ncreasng as overexplotaton of groundwater and poor water management lead to the droppng of water tables n some areas and ncreased water loggng and salnty n others (Harrngton et al., 1993; Pngal and Shah 001; Quresh et al., 003). In addton, tube-well rrgaton has rased producton costs n vew of the energy expenses ncurred (electrcty or desel) (Quresh et al., 003). Agrcultural technologes that can save water, reduce producton costs and mprove producton are therefore becomng ncreasngly mportant (Gupta et al., 00; Hobbs and Gupta, 003). The rce-wheat croppng zone of Punjab s the man producer of hghvalued and fne qualty basmat rce n Pakstan. The rce produced n ths area s famous for ts gran length and aromatc characterstcs. Beng an mportant export tem, rce contrbutes sgnfcantly to the natonal foregn exchange earnngs. Wheat s the other major crop of the rce-wheat system and beng the staple food s central to natonal agrcultural polces. Rce s grown on a vast area n ths zone durng Kharf mostly followed by wheat n the Rab season. Studes have shown that a large gap exsts between the potental and actual yelds realzed by the wheat growers of the area (Byerlee et al., 1984). Farmers practces regardng land preparaton for paddy, wheat plantng tme, and other conflcts endogenous to the rce-wheat based croppng system were dentfed as the major factors lmtng wheat yeld n the area. The flooded and puddled sols that are well suted for paddy producton as compared to well-draned condtons requred for wheat s such an example of the system conflcts. The farmers n the rce-wheat zone of the Punjab predomnantly grow basmat varetes, whch are late maturng as compared to coarse varetes of rce. Therefore, paddy harvest s generally delayed at most of the farms n ths zone. The late paddy harvest coupled wth poor sol structure and loose plant resdues create problems for preparaton of a good seedbed and plantng of wheat often gets late (Byerlee et al., 1984). The farmers also had to resort to the broadcast method for wheat sowng whch results n poor and patchy plant stands. Moreover, the occurrence of ran durng land preparaton operatons may cause a further delay of -3 weeks n wheat sowng (Aslam et al., 1993). Studes have reported that after the md-november a day s delay n plantng of wheat results n a yeld loss of one percent per hectare. Adopton of Zero Tllage Technology n Pakstan Farmers n the Indo-Pak sub contnent are rapdly adoptng zero-tllage for sowng wheat after rce due to unbelevable benefts from zero tllage.e. 1716

3 Journal of Agrcultural Technology 013, Vol. 9(7): most mportant tmely sowng of wheat, more yeld, cost effectveness, sol qualty and sgnfcant savng n water. The conventonal tllage practces after rce harvest nvolve extensve ploughng wth common cultvator and deep tllage mplements for preparaton of a fne seedbed for wheat plantng whch s tme consumng as well as costly. In order to save the sowng tme and the tllage costs, a new seed drll was ntroduced n early 1980s that made t possble to sow wheat n freshly harvested and untlled paddy felds utlzng resdual mosture. The drll was named as zero tllage drll and the method of wheat sowng wth ths drll s called as zero-tllage technology. The results of ths expermentaton showed that the crop stand s mproved for wheat sown wth zero-tllage drll as compared to that obtaned under conventonal system. Based on these fndngs a comprehensve zero-tllage plot producton program was ntated n 1990s to expand the usage of the technology n the rce-wheat zone of Punjab (Aslam et al., 1993). However, a perceptble use of the drll started by the On-Farm Water Management (OFWM) and got nvolved n drll promoton efforts. The zero-tllage technology s wdely mantaned as an ntegrated approach that can tackle the problem of wheat yeld stagnaton n the rce-wheat zone by mprovng plantng tme and enhancng fertlzer and water use effcency. It s observed that zero-tllage technology helps n reducng the Phalars mnor weed nfestaton and also enables tmely seedng of the wheat crop. Durng the past two years substantal wheat acreage was sown wth zerotllage drll. ZT n rce-wheat systems ranges from surface seedng to