DISCUSSION PAPER Report No.: ARU 34 THE IMPACT OF AGRICULTURAL EXTENSION: A CASE STUDY OF THE TRAINING AND VISIT SYSTEM (T&V) IN HARYANA, INDIA by Gershon Feder, Lawrence J. Lau and Roger H. Slade Research Unit Agriculture and Rural Development Department Operational Policy Staff World Bank May, 1985 The views presented here are those of the author(a), and they should aot be interprited as reflecting those of the World Bank. MRSiM M MMIRRmmm MniMMBMn!amu mnim m mmrwna mmseu m m Messrs. Feder and Slade are World Bank staff members. Professor Lau is a member of the Department of Economics of Stanford University. The World Bank does not accept responsibility for the views expressed herein which are those of the authors and should not be attributed to the World Bank or to its affiliated organizations. The findings, interpretations and conclusions are the results of research supported by the Bank (RP0672-29); they do not necessarily represent official policy of the Bank. The designations employed, the presentation of material and any maps used in this document are solely for the convenience of the reader and do not imply the expression of any opinion whatsoever on the part of the World Bank or its affiliates concerning the legal status of any country, territory, city, area or of its authorities or concerning the delimitation of its boundaries or national affiliation. "Whoever couLd make two ears of corn, or tzx bZzdes of grass, to grow upon a spot of ground where only one grew before, wou Zd deserve better of -an- kind, and do more essential service to his country, than the whoZe race of politiciar put together." Jonathan Swift, Gulliver's Travels, 1726. TABLE OF CONTENTS Section Page Summary ..............................................i Acknowledgement ....................................... I. Introduction ........................................... II. Background ............................................. 4 III. Econometric Analysis of Productivity Differentials ......................... 29 IV. Estimating Productivity Differentials in the Production of HYV Wheat ........................ 40 V. Estimating Productivity Differentials in HYV Rice ............................ 70 VI. Cost-Benefit Analysis ................................. 71 VII. Concluding Remarks .................................... 81 Annex I. A Note on Cost Estimation ........................ 83 Annex II. Estimating Productivity Differentials in the Production of HYV Rice, 1982 ............. 88 Annex III. Wheat Prices ..................................... 95 References .................................................. 96 THE IMPACT OF AGRICULTURAL EXTENSION A CASE STUDY OF THE TRAINING AND VISIT SYSTEM IN HARYANA, INDIA SUMMARY The paper presents results from a study designed to estimate the effect on agricultural productivity of Training and Visit (T&V) Extension. As part of the field work for the study sample surveys of farmers were conducted in 1982 and 1983 in two contiguous areas in India. One of the areas is Karnal district in the State of Haryana where the T&V system was introduced in late 1979 and the other is Kairana tehsil in the State of Uttar Pradesh where the older community development system of extension still operates. The two areas are environmentally and culturally very similar except for the presence in Karnal of the more intensive T&V extension system and some differences in the incidence of production problems. The farm level data show that in Karnal under the T&V system, and despite some organizational problems, the extent of farmer interaction with extension agents was greater than in Kairana and that farmers in Karnal viewed the change in the extension system favorably. Moreover, an examination of the rates of knowledge diffusion for a select set of practices for High Yielding Varieties (HYV) of rice and wheat suggests that T&V extension led, in the study area, to a noticeable increase in the rate of knowledge diffusion of several HYV wheat practices. The main thrust of the paper, however, is an analysis of productivity differentials between the two areas for two crops, HYV wheat and rice and, the extent to which any estimated differences can be attributed to the introduction of the T&V extension system. Using the farm level survey data from the post-project period (1982/83) and econometric estimation procedures production and supply functions are fitted with explicitly incorporate and hence control for, a number of variables which may cause productivity differences. The resulting estimates suggest that in 1982/83, after three years of T&V extension, HYV wheat yields in Karnal were about 9 per cent higher than in Kairana. This estimate however is gross of any differential that existed before the more intensive T&V extension system was introduced. The results for HYV rice were not statistically significant. Accordingly, the remainder of the analysis focusses only on HYV wheat, for which the baseline differential is next estimated. In the absence of comparable farm level survey data for the immediate pre-project period secondary data are utilized to estimate the productivity differential in HYV wheat yields between the two areas in.the baseline year (1979/80). These secondary data, mainly average yield levels, although available for a number of years are not compatible with the 1982/83 sample survey data in a number of important respects. In particular they do not distinguish between HYV and traditional varieties or between irrigated and unirrigated conditions. Hence a number of detailed -ii.- adjustments are undertaken to derive a measure of the baseline productivity differential that is consistant with the evidence derived from the detailed farm sample survey undertaken in 1982/83.. In the order in which they are undertaken these adjustments include, the estimation of mean yields of HYV wheat from the gross yield data, the estimation of a smooth time trend (to eliminate the effects of random disturbances) of the pre-project growth in average yields, the removal of the effect on productivity resulting from irrigation problems (which derive from differences in the canal irrigation systems and the incidence of tubewells in the two areas) and finally adjustments which account for systematic differences in the use of farm inputs and other variables. This last step is achieved Tith the aid of the results from the econometrically estimated production functions. These procedures establish that there was a baseline productivity differential of between 2 and 3 percent in favor of Karnal which must be subtracted from the post-project productivity differential. The final results show that, after three years of T&V extension there was a gain in productivity for HYV wheat of about 6-7 percent which can be attributed to T&V extension Subsequently a cost-benefit analysis of the incremental investment in T&V extension in Karnal is undertaken taking into account only the estimate of benefits for HYV wheat. Under several different assumptions about the profile of productivity changes over time, the results show, with a high degree of statistical confidence, that the internal rate of return is at least 15 percent. Several parts of the paper are technical and may not be of interest to all readers. In particular, pp. 29-40 can be skipped without serious loss of understanding. -iii- ACKNOWLEDGEMENT The authors gratefully acknowledge the sustained collaboration of Dr. H.C. Sharma, Dr. D.S. Nandal, Dr. U.K. Pandey, Dr. S.S. Guliani, Dr. A.C. Gangwar and Dr. K. Kumar all of the Haryana Agricultural University at Hissar without whom the study could not have been accomplished. Thanks are also due Mr. K.S. Sisodia who greatly assisted data collection in the field. Throughout the study Mr. Apparao Katikineni provided expert and dedicated computational assistance and Mr. A.K. Sundaram helped with the cost-benefit analysis. We also thank the numerous Bank staff in headquarters and the New Delhi office who provided consistent support and perceptive comments throughout the study. THE IMPACT OF AGRIGULTURAL EXTENSION: A CASE STUDY OF THE TRAINING AND VISIT METHOD (T&V) IN HARYANA, INDIA I. INTRODUCTION With dwindling land and water resources in many less-developed countries, there is increasing recognition that the major potential for increasing agricultural output and improving the lot of the rural population lies in raising the productivity of existing resources. A significant gap exists in many countries between known and fLasible agricultural technolo- gies and the actual practices of the majority of farmers. Attempts to close this gap have included programs that focus primarily on the delivery of physical inputs - fertilizer, water, seeds - as well as those aimed at improving farming practices. It has been asserted by some that even when equipment and material-dependent techniques are excluded, there are still substantial gains to be realized if cultivation practices are improved in accordance with the results of specific agro-climatic and agronomic research. A major channel for the dissemination of information on known and feasible technologies is the agricultural extension system. Extension agents are an important link between agricultural research institutions and the ultimate clients of agricultural know-how, the farmers. Upgrading the efficiency and the quality of the extension organization is thus considered by many as an indispensable element of an agricultural development strategy. Extension organizations, in various forms, are not new and in less-developed countries their record of effectiveness is mixed. While no comprehensive study is available, some of the major problems are well known. -2- Generally, insufficient extension manpower, due to inadequate budgetary allocations and a shortage of properly trained staff, leads to a high ratio of farmers to extension agents and to low effectiveness of extension efforts. In many areas agricultural extension agents are expected to perform multifarious duties, thus reducing the time available for informa- tion dissemination. Another common problem is the lack of routine updating of agents' agricultural knowledge and still another, the haphazard nature of agents' contacts with farmers. Extension projects supported by the World Bank aim to relieve the manpower shortage by funding appropriate numbers of extension staff at all levels and to assure the mobility of extension agents, and thus allow the coverage of wider areas and a larger number of farmers. To further increase the effectiveness of extension organizations, a large number of recent Bank projects have adopted the Training and Visit extension system (T&V). The essence of the T&V approach is a tightly structured work program for extension agents based on a strict schedule of regular and frequent visits to selected "contact" farmers; technical training and updating sessions for extension agents; a hierarchical organizational structure (with subject-matter specialists and supervisors to assure quality and efficiency) and an exclusive devotion to extension work. The T&V system of extension is comprehensively described in Benor and Baxter (1984). The T&V system was first tried in Turkey in the late sixties and was introduced in most Indian states between the period 1975-85. Similar projects are being implemented in Pakistan, Sri Lanka, Bangladesh, Nepal, Indonesia, Thailand, Kenya and other countries in Africa, Asia and South America. -3- Given that the T&V system of extension involves higher set-up and recurrent costs, it is only justified if it can increase farm productivity relative to other systems of extension, by more than the additional costs involved. However, the attribution of productivity effects to extension is not a simple matter, and to date there have been no rigorous studies of the productivity effects of the T&V system. The present study takes advantage of a situation whereby two neighboring and agro-climatically comparable areas in northwest India were subjected to different extension approaches. Data were collected from farmers in both areas through farm surveys designed and supervised jointly by two of the authors (Feder and Slade) and collaborators at the Haryana Agricultural University. Sample selection was based on a stratified two-staged random design. The first stage coasisted of the selection of extension circles and the second was a separate and equal selection of contact and non-contact farmers within selected circles. Farmers were interviewed twice in each season using structured questionnaires. At no point were extension personnel involved in data collection or analysis, and farmers were repeatedly assured that the surveys were not sponsored by nor connected with the extension organizations or the Departments of Agriculture of the respective states. The study benefitted from insights gained through the parallel work of Mr. B. Hoeper who was guided by Prof. P. von Blanckenburg of the Institute for Socio-Economic Agricultural Development at the Berlin Technical University. Heper's study involved inter alia detailed interviews with extension staff in part of our study area. -4- The purpose of this paper is to estimate the iacremental produc- tCivity effect, if any, which can be attributed to T&V extension over tradi- tional extension in the study areas, and to utilize the resulting estimate in a cost-benefit analysis. II. BACKGROUND The Areas Studied. The two study areas, Karnal district (in the state of Haryana) and Kairana tehsil (part of Muzaffarnager district in the state of Uttar Pradesh), are similar in many respects. They lie on opposite banks of the Ja-tuna river, are flat, have similar light alluvial soils and are connected by a solitary bridge which spans the river some distance to the south. Average annual rainfall in Karnal is 803 mm and in Kairana tehsil 897 mm. Both areas are heavily irrigated: in 1981/82 more than 95 percent of the wheat and rice in each of the two areas was grown under irrigation. The study areas are linguistically and ethnically similar. in the Rabi (dry) season, wheat is the dominant crop in both areas. In the Kharif (rainy) season, however, rice is the major crop in Karnal, sugarcane being of lesser importance. In Kairana tehsil, sugarcane is the preponderant Kharif crop while rice is of secondary importance. Kairana tehsil has better irrigation facilities than the remainder of Muzaffarnagar district, and is geographically the closest to Karnal district (see Figure 1). Both Karnal district and Muzaffarnagar district were covered by the Intensive Agricultural Development Plan (IADP), in the 60's and early 70's, which provided public funds for various investments and services. 立 -6- In the state of Uttar Pradesh the extension syscem at the time of the study was based on the traditional network of viilage level workers (VLW) administered by the Community Development Department. These workers are not only responsible for the provision of extension advice but also regulate the supply of inputs and credit' and provide links with the rural population for several other government agencies. In 1982/83 there were some 40 VLWs in Kairana tehsil: one worker for every 6.1 villages. By comparison the number of villages per Village Extension Worker (VEW) was 4.7 in Karnal. Alternatively, in 1981 there was one VEW for every 7,400 members of the rural population in Karnal and one VLW for every 11,500 rural pers(,ns in Muzaffarnagar district (separate figures for Kairana tehsil are not available). In Kairana VLWs are supplemented by staff from the Department of &griculture working mainly under the aegis of a series of special crop programs. In addition to agricultural extension VLWs are much concerned with the supply of farm inputs and the administration of specific subsidy and incentive schemes. . In Haryana, of which Karnal is one of eleven districts, the extension system was reformed along T&V lines beginning in December 1979. The reorganization reduced the ratio of villages to VEW from the initial position of 6.05 (similar to the situation in Kairana tehsil at the time of the study) to 4.7, and created higher level positions (supervisors, sub- ject-matter specialists) which did not exist before (See Table 1). Village level workers, who under the Community Development Department were performing duties in addition to extension, were transferred to the Department of Agriculture and were relieved of their non-extension duties. -7- Table 1 EXTENSION PERSONNEL IN HARYANA, 1978 AND 1983 Pre-T&V T&V Position (1978) (1983) Karnal Haryana Karnal Haryana Village Extension Worker 96a/ 1,136 124 1,490 Agricultural Extension Officer 0 0 16 186 Subject Matter Specialist 0 0 7 103 (Sub-District Level) Subject Matter Specialist 0 0 1 11 (District Level) Subject Matter Specialist n.a. 0 n.a. 4 (Zone Level) Sub-Divisional Agricultural 3 30 3 30 Officer Senior Management 1 11 1 14 a/ In the pre T&V period, village extension workers were mostly multi-purpose employees of the Community Development Department. Source: World Bank Reports. -8- By August 1981, 95% of village extension worker positions as well as 79% of the subject-matter specialist positions (technical specialists of intermediate rank) were filled. By March 1983 these figures had increased to 99% and 88%, respectively. However, in May 1982, 25% of agricultural extension officer positions (supervisors of VEWs) were still vacant and were not filled until April 1983. District Technical Committees, Zonal Workshops, and the State Technical Committee were either not convened or worked erratically until 1984. These bodies have inter alia important functions related to defining and programming technical recommendations. Their prolonged ineffectiveness constituted an important weakness in the T&V extension system in