Dairy development and milk cooperatives : the effects of a dairy project in India

WDP-15 15 E zWorld Bank Discussion Papers Dairy Development and Milk Cooperatives The Effects of a Dairy Project n India George Mergos Roger Slade FlLE COPY   115 U World Bank Discussion Papers Dairy Development and Milk Cooperatives The Effects of a Dairy Project n India George Mergos Roger Slade The World Bank Washington, D.C.  The World Bank 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First printingJuly 1987 Discussion Papers are not formal publications of the World Bank. They present preliminary and unpolished results of country analysis or research that is circulated to encourage discussion and comment; citation and the use of such a paper should take account of its provisional character. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. Any maps that accompany the text have been prepared solely for the convenience of readers; the designations and presentation of material in them do not imply the expression of any opinion whatsoever on the part of the World Bank, its affiliates, or its Board or member countries concerning the legal status of any country, territory, city, or area or of the authorities thereof or concerning the delimitation of its boundaries or its national affiliation. Because of the informality and to present the results of research with the least possible delay, the typescript has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. The publication is supplied at a token charge to defray part of the cost of manufacture and distribution. The most recent World Bank publications are described in the catalog New Publications, a new edition of which is issued in the spring and fall of each year. The complete backlist of publications is shown in the annual Index of Publications, which contains an alphabetical title list and indexes of subjects, authors, and countries and regions; it is of value principally to libraries and institutional purchasers. The continuing research program is described in The World Bank Research Program: Abstracts of Current Studies, which is issued annually. The latest edition of each of these is available free of charge from Publications Sales Unit, Department F, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A., or from Publications, The World Bank, 66, avenue d'1ena, 75116 Paris, rance. George Mergos is a lecturer at the University of Athens, a research associate at the Centre of Planning and Economic Research (Athens), and a consultant to the World Bank. Roger Slade is a senior economist in the Projects Department of the Bank's South Asia Regional Office. Library of Congress Cataloging-in-Publication Data Mergos, George, 1948- Dairy development and milk cooperatives. (World Bank discussion papers ; 15) Bibliography: p. 1. Milk trade--India--Madhya Pradesh. 2. Dairying, Cooperative--India--Madhya Pradesh. I. Slade, Roger, 1941- . II. Title. III. Series. HD9282.I53M286 1987 334'.68371'09543 87-15976 ISBN 0-8213-0932-3  ACKNOWLEDGEMENTS This study represents part of the results of a collaborative research project undertaken and financed by The World Bank and The Interna- tional Food Policy Research Institute. It could not, however, have been undertaken without the agreement of the Government of India, the cooperation of the Madhya Pradesh Dairy Development Corporation (MPDDC) and the collaboration and assistance of the Institute of Rural Management at Anand (IRMA) in India. We gratefully acknowledge the support provided by these bodies through, in particular, Mr. K.J.S. Bhatia, then Managing Director of MPDDC and Dr. R. N. Haldipur, then Director of IRMA. We also wish to thank Dr. K. Singh and Dr. J. Archarya of IRMA for sustained collaboration from the inception to the completion of this study. Thanks too are due to Dr. B. Thanka of MPDDQ whose support materially aided our field work. During the analysis and writing we received valuable comments and advice from many people. We are particularly grateful for the perceptive and detailed comments provided by Richard Just, Per Pinstrup-Andersen, Harold Alderman, Alan Gelb and Mogens Jul. Finally, we record our thanks to Devi Katikineni for extensive assistance during the early stages of the computer work and to Terry Gean who patiently typed and retyped the manuscript. -i~ii -   TABLE OF CONTENTS Page Summary xilt Chapter I. INTRODUCTION ......... ..................................... 1 The Context of the Study 1 Overall Objectives 2 The Evaluation Framework 3 II. BACKGROUND ...... ......... 7 Operation Flood 7 The Madhya Pradesh Dairy Development Project 8 The Sample Survey 12 DCS Activities and Farmer Attitudes in the Sampled Villages 16 Differences Among Project and Control Households 21 An Overview of the Methodology 23 III. ENHANCEMENT OF MILK PRODUCTION INCENTIVES ................ 28 Market Efficiency 27 Provision of Technology Enhancing Inputs 34 Concluding Remarks 37 IV. THE ADOPTION OF CROSS-BRED TECHNOLOCY .............. ...... 40 Adoption of Cross-bred Technology in India; Issues and Opinions 40 Factors Affecting Holdings of Cross-bred Milk Animals: Evidence from the Sampled Villages 43  -vi- An Empirical Model of Adoption of Cross-bred Technology 49 Concluding Remarks 56 V. THE STRUCTURE OF MILK SUPPLY RESPONSE .................... 60 Change in Milk Yields 60 The Size and Composition of the Milk Animal Population 66 Concluding Remarks 69 VI. MILK PRODUCTION AND PRODUCTIVITY ......................... 72 The Effect of the Project on Productivity 71 The Effect of DCS Membership 76 Effects of Village Specific Factors 76 Sources of Change in Milk Output 78 The Effect of the Project on Milk Production 80 Effects on Input Demand 82 Concluding Remarks 85 VII. INDIRECT EFFECTS ON CROP PRODUCTION ...................... 90 VIII. DISTRIBUTION OF BENEFITS TO PRODUCERS .................... 94 Participation of the Poor in Dairying 94 Benefits and Membership of the DCS Among Milk Producers 99 Determinants of DCS Membership and Use: An Empirical Analysis 103 Concluding Remarks 108  -vii- IX. IMPACT ON RURAL CONSUMERS . ............................. 110 Who are the Poor? ill The Framework of Analysis 112 Project Impact on Food Expenditure 113 Project Impact on Milk Consumption 115 Impact on Calorie Consumption 123 Concluding Remarks 128 X. CONCLUSIONS AND POLICY ISSUES . ......................... 131 Main Conclusions 131 Output Effects 131 Consumption and Nutrition Effects 132 Distribution of Benefits to Milk Producers 133 Public Policy in the Dairy Sector 133 Cooperative Policy in the Dairy Sector 134 References ............. ..... ........ 136 Note. Footnotes appear at the end of each chapter.  -viii- List of Tables Table Page Chapter 2 2.1 Project Milksheds in Madhya Pradesh ......................... 9 2.2 Project Dairy Development Activities ........................ 11 2.3 Village and Household Characteristics Without The Project ....... ..13 2.4 Probabilities of Sample Village Selection: A Logit Analysis ............................................ 14 2.5 DCS Performance in the Project and the Sample ............... 15 2.6 Characteristics of Sampled Households ....................... 15 2.7 Farmers' Opinions of the DCS and Their Reasons For Joining.. 18 2.8 Changes After DCS Establishment as Perceived By Farmers With Milk Animals in DCS Villages ........................... 19 2.9 Membership of the DCS and Use of Milk Markets (percent) ... 19 2.10 Need for and Source of Veterinary Services . . 20 2.11 Characteristics of Households With Milk Animals ............. 21 Chapter 3 3.1 A Typology of Changes in the Structure of the Milk Market When a DCS is Introduced .................................... 29 3.2 Project Impact on Milk Prices ........................ ....... 32 3.3 Need for and Ability To Obtain Veterinary Services (Logit Analysis). ................. 35  -ix- 3.4 Opinion About and Use of Artificial Insemination Among Milk Animal Owners (Logit Analysis) .. 36 Chapter 4 4.1 Holders and Non-Holders of Cross-Bred Cattle Classified By Size of Land Holding . ................................ . 44 4.2 Holders of Cross-Bred Cattle Classified By Milk Animal Ownership .............. ....... ........ 45 4.3 Holders and Non-Holders of Cross-Bred Cattle Classified By Training....... 45 4.4 Holders of Cross-Bred Cattle Classified By Caste . . 46 4.5 Cross-Bred Cattle Holders in DCS Villages According To The Age of the DCS....... 46 4.6 Farmers' Opinion of Artificial Insemination .47 4.7 Primary Reasons Given For Positive Opinion of A.I. 48 4.8 Primary Reasons Given For Negative Opinion of A.I. .48 4.9 Mean Values and Variance of Variables .51 4.10 Logit and Probit Analyses of Factors Affecting Adoption of Cross-Bred Technology.... 52 4.11 Participation of the Household in the Feed Market As a Net Purhsr r......, h.............e..... 55 Chapter 5 5.1 Definitions of Variables In Yield Functions ................ 62 5.2 Estimated Yield Functions For Local Cows ................... 64 5.3 Estimated Yield Functions For Buffaloes.................... 65  5.4 Response Functions For The Size and Composition of the Milk Animal Population ...................... .......... ......... .. . 67 5.5 Variable Definitions For Lactation Analysis ................. 68 5.6 Response Function For Share of Lactating Milk Animals ....... 69 Chapter 6 6.1 Milk Output, Numbers of Milk Animals and Input Use for Milk Animal Owners .......................................... 72 6.2 Definitions of Variables Used In The Econometric Analysis ... 74 6.3 Estimated Reduced Form Milk Output Supply Functions ......... 75 6.4 Availability of Public and Private Facilities In Sampled Villages ............................................ 77 6.5 Estimated Milk Production Function .......................... 81 6.6 Estimated Input Use Functions For Fodder and Concentrates ... 84 Chapter 7 7.1 Cropping Pattern In Sampled Villages . . 90 7.2 Variable Definitionsfs.. . 91 7.3 Use of Modern Inputs in Crop Farming (Logit Analysis) .92 Chapter 8 8.1 Ownership of Milk Animals in Study Area-................... 95 8.2 Perceived Constraints of Non-Milk Animal Owning Households in the Survey Area on Acquiring Milk Animals . .. .. 96  8.3 Project Impact on Farmers' Decision To Keep Milk Animals (Logit Analysis) ........... .................................. 98 8.4 Membership of the Dairy Cooperative Society and Sales of Milk ..100 8.5 Average Quantity And Price of Milk Sold. ............. 102 8.6 Characteristics of Households That Are Members of But Do Not Sell To The DCS .102 8.7 Membership of the DCSCS....... 106 8.8 Use of the DCS Milk Market ..107 Chapter 9 9.1 Consumption Characteristics of Sampled Household By Expenditure Class .112 9.2 Estimated Food Expenditure Functions .114 9.3 Average Per Caput Consumption of Milk and Milk Products 116 9.4 Households Who Produce And Consume Milk .117 9.5 Estimated Consumption Functions For Milk .118 9.6 Estimated Consumption Functions For Milk By Expenditure Class ..............122 9.7 Average Per Caput Consumption of Calories and Proteins..... 123 9.8 Distribution of Calorie and Protein Intake Per Caput Per Day............................. 124 9.9 Estimated Calorie Consumption Functions .126 9.10 Estimated Protein Consumption Function . ........... 127  -xii- List of Figures Figure Page 1.1 A Complete Evaluation Framework . ......... 3 2.1 Measurement of Project Impact .............................. 24 3.1 Effect of a Reduction in Marketing Cost on Supply and Demand for Milkilk............... * .................... 30 4.1 Business and Financial Risk of an Investment in Cross-Bred Milk Animals With Constant Risk Aversion And a Given Level of Debt .42 9.1 Milk Consumption at Different Expenditure and Milk Production Levels . 120  -xiii- SUMMARY Since the early 1970's the Government of India has pursued dairy development on a large scale through the execution of a series of discrete investment projects collectively known as Operation Flood. The aggregate investment programme has depended on large scale external aid provided by the Food Aid Program of the European Economic Community and agencies such as the World Bank. The projects comprising Operation Flood are intended to induce increased milk production in rural areas by creating an efficient milk collection and marketing organization, by delivering appropriate technical knowledge and services to dairy farmers, including artificial insemination services. The latter are designed to make it possible for farmers to increase the quality and productivity of their herds through cross-breeding. The entire programme is based on individual village dairy cooperative societies which are grouped into unions and federations of cooperatives. These larger groups own and operate the milk processing plant to which farmers through the village societies supply milk. Operation Flood is a controversial subject in India and elsewhere and opinion about its effects is sharply divided. The resulting debate is not, however, well founded on careful empirical research. Accordingly, farm household data from a group of villages in an Operation Flood project in Madhya Pradesh are analyzed in order to test a number of hypotheses about the effects of dairy development projects. The data were collected through sample surveys conducted in 1983 in 12 villages some with Dairy Cooperative Societies and some without. Apart from this difference the villages were agriculturally, culturally and ethnically similar. Using a variety of econometric techniques the analysis establishes, in the area of study, that the project improved the milk marketing system thereby increasing competition with the result that the average price of milk to producers rose about 8 percent. This, together with improvements in the provision of animal husbandry and veterinary services and an induced increase in the use of variable inputs raised milk production by about 17 percent over five years. These changes were, as far as can be ascertained, neutral with respect to crop and arable production. Nor did they depend on the adoption of cross-bred technology. Although the project increased the adoption of cross-bred cows this effect was so small as to have a quantita- tively insignificant effect on milk production.  -xiv- Further analysis shows that the resulting increase in incomes for milk producers led to increased food consumption. In particular, and after accounting for both price and income effects, milk producers increasd their consumption of milk slightly, about 2 percent on average, although the increase was about 8 percent amongst the poorest producers. There were similar consequences with respect to the consumption of calories and protein. Poor households who are not milk producers were found not to consume milk (a relatively expensive item) and hence towards them the project was neutral. The project did not, however, increase the number of milk animal owning households and the distribution of benefits in the rural areas was skewed towards those with the largest dairy herds and farms. However, the project did increase rural incomes on average and may therefore have had positive effects on the national distribution of income. Finally, the study suggests that future cooperative and dairy policy in areas such as that studied will benefit in the short-run from further intensification of animal husbandry and veterinary services to farmers to allow them to maximize output from their holdings of traditional cows. Only then should emphasis shift to cross-bred technology. Addition- ally, production might also be enhanced by increasing the ability of the dairy cooperatives to set milk prices. To the extent that this would allow differentials in marketing costs between cooperatives to be reflected in prices than there would probably be a gain in overall economic efficiency.  Dairy Development and Milk Cooperatives The Effects of a Dairy Project in India   I. INTRODUCTION The Context of the Study Dairy development is an important component of strategies to expand agricultural output in many developing countries. The use of freely available natural resources, the integration of crop and livestock systems, the increased use of family labor, improvements in the welfare of the poor and improved nutrition, especially of vulnerable groups (e.g. women and children) are often cited as reasons why dairy development should be encouraged. An important role for dairy development is suggested by the (rela- tively) high income elasticities of demand commonly observed for dairy products. 1/ Such high elasticities suggest a ready market, provided it can be reached economically, and opportunities to exploit linkages to other sectors of the economy. In the absence of increases in domestic dairy production, developing countries face increased imports or rapidly escalating prices as the demand for milk and milk products grows as fast as, or faster than, incomes. Moreover, price elasticities of demand are generally higher for dairy products than for food grains, and are often greater than unity. Hence, total expenditure on milk will increase if price falls. Declining prices may result from economies of scale, technological innovation in milk production, a reduc- tion in the costs of marketing and processing or some combination of these. The marketing of milk is more problematic than most other agricultural products. Milk is especially perishable and to be transported from the rural producer to the urban consumer requires considerable organization and capital investment in transport facilities and chilling and processing plant. For these reasons dairying is frequently promoted through collective marketing arrangements in developed and developing countries alike. One such endeavor is the Anand Milk Producers Union, Ltd. (AMUL) in Gujarat, India. The success of this producers' cooperative, founded in 1946, encouraged the Government of India to promote dairy development by replicating AMUL on a wide scale under two ambitious programs, Operation Flood I (1970-1981) and Operation Flood II (1979-1986). Interest in these programs, which have received large amounts of food aid, transcends the borders of India because a similar strategy has been advocated for other Southeast Asian and for African countries [Lipton, 1985]. Because of its importance in India's development strategy Operation Flood is vigorously debated not only within India but also internationally [see, e.g., George, 1986, Apte, 1986 and Terhal and Doornbos, 1983]. Operation Flood has been cited as an example of the successful use of commodity aid. But it has been criticized as having questionable production objectives, benefiting the already better off in the rural areas and having adverse nutritional conse- quences for households in areas that come under the project. The magnitude of its impact on milk production is at the center of the debate. The World Bank helped to finance dairy development projects under Operation Flood I in three states, Madhya Pradesh, Karnataka, and Rajasthan and  -2- has also helped to fund the National Dairy Development Project (Operation Flood II). Since 1974 the Bank has loaned India $366.1 million for dairy develop- ment. 2/ Terminal project completion reports on the Madhya Pradesh project [Singh and Pethiya, 1983] and the Karnataka project [Singh, Srinivasan and Raju, 19851 find these investments to have been highly successful. Since its establishment the Bank has lent $3.3 billion worldwide for livestock projects [World Bank, 1985a, Table 5.5]. Yet, Baum and Tolbert [1985, p. 113] in reviewing the performance of Bank assisted livestock projects in the sector at large (in India and elsewhere) characterize overall perfor- mance as dismal. Hence, further carefully collected and rigorously analyzed empirical evidence may be important in the design of future projects in the livestock sector. This paper provides statistical evidence of the effects of dairy development at the farm level in one rural area in India where such a project was introduced. Specifically, the paper reports the results of a study undertaken in 1983 in the state of Madhya Pradesh. The analysis is based on data drawn from nine project and three control villages with similar agro-climatic, economic and cultural conditions. The data were collected from farmers in two rounds (July-August and November-December 1983) through farm sample surveys designed and supervised jointly by the authors and collaborators at the Institute of Rural Management, Anand. A companion study was undertaken at about the same time in Karnataka and is reported in Alderman [19861. Overall Objectives The overall objectives of this study are to estimate (through a case study) the production and distributional effects of the Indian dairy projects at the household level and to expand knowledge about procedures for undertaking ongoing and ex-post evaluation of dairy development projects. 3/ Specifically the objectives of the study are: (a) to estimate the magnitudes of the direct economic effects produced by the projects; (b) to estimate how project benefits are distributed among population groups paying particular attention to the projects' impact on the poor; (c) to estimate how the projects affect food consumption and nutrition among the poor and energy-protein deficient population groups; and (d) to identify the key relationships and parameters determining the projects' effects on production, food consumption and nutrition and to empirically estimate the magnitudes of such parameters. 4/  -3- The Evaluation Framework Figure 1.1 demonstrates that the estimation of the effects (impact) of a dairy development (or any other) project requires a comparative approach along two dimensions. Rarely is this possible for either practical or administrative reasons. Sometimes, however, it is possible to work along only one dimension, most commonly, studies of the situation in the area where the project is undertaken both before and after (as well as during) implementation. Even this somewhat inadequate approach is often impractical, particularly in areas (projects) where monitoring and evaluation efforts were not made prior to the project and subsequent implementation covered the entire country or state or is otherwise sufficiently extensive to prevent the identification of com- parable areas that remain uninfluenced by the project. In these situations evaluation must remain, as it were, within a single cell of the matrix - the situation in the area of the project during and after implementation. Such a restricted form of evaluation, common in many project situations, is usually unable to yield definite answers about the effects of the project. It is, however, sometimes possible to achieve adequate results in this situation. This requires that the project be executed in such a way that after it has been implemented some areas remain unaffected and that it can be shown that these areas remain very similar, in all material respects, to the area affected by the project before it was implemented. This study was obliged to use this restricted framework. Figure 1.1: A COMPLETE EVALUATION FRAMEWORK : Before Project : After Project Without : The situation before the time the : The situation after the Project : project is introduced in an area : project has been introduced identical to that where the : in an area identical to : project is planned. : that where the project was undertaken. With : The situation in the area where : The situation in the area of Project : the project is planned before the : the project during and after : project is undertaken. : it has been implemented. Source: Feder and Slade [1986] p. 257. Within this framework the chain of cause and effect is assumed to follow closely that developed by Pinstrup-Andersen [1981] for the assessment of the nutritional effects of projects. Hence, the introduction of the project is assumed to affect nutrition at the household level through changes in food output, food prices and incomes. These changes interact to produce further changes in, the ability of the household to obtain food, the household's food preferences and the intra-household distribution of food. 5/ Within this general framework the study focuses on the testing of a set of hypotheses and  -4- the quantitative estimation of related parameters. These hypotheses are presented below in their positive form and are grouped into three categories. (i) Output Effects The project aims to significantly increase aggregate milk production by improving the output of existing milk animals and by introducing improved technology in the form of cross-bred cattle and increased fodder crop produc- tion. This suggests the following specific hypotheses: (a) milk yields have increased significantly as a result of the project; (b) the size and/or the composition of the dairy herd has increased or changed significantly as a result of the project; (c) the number of households owning cross-bred cattle in the project area has increased through purchases or through breeding as a result of project activities; (d) the primary source of expansion of milk production is increased yields rather than growth of the herd; (e) the project caused a significant increase in the acreage devoted to the production of fodder crops among project participants; and (f) the purchase and use of modern inputs for foodgrain production increased as a result of income changes induced by the project. (ii) Consumption and Nutrition Effects The project was not intended to adversely affect the consumption of milk or other products by rural households. However, views vary on whether the effects have been positive, negative or benign. This suggests the following hypotheses: (a) milk consumption by milk producing households increased due to the project. (The basis of this hypothesis is that the effect on milk consumption of any increase in the price of milk was more than offset by the increase in income); (b) milk consumption by non-milk producing households increased as a result of the project; (c) a large share of incremental household income generated by the project is spent on the purchase of additional food; and (d) the increase in the consumption of purchased and home grown calories and protein significantly exceeded the reduction in  -5- calorie and protein consumption resulting from the sale of milk and dairy products hitherto consumed. (This hypothesis tests whether the income effect on food consumption overrides the loss of calories resulting from sales of milk and milk products by milk producing households). (iii) The Distribution of Benefits to Milk Producers These hypotheses are related not only to the aggregate project impact on incomes but also to the way in which income changes are distributed among milk producers. Additionally, because many households had no cattle before the introduction of the project it is necessary to establish whether the proportion of households owning milk animals increased. (a) The proportion of households with milk animals increased as a result of the project; (b) Poor households started a dairy enterprise as a result of the project; and (c) Project benefits are skewed, being proportional to land and animal ownership. In this study all of the hypotheses noted above are tested using appropriate statistical techniques and farm-household survey data from the state of Madhya Pradesh. We begin in Chapter II by briefly describing Operation Flood and the Madhya Pradesh Dairy Development Project. This is followed by details of the study area and the household sample. Attention is also given to the extent that the control villages in the sample can be considered representative of the without project condition and whether the villages having dairy cooperative societies (DCS) are typical of the project at large. The chapter concludes with a brief statistical description of the DCS and control villages and an overview of the methodology used in the chapters that follow. In Chapter III changes induced by the project in the incentive struc- ture for milk production are examined. The project sought to increase the number of cross-bred milk animals in the dairy herd and this is analyzed in Chapter IV. Chapter V provides an examination of the structure of the milk supply response to the project. This is followed in Chapter VI by an estima- tion of the project impact on aggregate milk production. Possible indirect effects of the project on crop production are examined in Chapter VII. Sub- sequently, in Chapters VIII and IX respectively the distribution of benefits to producers and the project's impact on rural consumers are analyzed. Finally, in Chapter X the overall results of the study are assembled and summarized and some policy implications derived.  -6- FOOTNOTES (Chapter I) 1! For estimates of Indian expenditure and price elasticities for milk and milk products see Radhakrishna and Murty [1980]. 2/ This figure of $366.1 million is made up of $224.2 million for four free standing dairy development projects and $141.9 million - the estimated Bank share in the livestock components of a further nine projects. The total cost of these projects including the livestock components of the other projects was $833.9 million [World Bank, 1985b, Vol. II, Annex 5, Table 2, p.6]. 3/ The emphasis on project assessment at the household level stems from a concern about the welfare of the individuals who are affected by the project and, hence, a belief that the true test of the project is the examination of the benefits which accrue to those individuals (see also United Nations (1978) p. iii). 4/ The objectives as stated in the Research Proposal included intra-family food distribution in (c). This was not studied because the necessary data could not be collected. 