plantng wth seed drlls drawn by four-wheel tractors (Hobbs et al., 1997). In surface seedng, wheat seeds are broadcasted on a saturated sol surface before or after rce harvest (Trpath et al., 006). It s a smple technology for resource-poor farmers requrng no land preparaton and no machnery, but ts use s stll largely confned to low-lyng felds that reman too most for tractors to enter, partcularly n the eastern IGP. Mechanzed ZT has proven more popular n the IGP, but mples the need for a tractor-drawn ZT seed drll. Ths specalzed seedng mplement allows wheat seed to be planted drectly nto unplowed felds wth a sngle pass of the tractor, often wth smultaneous basal fertlzer applcaton. The use of ZT sgnfcantly reduces energy costs, manly by reducng tractor costs assocated wth CT methods, and also because water savngs reduce the tme that tube wells must be operated. The use of ZT also allows the wheat crop to be planted sooner than would be possble usng CT methods, whch sgnfcantly reduces turnaround tme. Ths s an mportant consderaton n many parts of the rce-wheat belt, where late plantng of wheat s a major cause of reduced yelds: termnal heat mples that wheat yeld potental reduces 1717

4 by percent per day f plantng occurs after 0th November (Ortz- Monastero et al., 1994; Hobbs and Gupta, 003). The zero-tllage technology s wdely mantaned as an ntegrated approach that can tackle the problem of wheat yeld stagnaton n the rce-wheat zone by mprovng plantng tme, reducng weed nfestaton, and enhancng fertlzer and water use effcency Hobbs et al. (1997). The current study was carred out wth the objectve to analyze the mpact of zero tllage adopton on household welfare n Pakstan. The welfare s estmated n terms of crop yeld, household ncome, labour demand and nput savngs. For that the rest of the paper s organzed as follows. In secton conceptual framework s presented. In secton 3 emprcal model s presented. In secton 4 data and descrpton of varables are presented. In secton 5 emprcal results are presented and paper fnally concludes wth some polcy recommendaton. Conceptual Framework Zero Tllage Technology Adopton Let the adopton of zero tllage technology be a bnary choce, where the farmers adopt the new technology when the net benefts from adopton are greater as compared to non adopton. The dfference between the net benefts from adopton and non adopton can be denoted as P* such that P*>0 ndcates that the net benefts from adopton are greater as compared to non adopton. Although P* s not observable, but t can be expressed as a functon of observable elements n the followng latent varable model * P, P 1 [ P * 0] (1) X Where P s a bnary choce varable that equals 1 for household n case of adopton and 0 otherwse, s a vector of parameter to be estmated, X s a vector of household and plot level characterstcs and s an error term assumed to be normally dstrbuted. The probablty of adopton of new technology can be represented as * Pr( P 1) Pr( P 0) Pr( X ) 1 F( X ) () Where F s the cumulatve dstrbuton functon for. Dfferent models lke probt and logt normally results from the assumptons, whch are made about the functonal form of F. As the adopton of zero tllage drll s expected to affect the demand for nputs such as labour demands as well as yelds and net returns. 1718

5 Journal of Agrcultural Technology 013, Vol. 9(7): To lnk the adopton decson wth the potental outcomes of adopton, we consder a rsk neutral farm that maxmzes net returns, subject to compettve nputs and outputs markets and sngle output technology that s quas concave n the vector of varables nputs, W. Ths can be expressed as / max RQ( W, X ) Y W (3) Where R s the output prce, and Q s the expected output level, Y s the column vector of nput prces, where W s a vector of nput quanttes and X represent farm and household characterstcs. The farm net returns can be expressed as a functon of technology choce P, output prce, varable nputs and household characterstcs as follows. ( P, Y, R, X ) (4) Applcaton of Hotellng s lemma to equaton (3) yelds reduced form equatons for nput demand and output supply W W( P, Y, R, X ) (5) Q Q( P, Y, R, X ) (6) The specfcaton n equaton (4)-(6) show that the choce of technology nput and output prces, as well as farms and household characterstcs tends to nfluence farm net returns, demand for nputs and level of farm output. Impact evaluaton problem The dscusson n the prevous secton shows that new agrcultural technologes can help ncrease productvty and farm ncomes, and as such, mprove the welfare of farm households. Although several other reasons can be advanced to explan why agrcultural technology may be crucal n mprovng the welfare of farm households, t s dffcult to smply attrbute the dfferences n welfare between adopters and non-adopters of the technology to adopton. In cases where expermental data are gathered through randomzaton, nformaton on the counterfactual stuaton would normally be provded, and as such the problem of causal nference can be resolved. However, when the data avalable are from a cross-sectonal survey, as the one employed n the present study, no nformaton on the counterfactual stuaton can be obtaned. An effectve way of addressng the problem s to resort to an nvestgaton of the drect effect of technology adopton by lookng at the dfferences n outcomes among farm households (Blundell and Costa Das, 000). Gven that the decson of households to adopt or not to adopt the new technology may be assocated wth the net benefts of adopton, the ssue of self-selecton s crucal. To show the sgnfcance of self-selecton, consder a reduced-form relatonshp between the technology choce and the outcome varable such as 1719

6 O 0 1P X (7) Where O represents a vector of outcome varables for household such as demand for nputs and farm out, X represents farm level and household characterstcs, s an error term wth ~ N(0, ). The ssue of selecton bas arses f unobservable factors nfluence both the error term of the technology choce n equaton (1) and the error term of the outcome specfcaton n equaton (7), resultng n a correlaton between the two error terms. When the correlaton between the two error terms s greater than 0, then OLS regresson technques tends to yeld based estmates. In the current paper to address the sample selecton problem, propensty score matchng approach s employed. Propensty Score Matchng Approach The emprcal analyss s carred out by employng the propensty score matchng approach. It follows that the expected treatment effect for the treated populaton s of prmary sgnfcance. Ths effect may be gven as I 1 E( P 1) E( O1 I 1) E( O0 P 1) (8) where s the average treatment effect for the treated (ATT), O1 denotes the value of the outcome for partcpaton n a partcular labour category and O s the value of same varable for non partcpaton. A major problem s that 0 we do not observe E ( O0 P 1). Although the dfference e [ E( O1 P 1) E( O0 P 0] can be estmated, t s potentally based estmator. In the absence of expermental data, the propensty score-matchng model (PSM) can be employed to account for ths sample selecton bas (Deheja and Wahba, 00). The PSM s defned as the condtonal probablty that a farmer partcpate n a partcular labour category, gven pre-adopton characterstcs (Rosenbaum and Rubn, 1983). To create the condton of a randomzed experment, the PSM employs the unconfoundedness assumpton also known as condtonal ndependence assumpton (CIA), whch mples that once X s controlled for, partcpaton s random and uncorrelated wth the outcome varables. The PSM can be expressed as, p( X ) Pr{ P 1 X} E{ P X} (9) where I = {0,1} s the ndcator for partcpaton and X s the vector of pre-partcpaton characterstcs. The condtonal dstrbuton of X, gven p(x) s smlar n both groups of partcpaton and non partcpaton. 170