Haryana throughout the period of study. There were also other problems. Between early 1980 and early 1984, there were six changes among the three most senior government civil servants responsible for the T&V system in Haryana and this resulted in periodic and marked fluctuations in the strength of managerial commitment to the T&V system. There were also widespread and disruptive staff trans- fers at the lower levels during 1982. Characteristics of Farmers and Extension Operations in the Study Areas Information on several characteristics of the sampled farmers is summarized in Table 2. These data, which are derived from a sample survey conducted in 1982/83, provide a backdrop for the later comparative analysis between the two study areas. Since the sampling procedure distinguished Table 2 CHARACTERISTICS OF SAMPLE FARMERS RABI 1983 Kairana Tehsil (Muzaffarnagar Karnal District District, (Haryana) Uttar Pradesh) Non-Contact Contact Farmers Farmers All Farmers a/ All Farmers (%) (%) (%) (%) Own or share tubewell 96 83 84 79 Own or share tractor 43 30 31 41 Own or share other farm machinery 72 52 54 59 Reported irrigation problems affected output 3 3 3 16 Reported pest/disease problems affected output 2 0 0 11 Reported other production problems affected output 49 52 52 42 Reported shortage of credit 6 7 7 8 Reported iaadequate supply of fertilizers 2 1 1 4 Reported inadequate supply of pesticides 6 6 6 0 Reported inadequate supply of weedicides 18 17 17 4 a/ The Farmers" figures for Karnal are weighted averages of the figures for che Contac: and Non-Contact Farmers with weights of 0.1 for Contact Farmers and 0.9 for Non-Contact Farmers. These weights reflect the actual proportions of each type of farmers in all farmers in Karnal district. -10- between contact and non-contact farmers in Karnal district (the former being regularly visited by T&V extension agents), the data are also broken down by extension status. The ratio of contact to non-contact farmers in Karnal district is approximately 1 to 9. The data confirm that the farming environments are, on the whole, quite similar except for the incidence of irrigation and pest/disease problems which were higher in Kairana tehsil and the existence of other production problems and shortages in weedicide supplies which were more common in Karnal. All of the indicators listed in Table 2 are explicitly incorporated in the econometric analysis in order to account for the effects of these differences on productivity at the farm level. Extension Operations in the Study Areas Before analyzing the incremental effects of the T&V extension system on productivity, it is necessary to verify that the change from the traditional to the T&V system was indeed implemented at the village level in Karnal district. For that purpose, data were collected on the interaction between extension agents and farmers during the .years 1981/82 and 1982/83 in Karnal. Comparable data for Kairana tehsil are available for the latter year only. Table 3 reports farmers' awareness of the change in the mode of operation of the extension system and their evaluation of the nature of the change. Almost all contact farmers in Karnal, who were frequently visited by extension workers, said they were aware of the change in extension operations. According to these farmers, the most beneficial aspect of the change was an increase in the usefulness and - 11 - Table 3 RABI 1983: FARMERS' AWARENESS OF CHANGES IN EXTENSION OPERATIONS IN COMPARISON WITH PRE-1979 SYSTEM Kairana Tehsil Karnal District (Haryana) (Muzaffarnagar Dist., Contact Non-Contact Uttar Pradesh) Farmers Farmers All Farmers ResDonse (N-160) (N=165) (N=90) (%) (%) (%) 1. Not aware of change 3 58 100 2. Extension delivers 71 36 0 more useful and timely information 3. information is not as 0 0 0 useful as before 4. Know the VEW or VLW 98 52 11 -12- timeliness of extension advice. This point is of much 4inportance in understanding the nature of extension impact in this region of fairly advanced farmers, and is discussed in some detail later. About a third of non-contact farmers were aware of the intensifi- cation of the extension system, and the overwhelming majority of these pointed to the improved timing and utility of extension information as the main benefit of the change. In the sample from Kairana tehsil, where no change in extension organization took place, farmers confirmed that they were not aware of any change. Another aspect of extension-farmer interaction is the extent of farmers' familiarity with the village extension worker assigned to their village. Clearly, if farmers do not know who the extension worker is they will not approach him with queries regarding production practices. While information will diffuse among farmers through communications with one another, the ability to get timely information to address specific produc- tion problems will be diminished if farmers do not have direct access to extension workers. The last line in Table 3 shows that more than half of the non-contact farmers in Karnal (who comprise about 90 percent of all farmers) knew the VEW, while only 11 percent of the farmers in Kairana tehsil knew the village level worker assigned to their village. As expect- ed, all contact farmers knew the VEW since they saw him frequently. The regular fortnightly visit by the VEW is one of the distinct features of T&V extension. While it is not realistic to expect extension workers to provide new farming information every two weeks in an area as agriculturally advanced as Karnal district, the regularity of the visit -13- serves another purpose: if farmers know that the extension worker is regularly available in the village (rather than some distant headquarters), they can easily approach him with queries about from specific or local production problems. Table 4 shows that visit frequency and regularity were significantly higher in Karnal compared to Kairana. In the latter there was only negligible interaction between extension agents and farmers during the period covered by the study. In the T&V system, apart from large scale seasonal workshops, VEWs are required to attend a fortnightly training session where the main messages and problems anticipated in the next two weeks are discussed. Because most practices are not new, the main value, to VEWs, of this frequent training would seem to lie in the access it provides to subject-matter specialists; enabling specific local production problems (e.g., pests) about which the VEW has insufficient knowledge, to be raised and discussed. Since the training sessions are frequent the feedback to farmers is fast, allowing them to utilize the advice before adverse effects become substantial. A recent study of extension workers in Karnal and Jind districts of Haryana (Hoeper, 1983), reported that of the VEWs who were employed in the pre-T&V extension system, 93 percent perceived the training they received under the T&V system to be of better quality and quantity than before the reform. Close to 50 percent of the VEWs interviewed considered frequent training sessions to be advantageous mainly due to the resulting enhanced ability to solve farmers' problems in time. A similar percentage suggested that problem-solving is the main reason justifying frequent visits to the village. II� 7'аЫе 4 VISH'!'S Ъ'0 6�AKMtHS !fY VК{д°5 ФFд ULlд°5 I Kurnafl D1и8гlсс, 1lапуаnи � Киlcana TehиL1 5саыоп Сип[ась Fьгтnгы fэоц-СопгыСt �игшеги N::cdfFarnagac D1ы[г1сг, UCCac Fгадеиlе $ V1ы1[ед 1п $ VAalceJ .$ V1ыlгед 1n Х У1ы1[ед Х V1ыАгед 1n Х VlыАгед Sam{г1е 8еfегепсе 1�uciug $ ц!и ц ед Sample Жеfегипсе Uuring Z Vle1[ед Saaple Кufereuce �urtng Х l7быlСед Slze А1ипСЬ Sеаиггп Не,gиlлгlу eJ 51ге МипгЬ Sеаиип Kegul.�rly а/ Slze ИопсЬ Seaauq lгeguLacly а/ Ь(hигАЕ В75 8l В3 6? 1ц9 44 51 з1 - и.и. п.и. а.а. В l981 ' . F-д .� EtabH 1bб J9 85 71 ftl4 l0 12 1l - n.a. п.а. n.a. В D4bb/Н2 khaclE !)7 75 Ч! 81 159 i4 7tl 14 Ч'L 4 1t 0 1Ч82 kдhl 160 87 tl6 75 Io6 27 "16 й1 '10 0 7 U !9(!2/ЧЭ n.в.: поt avatlaLЭc а/ Perenn[аце сыlсиlагед !q relatlun to rLuac vlslsиd durlub iЬс це�ыои. ' , • � ° � -15- Knowledge Diffusion in Study Areas Having confirmed that the reformed extension system in Karnal was functioning (albeit imperfectly),at a level of intensity (and quality) higher than that which prevailed before the reform, we turn next to an examination of the nature of the information delivered by the extension service. In the area covered by the study, the "green revolution" had largely run its course before the change in extension systems took place. Thus, the vast majority of farmers were growing high-yield varieties of rice and wheat by 1979, and almost all were using substantial amounts of nitrogenous fertilizert- At Lhe time of the study the extension messages for these two main crops consisted of practices which have less dramatic and visible effects on yields, such.as the appropriate timing and dosage for lesser-known fertilizers (phosphate, potas , zinc), optimal plant spacings, the timing of irrigations, treatment of seed, the choice of varieties best suited to the locally preferred time of sowing, the utilizat*ion of pesticides, etc. Even these practices were not new, and some farmers were aware of them before the change in the extension system. It may be expected, therefore, that knowledge of these practices would diffuse even without the intensification of extension. However, the pace o diffusion may be accelerated if extension agents increase their interaction with farmers and propagate these practices. Data were collected on about a dozen improved practices for wheat and rice for which it was relatively easy to establish farmers' awareness and knowledge. These are by no means the complete set of recommended prac- -16- tices, and should be viewed as no more than a partial set. of indicators of farmers' knowledge. Knowledge is difficult to measure without conducting a thorough examination of a respondent's understanding of all aspects of a given recommendation. For some practices this was possible but for others detailed testing was beyond the time and resources available. In such cases, however, it was still possible to establish the farmers' current and past awareness of the practice. Such awareness is an important indication of knowledge because a farmer who is unaware of a practice cannot be familiar with its detail. From the resulting data the levels of knowledge in 1978/79, the year before T&V extension was introduced in Haryana, and 1982/83, four years later, have been calculated. To increase the validity of comparisons between Karnal and Kairana tehsil contact farmers in Karnal have been excluded from the analysis because they received a disproportionate amount of direct extension advice and may also be different in other ways. Thus the results reported in Tables 5 and 6 are, for Karnal, based only on responses from non-contact farmers. From Table 5 it is apparent that the levels of knowledge of recommended high-yield variety wheat practices in 1979 (the year of the changeover of the extension system in Karnal district) were higher in Kairana tehsil than in Karnal district for eight out of the ten practices studied, suggesting that the farmers in Kairana tehsil might have been more knowledgeable about HYV wheat than farmers in Karnal district, at least in 1979. However, from Table 6 it is apparent that the levels of knowledge in 1978 of recommended high-yield variety paddy practices were higher in 17 Table 5 RABI 1982/83, LEVELS OF KNOWLEDGE OF RECOMMENDED HYV WHEAT PRACTICES AMONGST NON-CONTACT FARMERS IN KARNAL AND ALL FARMERS IN KAIRANA TEHSIL Karnal District, Kairana Tehsil, Haryana Uttar Pradesh Percentage Percentage Knowledgeable Knowledgeable (N=166) a/ (N'-92) a/ Practice 1979 1983 1979 1983 1. Varieties for Late Sowing 63 94 84 95 2. Seeding Rate Late Sown Varieties 28 47 89 100 3. Seeding Rate Normally Sown Varieties 55 87 28 30 4. Correct Spacing 42 71 77 80 5. Chemical Treatment Against Fungi 2 10 10 14 6. Chemical Treatment Against Termites 3 13 8 9 7. Method of Nitrogen Application 46 82 66 71 8. Utility of Phosphate 56 97 78 87 9. Utility of Pota7h 50 72 59 65 10. Utility of Zinc Sulphate 31 60 2 5 a/ All respondents actually grew HYV wheat. - 18 - Table 6 KHARIF 1982, LEVELS OF KNOWLEDGE FOR RECOMMENDED HYV PADDY PRACTICES AMONGST NON-CONTACT FARMERS IN KARNAL AND ALL FARMERS IN KAIRANA TEHSIL Karnal District, Kairana Tehsil, Haryana Uttar Pradesh Percentage Percentage Knowledgeable Knowledgeable (N=138) a/ (N=56) a/ Practice 1978 1982 1978 1982 1. Best Spacing 58 81 46 75 2. Number of Seedlings per Station 54 97 38 67 3. Chemical Treatment of Seed 23 29 0 2 4. Utility of Weedicides 19 38 5 14 5. Salt Treatment of Seed 12 14 34 48 6. Method of Nitrogen Application 62 78 45 73 7. Utility of Pesticides 22 41 9 13 8. Utility of Phosphate 51 73 34 61 9. Utility of Potash 14 24 16 21 10. Utility of Zinc Sulphate 49 75 32 61 a/ All respondents actually grew HYV Paddy. -19- Karnal district than in Kairana tehsil for eight out of chte ten practices, suggesting that farmers in Karnal district might have been more knowledge- able about HYV paddy than farmers in Kairana tehsil. This seeming incon- sistency may be partially explained by the fact that paddy rice is the major Kharif (rainy) season crop in Karnal district, being only of secondary importance in Kairana tehsil. Thus the extension workers in Kairana tehsil might have put less priority on the dissemination of know- ledge about HYV paddy practices and the farmers in Kairana tehsil might have had less incentive to acquire such knowledge. In any event, as far as the cultivation of HYV wheat is concerned, the evidence suggests that in 1979, the farmers in Kairana tehsil had at least equal, if not better, knowledge than the farmers in Karnal district. The levels of knowledge of different practices, however, are not expected to be stationary over time. Even in the absence of any extension effort, they may be expected to increase with time,_/ until eventually they reach or approach 100 percent, that is, every farmer is knowledgeable about them. By plotting the levels of knowledge of a given practice, measured as the percentage of farmers knowledgeable, in a given community at different times against time, what is known as a "diffusion path" of the practice is obtained. Diffusion paths may take many forms but generally they must exhibit the following properties: (i) before some initial time to, the level of knowledge of the practice is zero; (ii) the level of knowledge is non-decreasing with time, and 1/ It is assumed that knowledge, once acquired, is never lost. Otherwise the levels of knowledge of certain practices may decline. -20- (iii) after a sufficiently long period after to, the level of knowledge etther reaches its maximum value, unity, or approaches unity asymptotically from below. Subject to these three properties a diffusion path can take an infinity of forms. Figure 2 below displays the shape of two diffusion paths. K K + t 1 --- - 1 (a) (b) 0 -t 0 t Figure 2. The Shape of two Plausible Diffusion Paths Figure 2(a) describes a sigmoid diffusion path while Figure-2(b) illustrates a negative exponential path. Using formulae which describe these two paths, Feder and Slade (1985) show that the diffusion paths for almost all wheat practices in Karnal lie above those of Kariana. For rice however, such a clear superiority could not be discerned. These results require the a_2riori specification of the diffusion paths. Hence we pose -21- the following question. Without any a priori knowledge of the shapes of the diffusion paths of the different practices, can we infer anything about the relative effectiveness of the T&V extension system and the traditional extension system in disseminating knowledge about the different practices from observations at only two points in time - 1979 and 1983 (1978 and 1982 for paddy)? The answer is in the affirmative if certain relatively strong but nevertheless plausible assumptions are maintained. First, assume that the shapes of the diffusion paths corresponding to each particular practice would have been the same in both study areas under the traditional extension system except for a possible difference at the initial time to, that is, the time at which the level of knowledge first becomes positive. Note however, that the initial time to may differ in the two study areas for random or historical reasons. In other words, the diffusion path of a particular practice in Karnal district under the traditional extension system would have been the same as that in Kairana tehsil except for a possible translation of the time axis due to the difference in the initial time to. This assumption is plausible because the two study areas are agro-climatically similar. -This assumption, coupled with the monotonicity of the diffusion path, implies that if the level of knowledge of a particular practice was higher (lower) in Kairana tehsil than in Karnal district in 1979 (1978), it would have continued to be higher (lower) in 1983 (1982), if the extension system in Karnal district did not change. Second, assume that any deviation between 1979 (1978) and 1983 (1982) in the shape of the diffusion path for a particular practice in the -22- Karnal district from the corresponding path in Kairana tehsil can be fully attributed to the replacement of the traditional extension system by the T&V extension system in Karnal district. Such deviations in the shapes of the diffusion paths can be identified by reversals of the rankings of Karnal district and Kairana tehsil measured by the levels of knowledge of different practices between 1979 (1978) and 1983 (1982). Third, assume that the errors in the measurement of the levels of knowledge, based on samples of farms in Karnal district and Kairana tehsil, are