5/ See footnote 4/ above.  -7- II. BACKGROUND Overall dairy policy in India is founded in autarky. The principle instrument is a restrictive trade policy (large scale commercial milk imports are prohibited and capital imports for dairy development are given favorable excise treatment) which seeks to create conditions favorable to the expansion of domestic production. Capital for development derives substantially from domestic sources augmented by large scale food aid and concessional loans from IDA, the World Bank's soft loan affiliate. During the late 1970's three main alternative strategies for increas- ing domestic milk output were considered. The first of these would have stabilized the size of the national herd at its level in 1980 and increased feed inputs per milk animal. This was rejected on the grounds that it would have required farmers to feed their animals at levels which were not economi- cally optimal. The second strategy was the inverse and would have increased the size of the national herd but maintained feed inputs per animal at their level in 1980. This too was rejected because it would have greatly over- strained estimated feed supplies. The third and accepted strategy aims to maintain a national herd of constant size but to replace a large proportion by genetically superior cross-bred milk animals and at the same time encourage efficient feeding levels for both groups. This strategy of tech- nological change underlies all of the projects and schemes that, over the past decade and a half, have been launched under the umbrella of Operation Flood (OF). Operation Flood Operation Flood is the popular name given to a Government of India (GOI) scheme operating in almost all states of the Union and embracing a large number of discrete dairy development projects. The projects aim to enhance milk production in rural areas and establish institutions for the collection and processing of milk. Additional objectives are to ensure a fair return to the farmer for his milk and to increase the quantity and quality of milk supplies to consumers by providing milk, processed under hygienic conditions, to urban areas [GOI, 1971]. The mechanisms to attain these objectives consist of: (a) building an efficient system of collecting fresh milk in rural areas and marketing it in urban areas as fluid milk or milk products; (b) creating a system to efficiently deliver technical knowledge and services to farmers so that their milk production capacity is enhanced; and (c) developing a professionally managed cooperative organiza- tion as a means of implementation [Gupta, 1983]. The institutional forerunner of OF was the Anand milk cooperative which was started in 1946 in the Kaira district of Gujarat. The origins and history of the dairy cooperative movement in India are described at length in Singh and Kelly [1981] and summarized in Alderman, Mergos and Slade [1987]. Indian dairy development policy operating through OF aims to replicate the Anand pattern nationwide.  -8- In 1970 GOI sought World Food Program commodity assistace for OF. The National Dairy Development Board (NDDB) was initially the implementing authority but later the Indian Dairy Corporation (IDC), a public sector company, was set up to implement OF. The first phase of Operation Flood (OF I) created about 10,000 village level cooperatives involving 1.3 million producers during the period 1970 to 1981. When it was closed about Rs 4.85 billion had been spent. Of this about Rs 1.2 billion was spent by World Bank assisted projects in the States of Karnataka, Madhya Pradesh and Rajasthan. These projects created about 5,000 village level cooperatives with about 400,000 member producers. OF II which overlapped OF I involved an investment of about Rs 4.9 billion during the years between 1978 and 1985 and was financed with a World Bank loan of US$150 million (about Rs 1.65 billion) and commodity assistance from the European Economic Community. The objective of OF II was to cover 10 million producers. By 1984, 28,600 village level cooperatives had been organized with 3.1 million producer members. About another three million producers are thought to have been affected indirectly. The Madhya Pradesh Dairy Development Project The project covers about 7 million hectares containing about 10 million people (about one fifth of the population of the state of Madhya Pradesh). It is situated in the Western part of the state on the Malwa Plateau which is part of the semi-arid black soil belt of West-Central India. Agriculturally the project area comprises the most important part of the state. The project was organized on the basis of a series of milksheds (milk producing areas) surrounding the major urban centers of Bhopal, Indore and Ujjain (see Map I). These three milksheds spanned the administrative dis- tricts of Bhopal, Hoshangabad, Shajapur, Rajgarh, Raisen and Sehore (the Bhopal Milkshed); Indore, Dewas and Dhar (the Indore Milkshed); and Ujjain and Ratlam (the Ujjain Milkshed). Descriptive statistics of the three milksheds are given in Table 2.1. The project area experiences a hot season commencing in March and lasting to about the middle of June, a June to September monsoon, a post-monsoon season spanning October and November and a cold season from December to February. In 1974 the bovine population of the state was about 26 million cattle and 6.0 million buffalo. In 1982, the figures were 27 million and 6.5 million respectively. In that year about 8.5 million cows and 3 million female buffalo were of breeding age and some 50 percent of those were in milk. Most animals do not conform to any specific breed and milk production varies considerably according to breed, feeding and management practices. On average in 1982, the yield per lactation for cows was about 500 liters and 800 liters for buffalo [Singh and Pethiya, 1983].  -9- Table 2.1: PROJECT MILKSHEDS IN MADHYA PRADESH Project Bhopal Indore Ujjain Total Project as % of Characteristic Milkshed Milkshed Milkshed Area State Population (millions) - Rural 3.585 2.048 1.240 6.873 16.54 - Urban 1.322 1.208 0.660 3.190 30.12 - Total 4.307 3.256 1.855 10.062 19.28 Area (millions of hectares) - Cultivated 2.150 1.116 0.758 4.024 21.9 - Irrigated 0.154 0.140 0.074 0.368 16.5 - Total 3.995 1.903 1.098 6.996 15.8 Distribution of Land Holdings (thousands) 0 - 0.5 ha 106 17 14 137 19.6 0.5 - 1.0 ha 34 16 19 69 12.9 1.0 - 2.0 ha 63 30 31 124 18.1 2.0 - 3.0 plus ha 56 25 22 103 12.6 Source: Singh and Pethiya [1983] Tables 1 and 2. The most commonly used cattle feeds are by-products of agricultural crops, weeds, perennial pasture grasses, concentrates and cultivated fodder crops. For many households dairying is largely a subsidiary activity to crop farming, though in many instances it provides considerable cash income. Over 80 percent of the cows and buffaloes are kept in villages and the average herd size is 2-3 breedable animals per animal owning household. Before the introduction of the project the State of Madhya Pradesh had an undeveloped dairy processing and marketing system characterized by severe shortages of milk and marked price rises during the lean season [ibid]. Animal health care and technical services to farmers were also inadequate and the veteri- nary service concentrated on disease control and prevention [World Bank 19741. Moreover, before the project there was little institutional support for dairy development nor had adequate facilities for milk collection, processing and marketing been developed, although dairy cooperatives were launched in the project area in 1960 under an earlier Government scheme. Inappropriate price policies and other organizational shortcomings caused the cooperative movement to loose ground throughout the 1960s. By 1973, there were less than 100 cooperatives supplying milk to the state owned processing plants [ibid].  ",J .e,,g,0 CHINA 7 Or JOIALA WAR rO JMALA WAR To GVNA 7170 T r UNA PAKISTAN,,,",' Chhapera n g p ara |7 Sne= IchwrINDIA o </l ~ ~~~ ~ ~~~~~~~~~~~~~~ , 0, O0'e9aP3Sb"&s,an,@r Statpe Boundares ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.onmnoec,a,TETU,rewog inw teratioa Boundaries 1 i M~~~~~~~~~~~~~~~~~~~~hbow,dasch lo,,Sre 0 _ o < \ ~~~~~~~~~~7697* 9 150 2|0 3|0 40 50 "3 X ~~~~~TO :h;ANOWA j KILOMETERS N,  -11- The project was initiated in 1974 with the objectives of building an efficient system of collecting milk in rural areas and marketing it in urban areas. The system was also to be capable of efficiently delivering technical knowledge and services to farmers and was to be based on village level cooperatives. Total expenditure was estimated to be Rs 249.7 million spread over seven years (1974-1981). In the event the project was not declared complete until 1983 by which time Rs 205.3 million had been spent. Like other projects under Operation Flood the objectives of the Madhya Pradesh project were implemented through the development of a three- tiered cooperative structure. First, the project area was divided into a number of milksheds each under the control of a Cooperative Union, that is, a union of the village dairy cooperative societies (DCSs) in each milkshed. The Unions have the mandate to organize DCSs in their territory (milkshed). The DCS's are linked through a series of milk collection routes which are established mainly on the basis of the cost of transporting milk to the Union owned dairy plants located in the major urban centers. Milk routes are developed sequentially each building on its predecessors. Each route may therefore include villages with different levels of actual and potential milk production. The Unions also own and operate feed plants and provide services (artificial insemination, animal health, training and extension, milk collec- tion and milk marketing) to farmers through the DCS's. The Unions form the MP Dairy Development Corporation which coordinated project implementation, produced exotic breeding stock, provided training for DCS staff and planning services to the Unions. These institutions have continued to operate since the project ended. Summary statistics on project achievements are given in Table 2.2. Table 2.2: PROJECT DAIRY DEVELOPMENT ACTIVITIES /a Union Activity /a Bhopal Indore Ujjain Total Target DCS Organized 282 289 278 849 1,200 DCS Functioning 229 267 261 757 1,200 Average No. of Member Households per DCS 48 59 48 52 130 Milk Procurement ('000 l.p.d.) 32 54 71 157 300 Milk Procurement per DCS (l.p.d.) 140 202 272 207 250 Milk Routes Established 16 18 17 51 n.a. Mobile Veterinary Units Operational 5 5 7 17 34 DCS Served by Veterinary Units 228 274 260 762 n.a. No. of Artificial Insemination Centers 187 203 215 605 n.a. /a Up to March 1983. Source: Singh and Pethiya [1983] Table 8 and Exhibit 3.  -12- The Sample Survey Prior data on characteristics of interest among villages and households in the project area were not available. Hence it was not pos- sible, without undertaking extensive and expensive pre-survey data collec- tion, to devise an optimal sampling design within the resource constraints imposed by the study budget. Nevertheless, as the efficiency of a sample is not, as popularly supposed, a function of the size of the population [Scott, 1985, p. 13], it was necessary only to choose a sample of sufficient size to have a reasonable chance of capturing the heterogeneity of the population in both village and household behaviour and to be logistically manageable. Accordingly, to study the implementation and effects of the project a sample of twelve villages (nine DCS villages and three control villages) was drawn from all districts of the project. After a complete enumeration of all households in each village a sample of 869 farm-households was selected. The breakdown of the sample between DCS villages and control villages is given below. DCS Group Control Group Total Villages 9 /a 3 12 Households 604 265 869 /a The age breakdown of the DCS's was Age 7, 2 villages; Age 6, 1 village; Age 3, 3 villages; Age 2, 3 villages. The sampled households were interviewed twice in 1983; once (July-August) in the wet season and once (November-December) in the dry season. In the second round only 804 households were available for inter- views of which 561 were in the DCS group and 243 in the control group. Throughout both rounds of survey great care was taken to ensure that ques- tions put to respondents were not leading and that the questionnaires con- tained cross checks on the consistency of respondents answers. Moreover, the field work was conducted by a professional and experienced sociologist who also trained and supervised the field investigators. The factors considered in the selection of the DCS villages were: (a) the age of the DCS; (b) level of village social and economic development; (c) DCS performance; and (d) pre-project accessibility to milk markets. For the selection of control villages the criteria were: (a) they should repre- sent conditions typically found in villages before the project was started; and (b) they should be as free as possible of any direct or indirect influen- ces from the project. The selection of the DCS villages in accordance with these criteria was straightforward. The selection of the control villages was more dif- ficult since the project area is located in the Malwa plateau, the more developed part of the state. In order to find a group of villages with agro-economic conditions similar to those of the DCS group, the search was  -13- confined to the project area itself. Thereafter, villages outside the influence of the project were identified and three selected. Among these, however, it turned out that one was very close to an industrial city with substantial milk demand. Because of this strong urban demand this village had an active private milk market where the cooperative organization had failed to establish a DCS because of fierce competition from the private sector. Moreover, the general level of social and economic development in this village was above the level of the other two. This village is probably one of the most advanced in the area in all respects related to milk produc- tion and consequently, its selection as a control, although undesirable in principle, helps to ensure that project performance is measured against stringent standards. Table 2.3: VILLAGE AND HOUSEHOLD CHARACTERISTICS WITHOUT THE PROJECT 1973 1983 Project Control Characteristic Area Villages Households per village 83 /a 176 Household size (persons) 6.0 6.1 Landless Households (percent) 31 /b 29 Small and Marginal Farmers (percent) /c 23 7r 37 SC/ST Households (percent) /e 33 7 26 Literate (percent) 19 7 26 Average Size of Landholding (ha) 4.7 4.4 Average Size of Dairy Herd Owned (animals) /g 4.3 /b 3.9 Pasture area per animal (ha) 0.10 7- 0.08 Milk Animals in Milk (percent) 37 Th 47 Village Area Cultivated (percent) 54 rh 67 Distance to Road (kms) - *7 0.0 Households holding milk animals (percent) n.a. 63.4 /a Based on an estimated total of about 13,000 villages. Many of these are hamlets or satellites of large villages. If each principal village is assumed to have one satellite then the number of households per village becomes 166 in 1973 and 198 in 1982. /b Based on 9 of the 11 project districts in 1971. 