7 Journal of Agrcultural Technology 013, Vol. 9(7): Unlke the parametrc methods, propensty score matchng requres no assumpton about the functonal form n specfyng the relatonshp between outcomes and predctors of outcome. The drawback of the approach s the strong assumpton, of unconfoundness. As argued by Smth and Todd (005), there may be systematc dfferences between outcomes of partcpants and non partcpants even after condtonng because selecton s based on unmeasured characterstcs. However, Jalan and Ravallon (003) ponted out that the assumpton s no more restrctve that those of the IV approach employed n cross-sectonal data analyss. In a study by Mchalopoulos et al. (004) to assess whch non-expermental method provdes the most accurate estmates n the absence of random assgnment, they conclude that propensty score methods provded a specfcaton check that tended to elmnate bases that were larger than average. On the other hand, fxed effects model dd not consstently mprove the results. Average treatment effects After estmatng the propensty scores, the average treatment effect for the treated (ATT) can then be estmated as E{ O1 O0 P 1} E{ E{ O1 O0 P 1, p( X )}} E{ E{ O1 P 1, p( X )} E{ O0 P 0, p( X )} P 0} (10) Several technques have been developed to match adopters wth nonadopters of smlar propensty scores. In the current paper Nearest Neghbour Matchng (NNM) and Mahalanobs Metrc Matchng (MMM) methods are employed. Data and Descrpton of Varables The data and descrpton of varables s presented n table 1. The data was collected from 3 man dstrcts of rce wheat area.e. Gujranwala, Shekhupura and Hafzabad. About 46 percent farmers were ntervewed from Gujranwala dstrct, percent were ntervewed from Shekhupura dstrct and 3 percent were ntervewed from Hafzabad dstrct. In total 34 farmers were ntervewed. The mean dstance to the market was about 7 klometers from the household. The mean age of the farmers was about 45 year and the mean experence was about 4 years. The mean educaton level was about 6 years of schoolng. The average famly sze was about 7 persons per household. Informaton regardng a number of household assets was also collected. About 48 percent of the households have own refrgerator and 36 percent of the 171

8 households have own tractor. About 6 percent of the households have own bcycle. The 30 percent of the households have own motorcycle. Only 8 percent of the households have own zero tllage drll. About 9 percent of the households have own car. About 8 percent of the households have own tube well. About 30 percent of the households have own rado. Smlarly 6 percent of the households have own TV. About 58 percent of the households have own washng machne. About 76 percent of the households have avaled credt faclty. The area under rce was about 18 acres and average rce yeld was 3 maunds. The area under wheat was about 16 acres per households. The average wheat yeld was 7 maunds per household. Emprcal Results The emprcal results regardng technology adopton are presented n table. The dependent varable s dummy,.e. 1 f the farmers have adopted zero tllage technology and 0 otherwse. The dstance to market s negatve and sgnfcant at 10 percent level of sgnfcance ndcatng that more the dstance to the market less the adopton of zero tllage technology. The age coeffcent s negatve and non sgnfcant ndcatng that more the age, less the adopton of zero tllage technology and vce versa. The educaton coeffcent s postve and sgnfcant at 5 percent level of sgnfcance, ndcatng that more the educaton levels more the chances that farmers wll adopt the new technology. The famly sze coeffcent s negatve and sgnfcant at 5 percent levels of sgnfcance ndcatng that large famly sze households are less wllng to adopt the new technology and vce versa. The farmers organzaton membershp was also ncluded as dummy varable. The coeffcent s negatve and non sgnfcant. The land holdng s postve and hghly sgnfcant at 1 percent level of sgnfcance ndcatng that large land holders mostly adopt the new technology. A number of dfferent households assets were also ncluded n the model. The refrgerator ownershp s postve and sgnfcant at 10 percent level of sgnfcance. The tractor ownershp s postve and sgnfcant at 1 percent level of sgnfcance. The bcycle ownershp s postve and non sgnfcant. The car ownershp s negatve and non sgnfcant. The rado was ncluded as dummy varable to capture mpact of nformaton source regardng technology adopton, the coeffcent s postve although non sgnfcant. The farmers access to credt s postve and sgnfcant at 10 percent level of sgnfcance. The agrculture extenson servces role s postve although non sgnfcant. The dstrct dummes were also ncluded n the model to capture the regonal varaton. Although ndvdually the dstrct dummes are non 17