negligible. From Tables 5 and 6, the levels of knowledge of recommended high- yield variety wheat practices in 1983 were higher in Kairana tehsil for only four out of ten practices; the levels of knowledge of recommended high-yield variety paddy practices in 1982 were higher in Kairana tehsil for only one out of ten practices. Thus there were reversals in the ranking of the levels of knowledge of the two study areas in four high-yield variety wheat practices and one high-yield variety paddy practice between 1979 and 1983, all in favor of Karnal district. Based on the assumptions, these reversals provide prima facie evidence that the actual diffusion paths of these practices were higher in Karnal district than what they would have been had they followed the same paths as those in Kairana tehsil. We infer therefore that T&V extension which was instituted in Karnal district in 1979 must have been more effective than the traditional extension system would have been. Further corroboration of these results is provided by an examination of the reasons given by wheat growers in our sample for -23- not adopting pesticides and weedicides (Table 7). While Lack of knowledge was the main reason in Kairana tehsil, other reasons (such as unavailability or price) were dominant in Karnal district. However, even though knowledge about practices has increased, nothing can be said about the extent to which the new knowledge is either useful or profitable to farmers. Consequently, these data per se do not allow us to determine whether any gains in agricultural productivity resulted from the observed increases in knowledge, or whether such gains outweighed the incremental costs of T&V extension. Nevertheless, the results suggest that T&V extension in Karnal district, when compared to the traditional extension in Kairana tehsil, has led to a noticeable increase in the rate of knowledge diffusion for a sample set of recommended practices for HYV wheat. Increasing the rate of diffusion of knowledge of improved prac- tices is one way in which extension increases the growth of farm productiv- ity. The induced use of certain inputs is another way. But in areas as advanced as Karnal and Kairana, where most major inputs are already widely used, a major benefit of extension is the apparent ability of extension workers to tackle specific production problems which may be quite localized. As pointed out earlier, both extension workers and farmers indicated that the provision of timely answers to farmers' problems is an important aspect of the reformed extension system. Although we do not have detailed data on this aspect of extension activity, our econometric analysis will shed some light on this issue. -24- Table 7 REASONS FOR NON-ADOPTION (AMONGST FARMERS NOT ADOPTING BUT NEEDING THE PRACTICE), WHEAT RABI 1982/83 Pesticides Weedicides Karnal Karnal Karnal Karnal Reason (Contact (Non-Contact Kairana (Contact (Non-Contact Kairana Farmers) Farmers) Tehsil Farmers) Farmers) Tehsil N = 23 N =33 N = 55 N = 49 N = 54 N = 66 percent percent Too Expensive 13.3 15.0 7.3 20.4 29.6 9.0 No Knowledge 30.3 54.7 91.0 2.0 7.4 81.8 Not Available 43.3 30.3 0.0 59.2 53.7 6.1 Other 13.4 0.0 1.7 18.4 9.3 3.1 -25- An Overview of the Methodology The methodology adopted in this study consists of a comparison of the levels of productivity in two specially selected geographically adjacent areas - Karnal district and Kairana tehsil - which are similar in most agro-climatic respects, but which differed in their extension systems during the period under study. A detailed farm-level survey was conducted in 1982/83 in both areas, about 3-4 years (depending on the season considered) after the T&V extension system was introduced in Karnal district. Subsequently, econometric techniques which account for differences in the quantities of variable and fixed inputs, the types of soils, human capital, irrigation (both quantity and quality), and the production environment are used to estimate the percentage output differentials between the two areas for two major crops, high-yield wheat and rice varieties.'/. Only high-yield varieties of wheat and rice are considered, as traditional varieties are rarely grown in Karnal district, and their occurrence in our sample was practically nil. The residual productivity differential between Karnal and Kairana tehsil which is not accounted for by the extensive set of explanatory variables listed above would be attributable to differences in the extension system if there were / When differences in variable inputs are controlled, the effect of extension on input use, if any, is ignored. If differences in variable inputs are not taken into account then the output differential includes the effect of extension on farm efficiency as well as the effect on the use of inputs, provided that price differentials are also controlled. A complete accounting for the effects of extension should therefore take both types of effect into account. Our subsequent analysis attempts this but owing to inadequate price data the possible effect of extension on input use is eventually ignored. -26- no other systematic factors differentiating the two area.s, and if it can be assumed or established that the two areas were, in 1979 (1978), at the same productivity level, after similarly accounting for differences in the levels of explanatory variables in that year. This last assumption cannot be maintained and we therefore undertake detailed calculations to establish the baseline productivity differential. While the set of explanatory variables included in the analysis is quite exhaust-ve, there is always the possibility of the presence of systematic differences in micro-climatic conditions which cannot be measured or otherwise taken into account. It is to minimize this possibility that our control sample (the non-T&V case) was selected from that portion of Muzaffarnagar district (Kairana tehsil) adjacent to Karnal, rather than from the whole of Muzaffarnagar. Thus, the villages in Kairana tehsil from which the control sample was drawn are located at distances not exceeding 30 miles from the center of Karnal district. Given that the two areas are so close geographically, there is also a possibility of spillover effects. That is new information disseminated in Karnal through the reformed extension system, being transmitted to farmers in neighboring Kairana tehsil through inter-farmer communications. However, none of the messages propagated by extension regarding rice and wheat are of a revolutionary nature (the green revolution has already run its course in the study areas). Thus, there are no outstanding and visible patterns to attract a passing farmer's attention (such as dwarf varieties in areas where tall traditional varieties are grown). Recall also that, much of the extension effect is due to -27- problem-related advice, which may be relevant only to farmers in a limited area. Furthermore, our data show that the frequency of interaction across the Jamuna river of farmers in the two areas is quite limited: more than 50 percent of the sample from Kairana tehsil had not visited Haryana state since 1980. Visits are neither frequent nor regular, and spillover effects may therefore be expected to be minimal. If there were any unaccounted fixed and systematic differences between the areas or if, for historical reasons, there was more rapid diffusion of knowledge in one of the areas prior to the initiation of T&V extension, then the productivity levels in 1979, net of differences in explanatory variables, would not be equal. This initial productivity differential would have to be subtracted from the estimated productivity differential for 1983. Then, any remaining productivity differential in favor of Karnal district would indicate a positive T&V extension impact assuming that the rate of increase in productivity between the two areas in the absence of any change in the extension systems would have been the same. Ideally, if a similar detailed farm-level sample for 1979 were available, the econometric analysis which is applied to the 1982/83 sample could be replicated for the base year sample, and the residual productivity differential for 1979 could be directly estimated. In turn this would permit a test of the hypothesis that the 1982/83 residual productivity differential is larger than the 1979 differential. As above, any difference between these two levels would be attributable to T&V extension assuming that the raze of increase in productivity between the two areas in -28- the absence of any change in the extension systems wouli have been the same. Unfortunately, such detailed farm-level data from our study areas (Karnal and Kairana tehsil) for 1979 are not available. However, some data derived from seasonal crop-cutting estimates are available. These provide a time series of mean yields for wheat and rice for both Karnal and Kairana tehsil. These mean yields have a number of deficiencies. First, sample sizes for sub-districts are small, and thus the number of observations underlying the Kairana tehsil means is only 30-40. Second, the mean yields do not differentiate between irrigated and unirrigated conditions, or between high-yield and traditional varieties. Our 1982/83 sample, however, focuses on high-yield varieties under irrigated conditions only. Third, no information is available on the mean input levels (and other relevant explanatory variables) pertaining to the sample observations used to calculate the mean yields. Fourth. the mean yields in any given year include random elements which fluctuate over time, e.g., a severe pest problem in a given year, or an adverse micro-climatic condition such as hail. Thus, the differences in mean yields are not directly comparable with our estimates of productivity differentials. In order to overcome these deficiencies, we undertake a number of adjustments so as to derive from the available data mean yields for 1979 which are compatible with the evidence used in our detailed analysis of the 1982/83 data. We then utilize econometrically estimated values of the parameters associated with explanatory variables (e.g., inputs) to generate -29- an estimate of the residual productivity differential be.qeen Karnal and Kairana tehsil in 1979. This differential is then subtracted from the one estimated from the 1982/83 sample data to obtain the net differential in productivity attributable to T&V extension, if any. The gainas due to this net differential are then calculated under several alternative scenarios. These estimates of extension-induced gains are subsequently utilized in a cost-benefit analysis of the incremental investment required by T&V extension. This analysis requires assumptions regarding the dynamic pattern of productivity changes, and relies, in part, on the extrapolation of the pre-T&V trend of mean yields. III. ECONOMETRIC ANALYSIS OF PRODUCTIVITY DIFFERENTIALS The preceding section utilized the concept of a productivity differential somewhat loosely, and it is useful to provide a more precise definition at the outset of this section. In comparing the output obtained by a farmer with a given set of attributes in Karnal district with another farmer, with a possibly different set of attributes, in Kairana tehsil, we need to control for these differences in attributes (soil type,farmer characteristics, irrigation variables and production environment variables). Once the influences of these differential attributes are accounted for, any remaining differences in output between these two farmers when they apply the same amounts of physical production inputs is referred to as the disembodied productivity differential, since it does not depend on the increased use of physical production inputs. Rather, it involves better utilization and timing of inputs, the adoption of better -30- practices and more timely responses to problems encountered during the season. Thus, if intensive extension accelerates the diffusion of know- ledge related to better crop husbandry, it will cause a larger disembodied productivity differential than that which would have existed with a less intensive extension system. But differences in extension systems may have additional effects on output by inducing the increased use of physical production inputs. Any resulting increase in yield may be referred to as the embodied productivity differential. Figure 3 illustrates, for the case of one variable input, the distinction between embodied and disembodied differentials. Embodied differentials cannot be identified from an analysis of the relationship between the outputs and inputs alone. Their estimation requires an analysis of the supply of output and the demand for inputs as functions of exogenous variables including fixed inputs and the prices of output and inputs. The difference in output supply between farmers in the two areas with the same exogenous or predetermined variables (but not necessarily the same variable production inputs) represents the sum of the disembodied and embodied productivity differentials. It was our original objective to identify both the disembodied and embodied effects. As elaborated below, inadequate price data precluded the estimation of the embodied effect. Hence our subsequent cost-benefit analysis is based only on the disembodied effect. Our empirical measurement of the incremental effect of T&V extension on agricultural productivity involves the statistical estimation of both the production function and the reduced form output supply function using data from the 1982/83 sample of farm households in Karnal district (Haryana) and Kairana tehsil (Uttar Pradesh). -31- Figure 3: Embodied and Disembodled Productivity Differentials Output Disembodied Embodied Differential Differential Output in Area b -- - - - - - - - - -- - - - - - - Total Differential Output in Area a Input Input Inpur in in Area a Area b -32- The Production Function The production function is taken to be a technological relation- ship between the quantity of physical output and the quantities of physical inputs, the types of soils, the characteristics of the farm household, irrigation variables and production environment variables. In addition, the disembodied productivity differential between the area under T&V exten- sion and the area under traditional extension is captured by the coeffi- cient of a dummy variable which takes the value of unity for Kairana tehsil and zero for Karnal district. This interpretation is justified because agro-climatic and environmental conditions are approximately equivalent in these two contiguous areas, or are otherwise accounted for in the analy- sis. The soil type and irrigation variables provide additional control for variations in the environmental conditions between the two areas. The disembodied effect of the different extension systems, if any, is included in the coefficient of the Kairana tehsil dummy variable. The Physical Output and Inputs . The quantity of physical output is the quantity of gross output produced, as reported by the farmer. The physical inputs explicitly distinguished are: (1) Capital (5) Phosphates (2) Land (6) Potash (3) Labor (7) Zinc (4) Nitrogen -33- The Types of Soils Six types of soils are explicitly distinguished. They are: (1) Fine loamy and clayey, moderately drained, non-saline, non- alkaline. (2) Sandy loam to clay loam, well drained, non-salt affected. (3) Enti-soils: recent soils, sandy to laomy soils, slightly flood affected. (4) Incepti-soils: sandy loam to clayey loam, non to slightly salt affected. (5) Clay loam to sandy loam soil, moderately heavy soils, moder- ately to strongly salt affected. (6) Sandy to coarse loamy soils, non-salt affected, moderately drained. Each type of soil is represented by a dummy variable which takes the value of unity if the farm household is located in the area with the specified type of soil, and zero otherwise. In the estimation of the production function only the last five soil type dummy variables are used. The coefficient of the dummy variable for the first soil type is absorbed in the common constant term. The Characteristics of the Farm Household The characteristics of the farm household distinguished are: (1) Contact farmer: unity if the head or the farm household was previously, or is currently, a contact farmer, zero otherwise. (2) Highest level of education achieved by any member of the farm household, in years. -34- Irrigation Variables The irrigation variables distinguished are: (1) The timing of the first irrigation. (2) The number of irrigations measured by the natural logarithm of the number of irrigations. (3) The nuiber of tubewells owned. The Production Environment Variables The production environment is represented by the presence or absence of production problems. The production problems explicitly distinguished and identified are: (1) Pest problems: unity if pest problems were reported to have affected output, zero otherwise. (2) Irrigation problems: unity if irrigation problems were reported to have affected output, zero otherwise. (3) Other production problems: unity if production problems other than irrigation and pests were reported to have affected output, zero otherwise. The Functional Form The functional form selected for the estimation of the production function is the Cobb-Douglas production function: 7 6 2 4 2 (3) In Q = a + 1 .S. + y 1 6 .I. + A P1+ O 1 1 1 1 1 1 1 i 1 i1i=2 i=1 i=1 i=1 -35- where Xi's, i=1, ..., 7, are the physical inputs - capital, land, labor, nitrogen, phosphates, potash and zinc, respectively. The Si's, i=2, ..., 5, are the type of soil dummy variables, the Hi's, i=1, 2, are the farm household characteristic variables, the li's, i=1, ..., 3, are the irrigation variables, the Pi's, i=1, 2, 3, are the production problem variables and M (for Muzaffarnagar) is the Kairana tehsil dummy variable. The observed quantities of phosphates, potash and zinc in the sample are not always positive and the natural logarithm of zero is minus infinity. To overcome this difficulty natural fertility levels in terms of quantities of nutrients per hectare are introduced for phosphates, potash and zinc, say x5, x6, and x7. The total quantity of phosphates available for production, X5, is therefore the sum of the observed quantity of phosphates, X5,,and the natural level of phosphates, per hectare, x5, times the number of hectares, X2* X5 = X5 + x5 * 2 Similarly X. = X. + x. * X , i = 6, 7. 