78 Farmers with less than 2.1 hectares. 7T 1971 data. 7h SC/ST - Scheduled Caste and Scheduled Tribes. 7T Statewide figure 1971. E Includes cows and buffalo. Statewide figure, 1977. 71 Figure not available but the World Bank [1974] states that 60 percent of villages were easily accessible in 1973. Source: Compiled from information in GOI [1973], World Bank [1974], Singh and Pethiya [1983] and Dairy India [19851.  -14- Table 2.3 provides reasonable assurace that the control villages are acceptably representative of the situation prevailing in the entire project area before the project was implemented. Although the proportion of the population who are literate appears, not surprisingly, to have increased over the decade 1973-82. There also seems to have been a slight decline in the size of the average landholding and an increase in the proportions of farmers classified as small and marginal (holdings of less than 2.1 hectares). In other respects there are few differences between the two sets of figures. To establish whether there was bias in the selection of the project and control villages we tested whether the probabilities of a village being selected for the project or the control group were different. 1/ To do so we used the conventional logit model, Pr (y=1) = 1/[l + exp (x.b)] where y=l represents selection of the village for the project group and y=0 selection for the control group. Because there were only 12 villages in the sample the model was run repeatedly -- one characteristic at a time. The results are given in Table 2.4 and show that the probabilities of being selected for either the project or the control group are equal with respect to several important indicators, including, for example, distance from major urban centers, a factor which determines (as noted above) the design of milk collection routes. Within the limitations of this necessarily restricted analysis, we could not detect any bias in the selection of either the project or the control villages included in the sample. Table 2.4: PROBABILITIES OF SAMPLE VILLAGE SELECTION: A LOGIT ANALYSIS 2 Significance Characteristic Coefficient X value Level Distance from major city 0.56 0.94 0.36 Size of the village 40.63 1.61 0.20 (households) Fertilizer store -1.386 0.73 0.39 Post Office -2.063 0.28 0.62 Panchayat headquarters 0.916 0.43 0.51 Veterinary Clinic -0.918 0.41 0.53 Cooperative Credit Society -1.945 1.76 0.18 Cooperative Marketing Society -0.470 0.11 0.73 The cooperatives in the sampled villages were on average typical of those in the project area (Table 2.5).  -15- Table 2.5: DCS PERFORMANCE IN THE PROJECT AND THE SAMPLE Project Sample Average Average Members per DCS 53 63 Milk Procured per DCS ('000 litres per year) 59.7 65.7 /a Milk Procured per DCS member ('000 litres per year) 1.13 0.97 Share Capital per DCS member (Rs) 29.8 29.8 Feed Sold per DCS (Rs '000 per year) 9.5 9.3 Artificial Inseminations per DCS (per year) 14 19 A.I. Success Rate per DCS (percent) /b 20 21 /a Quantity procured was not available for four DCS's. Hence, expenditure on milk procurement was divided by the average price paid for milk to obtain these missing values. /b Success is defined as live births. Source: For project averages Singh and Pethiya [1983] Tables 7 and 8 and Exhibit 3. For sample averages DCS records in sampled villages. The criteria for the selection of the farm-households were: (a) to ensure sufficient variation with respect to the dairy enterprise; and (b) to ensure representativeness with respect to the agro-economic situation of the farm-household. Sampled households in both the DCS and control villages included those with and without land and with and without milk animals. Table 2.6: CHARACTERISTICS OF SAMPLED HOUSEHOLDS With Milk Animals Without Milk Animals Characteristic DCS Control DCS Control Household Size (persons) 7.31 6.84 5.58 6.30 Education Index /a 2.07 1.88 0.04 1.50 Low Caste Households (percent) 59 57 31 29 Land Owned (hectares) 4.1 3.6 1.4 1.4 Land Irrigated (hectares) 0.6 0.6 0.0 0.1 Bullocks Owned (Nos.) 1.85 2.23 1.33 1.32 Off-Farm Income (Rs per year) 1338 1344 1254 1387 Sampling Proportion .386 .796 .089 .291 /a This index was computed on the basis of the level of educational attain- ment of the household head. Attainment varied from none (illiterate) to five (university graduate). Table 2.6 demonstrates that the sampled households in the DCS and control villages are broadly similar. The cropping pattern was also found to  -16- be similar for all groups (not shown in Table 2.6) with most of the cul- tivated area devoted to grains and pulses with some, depending on the availability of irrigation water, to cash crops. Fodder cultivation was very rare in all groups. Very few households (less than one percent) cultivated fodder and even then only on a very small portion of their land. There are however some notable differences (Table 2.6). For example, households without milk animals in the DCS villages have a lower education index than their counterparts in the control villages. Whether this is a reflection of better educated households without milk animals being drawn into the project -- increased participation -- cannot be inferred from these comparative statistics. This question, however, is explicitly addressed in a later chapter. Similar remarks apply to the households owning bullocks. Other minor differences remain but as all the characteristics listed in Table 2.6 explicitly enter the analysis their influence is fully controlled. A final point remains. To ensure that sufficient households possess- ing milk animals were included in the sample such households in both the DCS and control villages were over-sampled. That is to say the sampling propor- tions exagerate the proportions of such households in the population. This is of no account in the bulk of the analysis that follows where interest centers on the effect of the project on households with milk animals. When, however, the question of increased participation (households acquiring milk animals when previously they had none) is addressed these divergent sampling proportions are taken into account. DCS Activities and Farmer Attitudes in the Sampled Villages The project's development activities are implemented at the village level through the Dairy Cooperative Societies (DCSs). Farmers may choose whether to become members or not. Formally, membership confers no particular benefits except an end-of-year bonus or dividend related to the quantity of milk sold to the DCS. Members may, however, gain preferential access to supplies and services provided by the DCS but this is not a formal privilege. Hence, non-members of the cooperative are able to obtain most project benefits including selling milk to the cooperative at the same price as members. Accordingly, is is not surprising that membership rates in the nine sampled DCS villages ranged from only 27 to 58 percent of the households that own milk animals. This membership rate is quite small compared to project expectations and may be related to the fact that non-members are not excluded from benefits and the apparently marginal nature of the additional benefits accruing to members. Although the overall performance of the Madhya Pradesh project was less than anticipated (Table 2.2) the DCSs in the sampled villages appear to have been functioning as well as any others (Table 2.5). Hence, the basic structure built by the project at the village level seems to be in place and operating adequately. Some project activities, however, operated better than others. For instance milk collection and marketing seem satisfactory in volume terms while the provision of artificial insemination (A.I.) and the provision of inputs (feeds, fodder seeds, etc.) is relatively unsatisfactory. However, even though these are broadly favorable observations Attwood [1985]  -17- in a related sociological study of the sampled villages notes that the DCS were established with little or no input from the farmers. He writes, "... In the Bhopal region...the village farmers have had almost no hand in the planning or implementation of the new dairy scheme. Enquiries in three (sampled) DCS villages...established that the DCSs were organized as follows: some officials from Bhopal came to each village, offering to show a film about the benefits of cooperative dairying and then urged the village leaders to sign up members for a DCS. The selection of which villages to include in this process, and when to approach them, was made entirely in the dairy offices in Bhopal. Moreover, the Bhopal Dairy Union (not to mention the Madhya Pradesh Cooperative Dairy Federation) was not formed by a process of unifying leaders and initiatives from among the village DCSs. Instead, it worked the other way around: the DCSs were formed at the behest of the Union and the Federation. Consequently, in the Union's annual reports, members of the Board of Directors, consisting of representatives from the component DCSs, are not even mentioned by name. They are probably irrelevant to the decision-making processes within the Union..." [p. 49]. The three sampled DCS villages studied by Attwood also had other problems. He notes that "...in terms of social structure, all three are similar, though...(two)... are dominated by a large Khati lineage along with two or three small lineages, while... (the other)... is dominated by two lineages about equal in size. In terms of politics they are very dif- ferent... (one)... is divided by lineage-based factions which hamper the performance of the DCS... (another)....is dominated by a small group, led by the sarparch who is committed to his own profit making in the private milk trade.... The other is led by a loose coalition in which the leaders of the DCS seem to be able to get along well with the leaders of the panchayat and the multipurpose cooperative" [p. 37]. 2/ Hence, it is not surprising that many farmers in the DCS villages have only a rather vague idea about the way the DCS functions, its legal status and the reason for it. Moreover, the authors field experience revealed that farmers sometimes confuse the cooperative structure with government departments. Among DCS members, when asked in the survey whether they were familiar with the by-laws of their Society, only one third answered positively. Nevertheless, and despite their limited knowledge of the legal status of the DCS most farmers still felt that they could rate DCS perfor- mance. Their responses are shown in Table 2.7 panel (a). Overall, it seems that more than half of all the farmers believe that the DCS operates adequately, but among DCS members this proportion rises to about 86 percent. Rather more than half of the farmers believed that the DCS confers benefits in addition to price and marketing improvements; Table 2.7, panel (b). This view is held more frequently by members than non-members. Nevertheless, there is, even among non-members an apparent widespread knowledge of the DCS and most believe it to be beneficial. This is consis- tent with the point made earlier that non-members are not denied access to the services provided by the DCS. Additional information (not shown in Table 2.7) indicates that about 80 percent of the respondents reported that operat- ing surpluses from the DCS had not been used, as yet, for the general benefit  -18- of the village. This may be because, in their early years, the Societies aim to accumulate surplus funds in order to build their own quarters. In one of the sampled villages the society had used surplus funds for this purpose. Table 2.7: FARMERS' OPINIONS OF THE DCS AND THEIR REASONS FOR JOINING a/ (a) Performance of (b) Additional Benefits Received the DCS (%) by Members (%) Members Non-members All Members Non-members All Good 43 28 34 Yes 67 52 58 Satisfactory 43 34 38 No 29 19 23 Bad 12 10 11 Do not know 4 29 19 Do not know 2 28 17 (c) Reason for Membership (%) (d) Reason for Non-Membership (%) Members Non-members To sell milk 76 Inability to buy share 17 To get bonus 9 No spare time 13 To obtain vet services 7 DCS not useful 19 Other reason 8 No specific reason 51 /a Households with milk animals in DCS villages. The reasons reported by members for joining the DCS are shown in Table 2.7 panel (c) and the reasons for non-members not joining in panel (d). The primary motive for joining seems to be the prospect of better milk marketing. With regard to input supply only veterinary services seem to play a role, and a small one, in farmers' decisions to become DCS members. About twenty percent of non-members do not find it useful to become members while about half of them declined to give a specific reason for not joining. Money and time constraints are always relative to expected benefits but seem, in about 30 percent of cases, to have restricted the desire of farmers to become members. Even in the control villages, however, opinion in favor of the DCS seems strong as 63 percent of survey respondents said they would join a DCS if one were established. Thus, it seems that the motives of the average farmer in joining the DCS are affected mainly by the change in the milk market rather than by changes in the provision of inputs. This of course, may reflect the relative success of the cooperatives in addressing and solving these problems rather than farmers intrinsic priorities. This is consistent with the evidence in Table 2.8 which shows that the changes perceived by farmers following the establishment of the DCS were mostly related to the milk market and to the price of milk. Changes in the provision of inputs were generally perceived to be of secondary importance.  -19- Table 2.8: CHANGES AFTER DCS ESTABLISHMENT AS PERCEIVED BY FARMERS WITH MILK ANIMALS IN DCS VILLAGES Primary (Z) Secondary (%) Increase in milk price 45 11 Better market for milk 27 14 Increase in availability of veterinary sevices 10 35 Regular payment for milk sold 5 14 Improved fodder supply 2 5 Other 11 21 Total 100 100 Table 2.9: MEMBERSHIP OF THE DCS AND USE OF MILK MARKETS (percent) Selling Milk to: DCS Private Only Traders Only Both Total Members of DCS 60.0 14.2 0.6 74.8 Non-members 12.6 12.6 0.0 25.2 Total 72.6 26.8 0.6 100.0 Note: Percentages are based on the number of farm-households (N = 175) that reside in sampled DCS villages, own milk animals and were selling milk either to the DCS or to private traders at the time of the survey. Nevertheless, the establishment of a DCS does not necessarily lead to the demise of the private milk trade. Indeed, as already noted, one of the control villages had a sufficiently strong and competitive private market that DCS establishment was unsuccessful. Accordingly, as Table 2.9 shows, among the sampled households selling milk the DCS is used by both members and non-members while the private milk market continues to operate; attracting sales from both non-members and members of the DCS. Besides operating in the milk market, the cooperative organization undertakes a number of activities that aim to increase milk production in the project area. Amongst these the provision of veterinary services was per- ceived by farmers to be the most successful. Evidence, presented in Table 2.10, shows that the operation of the project may be related to an increase  -20- in the awareness of the need for veterinary services (panel a) and an increase in the rate with which such needs are satisfied. This applies about equally to DCS members and non-members. In the project area the cooperative organization is the main source providing these services which are probably entirely additional, i.e., there was no reduction in the existing level of government or privately provided veterinary services. The increased provision of veterinary services is designed to comple- ment the expansion of artificial insemination and cross-bred technology. However, only 2.5 percent of the sampled households with milk animals own cross-breds (3 percent in the DCS villages and 1.5 percent in the control villages). That cross-bred technology seems to be of very limited attraction to farmers is demonstrated by about 87 percent of the sampled farmers choos- ing not to use it. Hence, even without widespread adoption of cross-bred milk animals the provision of veterinary services has proved useful and attractive to farmers. Table 2.10: NEED FOR AND SOURCE OF VETERINARY SERVICES (percent) (a) Need for Service DCS Villages Control Members Non-members All Villages Need for Veterinary Services /a 15 27 20 13 Need was satisfied /b 96 88 93 55 (b) Source of Service /c DCS Villages Control Members Non-members All Villages DCS (%) 64 45 60 - Private Veterinary Service /d 7 10 6 25 Government Veterinarians 29 45 34 75 Total 100 100 /a Households holding milk animals and needing veterinary services during the season preceding the interview. /b Households who obtained veterinary services as percent of those who needed them. /c Percentages represent the share of those who obtained services. 7_ Private veterinary services are not thought to be available in the project area. Hence this may reflect the use of 'traditional' experts. The provision of cattle feed by the DCS is quite limited and is confined mainly to concentrates (balanced feed) which is used to supplement the usual fodder, providing an additional source of energy and protein. There is evidence [Singh and Pethiya 1985 p.83] to show that societies sell only modest quantities of concentrate. This evidence is supported by survey data  -21- showing that while the private feed market is used by as many as 50 percent of the households, the DCS is used as a source of feed by only 28 percent of households. Other activities of the DCS (e.g. sale of fodder seeds) are very limited. Differences Among Project and Control Households From the evidence above it seems that the project has affected a substantial proportion of households in the sampled villages providing them with a better milk market and with an improved supply of dairy inputs. It is an empirical matter to measure the import of these changes for the average farm-household. This study seeks to determine the impact of the project on the following parameters at the household level: (a) milk output; (b) output of other agricultural products; (c) consumption of milk and other products and, hence, nutrition; and (d) the distribution of benefits. Comparisons of sub-sample means with respect to these parameters provide a first, impres- sionistic view, of differences between households in the DCS and control villages that may be attributable to the project. Table 2.11 provides such averages for the main characteristics of the sampled households in the project (DCS) and control villages. Table 2.11: CHARACTERISTICS OF HOUSEHOLDS WITH MILK ANIMALS DCS Control Characteristic Villages Villages Members of the DCS (X) 35.00 n.a. Milk Output (lt/day per household) 3.38 3.09 Milk Sales (Rs/day per household) 5.46 4.74 Concentrate Use (kgrs/day per household) 1.84 1.88 Households Producing Milk (x) 76.00 67.00 Number of Buffaloes (per household) 1.70 1.33 Number of Local Cows (per household) 2.70 3.01 Number of Cross-bred Cows (per household) 0.05 0.03 n.a. = not applicable. In contrast to the similarities among milk animal owning households with respect to their socio-economic characteristics (Table 2.6) the two groups in Table 2.11 present some appreciable differences in their dairy enterprise. First, and as expected, a substantial proportion of households with milk animals in the DCS villages are members of the DCS. It is some- times argued that a much higher membership rate should be expected (see Chapter VIII). However, since non-members are not excluded from using the project facilities membership is likely to underestimate the true number of households that benefit from the project. It was shown in Table 2.9, for example, that 73 percent of milk animal owning households in the DCS villages who were selling milk at the time of the survey sold at least some of their milk to the DCS. Revenue from milk sales is higher for the households in the DCS villages but whether this is the result of a higher price for milk or  -22- increased milk sales as a result of increased output or decreased consumption cannot be discerned from Table 2.11. Average household milk output is higher (test for statistical sig- nificance not shown) among households in the DCS villages. Part of this difference in output may, however, be attributable to differences in other household characteristics and cannot, without further analysis, be assigned to the introduction of the project. Potential sources of higher milk output are increased milk yields per animal and changes in the size and composition of the milk animal population. There are indications, as shown in Table 2.11 that these factors may contribute to the difference in milk output, however, their precise and respective contributions must be measured with more rigorous statistical techniques. Changes in milk yields per animal may be due to better feeding, better veterinary services or better management. The amount of concentrates used per household does not seem to be different between the two groups; however, more households use concentrates in the DCS villages than in the control villages, 48 percent compared to 31 percent. But the contribution of such factors can only be established with a higher order of analysis. Changes in the productivity of the dairy enterprise may also affect the level of resources farmers allocate to their other farm enterprises. Preliminary observations suggest that these links may not be very strong. As already noted little fodder is cultivated by the sampled farmers and there was no discernible difference between the amounts of land allocated to fodder cultivation by farmers in the DCS and control villages. This suggests no, or at least a weak, project effect on cropping patterns. Improved dairy manage- ment and increased levels and regularity of income from dairying, by improv- ing overall farm cash flow, may affect the level of inputs and the quality of management devoted to crop husbandry. This link too, may not be strong, as the frequency of fertilizer use in the two groups was about equal. This of course does not account for any change in the intensity or timing of fer- tilizer applications. Similar considerations apply to labor inputs. But factors such as these must all enter the analysis together in order to achieve a proper accounting. Project impact on consumption and nutrition, and on income distribu- tion as well, is difficult to measure by comparing group means. Since milk output and sales per household both seem to be higher in the DCS villages their incomes may also be higher with consequent changes in their consumption patterns, but the net effect of such consumption changes, especially when prices have also changed, cannot be easily determined by comparing simple group averages. In subsequent chapters we deal more rigorously with these and other issues. Before doing so, however, we provide below an overview of the methodology. An Overview of the Methodology The impression given by the foregoing description of project activities and by the rough comparison of project and control households is that the institutional structure created by the project is operating and that  -23- there may be positive project effects on milk output. The subsequent chap- ters of this study undertake the task of quantifying the project effects according to the hypotheses stated at the outset (cf. Chapter I). It will be helpful, however, to first briefly present the methodology used to estimate project effects and to outline the sequence in which the hypotheses are tested. We start by elaborating the evaluation framework presented in Figure 1.1 (Chapter I) through a discussion of the measurement of project impact on an indicator variable. The Madhya Pradesh Dairy Development Project, after its inception in 1975, expanded in annual increments, each covering a number of new villages. The sample survey reflects this fact and includes villages with DCSs estab- lished at different points in time as well as villages without a DCS - the control villages. This offers a particular advantage for impact evaluation: it creates a pseudo-time dimension in otherwise conventional cross-sectional data and hence allows the specificaticn of the project effect as a continuous (as time since the establishment of the DCS) as well as a dichotomous (with or without the project) variable. The advantage of a continuous specifica- tion is that it allows comparisons of effects not only between project and control groups but also between project groups with different periods of exposure to the project. In the analysis in the chapters that follow we use both specifications. An illustration of the incremental project effect on an indicator variable, Qm, at time ts (in this case the time of the sample survey) is shown in Figure 2.1. We assume that there is an underlying upward trend in Qm, represented by the trajectory OAEB, the slope of which is unknown. This obviously implies that there is an underlying secular trend attributable to exogenous factors, unrelated to the project, which applies equally to areas affected and not affected by the project. 3/ Thus we are concerned with additional change, attributable to the project, over and above the underlying secular trend. A dichotomous specification of the project impact would result in a discontinuous upward shift of AB and an estimate of CE, (the differential in Qm) at time ts. In contrast, a continuous specification of project impact, by admitting the time dimension, allows the estimation of the project effect as the change in Qm, for an increase in exposure to the project by one year at the margin. This allows an estimate of the area ACE which represents the incremental cumulative output in Qm attributable to the project at time ts.4/  -24- Figure 2.1: Measurement of Project Impact Qm _. D B 0X to ts time to: time of introduction of the project ts: time of sample survey Qm: indicator variable (quantity of milk output) ACE: shaded area indicates the incremental effect on the indicator variable Qm due to the project. The empirical analysis starts with the question of whether the project changed the economic environment within which milk producers operate. Without some identifiable change in economic incentives any output change attributed to the project may be spurious. The results of the project inter- ventions in the milk market (creation of an efficient milk collection system) and on input supply (provision of a technology package focussing on A.I., veterinary services, and feed inputs) are explicitly measured with reference to milk prices and farmers' awareness and use of inputs such as A.I. and veterinary services. Having established that changes did occur in the economic environment of the dairy farmer we proceed with the testing of the hypotheses on project impact. The introduction of new technology based on cross-breeding, by affecting the size, composition and productivity of the milk animal popula- tion, was perceived by the project to be the main instrument for increasing milk production. Hence, the adoption of the new technology is explicitly tested with respect to the effect of the project and other factors using a binomial model of choice to analyze farmers decisions about holding cross-bred milk animals.  -25- The testing for positive direct project effects on milk output is accomplished using a supply response framework testing sequentially for changes in milk yields and for changes in the size and composition of the milk animal population. Having established that such changes did occur we proceed (testing for aggregation bias) to estimate an aggregate milk produc- tion function and to analyze and quantify the project impact on milk output and productivity. Subsequently the sources of the increase in milk output are identified and the project impact decomposed by source. Had adequate data on the costs of the project been available it would have been appropriate, at this stage, to analyze the effects of the project in conven- tional economic terms by doing a cost-benefit analysis. Unfortunately, appropriate data were not obtainable. The indirect effects of the project on the output of other agricul- tural products are examined next. Crop and livestock production are integral parts of the farming system in the area studied. We hypothesized in Chapter I two links between the project and crop production, one operating through land allocation and the other through the use of modern inputs. We explicitly test for the presence of both. Given the interest in India and elsewhere in the question of whether project-induced output gains are distributed equitably we examine how project benefits are distributed. First, we look at the participation issue, i.e., whether poor non-producers are encouraged by the project to acquire milk animals and to begin a dairy enterprise. Second, we analyze the factors that affect membership of the DCSs created by the project and the use of their facilities. Distribution issues have been debated in India on the basis of DCS membership figures only. Our analysis attempts the identification of the groups that benefit most on the basis not only of group averages but also through an analysis of the factors that affect DCS membership and the use of DCS services. The empirical analyses are again based on binomial models of choice and explore farmers' decisions to participate (own milk animals), to be a member of the DCS (membership), and to market their milk through the DCS. Consumption and nutrition effects are measured using a consumption function analytical framework. First, the project impact on milk consumption is quantified distinguishing the effects of such factors as the price of milk, household expenditure, the level of own milk production and other household characteristics. Second, the project impact on calorie and protein consumption is quantified again distinguishing the effects of the factors noted above. Finally, in a concluding chapter, we present a synthesis of the empirical results as they relate to the hypotheses stated at the outset and deduce a number of implications for dairy development policy.  -26- FOOTNOTES (Chapter II) 1/ A major issue in the evaluation of social programs is the possibility of selection bias. The model commonly employed for such evaluation is Y = x + aI + p (1) where Y is outcome, X is a vector of exogenous characteristics and I is usually a dummy variable (I = 1 for participation in the program and 0 otherwise). In this model the effect of the program is measured by the coefficient a . However, the dummy variable I cannot be treated as exogenous if the decision to include an individual in the project is non-random. In such cases I is endogenous and equation (1) must be estimated by instrumental variable techniques. If, however, the assignment to the project and control groups is random (i.e. the probabilities of being selected in the control or the project group are equal) I is an exogenous dummy variable and equation (1) can be estimated using OLS. 2/ These characteristics of village social structure influence the participation of households in cooperative activities. This is more extensively and rigorously analyzed in chapter VIII. 3/ The implied assumption of a common dairy production environment for all villages in the sample is based on their geographic proximity, a common policy situation and the fact that we were unable to detect any bias in village selection. Moreover, our analysis is not a simple between group comparison but a multivariate analysis. Hence we are able to account for the influence of a number of other factors not explicitly controlled by the sampling design. 4/ In conventional time-series there are always complications due to the presence of other random effects (e.g. weather), the impact of which is not easy to control. In this sense, the pseudo time-series structure obtained from otherwise conventional cross-section data is superior. Moreover, such data is less expensive and quicker to collect.  -27- III. ENHANCEMENT OF MILK PRODUCTION INCENTIVES Part of the measurement of the impact of the project must be an examination of the changes that the project induces in the economic environ- ment within which dairy farmers operate. The project objectives are: (a) the building of an efficient system for collecting milk in the rural areas and for marketing it in urban areas either as fluid milk or milk products; and (b) the building of a system capable of efficiently delivering technical knowledge and dairy inputs to farmers so that their milk producing capacity is enhanced. The cooperatives organized by the project may be seen as providing the following advantages in milk marketing to farmers; economies of scale, vertical integration and bargaining power. The provision of such services should elicit a supply response from milk producers. Supply response, however, also depends on the supply of technology and of inputs [Krishna, 1982]. The cooperatives organized by the project do provide tech- nical knowledge and inputs (artificial insemination, veterinary services, feedstuffs etc.) that are intended to increase milk supply (output). In this section, however, the effectiveness of the project in increasing the efficiency of the milk market and in delivering inputs that enhance milk production capacity is examined. Market Efficiency The traditional milk marketing system depends on private traders who collect milk in rural areas and transport and sell it to nearby urban con- sumers. Such systems are often thought to operate in a manner that enables private traders to realize excess profits at the expense of both rural producers and urban consumers. If such exploitation exists it implies an inefficient system if not inefficient operations. An efficient marketing system is usually defined as one in which the movement of goods from producers to consumers takes place at the lowest cost consistent with the provision of the services consumers desire and are willing to pay for. Wanmali [1981] has shown, on the basis of extensive studies, that rural periodic markets in India are diverse and that there is little market integration. Traditional milk markets in India are no exception and are thought to be inefficient in several ways. The village level markets are uncertain and depend on the traders decisions about when to collect milk. Prices may also fluctuate considerably since milk production is highly seasonal. Moreover, there is evidence [Singh and Kelly, 1981, p. 15] sug- gesting that traders do not collect milk during the dry season in a village if they consider the volume to be inadequate. Additionally, when the number of traders is small there is always the risk of the traders operating col- lusively. Producers may however market surplus milk in the form of milk products such as ghee. The volume of milk that such a system can handle is limited not only by these factors, but also by the perishability of the product and the large capital investments required to establish processing and marketing facilities in the cities. Such capital investment is generally beyond the means of private traders.  -28- The inefficiencies enumerated above characterize most village level private markets for perishable products. Such markets are usually "... less organized and less competitive..." than markets for other food crops [Lele, 1981, p. 62]. Often, cooperatives and other public sector institutions in developing countries are seen by governments and the public as a means of improving conditions in those markets by replacing traditional systems and eliminating the middleman [Fox, 1979, Lele, 1981]. Nevertheless, the solu- tion to monopsonistic tendencies is not necessarily the replacement of the traditional marketing system [Lele, 1981]. Instead, additional marketing channels should be created in order to foster competition without eliminating the traditional ones [Fox, 1979, Lele, 1981]. Marketing policies and govern- mental actions should also be focussed more directly on the improvement of the conditions of private markets for agricultural products, inputs and rural consumer goods [Fox, 1979] in order to increase their efficiency. Neverthe- less, cooperatives are often seen as a means of enhancing the capacity of producers to market their product, thus reducing the cost of marketing and, hence, the inefficiency of the system. 1/ Lele points out specific measures that can be implemented through a marketing institution (such as a cooperative) which would help to improve rural markets. She suggests (a) the establishment and use of standard weights and measures; (b) the dissemination of information on prices prevail- ing in other producing and consuming centers; and (c) the construction of storage facilities, particularly for perishables, that allow delayed market- ing and thus help to prevent 'distress' sales [Lele, 1981, p. 63]. The dairy cooperatives established under the project undertake all of these actions. The cooperatives test all milk procured for fat content in order to dis- courage adulteration of milk and to offer farmers and consumers a fair price adjusted for quality. The milk collection system that links producer cooperatives also links the village milk markets to large urban markets and, in principle, allows better dissemination of information about prices. Finally, the establishment of milk collection routes and the building of milk chilling facilities overcomes the problem of perishability. Furthermore, dairy plants operated by the Unions undertake dehydration of surplus milk during the winter when supply exceeds demand and reconstitution during the summer when supply is short. Such efforts aim to diminish seasonal price fluctuations, reduce uncertainty and provide an integrated market in both space and time. If the DCS is to capture all or part of an existing private milk market it must be more efficient than the existing private operators. Using the villages in the sample, Table 3.1 provides a typology of the changes in the structure of the milk market resulting from the introduction of a DCS. This typology suggests that the DCS in most cases succeeds in capturing all or some of the previous private trade. An increase in market efficiency implies a reduction in marketing costs, hence, a possible change in the level of farm and retail prices. This is illustrated in Figure 3.1. The equilibrium pre-project condition is at points A and B for the farm and market level respectively, with QO milk quantity produced and marketed and Pf and Pm milk prices at the farm and  -29- market level respectively. The difference between milk prices represents the marketing cost (also called the marketing margin). This margin represents real economic cost since "... real economic resources are required to trans- form food commodities in space, time and form to food that consumers buy and eat..." [Timmer et al, 1983, p. 197]. To keep the presentation simple it is assumed in Figure 3.1 that the marketing margin is constant. Table 3.1: A TYPOLOGY OF CHANGES IN THE STRUCTURE OF THE MILK MARKET WHEN A DCS IS INTRODUCED Number of Pre-project Post-Project Villages Type Milk Market Milk Market in Sample I Milk market is limited Almost exclusive 4 except occasional dependence on the (project) selling of milk to DCSs which operate teashops, etc. and with varying degrees nearby weekly markets of success II An active private DCS captures most of 3 milk market the milk market and (project) eliminates almost all private trade III An active private DCS captures part 2 milk market of milk market and (project) results in competitive operation of private and DCS markets IV An active private DCS fails to become 1 milk market established and hence (control) to capture part of the milk market (no project activity) V A limited milk market No project activity 2 (control) Using Figure 3.1 an increase in market efficiency can be illustrated by a reduction in the marketing cost from P to Pf under pre-project condi- tions and from P , to PfI under post-project conditions. The reduction in marketing cost (alssuming a constant marketing margin) causes a shift of the milk demand function at the farm level, Df, to Df,. For the same reason the milk supply function, at the farm level, AMD shifts to Sm . The changes in prices to PM ,and P , that follow, in a competitive market, would induce a supply response by the producer from Q to Q1 and finally a new equilibrium position will be attained indicated by C for the farm level and D for the market level. The new price at the farm level is higher while that at the  -30- Figure 3.1 EFFECT OF A REDUCTION IN MARKETING COST ON SUPPLY AND DEMAND FOR MILK Sm Milk s Price m Sf PM? PfX Pf DD l l I D~~~~~~f ; ~~~~~~~~~~~~I I Q0 Qi Milk Quantity Sm' Sm' Milk supply at retail market level Sf: Milk supply at farm level Dm: Milk demand at retail market level Df, Df' : Milk demand at farm level P~f : Marketing cost under pre-project conditions Pm-Pf: Marketing cost under post-project conditions PM, PM Milk price at (retail) market level Pf, Pf': Milk price at farm level  -31- market level is lower than under pre-project conditions; indicating that the gain from the increase in market efficiency is shared by both producers and consumers. The actual shares depend on the magnitude of the relevant elas- ticities of supply and demand. This illustration suggests that the introduc- tion of the project may shift the farm level milk price upwards. 2/ Such a possibility suggests the following null hypotheses: (i) the milk price offered by the private trade in the project area is not different from the corresponding price in the control area; (ii) the average price of milk in the project area (offered by either the DCS or the private trade) is not different from the corresponding price in the control area; and (iii) within the project area the milk price offered by the DCS is not different from the corresponding milk price offered by the private trade. (Competition does not suggest a divergence of private and DCS prices in the project area. Rather, it is likely that the private trade would adjust its prices to match those of the DCS. Thus, this is a test for competition in the project area.) These hypotheses are tested using the a model which relates the farm level milk price to the distance from the traditional local weekly market, that distance squared (to capture any non-linearities) and the presence of the project (DCS). Milk is an homogenous commodity except for its fat content.3/ Hence, differences in the market price of milk reflect, in fact, differences in fat content irrespective of whether it is buffalo or cow milk. The prices used in the estimation are implicit prices derived from data on the quantity of milk sold and total sales receipts without distinguishing between cow or buffalo milk. Farm level data on prices are usually quite difficult to obtain and there are valid criticisms of their usefulness when they are cross-sectional. Thus, the poor fit and low R-Squares of the estimates in Table 3.2 are, perhaps, excusable. Given the much higher price that milk with a high fat content commands in the market and the absence of data on the fat content of marketed milk we use a proxy variable in the analysis - the composition of milk produced (SHRMLKC). The estimated coefficient is statis- tically significant and implies, for example, in equation (3) of Table 3.2, that the price of cow milk is lowef than that of buffalo milk by Rs 0.25 per litre, calculated as [exp(-0.105) ] . With an average price of Rs 2.55 per litre and an average share of cow milk of 0.48 we arrive at prices for cow and buffalo milk of Rs 2.42 and Rs 2.67 per litre respectively, and a coeffi- cient, to transform buffalo milk to cow milk equivalent, of 1.10.  -32- Table 3.2: PROJECT IMPACT ON MILK PRICES Dependent Variable. Variable LPRICE1 LPRICE2 LPRICE3 LPRICE3 LPRICE3 (1) (2) (3) (4) (5) Intercept 1.388 1.946 1.637 1.557 1.372 (4.39) (5.91) (7.16) (6.68) (4.74) LDIST -0.328 -0.812 -0.545 -0.462 -0.314 (1.23) (2.85) (2.83) (2.33) (1.29) LDISTSQ 0.054 0.148 0.097 0.080 0.051 (1.03) (2.49) (2.50) (1.99) (1.06) SHRMLKC -0.116 -0.091 -0.105 -0.109 -0.106 (2.75) (2.27 ) (3.35) (3.49) (2.95) DCSDUM -- -0.085 -0.078 -0.045 -- (2.55) (2.41) (1.21) PRIV -- -- -- -0.041 -0.045 (1.64) (1.67) B2 0.09 0.27 0.12 0.13 0.10 No. Observations 151 91 241 241 199 Comments -- -- -- -- Project group Note: Variables were defined as follows; LPRICE1 = Log of price offered by the DCS LPRICE2 = Log of price offered by private trade LPRICE3 = Log of average price (DCS or private) LDIST = Log of distance from the local weekly market (km) LDISTSQ = Log of distance squared SHRMLKC = Share of cow milk on total milk produced. DCSDUM = Dichotomous variable taking the value 0 for households in the DCS villages and 1 for those in the control villages PRIV = Dichotomous variable taking the value of 1 if price is from the private trade and 0 otherwise The signs and significance of the coefficients yield some interest- ing results with respect to the hypotheses noted above. The effect of dis- tance as represented by the results in columns (1) and (2) is particularly convincing since it indicates that the price for milk offered by the private trade is significantly affected by distance while, and as expected, this is not so for the price offered by the DCS. On the other hand, a statistical test for no differences between all coefficients, i.e., including the dis- tance coefficients, in (1) and (2) was not rejected (the F-value with 5 and 236 degrees of freedom is 0.8698; well below the tabulated F-value of 2.21 for a 0.05 percent level of significance). Hence, pooling the data as in the  -33- estimation of specification (3) is statistically permissable. A similar test for structural difference between the project and control groups was also rejected. Hence (3) is the maintained specification. Column (2) of Table 3.2 indicates that the milk prices offered by the private trade are lower in the control area than in the project area. Hence, we reject the first hypothesis. The specification in column (3) of Table 3.2 combines the milk prices offered by the private trade with those offered by the DCS and the results confirm that the average price in the control area is lower than the cor- responding price in the project area. Hence, we also reject the second hypothesis. Columns (4) and (5) in Table 3.2 test the hypothesis that within the project area the milk price offered by the DCS is not different from that offered by the private trade. The coefficients corresponding to the project and the private trade variables are individually statistically insignificant but jointly significant, hence (4) is the maintained functional form. To further test the hypothesis we restrict the sample to the project group, as in column (5). The results indicate that we cannot statistically reject the hypothesis that private trade and DCS prices are equal. This implies that the milk marketing channel provided by the DCSs increased competition in the rural milk markets. In other words the private trade responded to the intro- duction of the DCSs by raising its price above that offered in non-DCS vil- lages. The village level cooperatives are also active in the feed market but to a more limited extent. Cattle feed sold by the DCSs is mainly con- centrates (balanced feed), which is used by farmers to supplement the usual fodder. The sample data reveals that more farmers use concentrate feed in the DCS villages than in the control villages (48 percent versus 31 percent). Hence, the more widespread use of concentrate feed may be related to the presence of the DCS. The use, however, of the DCS feed market is quite limited as DCSs are the regular supplier to only 28 percent of farm households owning milk animals while the private market supplies 50 percent of the households. Nevertheless, the DCS feed marketing channel by adding to supplies may remove input supply problems for some farm households. As noted above farm level price data are usually of limited useful- ness. This is even more so when the commodity (e.g. livestock feed) is mainly produced and consumed on the farm. Nevertheless, it is possible to test for any systematic difference in the prices of feedstuffs between the project and control groups using essentially the same model as that used for milk prices. The results revealed no project impact on feed prices. Given what was said above about the DCS activities in the input markets this result is not surprising. In short, farmers' opinions, the DCS market share and an analysis of feed prices all show that the DCS does not have an impact on the feed market. Whether this is related to a policy choice or a comparative disadvantage of the DCS could not be addressed within this study.  -34- To sum up, despite the well-known limitations of cross-sectional price data, it seems that the differences in milk-prices observed in the sample conform with the expectations derived from the market efficiency model illustrated in Figure 3.1. In short, it seems that the project has shifted the price of milk upwards in each of the post-project marketing situations involving the DCS in competition with the private trade depicted in Table 3.1. In contrast, there is no project impact on feed input markets. This suggests that market efficiency was improved only in the milk market where the intervention of the DCS was stronger. Finally, note that the milk prices included in the analysis are fluid milk prices and these are generally higher than the (reservation) prices obtained by milk producers when, in the absence of a market for fluid milk, they sell their milk in the form of milk products such as ghee. 4/ Provision of Technology Enhancing Inputs Traditional farming systems in India emphasize crop (mainly grain) production while dairying is usually considered to be an auxiliary source of income and employment. Cow milk is considered by many farmers to be a by-product in the production of bullocks for draught power; the latter are regarded by many farmers as an indispensible input in crop production. When dairying is included in the traditional farm activity mix, the milk stock consists mainly of buffaloes because of their higher milk yields and the higher fat content of their milk. Under the project, each village level dairy cooperative society, besides its activity in milk marketing, is designated as a distribution center for inputs into dairy production, artificial insemination services and technical knowledge designed to increase milk production by farmers. These inputs and services were described in Chapter II where evidence from farmers suggested that veterinary services are quite effectively provided by the project. To quantitatively confirm this observation we test the following null hypotheses: (i) the probability of farmers seeking veterinary services has not increased as a result of the project. (This hypothesis also indirectly tests the assumption that farmers are more aware of the veterinary needs of their animals as a result of information about good animal husbandry practices disseminated by the cooperatives.) (ii) the probability that farmers who need veterinary services do obtain them has not increased as a result of the project. To do so we employ a standard Logit model 5/ conditional on ownership of milk animals to test the first hypothesis and conditional on the need for veterinary services to test the second. The variables used in the model and the results are presented in Table 3.3.  -35- Table 3.3: NEED FOR AND ABILITY TO OBTAIN VETERINARY SERVICES (Logit Analysis) Need for Ability to obtain Variable Veterinary Services a/ Veterinary Services Constant -1.357 0.198 (7.83) (0.50) DCSAGE 0.021 0.164 (1.09) (2.99) COWL 0.058 - (2.34) BUFF 0.061 - (2.28) COWC 0.464 - (1.83) BULLOCK -0.007 - (0.17) LANDC - 0.013 (1.45) EDUC 0.027 0.057 (0.59) (0.46) CASTE -0.099 0.048 (0.75) (0.15) Summary Statistics x2 135.43 11.27 Log-likelihood -288.53 -47.03 Observations 691 129 Cases correctly classified 582 110 a/ This model was also estimated including interaction terms which operate to change the slope of the estimated function. The results obtained were very similar to those reported above. Note: Variables were defined as follows: DCSAGE = Time (in years) since the establishment of the DCS COWL = Number of local cows owned by the household BUFF = Number of buffaloes owned by the household COWC = Number of cross-bred cows owned by the household BULLOCK = Number of bullocks owned by the household LANDC = Area of households' landholding (in bighas, I hectare = 4.32 bighas) EDUC = Educational attainment (coded 0 for illiterate, 1 for some schooling, 2 for elementary etc.) CASTE = Dichotomous variable, coded 1 for low caste, 0 otherwise IRR = Area of land under irrigation (bighas)  -36- The results conform with ex ante expectations. The need for veteri- nary services is, predictably, mainly determined by the number of milk animals owned. The effect of the project (coefficient of DCSAGE) is positive but not statistically significant indicating that there was some increase in awareness among milk producers of the need for veterinary services which probably implies some improvement in the standard of animal husbandry. The first hypothesis, however, cannot be formally rejected. The effect of the project on the ability of farmers to obtain the veterinary services they needed is, however, positive and highly significant indicating a much better supply of this critical input in the project villages compared to the control villages. Therefore, the second hypothesis can be rejected. Table 3.4: OPINION ABOUT AND USE OF ARTIFICIAL INSEMINATION AMONG MILK ANIMAL OWNERS (Logit Analysis) Opinion of Success of Use of Artificial Variable Artificial Insemination Insemination Constant -0.980 -2.94 (4.34) (8.24) DCSAGE 0.140 0.150 (5.52) (4.11) EDUCATION 0.221 0.164 (3.39) (1.79) LAND 0.004 0.004 (0.54) (0.49) IRR 0.029 0.014 (1.21) (0.43) CASTE 0.236 -0.061 (1.36) (0.23) BUFF -0.030 -0.029 (0.70) (0.49) COWL -0.041 0.013 (1.21) (0.27) Summary Statistics x2 45.14 23.66 Log-likelihood -454.00 -241.34 Observations 691 659 Cases correctly 429 574 classified Note: Variables were defined as in Table 3.3. Among the inputs provided by the project sponsored cooperative sys- tem, artificial insemination, by introducing exotic blood, is the main instrument for improving the quality of the milk animal population. The  -37- aggregate data reported in Chapter II show that although the overall provi- sion of artificial insemination was broadly satisfactory this service was not greatly used. This observation, however, should be more rigorously tested. To do so we formulate the following null hypotheses and test them with data collected from farmers during the sample surveys. (i) there is no difference between the DCS villages and the control villages in the probability with which farmers who own milk animals expected good results from artificial insemination. (ii) there is no difference between the DCS villages and the control villages in the probability of farmers who own milk animals using artificial insemination. Again, using a standard Logit model, conditional on ownership of milk animals, the results set out in Table 3.4 were obtained. They show that the probability with which farmers have a good opinion about artificial insemina- tion and the probability that they will use it for their milk animals are both positively and significantly affected by the project. We thus reject both null hypotheses. As the sample is acceptably representative of DCSs and households in the project area these results imply that farmers in the project area are beginning to appreciate the potential benefits of using artificial insemination for upgrading their milk stock. They may also imply that there is an increasing awareness among farmers in the project area of the effects of better animal husbandry that were not measured or observed in the sample survey. Concluding Remarks To conclude, the foregoing analysis suggests that the project has positively changed the environment in which dairy farmers operate. The presence of the DCS appears to have increased market competition and shifted the producer price of milk upwards, even for producers not selling to the DCS, expanded the use of concentrate feed, stimulated and met the demand for veterinary services and provoked an increased awareness of the utility of artificial insemination. These positive results suggest that there may have been an (unmeasured) improvement in farmers knowledge and an increase in the adoption of improved animal husbandry.  -38- FOOTNOTES (Chapter III) 1/ Any reduction in inefficiency depends crucially on two factors. First, the extent to which competition in marketing is increased. Second, the degree to which marketing costs, especially transport costs, are based on marginal pricing policies. In the first case, if a DCS merely replaces a private monopoly by a cooperative one it is not clear whether there will be any improvements in marketing efficiency. In the second, average cost pricing in transport may encourage inefficiency by inducing increased output in areas that lack comparative advantage. These issues are not, however, addressed in this analysis. 2/ Dairy development projects have been criticized as depressing milk producer prices by as much as 10 percent because of the offloading of subsidized dairy products onto the urban markets [Lipton, 1985, p. 2]. Such an effect cannot be captured in a farm level analysis because it has a similar effect on both project and control farms. In contrast, a farm level analysis can examine the effects of the introduction of a new marketing system or technology shifts on milk supply. 3/ Strictly, this is an oversimplication as the protein content of milk can also vary. However, in India the fat content is the primary determinant of price. 4/ A reservation price (Pr) is the equivalent price of selling milk in the form of milk products like ghee. Depending on the fat content of the milk about 20 kilos (of milk) are required to produce one kilo of ghee. With the price of fluid milk around Rs 2.5 per kilo and of ghee around Rs 40 per kilo, Pr is about Rs 2.0 per kilo of milk, i.e., about 20 percent lower than the fluid milk price. In addition there are high transaction costs associated with the marketing of milk products. For example, a minimum quantity of milk is required before processing into ghee can begin and this involves the collection and preservation of milk for several days. Furthermore, substantial labor costs are also incurred in processing and marketing ghee as a trip to the local weekly market in a nearby town is usually required. Given such demands and considering the transportation conditions of the area it seems safe to conclude that the magnitude of the transaction costs are inversely related to the distance of the milk producer from the local weekly market. 