9 Journal of Agrcultural Technology 013, Vol. 9(7): sgnfcant but collectvely they are sgnfcantly dfferent from zero. The value of R s The hgh value of R ndcates that maxmum varaton n dependent varables s due to ndependent varables ncluded n the model. The LR s also sgnfcant at 1 percent level of sgnfcance ndcatng the robustness of varables ncluded n the model. After estmatng the adopton of zero tllage technology, n order to estmate the extent of adopton of zero tllage technology censored least absolute devaton (CLAD) v model s estmates and the results are presented n table 3. Snce n the face of heteroskedatcty or non-normalty, the Tobt model produces based estmates. In contrast, snce the censored least absolute devaton (CLAD) estmator does not depend on dstrbutonal or homoskedastcty assumptons of the errors and s robust to censorng, t produces consstent estmates even n the face of heteroskedastcty, nonnormalty and censorng. As the CLAD estmator mposes the weakest stochastc restrctons on the error terms, t results n the most precse estmates of the polcy effects v.in the model the dependent varable s numbers of acres under zero tllage technology. A set of ndependent varables are ncluded n the model. The market dstance s negatve and sgnfcant at 5 percent level of sgnfcance ndcatng that as dstance to market ncreases farmers allocate less area under zero tllage technology and vce versa. The results for age coeffcent are negatve and sgnfcant at 10 percent level of sgnfcance ndcatng that mostly the younger farmers allocate more area under zero tllage technology and vce versa. The results for educaton coeffcent are postve and hghly sgnfcant at 1 percent level of sgnfcance. The results for educaton coeffcent ndcates that mostly the educated farmers allocate more area under zero tllage technology. The results for famly sze are postve and non sgnfcant. The results for land holdng are postve and hghly sgnfcant at 1 percent level of sgnfcance. The results for tractor ownershp are also postve and sgnfcant at 1 percent level of sgnfcance, ndcatng that households havng own tractor allocate more area under zero tllage technology. The results for refrgerator ownershp are postve and non sgnfcant. The results for zero tllage drll ownershp are postve and sgnfcant at 1 percent level of sgnfcance ndcatng that households havng own zero tllage drll allocate more area under zero tllage technology and vce versa. The results for car ownershp are negatve and non sgnfcant. The results for tube well ownershp are postve and non sgnfcant. The dstrct dummes were also ncluded n the model to capture the regonal varaton. The ntal sample sze was 34 and fnal sample sze was 179. The value of R s

10 Propensty score matchng results for average treatment affect for the treated (ATT) are presented n table 4. Two dfferent matchng algorthms.e. Nearest Neghbour Matchng (NNM) and Mahalanobs Metrc Matchng (MMM) are reported. The mpact of zero tllage technology s estmated on Rce Yeld, Wheat Yeld, Labour demand, household ncome and poverty status. The results for rce yeld are postve and non sgnfcant both n case of NNM and MMM, ndcatng that zero tllage technology adopton leads to hgher rce yelds. The results for wheat yeld are postve and sgnfcant both n case of NNM and MMM matchng algorthms, ndcatng that zero tllage adopton leads to hgher wheat yelds. The postve and sgnfcant results ndcates the mportance of zero tllage technology adopton. The results for labour demand are also postve and sgnfcant at 10 percent level of sgnfcance ndcatng that zero tllage technology adopton decreases the labour demand. The results for household ncome are also postve and sgnfcant at 1 percent level of sgnfcance ndcatng that zero tllage technology adopton can help to ncrease household ncome n the range of rupees 6 to 456. The zero tllage technology adopton mpact on poverty (head count ndex) was also estmated. The poverty was negatve and sgnfcant both n case of NNM and MMM ndcatng that zero tllage adopton can decrease the household poverty levels from 8-10 percent. The results regardng