1 1 1 2' ,7 The x. 's are, however, unknown. A systematic attempt was made to estimate 1 the 'natural' levels of the different nutrients in the soil by nonlinear -36- estimation techniques. Unfortunately, none of the runs, each with differ- ent initializing values of the parameters, resulted in a convergence of the parameter estimates. Typically the 'natural' fertility level parameters would increase without limit, overwhelming the 'variable' components of total quantities of nutrients per unit of land, the production elasticity corresponding to the particular nutrient would also increase and the constant term of the production function would decrease in an offsetting manner while the estimates of the other parameters would remain approxi- mately the same. The net result was that the residuals and hence the goodness of fit statistics hardly changed. It was determined that the sample likelihood function was quite flat with respect to these parameters and the effort to estimate the 'natural' levels of nutrients was abandoned. The 'natural' levels xS, x6, and x7 were therefore all set equal to unity. The Reduced Form Output Supply Function It is assumed that within the production period capital and land are fixed inputs; labor, nitrogen and other chemical fertilizer inputs are variable inputs. The hypothesis of profit maximization is not maintained in this study. However, it is still assumed that the prices of output and the variable inputs, or more precisely the normalized prices of the variable inputs'/, may affect the output and variable input decisions. Thus, the reduced form output supply function depends on all the arguments of the production function except for the quantities of the variable inputs. In addition it also depends on the normalized prices of the 'I The normalized price of an input is its nominal price divided by the nominal price of the output. -37- variable inputs, the quantity of land owned (as a proxy variable for wealth) and problems in the supplies of certain variable inputs. The input supply problem variables explicitly distinguished are: (1) Credit problem, unity if shortage of credit reported, zero otherwise; (2) Fertilizer supply problem, unity if supply of fertilizer inadequate, zero otherwise; (3) Pesticide supply problem, unity if supply of pesticides -inadequate, zero otherwise; and (4) Weedicide supply problem, unity if supply of weedicides inadequate, zero otherwise. Finally, as in the case of the production function, the differential effect between T&V extension and traditional extension on output supply is includ- ed by the presence in the reduced form output supply function of a dummy variable M which takes the value of unity for Kairana tehsil (Uttar Pradesh) and zero for Karnal district (Haryana). The coefficient corresponding to this dummy variable in the output supply function may be interpreted to be the sum of the disembodied and embodied productivity differentials between the two study areas._/ However, this interpretation is strictly correct only if all the other variables -- fixed physical production inputs, the types of soils, the characteristics of the farm household, the irrigation variables, the production problem variables, the quantity of land owned, the normalized prices of the variable input and the / It is implicitly assumed that none of the differential effect of the T&V extension system is embodied in the capital-land ratio. -38- input supply problem variables - have independent variaLions within the sample. This latter condition is not fulfilled in the case of the Rabi (dry) season, 1983 data. Little variation was observed in the prices of output and variable inputs within either Karnal district or Kairana tehsil. There was, however, a significant difference in the prices between the two areas. Thus, the normalized price of labor and chemical fertilizer variables for Rabi(dry) season are highly collinear with the Kairana tehsil dummy variable - creating an identification problem. The reduced form output supply function for the Rabi (dry) season can be estimated only if the normalized input price variables are dropped. The resulting coefficient corresponding to the Kairana tehsil dummy variable thus captures the effects of differences in the normalized prices of the variable inputs as well. It represents the sum of the disembodied and embodied productivity differentials (which may in part be attributable to differences in extension), and normalized price effects. The functional form selected for the reduced form output supply function is the double-logarithmic form: 2 6 2 4 3 InQ = a + InX. + S. + I y. H. nW + 6. I + 1n. P. S 1 1 . 1 1 1 3 1 13 . + i=1 i=2 i=1 i=1 i=1 SSS + p. C. + or. p. + ( M (4) i=1 i=11 where W is the quantity of land owned in hectares, the Ci's, i=1, ..., 4, are the input supply problem variables and the Pi's, i=1, 2, are the normalized prices of labor and chemical fertilizer inputs, respectively. -39- The Reduced Form Input Demand Functions The reduced form input demand functions for labor or other chemical fertilizer inputs may be similarly defined. The parameters of the reduced form output supply function and the parameters of the reduced form input demand functions should be related through the production function. However, no attempt was made to impose these constraints on the parameters across the equations. Statistical Models The production function is estimated by ordinary least-squares, with a modification to allow for the possibility that the farms with irrigation problems may have a variance for their stochastic disturbance terms different from farms without irrigation problems. When both sets of farms are included in a sample a generalized least-squares estimator with an estimated variance-covariance matrix of the disturbance terms is used. Such a generalized least-squares estimator can be shown, under the standard assumption that the independent variables in the production function and the stochastic disturbance term are uncorrelated, to be consistent, asymp- totically efficient and asymptotically normal. Thus, it is possible to use statistical inference based on large sample theory. Our assumptions on the statistical model are open to the criti- cism that the quantities of output and variable inputs are simultaneously determined and hence the stochastic disturbance term in the production function may be correlated with the observed quantities of the variable inputs. This implies that the ordinary least-square estimator as well as -40- the generalized least-squares estimator will be subject LO possible simul- taneous equations bias. The conventional response to this argument is to suppose that the farmer determines the levels of the variable inputs before the stochastic disturbance term is realized or observed. For example, the farmer may maximize expected profit rather than actual profit.'/ Such an assumption will ensure that the quantities of the variable inputs will be uncorrelated with the stochastic disturbance term. Under the standard assumptions on the stochastic disturbance term in the reduced form, ordinary least-squares may be used to estimate the reduced form output supply function. Similarly, the reduced form demand functions of the variable inputs, labor and nitrogen, may be defined and estimated by ordinary least-squares. While nitrogen is used by practically all farmers in the sample, other fertilizers are used by only a portion of the sample, necessitating the use of Tobit2/ analysis for the estimation of reduced form demand functions for these inputs. IV. ESTINATING PRODUCTIVITY DIFFERENTIALS IN THE PRODUCTION OF HYV WHEAT Our analysis of wheat production pertains only to farmers growing high-yield varieties under irrigation (accounting for 96 percent of the wheat area in Karnal district). Our sample for the Rabi (dry) season 1982/83 contains only a small number of wheat growers who reported 1/ See, for example, Zellner, Kmenta and Dreze (1966). / See Tobin (1958) for a discussion of Tobit analysis. -41- experiencing irrigation problems (23 of 388). However, there is reason to believe that the production function of farmers with irrigation problems may be different from the production function of farmers not experiencing such problems. We therefore estimate the production function based on the sample of wheat farmers who did not experience irrigation problems. The production function was estimated under the assumption of constant returns to scale in the physical production inputs.'/ The esti- mated parameters of the production function are presented in column 1 of Table 8. Most of the estimated coefficients have the expected signs. It is noteworthy that the presence of production problems accounts for a substantial loss of output. As explained above, the parameter of the Kairana tehsil Jummy variable in the production function measures the extent of the disembodied productivity differential between the two study areas. The estimate reported in column I of Table 8 indicates that, when all variable inputs and other independent variables are held constant, a farmer in Kairana tehsil would obtain 8.92 percent less output per hectare than a farmer in Karnal district, if neither of them experienced irrigation problems. More precisely, other things being equal, the output of a wheat farmer in Kairana tehsil would be e-0. 0892 = 0.9147 times that of a wheat farmer in Karnal district. Thus, a wheat farmer in Karnal district would have (1/0.9147 - 1) x 100% = 9.33% higher output. This _ The assumption of constant returns to scale in the physical production inputs does not appear to be restrictive - the unrestricted estimate of the returns to scale for the sample of farms not experiencing irrigation problems is 1.004. -42- Table 8: ECONOMETRIC RESULTS FOR HYV WHEAT (FARMERS NOT EXPERIENCING IRRIGATION PROBLEMS) (1) (2) (3) ( 3)() (6) La Output !. Iutpuc (Engineering Ln Output L Nitrogen im Labor La Phosphate (Engineering Production (Reduced (Reduced (Reduced (Reduced Production Deendent Variable Function) Form) Form) Firm) Form) Function) Soil Type Nu=nw Variables 5, -0.0397 -0.0772 -0.0388 -).0232 -0.3221 -0.03411 (0.8823) (1.658) (0.6149) (3.9006) (1.07) (0.7398) S3 0.0777 0.0289 -0.0271 -0.0718 -0.0308 0.29833 (1.874) (0.684) (0.477) (2.527) (0.114) (2.299) S, -0.0224 -0.0523 0.0206 0.0851 -0.3830 0.02397 (0.4760) (1.169) (0.3439) (2.332) (3.05) (0.4860) S5 -0.0403 -0.1285 -0.1172 -0.0229 -1.782 -0.0171 (0.9319) (2.225) (2.263) (0.8851) (6.379) (0.3811) S5 -0.00558 -0.0249 -0.0181 -3.0138 -0.1583 Q.00173 (0.1333) (0.5781) (0.3127) (0.4758) (0.375) k0.2412) Production Problems Pest Problems -0.2289 -0.2109 - -0.0025 - -0.2149 (2.894) (2.549) - (0.0455) - (2.728) Other Problems -).1161 -0.1465 -3.0304 -).0393 -.4341 -0.1287 (4.709) (5.76) (0.8909) (2.298) (2.648) (5.062) Irrigation Variables Timing of First Irrigation -).00025 -.0121 0.2023 -0.0633 0.6101 -3.0306 (0.0049) (0.2236) (2.764) (1.745) (1.728) (0.5387) La Number of Irrigations 3.1137 - - - 0,14784 (2.496) - - - (2.952) Number of Tubewells 0.0376 0.0400 3.0557 -).0011 0.1027 2.24605 (3.252) (2.741) (2.822) (0.tbo) (1.099) (3.781) Fuman Caoital Education 0.00958 0.097 -0.00009 -).3014 0.0377 1.01022 (3.089) (2.975) (0.0197) (0.6551) (1.783) (3.221) Contact Farmer (Dummy) -4.00664 3.0001 0.0104 0.00002 -0.0283 -).0172 (0.2473) (0.1611) (0.2755) (0.0012) (.154) (0.6226) Owned Land - -0.0094 -3.0512 -3.3470 0.2127 - - (0.1918) (1.982) (3.63) (1.950) - Regional Effect Du=my for Kairama Tehsil -3.0892 -.1302 -).0508 -.0977 -.1755 -0.1169 (2.086) (2.928) (0.8588) (3.259) (0.619) 12.026) Inouts Labor 0.3282 - - - - 0.3231 (4.281) - - - - (4.135) Capital 3.0228 0.0294 ).0219 0.0090 0.1638 ).32034 (1.920) (2.371) (1.302) [..)847) (2.033) (1.67) Land '.3396 0.9706 ..9781 3.991 0.8362 ).3517 (6.256) (78.27) (58.23) '118.90) (10.380) (6.2M5) Nitrogen 0.0679 - - - - 9.723 (1.658) - - - - (1.738) Phosphate 0.0201 - - - - 0.01569 (1.801) - - - - (L.350) Potash -0.00224 -.- - - -.0068 (0.178) - - - - (0.5282) Zinc 0.0237 - - - 0.0236 (1.204) - - - - (1.140) Sucolv ?roblems '.'redit Shortage - -0.044 -).0886 -3.0369 0.2605 - - (0.804) (1.195) (1.003) (0.729) Fertilizer Supply Problem - 0.0163 -..2378 ).3041 -.3705 - - (0.165) -(1.356) (0.0623) (0.572) - Pesticide Supply Problem - -0.00951 - -0.0252 - - - (0.1454) - (0.5726 - - Ieedicide Supply Problem - -0.0095 - -).0105 - - (0.249) - (03.4085) -- CONSTANT 0.6626 2.173 3.338 3.410 -.0283 ).657 (2.265) (19.582) (22.44) (45.73) (0.155) 1Z.d8) 2 0.94 0.9367 1.8823 J,9647 n.a. .94 Number of Observations 365 365 365 365 365 341 Conts Constant Constant Constant Constant Constant Constant Returns Returns Returns Returns Returns Returns to Scale to Scale to Scale to Scale to Scale to Scale Tabit Analysis Tubemell Owners Only -43- order of magnitude is consistent with the opinion of loci agronomic experts that the difference in productivity between best practice and existing practice is approximately 20 percent. The estimates presented in column 1 of Table 8 are derived from a production function, where the quantities of the variable inputs are treat- ed as predetermined. In contrast we estimate the productivity differential utilizing the reduced form output supply function, in which all decision variables (endogenous variables) are omitted and additional exogenous variables which are hypothesized to influence farmers' input decisions are included. These variables include indicators of input supply problems faced by farmers in obtaining credit and supplies of various inputs, as well as an indicator of wealth (measured by the quantity of land owned), which may be a proxy for a host of other unobserved variables. Price data were available for only part of the sample'/ and are almost identical within each area. Farmers were asked towards the end of the season to give expected prices for wheat, but while these estimates varied across farms they turned out to be poor correlates for the expected prices that had prevailed earlier in the season (and affected most input decisions). As a result, our reduced form estimate does not explicitly incorporate price variation, although price effects are included indirectly through the Kairana tehsil dummy variable. 1/ Farm-level price data must be used since study area-wide average prices would correlate perfectly with the dummy variable for Kairana tehsil. -44- Maintaining the assumption of constant returns to scale (which implies that the output of wheat per hectare is independent of the quantity of wheat land) the reduced form estimate of the output supply function (Column 2 of Table 8) suggests that the 'disembodied' and 'embodied' productivity differentials combined with price effects generate a differ- ence of about 13% in wheat yields between Kairana tehsil and Karnal district. More precisely, the exogenous variables being equal, the output per hectare of a wheat farmer in Kairana tehsil would be e-0. 1302 = 0.8779 times that of a wheat farmer in Karnal district. Thus, a wheat farmer in Karnal district would have (1/0.8779 - 1) x 100% = 13.91% higher output. As with the production function, most other estimated coefficients have the expected signs. One way to calculate the net productivity differential (that is, net of price effects) implicit in the coefficient of the Kairana tehsil dummy variable estimated from the reduced form output supply function is to assume that full adjustment to prices has taken place in accordance with the hypothesis of profit maximization. This allows the calculation of own and cross-price elasticities of the variable inputs following the procedure of Lau and Yotopolous (1979). These elasticities are presented in Table 9, and are used to calculate the effect of price differentials on output supply_/. The normalized peak season labor wage in Karnal is lower, by 6.9%, than the comparable wage in Kairana tehsil. The normalized nitrogen price is lower by 6.5% and this differential is taken to be a proxy for the V For the procedure, see Jamison and Lau (1982), pp. 152-154. -45 - Table 9 IMPUTED PRICE ELASTICITIES FOR HYV WHEAT (BASED ON SAMPLE WITHOUT IRRIGATION PROBLEMS) Change in Change in Price of Labor Nitrogen Phosphate Quantity of Labor -1.5836 -0.1207 -0.0357 Nitrogen -0.5836 -1.1207 -0.0357 Phosphate -0.5836 -0.1207 -1.0357 -46- relative prices of other fertilizers as well. Under the above assumptions these price differences imply that farmers in Kairana tehsil who utilize these inputs use 12.22% less labor per hectare and 11.82% less nitrogen and other fertilizers per hectare than farmers in Karnal district. Accounting for these effects, the impact on output per hectare is calculated as the sum of the price effects on the various inputs weighted by their respective production elasticities as estimated in the production function. Potash and zinc are not included in this calculation as they are used only by a minority of farmers. The result implies that 5.08% of the output differ- ential between Kairana tehsil and Karnal district, as estimated in the reduced form output supply function, can be attributed to lower normalized input prices in Karnal in the Rabi (dry) season, which induced higher input use. This leaves a yield differential of about 8%, which in part may be attributed to the more intensive extension system in Karnal. This figure is very similar to the disembodied productivity differential estimated directly from the production function, implying that the 'embodied' produc- tivity differential may be small. As noted above this calculation assumes both profit maximization and full adjustment to prices. An alternative approach is to estimate input demand functions, again with a dummy variable to account for differences between the two areas. This dummy variable will thus account for both true productivity differential effects and price effects on input uses in HYV wheat. Subtracting these effects, properly weighted by their production elasticities, from the total yield differential estimated in the reduced form for output supply leaves the net 'disembodied' productivity differential. -47-- The estimates of the reduced form input demand functions for nitrogen, labor and phosphate are presented in columns 3, 4 and 5 of Table 8. The estimates reveal that the differential in input use between the two areas is somewhat smaller than the magnitudes predicted by the price elasticities in Table 9. As a result, the effect on output per acre attributable to the sum of 'embodied' and price effects is smaller, about 3.90% and the implied net productivity difference between Karnal and Kairana tehsil (13.02% - 3.90% = 9.12%) is somewhat larger. This figure of 9.12% for the net disembodied productivity differential is quite similar to the one estimated from the production function. The conclusion of the above analysis is that after three years of an intensive extension system in Karnal district, a yield difference between Karnal and Kairana tehsil of about 9% is observed for HYV wheat growers when all inputs and exogeneous variables are held constant. A yield difference of about 13% is observed when input levels are not held constant but all exogeneous variables except prices are controlled. This latter estimate contains, in part, yield differentials induced by the lower levels of input prices in Karnal. We turn next to a number of additional estimates made in order to assess the robustness of the results with respect to certain other characteristics of the sampled farmers. An issue of interest is whether productivity differentials are positively related to wealth. It has been argued (Moore, 1984) that the T&V extension system is often biased towards wealthy farmers and that they -48- receive a disproportionately large share of extension a-nns' attention. In order to observe whether the incremental extension effect, if any, is positively related to wealth, wealth variables for the farmers in Karnal and Kairana tehsil must be introduced to capture the separate wealth effects in the two areas, if any. Then, tests can be made to establish whether the two separate wealth effects are the same. Similar wealth effects would imply that the productivity differential between the two areas is constant and independent of wealth even though the effects of both extension systems may, in fact, be positively related to wealth. The results are reported in Table 10. It is apparent that no positive systematic relation exists between the yield differential and wealth. If anything, the point estimate indicates that the differential declines with wealth. The results are qualitatively the same for the reduced form. The next experiment eliminated from the sample all farmers who did not own tubewells, on the theoretical argument that farmers who own tubewellB have much better control over irrigation timing and may thus operate on a different production function. The hypothesis cannot be tested directly because the number of farmers without tubewells is too small to allow a credible separate production function to be estimated. It may, however, be argued a priori that since farmers with irrigation problems have already been excluded from the sample, and given that the number of tubewells owned is already included among the explanatory variables, that the results should not be much affected by considering - 49 - Table 10 ECONOMETRIC RESULTS FOR HYV WHEAT (WEALTH EFFECTS) Variable Production Function Output Supply Function Karnal* 0.0051 0.0199 In Land Owned (0.2530) (0.9502) Kairana* 0.048 0.0289 In Land Owned (1.326) (0.7689) Kairana Tehsil -0.1895 -0.2445 Dummy (-1.868) (-2.327) -50- tubewell owners only. This indeed is apparent from a com parison of columns I and 6 of Table 8. The coefficient of the Kairana tehsil dummy variable is higher by about two percentage points (in absolute value), suggesting that the disembodied yield differential may be somewhat lower for farmers who do not own tubewells. The final experiment was designed to gain insight into the sources of the disembodied effects resulting from the more intensive extension system. The samples were separated into two sub-groups, the first comprised farmers who have not experienced any production problems, and the second farmers who complained that their yield was affected by at least one type of production problem. Production functions were estimated separately for both groups. Comparing the productivity differential estimate for each sub-group (as represented by the Kairana tehsil dummy variable), we note that farmers with production problems (other than irrigation problems) in Karnal have a 14% yield advantage over farmers with production problems in Kairana tehsil, while farmers without production problems-in Karnal have only an 8% yield differential. Following the earlier discussion of the ways in which an intensive extension system may affect productivity in an area which is already quite advanced, these estimates may suggest that farmers facing production problems in an area with better access to expert advice suffer a smaller yield loss thanks to their ability to respond to these problems in a manner based on scientific knowledge. We cannot, however, draw a firm conclusion since there is no information on the yield differential between similar sub-groups of farmers in the two areas before the reform of the extension system in Karnal. -51- Establishing the Baseline Productivity Differential in Wheat In the absence of detailed farm-level data for the base year 1979/80, we utilize information on mean wheat yields derived from crop-cutting experiments conducted annually in both Karnal district and Kairana tehsil. These averages (see Table 11) do not distinguish between HYV and traditional (low-yield) varieties, between irrigated and unirri- gated areas, or between farmers with and without irrigation problems. In addition, annual mean yields in both areas tend to fluctuate over time due to random factors (e.g., micro-climatic variations). In any case, the baseline differential cannot be established by inspecting the'difference in a single year. Accordingly we make a number of adjustments to the available data. The first is for the composition of the wheat crop. The means reported in Table 11 combine high-yield and traditional varieties of wheat, while our sample pertains only to HYV wheat. This distinction would not matter if the proportion of wheat area devoted to HYV were the same in the two areas and over time. But as the following calculation for 1979/80 demonstrates, differences in the share of HYV wheat area can account for a significant portion of the difference between unadjusted mean yields. In Karnal the share of HYV wheat in total wheat area in 1979/6J was 96.18 percent. In Muzaffarnagar district, where Kairana tehsil accounts for about one third of the wheat area, the share of HYV wheat in total wheat area in 1979/80 was 82.1 percent. While a separate figure for Kairana tehsil is not available, we note that in 1979/80, the mean wheat yield in Muzaffarnagar district is almost identical to the mean wheat yield in - 52 Table 11 MEAN WHEAT YIELDS All Wheat RYV Wheat (1) (2) (3) (4) Year Karnal District Kairana Tebsil Karnal District Kairana Tehsil Kgs/Ha Kgs/Ha Kgs/Ha Kgs/Ha 1974/75 2,079 1,598 2,114 1,677 1975/76 2,150 2,024 2,150 2,129 1976/77 2,329 1,838 2,360 1,933 1977/78 2,189 1,900 2,227 1,999 1978/79 2,434 1,860 2,464 1,959 1979/80 2,371 2,068 2,404 2,210 1980/81 2,445a/ 2,253 - 2,335 1981/82 2,540a/ 2,196 2,287 1982/83 2,671a/ 2,564 2,660 Sampl e size: N.A. 30-40 a! Yields in years after the introduction of T&V extension in the district. Source: Column 1 (Karnal District), official published statistics of the novernment of Haryana. Column 2 (Kairana Tehsil), unDublished data supplied by Directorate of Agriculture, Lucknow. Columns 3 and 4 were calculated using the procedure explained in the text. The procedure utilizes data on the share of HYV in total wheat area based on district level statistics (tehsil level statistics are not available). For Muzaffarnagar district, reliable data on HYV wheat shares for the years 1975/76-1977/78 and 1982/83 were not available, thus these shares were obtained by extrapolation of the data from the years 1974/75, 1978/79-1980/81, and 1981/82. Data on HYV areas for Karnal district were obtained from Haryana official statistics. Data for Muzaffarnagar district were obtained from "Agricultural Development in Muzaffarnagar", Office of Agriculture, Muzaffarnagar, 1982/83. -53- Kairana tehsil (a difference of only 14 kg/ha which amounts to 0.67%). In the same year the percentage area irrigated by tubewells (closely related to HYV adoption) in Kairana tehsil was higher by only 3 percent than the comparable Muzaffarnagar district figure. It is therefore plausible that the proportion of HYV wheat area in Kairana tehsil was similar to that observed in Muzaffarnagar district as a whole in that year (1979/80). Publications of the Haryana Agricultural University point out that the yield of the most common HYV wheat variety in Karnal district (HD-2009) is 55.4% greater than the yield of traditional varieties.1/ This implies that the ratio of the yield of traditional varieties to HYV yield is .64. This number is compatible with ratios observed over the years 1978/79-1981/82 in various districts of Haryana, reported in Table 12, which generate a mean traditional to HYV yield ratio of .62. We use the more conservative estimate of .64. Assuming that the same ratio applies to both Karnal district and Kairana tehsil, the HYV wheat yield of the two areas in 1979/80 may be estimated as, Karnal district: 2,371 HYV wheat yield= 0.9618 + [(0.0382) x (0.64)] = 2,404. Kairana Tehsil: 2,068 HYV wheat yield = 0.821 + [(0.179) x (0.64)] = 2,210. / "Package of Practices for Rabi Crops 1982/83", Haryana Agricultural University, Hissar, 1982, Table 2, p. 3. -54 - Table 12 RATIOS OF MEAN YIELDS OF TRADITIONAL WHEAT VARIETIES TO MEAN YIELDS OF HYV WHEAT VARIETIES IN HARYANA Year 1978/79 1979/80 1980/81 1981/82 Mean District Ambala .5765 .5332 .7619 .6441 Sonepat n.a. .6569 .6129 n.a. Rohtak .6825 .5314 .5926 n.a. Jind .6657 .6386 n.a. n.a. Bhiwani n.a. n.a. n.a. .5567 Guragaon .6445 n.a. n.a. .5556 Mean .6181 Source: "Package of Practices for Rabi Crops", various issues, 1979/80-1982/83, Haryana Agricultural University, Hissar. -55- Thus, this adjustment for composition alone reduces the [979/80 yield difference between the two areas from 12.78 percent (2,068/2,371) to 8.07 percent (2,210/2,404), or, in logarithmic terms, from -0.1367 to -0.0841. We calculated the HYV wheat yields for other years in the same way and they are reported in columns 3 and 4 of Table 11. In principle an adjustment for the proportion of the wheat area which is irrigated should be made. However, because HYV wheat is generally grown only in irrigated areas, such an adjustment is unnecessary for estimating HYV wheat yields. The second adjustment is to smooth the annual fluctuations in the estimated mean HYV wheat yields caused by random factors. This is accomplished by the regression of the logarithm of HYV wheat yield on the time trend separately for Karnal district and Kairana tehsil. For Karnal district, data for the years prior to the introduction of the T&V extension system are used. The predicted 1979/80 HYV wheat yields for Karnal district and Kairana tehsil which result from the separate regressions (Table 1-3), are taken to be the "normal" HYV yields and their difference the 1979/80 baseline differential. The normalized (predicted) values for the year 1979/80 are 2,451 kg/ha and 2,207 kg/ha, for Karnal district and Kairana tehsil, respectively, implying a 9.96 percent difference or, in logarithmic terms, a differential of -.1049. It is assumed that this same "normal" differential would have prevailed in 1982/1983 had the extension system in Karnal district remained unchanged, since there is no statistically significant difference between the growth parameters - 56 - Table 13 PARAMETERS OF THE TREND OF NORMALIZED HYV WHEAT YIELDS a/ Parameter (t-statistic) Variable Karnal Kairana Tehsil Constant 7.6337 7.44550 (218.837) (146.924) Time .02839 .04230 (3.1697) (4.6977) R .72 .76 No. of Observations 6 9 a! The trend line was estimated using the logarithmic formulation In yt = In yo + t*[1n(1 + r)] where t denotes time and r is the growth rate. The parameter of time is therefore In(I + r). -57- estimated in the two separate regressions.1! The third adjustment is to account for the inclusion of farms with irrigation problems in the samples generating the mean yields. Irrigation problems can cause a substantial loss of yield and such problems are systematically more prevalent in Kairana tehsil (and in Muzaffarnagar district in general) than in Karnal district. This follows from two phenomena. First, while the extent of irrigation is similar in the two areas, a much higher proportion of the irrigated area in Karnal district is served by tubewells (76.3 percent compared to 58.9 percent in Kairana tehsil, 1979/80). Our sample from Kairana tehsil indicates that farmers without access to tubewells are more likely to experience irrigation problems than farmers with tubewells: 59 percent of farmers without tubewells had irrigation problems, compared to 9 percent of farmers with tubewells. Farmers without tubewells are dependent on public canal water, which is subject to more frequent supply disruptions, and this explains the higher incidence of irrigation problems reported by such farmers. . The second phenomenon is related to the differences in the operation of the public canal system in Karnal and Kairana tehsil. The public irrigation systems on either side of the Jamuna river, whilst both part of the Jamuna Canal system, are controlled at the distributary canal by different control devices. At the field level both areas operate local The varietal composition adjustment is made before the smoothing adjustment because the change in the composition of the wheat crop over time is more likely to be systematic than stochastic. Therefore such known systematic changes should be removed before smoothing. -58- variations of the warabandi or rotational system, whereby fields (farmers) secure water in a strictly predetermined order. The difference in the control devices is crucial. In Karnal, distributary outlets are equipped with Adjustable Proportionate Modules (APM's) which are devices which once set (calibrated to the area they command) are relatively tamper-proof, insensitive to the head of water in the distributary, as long as the distributary is flowing, and release a more or less constant volume of water. In Kairana tehsil on the opposite bank of the river APM's have not been introduced. There, the distributary outlets have a simple traditional construction (a pipe thrust through the canal bank) are prone to accidental damage and deliberate interference and the discharge is highly sensitive to the head of water in the distributary. Hence, in Kairana water discharges at the distributary level are less constant than in Karnal. In short, compared to Karnal district, Kairana tehsil has an irrigation system, controlled in such a way, that it is characterized by systematic water delivery problems at the farm level. Since these phenomena give rise to a persistently higher inci- dence of irrigation problems in Kairana, they exert a systematic negative effect on mean yields. Precise estimates of the proportion of farmers in the two areas afflicted by irrigation problems in the base year 1979/80 are not available. However, the differences in problem frequency between Karnal district and Kairana tehsil have most probably narrowed since 1979/80, as the number of tubewells in Kairana Tehsil increased more rapidly than in Karnal (the subject of tubewells is further discussed below). We thus make a conservative assumption that the incidence of irrigation problems in the two areas in 1979/80 is the same as that -59- observed in our 1982/83 sample (3.2 percent and 16.45 percent, in Karnal district and Kairana tehsil, respectively). Thus our third adjustment renders the estimated mean yields for 1979/80 properly comparable to those computed from the 1982/83 sample-data which pertain only to farmers without irrigation problems. Our approach is to calculate the effect of irrigation problems on the aggregate mean yield, given the assumed frequency of farmers with such problems, and given our knowledge of the yield differential between farmers with and without irrigation problems derived from our 1982/83 sample: Karnal District Kairana Tehsil Ratio of mean HYV wheat yield 1.038 .6537 of farmers with irrigation problems to mean yields of farmers without irrigaton problems (1982/83) Percent of farmers reporting 3.24 16.45 irrigation problems (1982/83) Since we have no evidence that irrigation problems in Karnal district cause yield loss, it is not necessry to adjust the 1979/80 normalized mean HYV wheat yield for 1979/80 (i.e., 2451 kg/ha) for irrigation