technology adopton are n lne wth the prevous studes lke Al and Abdula (010). Dfferent calpers were employed for estmaton of results as table 3 ndcates. The crtcal level of hdden bas s also reported n the table. The crtcal level of hdden bas ranges from to a maxmum of the value of 1.70 ndcates that adopters and non adopters dffers n ther odds of technology adopton upto 70 percent. The number of treated and number of control are also reported n the table. A number of dfferent balancng tests were employed to check the matchng qualty lke medan absolute bas before and after matchng and percentage bas reducton and the results are presented n table 5. The bas before matchng s qute hgh both n case of nearest neghbor matchng and mahalanobs metrc matchng. Before matchng the bas s n the range of After matchng the bas s qute low and s n the range of The percentage bas reducton s n the range of to The percentage bas reducton ndcates that after matchng the covarates have been balanced and there s not much dfference between adopters and non adopters of zero tllage technology. Another test employed to check the matchng qualty s the value of R before and after matchng. The value of R should be qute hgh before matchng and should be qute low after matchng. In the current analyss the value of R before matchng qute hgh n the range of

11 Journal of Agrcultural Technology 013, Vol. 9(7): After matchng the value s close to zero ndcatng that after matchng the covarates have been balanced and there s no systematc dfferences between adopters and non adopters. The jont sgnfcance of covarates s an ndcator of covarates balancng. The jont sgnfcance of covarates should always be accepted before matchng and should always be rejected after matchng. In the current analyss the p-value s qute low before matchng and s qute after matchng hence ndcatng that after the adopters and non adopters are not systematcally dfferent from each other. The results regardng covarates balancng are n lne wth the prevous studes lke Faltemer and Abdula (009). Table 1. Data and descrpton of varables Varable Descrpton Mean Std. Dev Dstrct1 Gujranwala Dstrct Shekhupura Dstrct 3 Hafzabad Market dstance Age Experence Educaton Famly sze Refrgerator Tractor Bcycle Motorcycle Zt drll Car Tube well Rado TV Washng machne Credt (dummy) Extenson Contact Rce area (acres) Rce yeld (maunds) Wheat area (acres) Wheat yeld (maunds) 1 f farmer belongs to Gujranwala dstrct, 0 otherwse 1 f farmer belongs to Shekhupura dstrct, 0 otherwse 1 f farmer belongs to Hafzabad dstrct, 0 otherwse Dstance of market n klometers Age of farmer n number of years Experence of farmer n number of years Educaton of farmer n number of years Total number of famly members n the household 1 f household owns a refrgerator, 0 otherwse 1 f household owns a tractor, 0 otherwse 1 f household owns a bcycle, 0 otherwse 1 f household owns a motorcycle, 0 otherwse 1 f household owns a Zt drll, 0 otherwse 1 f household owns a car, 0 otherwse 1 f household owns a tube well, 0 otherwse 1 f household owns a rado, 0 otherwse 1 f household owns a TV, 0 otherwse 1 f household owns a washng machne, 0 otherwse 1 f household have access to credt faclty, 0 otherwse 1 f farmer have contact wth extenson servces, 0 otherwse Area under rce n acres Yeld of rce n maunds Area under wheat n acres Yeld of wheat n maunds Source: Own Calculatons. 175

12 Table. Propensty score matchng estmates regardng zero tllage adopton (Pobt estmates) Varable Coeffcent t-values Market dstance Age Educaton Famly sze Membershp Land holdng Refrgerator Tractor Zt drll Bcycle Car Tube well Rado Credt Agr. Extenson Constant Dstrct dummes Gujranwala Hafzabad Number of observatons * ** -0.64** *** 0.031* 0.34*** 0.31*** * Pseudo R LR Prob> Note: The results are sgnfcantly dfferent from zero at ***, **,* 1, 5 and 10 percent levels respectvely. Table 3. CLAD Results for number of acres under zero tllage technology Varable Coeffcent t-values Market dstance Age Educaton Famly sze Land holdng Tractor Refrgerator Ztdrll Car Tube well Dstrct Dummes Gujranwala Hafzabad Intal Sample Sze Fnal Sample Sze ** -0.11* 0.75*** *** 0.383*** *** Pseudo R Note: In CLAD bootstrap replcatons are 100. The results are sgnfcantly dfferent from zero at ***1%, **5% and *10%, respectvely. 176