problems. But the 1979/80 "normal" HYV wheat yield in Kairana tehsil for farmers without irrigation problems is signficantly higher than the unadjusted figure, as demonstrated by the calculation below: 2207 .8355 + [(.1645) x (.6537)] -60- This adjustment further reduces the 1979/80 normalized 2 ld difference between the two areas from 9.96 percent to 4.52 percent, or, in logarithmic terms, from -.1049 to -.0463. If however, we assume that a higher frequency of irrigation problems prevailed in Kairana tehsil in 1979/80, compared to that observed in 1982/83 (which is plausible given the fact that tubewell intensity increased substantially), then the calculated 1979/80 yield differential between Karnal and Kairana tehsil would be smaller (in absolute terms). In that case the productivity gain attributable to T&V extension would be higher than the estimate generated by the procedure utilized above._/ With these three adjustments, the resulting mean yields are finally comparable to the mean yields thtic can be calculated from the 1982/83 sample, with the possible exception of the sizes of the samples from which the mean yields are derived. Even with these adjustments, however, the differential between the two areas in the mean yields for 1979/80 may still differ from the same differential for 1982/83, for reasons.other than those related to the differences in the extension systems during the intervening period. For example, the differential in the mean yields may differ between 1979/80 and 1982/83 if the differential in the use of inputs per unit of land changes over time. In order to measure the change in the differential of the mean yields due to the change For instance, if the growth of 6.3 percent in the area irrigated by tubewells in Kairana tehsil between 1979/80 and 1982/83 is taken as an estimate of the increase in the proportion of farmers who own tube- wells, then the incidence of irrigation problems in 1979/80 can be calculated using the differential probability of such problems for farmers with and without tubewells (.09 and .59, respectively). The calculation yields an estimate of 19.6 percent irrigation problems in 1979/80. This, in turn, implies a baseline yield differential (in logarithm) of -.0345 rather than the -.0463 used in the text. -61- in the extension system (assuming that no extension effect is embodied in inputs), the changes in the differential in the mean yields caused by changes in the differentials in the inputs and other explanatory variables must first be taken into account. To do this, we utilize the results.of our econometrically estimated production function. From equation (3), the logarithm of yield is given by: 7 Iny = InQ - InX2 = a0 + alln(X1/X2) + I a. ln(Xi/X2) i=3 6 2 4 + . Y .H. + S.I. i=2 i=1 i=1 2 + .P . + 1 (5) i=1 The differential in the logarithm of yield between two areas, a and b, in any given year is therefore given by: Inya - lnyb 7 = lln(Xa/Xa2) - ln(Xb/Xb) + ) - bn(X /Xb i=3 6 2 4 + I B.(Sa _ Sb i(Ha - Hb) + 6.(a _ b i=2 1 1 1 =1 =1 1 2 'a b + y (P- PI) + ( (6) i=1 where C may of course be different between 1979/80 and 1982/83. The 1979/80 value of C can be estimated from the above equation. -62- Although specific data are not available to indlzate the levels of most explanatory variables in the base year 1979/80, we utilize second- ary sources which allow us to infer relevant values or which support conservative assumptions. The purpose of these adjustments is to allow an estimate of the net productivity differential (if any) which prevailed in the base year 1979/80. Several of the explanatory variables can safely be assumed to be fixed over the three year period 1979/80-1982/83, namely: soil types and farmer characteristics./ To estimate the base-year difference in the input of capital per hectare, we utilize data on the changes in the ratio of tractors per wheat hectare. This implicitly assumes that the composition of the tractor stock is the same in both areas. The difference in the logarithm of tractors per hectare between the years 1982/83 and 1979/80 for each of the study areas is taken as a measure of the change in the logarithm of capital/ha. Thus, utilizing the mean values of capital/ha from our 1982/83 sample, estimated values for capital/ha for 1979/80 can be calculated (Table 14). Labor inputs per hectare of dwarf wheat increased substantially in Karnal district between 1979/80 and 1982/83 (24 percent), but no Information is available regarding labor use in Kairana tehsil. We assume that differences between the two areas in labor use in the base year were the same as those observed in 1982/83. 1/ For Karnal the means for these variables and for other explanatory variables were calculated as weighted averages between contact and non-contact farmers, with corresponding weights of .1 and .9, reflect- ing their shares in the population. Table 14 CALCULATION OF DIFFERENCES IN CAPITAL PER HECTARE IN 1979/80 Karnal District Kairana Tebsil No. of Wheat Area Tractors per No. of Wheat Area Tractors per Year Tractors ('000 ha) '000 ha Tractors ('000 ha) '000 ha (1) 1979/80 6,589 206.9 31.85 2,626 48.7 53.92 (ii) 1982/83 9,581 225.0 42.58 4,259 48.2 88.36 (iii) Difference in Logarithm .2903 .4939 of Tractor/ha between 1982/83 and 1979/80 a/ (iv) Mean Logarithm of 8.2147 9.0526 Capital/ha in 1982/83 Sample (v) Estimated Logarithm of 7.9244 8.5587 Capital/ha in 1979/80 b/ (vi) Estimated Difference in -.6343 Logarithm of Capital/ha between Karnal and Kairana Tehsil in 1979/80 a/ Line (ili) is obtained by calculating the difference between the logarithm of lines (i) and (i). b/ Line (v) is obtained by deducting line (iii) from line (iv). -64- Specific data on fertilizer use in HYV wheat for 1979/80 in Karnal district are reported in a Haryana Agricultural University report._/ Such data are not available for Kairana tehsil so we utilise aggregate tehsil-level data on fertilizer use (by type) to infer the wheat-specific input levels in the tehsil. First, the growth in the ratios of different fertilizers to total cropped area in Kairana tehsil between 1979/80 and 1982/83 are calculated. Second, assuming these rates of growth to be identical across all crops, the relevant average fertilizer per hectare ratios for the base year in Kairana tehsil are imputed (see Table 15). Next we examine irrigation variables. The number of tubewells in Karnal district increased between the years 1979/80 and 1982/83 by only 2 percent, while the area irrigated by tubewells grew by 17 percent and the net irrigated area by 18.1 percent. In Kairana tehsil, the comparable figures are 24 percent, 6 percent and 1.6 percent. Clearly, in Karnal the average number of tubewells per irrigated hectare declined during the period considered, while in Kairana tehsil the average increased. While these figures refer to cultivation in general, rather than to wheat specifically, we know that wheat comprises the bulk of the cropped area in the Rabi season. It is also known that HYV wheat is grown only under irrigated conditions. Since the composition of holdings (in terms of farm sizes) could not have changed much in the three years under consideration, it is clear that the ratio of tubewells per farm declined in Karnal after / "Package of Practices for Rabi Crops, 1980-81", Haryana Agricultural University, Hissar, 1980, p. 138. The data were derived from a random sample of Karnal farmers included in a "Cost of Cultivation" study. -65- Table 15 ESTIMATION OF FERILIZER INPUTS PER HECTARE IN THE BASE YEAR, 1979/80 Karnal District Kairana Tebsil Year Nitrogen Phosphate Potash Nitrogen Phosphate Potash (i) Fertilizer/Total Cropped 59.36 10.61 3.86 Area in 1979/80 (kg/ha) (ii) Fertilizer/Total Cropped 78.10 16.77 5.88 Area in 1982/83 (kg/ha) (iii) Difference betwen Line .2744 .4578 .4209 (i) and Line (ii) in Logarithn (iv) Logarithn of 3.6051 2.2843 .9861 Fertilizer per Wheat Acre from 1982/83 Sample (v) Estinated Logarithn 3.6426 2.6847 1.9746 3.3307 1.8265 .5652 Fertilizer per Weat Acre a/ 1979/80 Nitrogen Phosphate Potash Zinc (vi) Differences .3119 .8582 1.4094 .1110 Betwen Karnal and Kairana 1979/80 (Logarithn) a/ For Kairana Tebsil line (v) is obtained by subtracting line (iii) from line (iv). For Karnal, the figures wre obtained from "Package of Practices for Rabi Crops 1980/81", H.A.U., 1980. These figures pertain to a sample of farmers interviewed in 1979/80. b/ In the absence of aggregate information for Kairana Tehsil, the difference in zinc input per wheat hectare between Karnal and Kairana Tebsil in 1979/80 is assumed to equal the 1982/83 difference calculated from the farm level sample. -66- 1979/80, while it increased in Kairana tehsil. In order to calculate the mean ratio of tubewells per farm for 1979/80, we use our 1982/83 sample means as a starting point. The proportionate change in the ratio of tubewells per irrigated hectare in both Karnal and Kairana between 1979/80 and 1982/83 is assumed equal to the proportionate change in the ratio of tubewells per farm. Applying this rate of change to the 1982/83 sample mean, the 1979/80 figures are calculated (Table 16). In spite of the observed trend in tubewell intensities, we conservatively assume that differences in the timely availability of water for the important first irrigation and in the average number of irrigations per acre (which in 1982/83 were higher in Karnal) remained unchanged, even though these differences were probably higher in the base year, given that relative tubewell availability in Kairana tehsil in that year was lower than in 1982/83. We do not have evidence to show that differences in the incidence of production problems (pests or other) are persistent over time, or that they follow a systematic trend. We therefore perform two alternative calculations, one assuming that the incidence of such problems was the same in Karnal district and Kairana tehsil in the base year. The other assumes that differences in the frequency of occurrence are constant over time. Table 17 summarizes the various adjustments to the original base- line yield differential which result in an estimate of the baseline differ- ential C compatible with our econometric analysis. The estimated produc- Table 16: CALCULATION OF TUBEWELL TO FARM RATIOS FOR 1979/80 Karnal District Kairana Tehsil No. of Net Irriga- Ratio of Tube- No. of Net Irriga- Ratio of Tube- Tubewells ted Area wells/Irr. Area Tubewells ted Area well LIrr. Area i) 1979/80 65,297 270,000 .2418 11,643 82,534 .1411 ii) 1982/83 66,383 319,000 .2081 14,431 83,301 .1732 iii) Ratio of 1979/80 to 1982/83 level a/ 1.1619 .8081 iv) Mean ratios tubewell/ farm from 1982/83 sample 1.3601 1.4394 4 v) Estimated mean tube- well/farm ratios in 1979/80 b/ 1.5803 1.1631 Difference between Karnal district and Kairana Tebsil in 1979/80 .4172 a/ Line (iii) is obtained as the ratio of line (i) to Line (ii) b/ Line (v) is obtained as a product of line (iii) and Line (iv). - 68 - Table 17 HYV WHEAT: CAI=JLATION OF BASELINE HYV WHEAT PRODUCTIVIY DIFFERENTIAL BETWEEN KARNAL DISTRICT AND KAIRANA TERSIL, 1979 80. Disembodied Differential (Using Production Function Parameters fra Table 8) (i) Gross Difference between Wheat .1367 Yields (Unadjusted, 1979/80) (ii) Gross Difference between HYV .0841 Wheat Yields (iii) Gross Difference beteen .1049 Normalized HYV Wheat Yields (iv) Gross Difference between .0463 Normalized HYV Wheat Yields for Farmers Without Irrigation Problems Adjustments for Systematic Differences in Explanatory Variables Soil Types .0251 Irrigation Inputs a! -.0175 Human Capital b/ .0348 Capital .0145 Labor -.0356 Fertilizers -.0379 (v) Sub-Total (Systematic Adjustments) -.0166 (vi) Estima.ted Baseline Productivity .0297 Differential c/ (Accounting for Systematic Factors) (vii) Adjustment for Transitory Differences Production Problem d/ -.0128 (viii) Estimated Baseline Productivity .0169 Differential e/ (Accounting for All Factors) a! Includes timing of first irrigation, nunber of tubewells and number of irrigations. b/ Includes education. L/ Line (vi) is obtained as line (iv) plus line (v). d/ Includes pest problems and other problems. e/ Line (v) is obtained by adding line (vii) to line (vi). Note: All differences are calculated in logaritins. - 69 - Table 18 HYV WHEAT: CALCULATION OF PRODUCTIVITY GAINS ATTRIBUTABLE TO T&V EXTENSION Disembodied- Farmers Without Irrigation Problems (i) Estimated Productivity .0892 Differential in 1982/83 (ii) Estimated Baseline (1979/80) .0297 Productivity Differential (Accounting for Systematic Factors Only) (iii) Estimated Baseline (1979/80) .0169 Productivity Differential (Accounting for All Factors) (iv) Productivity Gain .0595 (With Systematic Factor Adjustments Only)a/ (v) Productivity Gain .0723 (With All Adjustments)b/ a/ Line (iv) is obtained by subtracting line (ii) from line (i). b/ Line (v) is obtained by subtracting line (iii) from line (i). -70- tivity gain attributable to T&V extension ranges from 6 percent to 7.2 percent._/ With plausible assumptions about the time pattern of productivity gains over a typical project lifetime it can be shown that an extension-induced gain of only one percent, in HYV wheat yields alone, achieved by the third year of the extension project is sufficient to generate a very large rate of return. There is, therefore, little merit in calculating the rate of return corresponding to the point estimates of T&V-induced gains reported in Table 18. Instead, we use a benchmark level of the gain in productivity which is compatible with a satisfactory rate of return, and provide an estimate of the confidence level at which the actual rate of return is at least as high as the benchmark level. V. ESTIMATING PRODUCTIVITY DIFFERENTIALS IN HYV RICE Our econometric analysis of productivity differentials between Karnal and Kairana tehsil in the production of rice parallels the one reported for wheat. While a disembodied and total productivity differen- tial of nearly 7.5 percent is estimated for the 1982/83 sample, the estimate has a large standard error which makes it impossible to state with a reasonable degree of confidence that the 1982 productivity differential is larger than any positive baseline productivity differential. We did not therefore perform the tedious set of adjustments required to establish the magnitude of the baseline differential, and the cost-benefit analysis of These figures represent only the disembodied effect and ignore any possible embodied effect. Hence they may be considered as conservative estimates. -71- the investment in T&V extension excludes any benefits which may have been realized in rice. Annex II contains details of the econometric analysis for rice. VI. COST-BENEFIT ANALYSIS The incremental effect of T&V extension on the output of wheat which we estimated in Section IV must be set against those incremental costs that were incurred to make the additional output possible. To do so we have chosen the familiar framework of cost-benefit analysis. Although we undertake the analysis from an ex post position we do not have a complete series of either costs or benefits for any of many possible periods of years which might constitute the project life. Hence, we are obliged to make a number of assumptions which are redolent of an ex ante analysis. It is to these assumptions and our methodology for handling them that we now turn. The Pattern and Structure of Costs Acceptably complete details of the incremental costs of introduc- ing T&V extension in Haryana are available from 1979/80 to 1982/83. These costs, however, are available only for the state as a whole and hence, in order to determine the share attributable to Karnal district (our area of study) we have allocated total state incremental costs to Karnal in direct proportion to the number of VEWs in Karnal relative to the number of VEWs in Haryana as a whole. Full details of the allocation procedure can be found in Annex I. The results are set out in the first four columns of -72- Table 19. Inevitably these costs are in current prices and must be trans- formed into constant prices in order for the analysis to be meaningful. The remainder of Table 19 describes how this has been done. We elaborate only two points. First, costs have not been calculated net of taxes and duties. This is partly a reflection of inadequate information with which to do so (the available cost data are insufficiently disaggregated) and partly a recognition that taxes and duties typically appear to comprise about four percent of total costs in extension projects and hence, their subtraction would make little difference to either the level or pattern of costs. Moreover, leaving the cost stream gross of taxes and duties provides a small cushion against attributable costs that should have been charged but have not. Second, the analysis is undertaken in terms of Indian Rupees and hence exchange rate adjustments are ignored. Had they been undertaken the effect of such adjustments would, however, have been very small. Estimation of Benefits Our analysis dealt with the estimation of the incremental produc- tivity gain attributable to T&V extension in the third year of the project. However, it is reasonable to expect that gains continue to accrue over time, but in the absence of data with which to estimate such gains we are obliged to construct a dynamic model to simulate the evolution, over time, of productivity change with and without T&V extension. The model is described below: Table 19: EXTENSION COSTS IN KARNAL DISTRICT (Thousands of Rupees) Actual & Projected Costs in Actual Costs in Current Prices Constant 1979-80 Prices /a Civil Equipment Other Civil Equipment Other Year Works & Vehicles Staff Costs Works Ab & Vehicles /c Staff Id Costs Le Total 1979-80 -- 539.9 405.9 581.7 -- 539.9 405.9 581.7 1,527.5 1980-81 964.8 94.9 483.2 829.7 816.2 85.6 430.7 662.7 1,995.2 1981-82 369.8 91.9 473.2 1,179.9 286.2 74.7 381.0 781.4 1,523.3 1982-83 152.8 -- 580.8 1,332.7 115.3 -- 428.0 828.3 1,371.6 1983-84 -- -- 585.1 893.6 If -- -- 396.4 Lg 522.0 918,4 1984-85 -- -- -- -- -- -- 396.4 522.0 918.4 1985-86 -- -- -- -- -- 539.9 396.4 522.0 1,458.3 1986-87 -- -- -- -- -- 85.6 396.4 522.0 1,004.0 1987-88 -- -- -- -- -- 74.7 396.4 522.0 993.1 1988-89 -- -- -- -- -- -- 396.4 522.0 918.4 1989-90 -- -- -- -- -- -- 396.4 522.0 918.4 1990-91 -- -- -- -- -- -- 396.4 522.0 918.4 1991-92 -- -- -- -- -- 539.9 396.4 522.0 1,458.3 1992-93 -- -- -- -- -- 85.6 396.4 522.0 1,004.0 1993-94 -- -- -- -- -- 74.7 396.4 522.0 993.1 1994-95 -- -- -- -- -- -- 396.4 522.0 918.4 1995-96 -- -- -- -- -- -- 396.4 522.0 918.4 1996-97 -- -- -- -- -- Lh -- 396.4 522.0 918.4 1997-98 -- -- -- -- -- 539.9 396.4 522.0 1,458.3 1998-99 -- -- -- -- -- 85.6 396.4 522.0 1,004.0 1999-00 -- -- -- -- -- 74.7 396.4 522.0 993.1 2000-01 -- -- -- -- -- -- 396.4 522.0 918.4 2001-02 -- -- -- -- -- -- 396.4 522.0 894.9 /a Unadjusted for incidence of taxes and duties. /b. Deflated by Wholesale Price Ind-:x for all commodities. Ic Deflated by Wholesale Price In6ex for mechanical and transport equipment. Replacement assumed after, six years. /d Deflated by Cost of Living Price Index for Urban Non-manual Employees (Chandigarh, Haryana). Ie Deflated by Wholesale Price Index for fuel, power, light and lubricants. I Average of 1982-83 and 1983-84. /g Assumed constant from 1983-84 onwards. lh If project life is 16 years then the residual value of civil works and equipment in 1995-96 is Rs.355.5 thousand assuming straightline depreciation. Hence total costs for that year become Rs. 562.9 thousand. Similarly, with project life of 9 years and 6 years, residual values of 1,015 thousands and 1,359 thousands need to be deducted from the cost of the respective terminal years. -74- Notation: a = annual rate of productivity change in the absence of T&V extension. 6 = productivity gain in year 3 which is due to T&V extension. T* = number of years until extension benefits dissipate, i.e., until productivity reaches the same level it would have attained without T&V extenoion. Y0 = yield level in the base year (1979/80). y= yield level in year t. In the absence of T&V extension, the yield in year t is given by ln yt = ln yo + a*t With the enhanced extension system, yield in year 3 is in y3 = In yo + 3 * a+ 6 We assume that growth up to year 3 is exponential, thus the annual rate of productivity growth in the years 1-3 is given by a + (6/3). If year T* is taken as the year in which yields with and with- out T&V extension will be equal, the exponential rate of productivity change between year 4 and year T* (say 5) can be derived from the equation a * T* = 3a + 6 + $*(T* - 3) -75- which implies a= - [6/(T* -3)] Yields at time t after T&V extension has been implemented, are thus given by the following equation: e[a+ (6/3)]*t < t < 3 y- 3 /(T* - 3) + 8*(t - 3) yt 0* e 3 < t < T* eat T* <t The model is illustrated in Figure 4. Since yields would be growing at a rate of a without T&V exten- sion, the benefits per acre (say b t) are given by the following [a+ 6/3]*t at e - e 1 < t < 3 bt P3a - 6/(T* - 3) + *(t - 3) at bt t 0* e - e 3K<t <T* 0 T* < t where pt is the price of wheat. Prices of wheat are reported in Annex 3. Given HYV wheat acreage in years 1-3 and assuming it remains unchanged thereafter, the calculation of the stream of total benefits over time is straightforward for any set of model parameters. - 76 - Figure 4: Time Pattern of Productivity Change ln y (log. yield) With T&V Without T&V tim 79/80 82/83 T* = Productivity gain due to extension in 1982/83. -77- The initial yield level is based on the normnlized mean Karnal yield for 1979/80, adjusted to reflect only HYV wheat. The rate of productivity growth which would have prevailed without T&V extension is assumed to be the same as the rate estimated in the time-trend regression for the pre-T&V period 1974/75-1979/80 in Karnal, i.e., .0284 percent. We consider three alternative values of T*, namely 7 years, 10 years and 17 years. These imply that the project life is 6 years, 9 years and 16 years respectively, assuming that the project terminates when gross benefits are zero. It can be shown that with these specifications, extension-induced productivity gains (i.e.,6) in HYV whear yield of 0.46, 0.4105 and 0.3446 percent (corresponding to a project life of 6, 9 and 16 years, respectively) in the third year of the project (1982/83) will generate an internal rate of return of 15 percent. The gross benefit streams corresponding to these three scenarios are presented in columns (1), (2) and (3) of Table 20. It is clear, however, from a comparison of Tables 19 and 20 that in the last few years under each scenario the annual costs exceed the corresponding annual benefits. The assumption that the project is maintained in the last few years can be rationalized only if institutional rigidities prevent closing of the project before gross benefits decline to zero. There is, however, for each given value of T*, an efficient project life which is determined by the length of time taken for marginal costs to equa.l or exceed marginal benefits. Terminating project life at the efficient point for a given value of T* would raise the rate of return above 20 percent. -78 - Table 20: INCREMENTAL GROSS BENEFITS IN KARNAL DISTRICT FOR DIFFERENT LENGTHS OF PROJECT LIFE /a (Constant 1979/80 Prices: Millions of Rupees) Project Life (1) (2) (3) Year 6 years 9 years 16 years 1979-80 -- -- -- 1980-81 .865634 .772421 .648348 1981-82 1.877716 1.675381 1.406113 1982-83 3.084684 2.758063 2.309497 1983-84 2.378770 2.426133 2.206020 1984-85 1.630578 2.079391 2.094712 1985-86 .838286 1.710918 1.975207 1986-87 -- 1.319755 1.847127 1987-88 -- .904910 1.710079 1988-89 -- .465348 1.563654 1989-90 -- -- 1.407426 1990-91 -- -- 1.240954 1991-92 -- -- 1.063779 1992-93 -- -- .875423 1993-94 -- -- .675393 1994-95 -- -- .463172 1995-96 -- -- .238226 Rate of Returr 15% 15% 15% /a Benefits calculated assuming yield differentials due to extension in 1982-83 are .0046 (i.e., 0.46 percent), .004105 and .003446 for project life of 6, 9, and 16 years respectively. -79- The above discussion established benchmark levels for produc- tivity gains attributed to extension which are compatible with lower-bound rates of return. The point estimates of gains due to T&V extension reported in Table 18 (lines (iv) and (v)) are obviously much higher than the benchmark levels, implying that the rates of return corresponding to the values in Table 18 are also much higher. Taking the baseline (1979/80) productivity differentials between Karnal and Kairana tehsil (lines (vi) and (viii) of Table 17) as given (non-stochastic), we can calculate a conditional confidence level for the lower-bound rates of return corresponding to Table 20, on the assumption that our point estimates for productivity differentials between Karnal and Kairana in 1982/83 (reported in Table 8) are normally distributed./ The conditional confidence levels corresponding to each of three assumed scenarios are reported in Table 21. They imply that when disembodied productivity gains only are considered, there is a probability of approximately .90 that the rate of return was at least 15 percent for each of the three project life scenarios. The confidence level is higher (approximately .95) if the calculation of the baseline takes into account systematic as well as transitory adjustments. We emphasize that the probabilistic statements are conditional on the assumptions made at different stages of the analysis. 1/ They are asymptotically normally .distributed in any case. -80- Table 21 CONDITIONAL CONFIDENCE LEVELS FOR BENCHMARK RATE OF RETURNL/ Confidence Level Project Life Type of Baseline 6 Years 9 Years 16 Years Adjustments for .900 .903 .905 Fixed Factors OnlyL/ Adjustments for .943 .944 .946 All Factors/ a/ The benchmark rate of return is 15% for all three assumed project lives. b/ Baseline estimate taken from line (vi), Table 17. c/ Baseline estimate taken from line (viii), Table 17. -81- VII. CONCLUDING REMARKS Several earlier studies of the effects of agricultural extension on productivity have shown that investment in extension can produce a very high rate of return (Orivel, 1982). Most of these studies are based on the experience of developed countries. The present study reflects the experience of one of the more advanced agricultural areas of India with a particular form of extension, namely, the Training and Visit System. The results suggest that there is a high probability of an acceptable rate of return to intensified extension. The benefits appear to result from an improvement in the overall management of farm operations (in at least one major crop) rather than from the induced use of more (or new) inputs. More specifically, the study suggests that the increased availability of extension field personnel, and the access that these staff (and, through them, farmers) had to subject-matter specialists and hence to the research establishment, contributed to the ability of farmers to respond optimally to local farming problems. This ability apparently contributed much of the gain attributed to the change in the extension system in the study area, where most farmers had already adopted high yield varieties and were using high levels of nitrogen before the extension system was reformed. This finding may be compared with the results of several studies cited by Herdt and Capule (1983, p.31), whic'- showed that the quality of extension is a major factor in farmers' adoption of modern varieties and inputs. An issue of much operational importance concerns the implica- tions of these findings for areas which are not as advanced, where much can be gained by accelerating the adoption of high yield varieties (or -82- new crops) and fertilizers. It has been observed that profitable inno- vations spread quickly in the absence of intensive extension (e.g., the green revolution in northwest India). On the other hand, as demonstrated in the present study, even a minute gain (less than half of one percent) in one major crop, induced by intensifying extension, is sufficient to provide returns that justify the extra cost. In their review Herdt and Capule cite several studies that show that extension can accelerate the diffusion of innovations such as high yield varieties. Once the more straightforward components of a new technology are diffused the important element seems to be the availability of trained professionals (subject-matter specialists) whose knowledge is based on up-to-date research results and who are linked to farmers by abundant and well-managed field staff. Simple and profitable technology will spread fast naturally, but the absorption of more complicated practices, and adaptation to situations or circumstances which are not standard, requires that farmers have ready access to more specific and detailed advice. Where appropriately qualified personnel to fulfill this role are not available in sufficient numbers, the effect of extension is likely to be much smaller. -83- Annex I A NOTE ON COST ESTIMATION Two main sources of data on costs relating to T&V extension in Haryana were used: (i) Staff Appraisal Report of the Composite Agricultural Extension Project, Report No. 2171a-IN, World Bank, November 1978; (ii) Supervision Reports, World Bank, 1979 through 1984. The former is the source document for expected project costs, while the latter reports record, in an evolutionary way, actual expenditures. Using the supervision reports specific costs were derived as follows: (a) Civil Works: Based on information on disbursements made by project authorities to construction agencies; (b) Equipment and Vehicles: Information on the amount left undisbursed in this account was subtracted from projected annual expenditures on this account; (c) Staff: Bi-annual data on current vacancies for various categories of staff were available. Treating per-employee salaries (disaggregated by category) in the project appraisal report as averages, the actual numbers of occupied staff positions were multiplied by the appropriate -84- average and then summed to derive the total; incencive payments and other allowances related to salary were included; (d) Operating Costs were derived as a residual using data on aggregate actual annual expenditure from the supervision reports and (a), (b) and (c) above. Since detailed cost data were available only up to July 1983 (the project continued through December 1984) expenditures for 1983-84 were derived as residuals - i.e., total estimated project costs less actuals for the period 1979-83. This estimate is, however, expected to be fairly accurate, since project funds were not increased (on the contrary, some physical targets such as VEW housing were reduced) and the projected inter-category distribution of expenditures was maintained except for a negligible transfer from the Equipment and Vehicle account to the Civil Works account. The cost estimates that resulted from these calculations for Haryana as a whole are shown in Table A-2 whilst comparably arranged projected expenditures based on the Appraisal Report are shown in Table A-1. To estimate the share of total state expenditure attributable to Karnal District, the ratio of the number of operational VEW's in Karnal to the total for Haryana State was used as the basis for allocation. This ratio in March 1984 was 96/1194 = 0.0804. The costs thus derived for Karnal District, are shown in Table A-3. - 85 - Table A-1 INDIA -- COMPOSITE AGRICULTURAL EXTENSION PROJECT (Cr. 862-IN) - HARYANA COMPONENT PROJECT COST ESTIMATES AT APPRAISAL (Rs. Millions) COST ITEM 1979-80 1980-81 1981-82 1982-83 1983-84 TOTAL CaDital Civil Works 2.790 10.645 4.837 -- -- 18.272 Equipment and Vehicles 6.715 1.180 1.140 -- -- 9.035 Recurrent Staff 7.221 7.394 7.548 7.702 7.856 37.721 Operating, Training, Contingencies 7.794 9.655 10.302 9.823 10.928 48.502 TOTAL 24.520 28.874 23.827 17.525 18.784 113.530 Source: Staff Appraisal Report, Annex 5, p. 95. 86 Table A-2 INDIA -- COMPOSITE AGRICULTURAL EXTENSION PROJECT (Cr. 862-IN) -- RARYANA COMPONENT ACTUAL EXPENDITURES (Rs. Millions) COST ITEM 1979-80 1980-81 1981-82 1982-83 1983-84 TOTAL Capital Civil Works -- 12.0 4.6 1.9 -- 18.5 Equipment and Vehicles 6.715 1.180 1.140 -- -- 9.035 Recurrent Staff 5.049 6.010 5.885 7.224 7.277 31.445 Operating, Training, Contingencies 7.235 10.320 14.675 16.576 5.653 54.459 TOTAL 19.0 29.6 26.3 25.7 17.93 113.530 Source: Supervision Reports 1979-84. - 87 - Table A-3 INDIA -- COMPOSITE AGRICULTURAL EXTENSION PROJECT (Cr. 862-IN) -- HARYANA COMPONENT ACTUAL EXPENDITURES - KARNAL DISTRICT (Rs. Thousands) ITEM 1979-80 1980-81 1981-82 1982-83 1983-84 TOTAL Civil Works -- 964.8 369.8 152.8 -- 1487.4 Equipment and Vehicles 539.9 94,980 91.7 -- -- 726.5 Recurrent Staff 405.9 483.2 473.2 580.8 585.1 2528.2 Operating, Training, Contingencies 581.7 892.7 1179.9 1332.7 454.5 4378.5 TOTAL 1527.5 2372.6 2114.6 2006.3 1039.6 9120.6 -88- Annex II ESTIMATING PRODUCTIVITY DIFFERENTIALS IN THE PRODUCTION OF HYV RICE,1982 One of the principal differences between the two samples is that rice is grown by only half of the Kairana sample, about 46 farmers. Unlike the wheat-growing season the incidence of irrigation problems was more frequent in Karnal district than Kairana tehsil. These variables as well as other indicators of the production environment are explicitly considered in the statistical analysis reported below. The variables included in the estimates of the production function for high yielding varieties of rice are similar to those used for HYV wheat. The only exception is the omission of the timing of the first irrigation, a variable which is not applicable to rice cultivation. . Since many farmers reported irrigation problems in both districts, it was possible to formally test whether the production functions for farmers reporting and not reporting irrigation problems are identical up to a proportionate shift. The straightforward Chow test for differences between regressions assumes that the variances of the stochastic disturbance terms of the two sub-samples (reporting and not reporting irrigation problems) are identical. Not wishing to impose such a restriction we employed the generalized least-squares estimator with an estimated variance-covariance matrix. In effect the observations are first -89- weighted by the standard deviations of the stochastic disturbance terms based on the separate sub-samples. The weighted model is then estimated under the null hypothesis using ordinary least-squares. The result,3 of the Chow-like test indicate that at a level of significance of 0.02 there'is no statistically significant difference between the production functions for farmers experiencing irrigation problems and those not experiencing irrigation problems apart from a multiplicative scalar factor. For subsequent analyses the two sub-samples were pooled and assumed identical except for an irrigation problem dummy variable. Conditional on the identity of the two production functions up to a multiplicative scalar factor, the hypothesis of constant returns to scale is tested on the pooled sample of estimated-standard-deviation-weighted observations. At a level of significance of 0.02, the hypothesis of constant returns to scale cannot be rejected. The estimate of the production function subject to the constant-returns-to-scale hypothesis is reported in column 1 of Table Bl. We note from the results that irrigation problems cause a significant loss of output, as do pest problems and other problems. We also note, holding inputs constant, that contact farmers seem to have a substantially higher yield than other farmers. The estimated disembodied productivity differential of 7.39% (represented by the estimated coefficient of the Kairana tehsil dummy variable) is somewhat smaller than the estimate for wheat. It is not statistically significantly greater than zero at a level of significance of 0.05 although the sign is in the hypothesized direction. However, the estimated coefficient of labor seems a bit too low and the estimat2d coefficient of land a bit too high. -90- Table Bl: ECONOMETRIC RESULTS FOR HYV RICE (1) (2) (3)()() rI Output (Engineering In Output :n lit.cgen in ::or In Phasahate Production (Reduced (Reduced (Re: (Reduced Deoendent Variable Function) Form) For.: r:) _ _ ) S5i1 T9oe Duv Variablas S, .0491 .0164 .299 .362 3.006 (1.033) (.312) (2.575) (1.579) (3.297) si -.0451 -.0543 178 .082 2.784 (.871) (1.082) (2.563) (.221) (3.412) S, -.0660 -.1026 .1764 .3394 2.694 (1.329) (1.91) (2.336) (.989) (2.983) 55 -.1722 -.1882 .1163 .342 2.1 (3.940) (4.187) (Z..37) (1.282) (3.717) -.3548 -.047 .1214 -.292 1.573 (1.254) (1-39) (l.988) (.92) (2.149) Production Environcent Pest Problems -.1337 -.128 - -.0631 - (4-206) k.179) - (2.72) - Jther Problem -.IL1 -.1322 -.2683 -.0362 -.293 (1.656) (2.32) (2.775) (.743) (.236) Irrigation Problems -.0810 -.1W44 .016 -.037 -L.