13 Journal of Agrcultural Technology 013, Vol. 9(7): Table 5. Indcators of covarates balancng before and after matchng Matchng bas Algorthm NNM MMM Outcome Medan Medan % absolute absolute bas bas before after matchng matchng Rce Yeld Wheat Yeld Labour Demand Household Income Poverty Rce Yeld Wheat Yeld Labour Demand Household Income Poverty Value Value p-value p-value reducton before of R before matchng after R of after matchng of jont sgnfcance of covarates matchng of jont sgnfcance of covarate matchng Table 4. ATT results for rce yeld, wheat yeld, labour demand, household ncome and poverty status of the household Matchng Algorthm Nearest Neghbor Matchng (NNM) Mahalanobs Metrc Matchng (MMM) Outcome Calper ATT t-value Crtcal level of of hdden bas Rce Yeld Wheat Yeld Labour Demand Household Income Poverty Rce Yeld Wheat Yeld Labour Demand Household Income Poverty * -1.34* *** -0.08* ** -1.4* 61.81*** -0.10** Number of treated Number of control Note: ATT stands for average treatment affect for the treated. The results are sgnfcantly dfferent from zero at ***, **,* at 1, 5 and 10 % levels respectvely. Concluson Ths can be concluded from the emprcal results that adopton of zero tllage technology has postve and sgnfcant mpact on household welfare. The adopton of zero tllage technology helps n ncreasng the crops yeld, partcularly the wheat crop yeld. The ncreased crops yeld leads to hgher

14 household ncomes whch n turn can help n reducng the household poverty levels. As the current study results ndcates that household havng adopted zero tllage technology were havng less poverty levels as compared to non adopters of the technology. Besdes ths there are many other benefts of zero tllage technology, as the zero tllage adopton can results n water and labour savng. In addton the zero tllage technology s also good for the envronment. The adopton of zero tllage technology needs to be encouraged among farmng communty. References Abrol, I.P. (1999). Sustanng rce wheat system productvty n the Indo Gangetc Plan: Water management related ssues. Agrcultural Water Management 40 (1): Ahmad, M.D., Turral, H., Mash, I., Gordano, M., Masood, Z. (007). Water savng technologes: Myths and realtes revealed n Pakstan s rce-wheat systems. IWMI Research Report 108. Colombo, Sr Lanka: Internatonal Water Management Insttute. Al, A. and Abdula, A. (010). The adopton of genetcally modfed cotton and poverty reducton n Pakstan. Journal of Agrcultural Economc 61(1): Aslam, M., Majd, A., Hashm, N.I. and Hobbs, P.R. (1993). Improvng wheat yeld n the rce-wheat croppng system of the Punjab through zero tllage. Pak. J. Agrc. Res. 14(1):8-11. Blundell, R. and Costa Das, M. (000). Evaluaton methods for non-expermental data, Fscal studes, Vol. 1, pp Byerlee, D.; Al, M.; Sddq, A Sustanablty of the rce-wheat system n Pakstan s Punjab: How large s the problem? In Improvng the productvty and sustanablty of rce-wheat systems: Issues and mpacts, ed. Ladha, J. K., Hll, J. E., Duxbury, J. M. Byerlee, D.; Shekh, A. D.; Aslam, M.; Hobbs, P. R Wheat n the rce-based farmng system of the Punjab: Implcatons for research and extenson. NARC/CIMMYT Reports Seres No. 4. Islamabad: NARC/CIMMYT (Natonal Agrcultural Research Center [Pakstan]/Centro Internaconal de Majoramento de Maze y Trgo [Mexco]. Deheja, R. H. and Wahba, S. Propensty score-matchng methods for nonexpermental causal studes, The Revew of Economcs and Statstcs, Vol. 84, (00) pp Faltermeer, L. and Abdula, A. The mpact of water conservaton and ntensfcaton technologes:emprcal evdence for rce farmers n Ghana, Agrcultural Economcs, Vol. 40, (009) pp Gupta, R. K., Buresh, R. J., ASA Specal Publcaton Number 65. Madson, Wsconsn, USA: ASA-CSSA-SSSA. Gupta, R.K.; Naresh, R.K.; Hobbs, P. R.; Ladha, J.K. (00). Adoptng conservaton agrculture n the rce-wheat system of the Indo-Gangetc Plans: New opportuntes for savng water. In Water wse rce producton. Proceedngs of the nternatonal workshop on water wse rce producton, Aprl 8-11, 00, Los