å6 (2.884) (3.761) (.j42) (1.855) (3.574) Irrizacion Variables In Number of krrigacions .0334 - (1203)---- Number of Tubewells --316 -.0147 .0531 -.008 .7 (1.333) (1.078) (2.57) (.788) (.328) Human Caciral Education .~028 .004 .00004 .012 .309 1.831) (.5)(.-09) (.320) (.335) Contact Farmer (Du=-y) .3690 .0606 -.0538 .0352 .879 (2.420) (2.149) (1.336) (1.693) (1.869) 0wnd Land .0049 -.0133 -.0204 - - (.237) (.436) (1.351) k.581) 3a4onl Efkl-t Du~y for Kalrana rehsil -.3739 -.3754 -.9333 .1188 -2.883 ( .415) (1.352) <4.303) (2.826) (3.9) rnouts Labor .3115 (.3822) - - Capital .3090 .3128 .024 .3051 .279 (.841) k1.179) (1.566) (.643) <1.515) Land .9248 .9872 .9756 .9949 .721 (10.303) '90.57) ,oZ.66) (124.02) (3.918) Nitrogen .j65 - (1.589) - - - - Phosphate .JL33 - "_.197)- -- Potash -.0165 - - (.755) - ZIna -.0071 -. (.540) - Supoly Constraint3 Credit Shortage .L190 .0861 .0337 .391 (.370) (L.13) (.877) (.105) Fertilizer Supply Problem - .3623 -.3403 .3208 -1.822 - (.692) (2.012) (.312) (.950) Pesticide Supply Problem - -.2172 - -.304 - - (1.492) - (.337) - Weedicide Supply Problem - -.1826 - .1103 - - (2.326) - (1.806) - Norma&lzed Prices Nitrogen Price - -.3341 -.1136 1.2131 9.198 -- (1.479) (.338) (1.266) (3.412) Labor Wage - -.0612 .3103 -.0251 -1.157 - (.773) (.91) (.431) (.875) CONSTANT 2.3933 2.754 3.521 3.738 -6.035 (-.848) (14.433) (12.857) (26,715) (1.86) R2 .9686 .9683 .938 .980 n.a. NLumber 2f ObservaCiona 305 305 305 305 305 CCKMMS Consrant Constant Constant Conatant Constant Returns Returna Returne Raturns Racurns to Scala ro Scala- to Scale to Scala to Scala T,hIp XnI-I«u -91- The reduced form estimate for rice supply output (again using the pooled weighted sample) is reported in column 2 of Table B1 and also includes the normalized prices for labor and nitrogen which were available for all observations. The estimated productivity differential, which-in the reduced form includes both the 'disembodied' and 'embodied' effects, is slightly higher (7.54%) than the estimated disembodied effect from the production function. We note that the estimated coefficients on the normalized input prices have the expected (negative) sign in the reduced form output supply equation. An alternative way of calculating the 'embodied' productivity differential in rice involves the estimation of input demand functions. As price effects are explicitly incorporated, the Kairana tehsil dummy variable would account for only the effects of productivity differentials on input use. These effects, weighted by the corresponding production elasticities, are added together to generate the total disembodied effect. The estimates reported in columns 3 and 4 of Table B1 show that while the productivity differential induced a higher use of fertilizers in Karnal, its effect on labor use was negative, possibly due to the adoption of labor saving practices (e.g., weedicides). The total 'embodied' effect is therefore: For Farmers Utilizing For Farmers Utilizing Nitrogen Only Nitrogen & Phosphate (%) (%) Labor -.14 -.14 Nitrogen 2.16 2.16 Phosphate -- 3.83 Total 2.02 5.85 -92- From these results the 'disembodied' productiv-cy differential may be estimated to be 5.52% (=7.54% - 2.02%) for farmers utilizing nitrogen only and 1.69% (=7.54% - 5.85%) for farmers utilizing nitrogen and phosphates. However, given that the estimated disembodied effect from the production function (7.39%) has an estimated standard deviation of about 5%, the results are quite compatible. Finally, we test the hypothesis that the incremental extension effect is not positively related to wealth. The results are reported in Table B2. At a level of significance of 0.01, we reject the hypotheses that the wealth effects in Karnal district and Kairana tehsil are the same in either the production function or the reduced form. How should these results be interpreted? Consider first a farm in Karnal. Because of the small Karnal wealth coefficient of -0.0076 output cannot vary much with respect to the quantity of land owned. In other words, the productivity level in Karnal does not seem to be much different for wealthy or poor farms. Now consider a farm in Kairana tehsil. Because of the large Kairana wealth coefficient of 0.1290, output must vary significantly with respect to the quantity of land owned. A farm owning twice as much land (other things, including all physical production inputs, being equal) will have a 9.4% higher yield. How should we interpret the large coefficient for the Kairana dummy variable? It does not mean that farms in Karnal are 40% more efficient than farms in Kairana. Instead, we take a representative level of wealth, say 10 acres owned, as a starting point. At 10 acres owned the differential between Karnal and Kairana is given in logarithmic terms by: - 93 - Table B2 ECONOMETRIC RESULTS FOR HYV RICE (WEALTH EFFECTS) Variable Production Function OutDut Supply Function Karnal (In Land Owned) -0.0076 -0.0165 (-0.354) (-0.733) Kairana (In Land Owned) 0.1290 0.1395 (2.654) (2.962) Kairana tehsil -0.4208 -0.4576 Dummy (-3.014) (-3.451) -94- -0.0076 In 10 - 0.1290 In 10 + 0.4208 = 0.1063 or approximately 11.2%. For a farm owning 20 acres the differential declines to approximately 1.2%. For a farm owning 5 acres the differential rises to 22.3%. The average differential over all wealth classes is probably around 10%. The dispersion of the differential is likely to follow the dispersion of wealth which may be quite high. Similar estimates can be obtained from the reduced form output supply function suggesting that the embodied productivity differential is probably quite small. These results suggest that the incremental effect of T&V extension in the production of rice tended to achieve a more equal distribution of the gains rather than an increase in the level of the gains. Another possibility is that extension-induced gains in productivity have not yet filtered through from those farmers with direct exposure to extension agents to other farmers. This is compatible with the estimate of a positive (and statistically significant) productivity gain of between 6% and 7% obtained for contact farmers (columns 1 and 2 of Table BI). -95- Annex III WHEAT PRICES Table Cl: WHEAT PRICES IN KARNAL DISTRICT Year Current Prices 1/ Constant 1979/80 Prices 2/ (Rupees per Ton) 1979-80 1,169.8 1,169.8 1980-81 1,300.0 1,113.0 1981-82 1,420.0 1,144.2 1982-83 1,510.0 1,176.9 1983-90 -I,176.9'/ 1/ Harvest price. 2/ Deflated by Consumer Price Index for agricultural laborers in1 Haryana. 3/ Assumed constant after 1982-83. -96- REFERENCES 1. D. Benar and M. Baxter, Training and Visit Extension, Washington, D.C.: The World Bank, 1984. 2. G. Feder and R. H. Slade, "A Comparative Analysis of Some Aspects of the Training and Visit System of Agricultural Extension in India", Journal of Development Studies, forthcoming, 1985. 3. B. Hoeper, "Selected Results of A.D.O. and V.E.W. Survey in Jind, Karnal and Mahendragarh District of Haryana, India", Working Note No. 1, Institute of Socio-Economics Agricultural Development, Berlin, October 1983. 4. R. W. Herdt and C. Capule, Adoption, Spread and Production Impact of Modern Rice Varieties in Asia, International Rice Research Institute, Los Banos, 1983. 5. D. T. Jamison and P. R. Mook, "Farmer Education and Farm Efficiency in Nepal", World Development, Vol. 12, No. 1, 1984. 6. M. Moore, "Institutional Development, The World Bank and India's New Agricultural Extension Programme", Journal of Development Studies, Vol. 20, No. 4, 1984. 7. F. Orivel, "The Impact of Agricultural Extension: A Review of the Literature", in Basic Education and Agricultural Extension", edited by H. Perraton, D. Jamison, J. Jenkins, F. Orivel and L. Wolff, World Bank Staff Working Paper No. 564, 1983. 8. J. Tobin, "Estimations of Relationships for Limited Dependent Variables", Econometrica Vol. 26, No. 1, 1958, pp. 24-36. 9. P. A. Yotopoulos and L. J. Lau (1979), Resource Use in Agriculture: Appli- cations of the Profit Function to Selected Countries, Food Research Institute Studies, Vol. 17, No. 1, pp. 1-120. 10. A. Zellner, J. Kmenta and J. Dreze, "Specification and Estimation of Cobb- Douglas Production Function Models", Econometrica, Vol. 34, No. 4, 1966, pp. 784-95. 11. Government of Haryana, Statistical Abstracts of Haryana. 12. Haryana Agricultural University, Package of Practices for Rabi Crops, various years. DISCU3STN ?k?ERS AGRResearch ni: Reott No.: ARU I Agricultural Zechanization: A Comparative Historical Perspective by Hans P. Binswanger, October 30, 1982. Report No.: ARU 2 The Acquisition of Information and the Adoption of New Technology by Gershon Feder and Roger Slade, September 1982. Report NTo.: ARU 3 Selecting Contact Farmers for Agricultural Extension: The Training and. Visit System in Haryana, India by Gershon Feder and Roger Slade, August 1982. Report No.: ARU 4 The Impact of Attitudes Toward Risk on Agricultural Decisions in Rural India. by Hans P. Binswanger, Dayanatha Jha, T. Balaramaiah and Donald A. Sillers May 1982. Report No.: ARU 5 Behavioral and Material Determinants of Production Relations in Agriculture by Hans P. B8nswanger and Mark R. Rosenzweig, June 1982, Revised July 22, 1985 Recor No.: ARU 6 The Demand for Food and Foodgrain Quality in India by Hans P. Sinswanger, Jai=e 3. Quizon and Gurushri Swamy, November 1982. Reoort No.: ARU 7 ?olicy Implications of Research on Energy Intake and Activity Levels with Reference to the Debate of the Energy Adequacy of Existing Diets in Development Countries by Shlomo Reuclinger, May 1983. Recort No.: ARU 8 More ffective Aid to the World's Poor and Hungry: A Fresh Look at United States Public Law 480, Title It Food Aid by Shlomo Reuclinger, June 1983. Reoort No.: _RL' 9 Faccor Gains and Losses in the Indian Semi-Arid Trooics: A Didactic Approach to Modeling the Agricultural Sector by Jaime 3. Quizon and Hans P. Binswanger, Sepce:.ber :983, Revised May 1984. Recor No.: ARU 10 the Distribution of Income in India's Northern Wheat Region by Jai=e B. Quizon, Hans P. Sinswanger and Devendra Gupta, Augus- 1933. Revised June L984. Renort No.: ARU 11 Population Density, Farming Intensity, Patterns of Labor-Use and hechanization by Prabhu L. Pingali and Hans P. 3iiswanger, September 1983. Recort No.: ARU 12 The Nutritional Impact of Food Aid,: Criceria for the Selection of Cost-Effective Foods by Shlomo Reutlinger and Judit Kacona-Apte, September 1983. nisc"ssion ?apers (Coant'd.) Remor. No.: ARU 13 Project Food Aid and Equitable Growth: lacome-Transfeer Efficiency Firsc! by Shlomo Reutlinger, August 1983. Report No.: ARU 14 Nuctricional Zmpact of Agricultural Projects: A Conceptual Framework for Modifying the Design and tmplementation of Projects by Shlomo Reutlinger, August 2, 1983. Report No.: ARU 15 Pacterns of Agrtcultural Protection by Hans P. Siaswanger and Pasquale L. Scandizzo, November L5, 1983. ,Report No.: ARU 16 Factor Costs, Lacome and Supply Shares in rndian Agriculture by Ranjan Pal and Jai=e Quizon, December 1983. Report No.: ARU 17 3ehavioral ana Macerial Decarminanes of Production Relations in Land Abundanc 7rpical Agriculture by Hans ?. 3inswanger and John Mctacire, January 1984. Repor No.: ARU 18 The Relacion Between Farm Size and Farm ?roduc:ivi:y: The Role of Faaily 'abor, Supervision and Credit Constraiats* by Gershon Feder, December 1983. Report No.: ARU 19 A Comparacive Analysis f 3ome Aspects of che T:a.iag and 7Lsic System or Agricul:ural Excension La ladia by Gershon Feder and Roger Slade, February 1984. Report No.: ARU 20 Distributional Consequences of Alternative Food Policies in India by Hans P. Binswanger and Jaime B. Quizon, August 31, 1984. Renort No.: ARU 21 Income Distribution in India: The Impact of Policies and Growth in the Agriculurai1 Sector, by Jaime B: Quizon and Hans P. Binswanger, November 1984, Report No.: ARU 22 Population Density and Agricultural Intensification: A Study of the Evolution of Technologies in Tropical Agriculture, by Prabhu L. Pingali and Hans P. Binswanger, October.17, 1984. Report No.: ARU 23 Tae Evolution of Farmining Systems and Agricultural Technology in Sub-Saharan Africa, by Hans P. Binswanger and Prabhu L. Pingali, October 1984. Report No.: ARU 24 Population Density and Farming Systems - The Changing Locus of Innovations and Technical CLange, by Prabhu L. Pingali and Hans P. Binswanger, October 1984. Report No.: ARUY 25 The Training and Visit Extension System: An Analysis of Operations and Effects, by G. Feder, R.H. Slade and A.K. Sundaram, November 1984. - 3 - Report No.: ARU 26 The Role of Public Policy in the Diffusion of New Agricultural Technology, by Gershon Feder and Roger Slade, October 1984. Fertilizer Subsidies: A Review of Policy Issues with Special Emphasis on Western Africa, by Haim Shalit and Hans P. Binswanger, November 1984. Revised August 1985. Report No.: ARU 28 From Land-Abundance to Land-Scarcity: The Effects of Population Growth on Production Relations in Agrarian Economies, by Mark R. Rosenzweig, Hans P. Binawanger, and John Mcntire, November 1984. Report No.: ARU 29 The Impact of Rural Electrification and Infrastr-icture on Agricultural Changes in India, 1966-1980, by Douglas F. Barnes and Hans P. Binswanger, December 1984. Report No.: ARU 30 Public Tractor Hire and Equipment Hire Schemes in Developing Countries (with Special Emphasis on Africa). A study prepared by the Overseas Division, National Institute of Agricultural Engineering (OD/NIAE), by P.J. Seager and R.S. Fieldson, November 1984. Report No.: ARU 31 Evaluating Research System Performance and Targeting Research in Land Abundant Areas of Sub-Saharan Africa, by Hans P. Binswanger, January 1985. Report No.: ARU 32 On the Provision of Extension Services in Third World Agriculture, by Alastair J. Fischer (Consultant), January 1985. Re0ort No.: ARU 33 An Economic Appraisal of Withdrawing Fertilizer Subsidies in India, by Jaime B. Quizon, April 1985. Revised August 1985. Report No.: ARU 34 The Impact of Agricultural Extension: A Case Study of the Training and Visit Method (T&V) in Haryana, India, Gershon Feder, Lawrence J. Lau and Roger H. Slade, March 1985. Report No.: ARU 35 Managing Water Managers: Deterring Expropriation, or, Equity as a Control Mechanism, by Robert Wade, April 1985. Report No.: ARU 36 Common Property Resource Management in South Indian Villages, by Robert Wade, April 1985. Report No.: ARU 37 On the Sociology of Irrigation: How de we Know the Truth about Canal Performance? by Robert Wade, May 1985. Report No.: ARU 38 Some Organizations concerned with Animal Traction Research and Development in Sub- Saharan Africa, by Paul Starkey, April 1985. 4- Report No.: ARU 39 The Economic Consequences of an Open Trade Policy for Rice in India, by Jaime Quizon and James Barbieri, June 1985. Report No.: ARU 40 Agricultural Mechanization and the Evolution of Farming Systems in Sub-Saharan Africa, by Prabhu L. Pingali, Yves Bigot and Hans P. Binswanger, May 1, 1985. Report No.: ARU 41 Eastasian Financial Systems as a Challenge to Economics: The Advantages of 'Rigidity', with particular reference to Taiwan, by Robert Wade, June 1985. Peport No.: ARU 42 Education, Exnerience and Imperfect Processing of Information in the Adoption of Innovat-Lons, by Alastair J. Fischer, June 1985. Report No.: ARU 43 A Review of the Literature on Land Tenure Systems in Sub-Saharan Africa, by Raymond Noronha, July 19, 1985. Report No.: ARU 44 Policy Options for Food Security, by Shlomo Reutlinger, July 1985. Report No.: ARU 45 Credit Markets in Rural South India: Theoretical issues and Empirical Analysis, by Hans Binswanger, Balaramaiah, V. Bashkar Rao, 1.J. Bhende and K.V. Kashirsagar, July 1985. Report No.: ARU 46 The Impact of Agricultural Extension: The Training and Visit System in India, by Gershon Feder and Roger Slade, June 1985. Report No.: ARU 47 Methodological Issues in the Evaluation of Extension Impact, by Gershon Feder and Roger H. Slade, July 1985. Report No.: ARU 48 Estimation of Aggregate Agricultural Supply Response, by Hans Binswanger, Yair Mundlak, Maw-Cheng Yang and Alan Bowers, August 1985.
Группа Всемирного банка · Working Paper (Numbered Series)
The impact of agricultural extension : a case study of the training and visit system (T & V) in Haryana, India
Открыть оригинал документа
Полный текст размещён на сайте публикующей организации. lawenc.com индексирует метаданные и ведёт на официальный источник.
Вернуться к постатейному просмотруПолный текст
Основные сведения
Организация
Группа Всемирного банка
Тип документа
Working Paper (Numbered Series)
Страна
Индия
Источник
Всемирный банк