Banos, Phlppnes, ed. Bouman, B. A. M.; Hengsdjk, H.; Hardy, B.; Bndraban, P. S.; Tuong, T. P.; Ladha, J. K Los Banos, Phlppnes:Internatonal Rce Research Insttute. Gupta, R. K.; Naresh, R. K.; Hobbs, P. R.; Jaguo, Z.; Ladha, J.K. (003). Sustanablty of post-green revoluton agrculture: The rce-wheat croppng systems of the Indo-Gangetc Plans and Chna. In Improvng the productvty and sustanablty of rce-wheat systems: Issues and mpacts, ed. Ladha, J. K.; Hll, J. E.; Duxbury, J. M.; Gupta, R. K.; Buresh, R. J ASA Specal Publcaton Number 65. Madson, Wsconsn, USA: ASA-CSSA-SSSA. Harrngton, L. W.; Fujsaka, S.; Hobbs, P. R.; Sharma, H.C.; Sngh, R.P.; Chaudhary, M.K.; Dng, Y. and K. Schoengold The mpact of weather extremes on agrcultural producton methods: Does drought ncrease adopton of conservaton tllage practces? Amer. Agrc. Econ.Assoc. Annual Meet. Portland, Oregon, USA. July 9 - Aug. 1,

15 Journal of Agrcultural Technology 013, Vol. 9(7): Hobbs, P. R.; Gr, G. S.; Grace, P Reduced and zero tllage optons for the establshment of wheat after rce n South Asa. Rce-Wheat Consortum Paper Seres. New Delh, Inda: RWC. Hobbs, P. R.; Gupta, R. K. 003a. Resource-conservng technologes for wheat n the rce-wheat system. In Improvng the productvty and sustanablty of rce-wheat systems: Issues and mpacts, ed. Ladha, J. K.; Hll, J. E.; Duxbury, J. M.; Gupta, R. K.; Buresh, R. J ASA Specal Publcaton Number 65. Madson, Wsconsn, USA: ASACSSA-SSSA. Kjne, J. W.; Barker, R.; Molden, D., eds Water productvty n agrculture: Lmts and opportuntes for mprovement. Wallngford, UK: CABI Publcaton. Kumar, P.; Jha, D.; Kumar, A.; Chaudhary, M. K.; Grover, R. K.; Sngh, R. K.; Mtra, A.; Josh, P. K.; Sngh, A.; Badal, P. S.; Mttal, S.; Al, J. 00. Economc analyss of total factor productvty of crop sector n Indo-Gangetc Plan of Inda by dstrct and regon. Agrcultural Economcs Research Report. New Delh, Inda: Indan Agrcultural Research Insttute. Ortz-Monastero, J. I.; Dhllon, S. S.; Fscher, R. A Date of sowng effects on gran yeld and yeld components of rrgated sprng wheat cultvars and relatonshps wth radaton and temperature n Ludhana, Inda. Feld Crops Research 37 (3): Mehla, R. S.; Verma, J. K.; Gupta, R. K.; Hobbs, P. R Stagnaton n the productvty of wheat n the Indo-Gangetc Plans : Zero-tll-seed-cum-fertlzer drll as an ntegrated soluton. Rce-Wheat Consortum Paper Seres 8. New Delh, Inda: RWC. Mchalopoulos, C., Bloom, H. S. and Hll, C. J. Can propensty score methods match thefndngs from a random assgnment evaluaton of mandatory welfare-to-work programs? Revew of Economcs and Statstcs, Vol. 86, (004) pp Pngal, P. L.; Shah, M Polcy re-drectons for sustanable resource use: The rce-wheat croppng system of the Indo-Gangetc Plans. Journal of Crop Producton 3 (): Quresh, A. S.; Shah, T.; Akhtar, M The groundwater economy of Pakstan. IWMI Workng Paper 64. Pakstan Country Seres No. 19. Lahore, Pakstan: Internatonal Water Management Insttute. Tmsna, J.; Connor, D. J Productvty and management of rce-wheat croppng systems: Issues and challenges. Feld Crops Research 69 (): Trpath, J.; Adhkar, C.; Lauren, J. G.; Duxbury, J. M.; Hobbs, P. R Assessment of farmer adopton of surface seeded wheat n the Nepal tera. Rce-Wheat Consortum Paper Seres 19. New Delh, Inda: RWC. Jalan, J. and Ravallon, M. Does pped water reduce darrhea for chldren n rural Inda? Journal of Econometrcs, Vol. 11, (003) pp Rosenbaum, P. R. and Rubn, D. B. The central role of the propensty score n observatonalstudes for causal effects, Bometrka, Vol. 70, (1983) pp Smth, J. and P. Todd (005). Does Matchng Overcome LaLonde s Crtque of Nonexpermental Estmators? Journal of Econometrcs, 15. (1-): (Receved 7 September 013; accepted December 013) The parametrc methods for mpact evaluaton ncludes Heckman Selecton Estmator, Instrumental Varable Approach and Swtchng Regresson Methods. One maund s equal to 40 kgs. v The CLAD estmator s a generalzaton of the least absolute devaton (LAD) estmator. Unlke the standard estmators of the censored regresson model such as Tobt or other maxmum lkelhood approaches, the CLAD estmator s robust to heteroskedastcty and s consstent and asymptotcally normal for a wde class of error dstrbuton. v 179