5/ See Judge et. al., [1985, Chap. 18] for an exposition of the Logit and other models with dichotomous dependent variables.  -39- IV. THE ADOPTION OF CROSS-BRED TECHNOLOGY A central objective of Operation Flood is to increase milk production by improving the genetic quality of the national milk stock. 1/ This is to be achieved by using artificial insemination (AI) to cross exotic breeds with local animals. Through this method it is planned to gradually upgrade the national herd while simultaneously maintaining desirable genetic properties found in local stock. As planned in 1980 Operation Flood aimed to create nationally a herd of 11.2 million highly productive and environmentally well adapted cross-bred cows by 1985. In Madhya Pradesh the project, starting in 1975, planned to replace all local cows with cross-bred cows during the subsequent 10 years [World Bank, 1974]. This was not achieved. It is estimated that by 1983 only one percent of the cows in the project area were crossbreds [Singh and Pethiya, 1983]. Data from the survey show small holdings of cross-bred cows amongst households in the sampled villages; three percent of farm households holding milk animals had cross-breds and their average holding was about 1.6 animals. There was also a low utilization rate for A.I.; about 13 percent of farm-households holding milk animals. These figures suggest that the objec- tives of the project will probably not be realized in the near future. 2/ What are the reasons for such a poor record? Are any of them curable so that the rate of adoption of cross-bred animals can accelerate? It may be argued that because of the low holding rate generalizations based on the sample are unlikely to be possible. On the other hand, it is the initial stages of the adoption process that are of critical importance and much can be learned by studying them. When an innovation has been adopted by a large proportion of the population further diffusion is mainly a matter of time. This chapter, therefore, examines the adoption of cross-bred technology using evidence from the studys' sample survey. The literature on adoption of cross-bred milk animals is quite extensive but limited in scope [Alderman, Mergos and Slade, 1987]. Existing opinions about the adop- tion of cross-bred milk animals, particularly in India, are not formulated as testable hypotheses. Rather, they are simple conjectures and no empirical evidence exists with which to assess their reliability or to offer guidance to policy makers so that they can focus selectively on the most important factors. This chapter first reviews the issues in, and the opinions about, the adoption of cross-bred milk technology in India. Subsequently the sample survey data are used to provide a brief description of factors that appear to influence the adoption of cross-breds in the project area. Finally, an empirical model of adoption is estimated and the results critically dis- cussed.  -40- The Adoption of Cross-bred Technology in India; Issues and Opinions As elsewhere in India, a cross-breeding program with all the associated activities of artificial insemination, provision of concentrate feeds, fodder development on arable lands, improved animal health services, better milk marketing and processing, education, training and research is the chosen mechanism for expanding milk production in the project area. The provision of the new technology as a part of a package of complementary inputs is designed to make the technology profitable and to reduce risks. The cross-breeding program is based on the results of technical research that indicate that local cattle are a suitable base population for crossbreeding with European bloodstock and that the resulting cross-breds have a superior growth rate, milk yield and breeding efficiency compared to local milk cows [NDDB, 1980]. Crossbred cattle have also been shown to be technically superior to buffaloes. Buffaloes are however superior to cattle in some ways, yet they are more difficult animals to manage on account of their need for shade and ample water. Moreover, they are heavier and have longer intercalving periods and therefore, require more nutrients per kilogram of fat corrected milk produced than cattle of similar performance [Groenewold, 1983]. Despite the research evidence demonstrating the superiority of crossbred cattle, farmers in the project area have been reluctant to adopt crossbred technology. Technically there are no problems associated with keeping cross-bred milk animals provided that certain conditions are met. For example some authorities claim that even pure-bred high yielding milk animals can be kept anywhere in the world provided they receive proper care [Groenewold, 1983, p.16]. Yet the survey data from the project area show that few milk animal owning households keep cross-breds. Although logistical problems in the provision of A.I. in the initial stages of the project may be partly to blame, behavioral factors may also be responsible for the slow rate of adoption. 3/ What then are the factors that determine the rate of adop- tion of crossbred technology? Convincing empirical evidence is altogether lacking on factors affecting producers decisions to adopt cross-breds [Alderman, Mergos and Slade, 1987]. Although the issue has been conspicuously neglected in the empirical literature, several opinions have been expressed by policy makers and development practitioners on the constraints that limit the adoption of cross-breds [see for example Rajapurohit, 1979]. These may be summarized as follows: (a) cattle in India are kept mainly for draught power and the male cross-bred calf if not considered suitable for such work and has little or no market value; (b) the feed requirements of cross-bred cattle exceed feed availability under farm conditions;  -41- (c) cross-bred cattle are expensive assets and a capital (or credit) constraint may operate under farm conditions (even for own produced cross-breds because they can be sold in the market for cash); and (d) the keeping of cross-bred cattle involves an unacceptably high level of risk under farm conditions. With respect to the hypothesized draught power constraint farmers' behavior may have two possible causes. First, farmers may want to be able to produce their own draught power assets (bullocks) within the farm. If this is so, such farmers may be unwilling to switch to cross-bred technology. In this case there should be a negative relationship between the ownership of bullocks and the adoption of cross-breds. Second, most commentators claim that farmers consider the male calves of local cattle to be marketable assets (as bullocks), and that the market proves that the cross-bred male calf is not similarly marketable. If this is true, studies of the productivity differences between local and cross-bred cattle that do not include the off-spring in the calculation of returns do not represent the complete picture of costs and returns (see Kumar and Singh, [1980] for a study focuss- ing on milk production only). Such studies by failing to recognize that the expected revenue from keeping cross-bred milk animals may not be sufficient (when the off-spring are included in the calculation even though cross-bred milk yields are higher) to offset other considerations are misinterpreting an inadequate productivity differential as a draught power constraint. Constraints (b) and (c) are easier to accept. There is evidence to show that the milk requirement of cross-bred calves exceeds the production of their local mothers [NDDB, 1980]. It is also possible that the feed require- ments of cross-breds exceed feed availability at the household level. The capital or credit constraint may also be a real one in situations where the credit market is imperfect and some sections of the rural population have restricted access to credit, especially longer term institutional money. Risk constraints, however, need to be made more explicit. Suscep- tibility to disease and the reliability of the supply of critical inputs that influence yields are objective risks that must affect a farmers' decision to adopt cross-breds. Following a portfolio selection approach the risk-return trade-off is illustrated in Figure 4.1 (for a risk-averse farmer with con- stant risk aversion and a given level of debt and equity). The risk averse farmer would undertake a particular non-risky investment if its return is higher than ro. Introducing business risk, the required minimum rate of return for undertaking the investment is rl. Introducing, additionally, financial risk (for debt-financed investment) the required rate of return becomes r2. The risk-return preference curve for an individual farm-operator, thus, represents the lower bound to a set of acceptable expected rates of return. The curve, of course, may be convex (increasing aversion) or concave (decreasing aversion) depending on the risk preferences of the individual farm operator. A farmer bases his choice of a local cow, a buffalo, or a cross-bred cow not only on the expected costs and returns but also on the variances of  -42- Figure 4.1: BUSINESS AND FINANCIAL RISK OF AN INVESTMENT IN CROSS-BRED MILK ANIMALS WITH CONSTANT RISK AVERSION AND A GIVEN LEVEL OF DEBT Required Rate of Return Constant risk aversion r2 ~~~~I I / ~~~I I rO ~~~~~~~~~~~I I I l r I I I .. CV1 CV2 Risk rO = minimum required expected rate of return for a non-risky investment r, = minimum required expected rate of return for an investment with returns having a coefficient of variation CV1 r2 = minimum required expected rate of return for an investment that is partially or totally debt financed and where the net cash flow has a coefficient of variation CV2 rl-ro = business risk premium r2-r, = financial risk premium  -43- those costs and returns. The business risk (variance) when cross-bred animals are introduced into the farm activity mix is two-fold: (i) yield risk, that is the risk that the expected milk yield of the cross-breds will not be realized because of uncertainty in the supply of critical inputs (e.g., green fodder in the dry period of the year); and (ii) cost risk, that is the risk that actual costs will substantially exceed expected costs either because of the need for more intensive veterinary care or, in extreme situa- tions, because the animal dies. 4/ The financial risk that the farmer accepts when he takes a loan to finance the purchase of a cross-bred milk animal is the risk that the returns may not be sufficient to service the loan. Hence, and returning to Figure 4.1, a risk premium r2 - rl is added to the required expected rate of return rl when the acquisition of a cross-bred cow is a debt financed investment. The three dairy activities (local cows, buffaloes and cross-bred cows) are usually ranked in terms of expected profitability (under certainty) from higher to lower as follows: cross-bred cows, buffaloes, local cows. In terms of risk (revenue and cost risk) the same activities can be ranked from higher to lower as: cross-bred cows, buffaloes, local cows. The choice of the activity is, then, a portfolio selection with a risk-return trade-off where a risk averse farmer might select the least profitable but also the least risky activity. Therefore, a cross-bred cow despite its expected profitability may not be adopted instead of the buffalo or even the local cow because of the high revenue and cost risks involved even when other con- straints are not binding. It follows that empirical work related to the adoption of cross-breds that focuses on determining the productivity difference through farm-management studies is, probably, not useful. Nevertheless, most of the available empirical evidence in the litera- ture on the adoption of cross-bred technology in India concentrates on examining the difference in productivity between local and cross-bred cattle [Alderman, Mergos and Slade, 19871. There is a consensus that if cross-bred milk animals receive proper feeding and care then they can achieve substan- tially higher milk production levels than local cattle. On the other hand, although buffalo are the main dairy animals, the empirical evidence on productivity differences between cross-bred cattle and buffaloes is rather mixed [See Acharya and Pawar, 1980 and Kalyankar, 1980]. Based on expert opinion, [Groenewold, 1983] cross-bred cattle are thought to be more efficient in nutrient utilization than both local cattle and buffaloes. 5/ Factors Affecting Holdings of Cross-bred Milk Animals: Evidence from the Sampled Villages The survey data on farm households that keep milk animals includes information on several aspects of the farm household as well as on the farmers attitudes to and perceptions of artificial insemination (A.I.). As noted above the latter is a prerequisite for a successful cross-breeding program. In this section these data are used to exemplify factors that may be related to the adoption of cross-bred technology.  -44- In Table 4.1 the sampled households are classified according to whether they have adopted or not adopted cross-bred technology (that is hold a cross-bred cow of any age) and according to the size of their land holding. The results suggest that adoption is positively related to the size of land holding. Land ownership is also an important determinant of feed and fodder availability. Under existing technological conditions the main source of feed for dairy animals is crop residues -- supply is obviously related to the size of land holding. In most areas of India the production of fodder crops requires irrigation and thus the amount of irrigated land may also be an important determinant of feed availability. Table 4.1: HOLDERS AND NON-HOLDERS OF CROSS-BRED CATTLE CLASSIFIED BY SIZE OF LAND HOLDING /a Area of Land Holding Holders (bighas) /b Non-holders Holders Total in Class -----------------percent----------------------- Landless 24.62 0.42 25.04 9.4 1-10 32.18 0.55 32.73 9.4 11-20 16.15 0.55 16.70 18.3 21-30 8.36 0.42 8.78 26.7 31 or more 15.69 1.09 16.78 36.1 Total 96.98 3.02 100.00 100.0 No. Observations 707 22 729 22 /a Milk animal owning households only. /b Land area in equivalent unirrigated bighas. 1.00 bigha = 0.23 hectares. Cross-bred ownership may be correlated with the number of milk animals owned. The appropriate classification of households is given in Table 4.2. As for land holdings the results suggest a positive relationship between the adoption of cross-breds and the number of milk animals owned. Experience in dairying, of which the number of milk animals owned may be one indicator, may be an important factor affecting adoption of crossbred technology. But the buffalo is the main milk producing animal, therefore, the number of buffaloes or the share of buffaloes in the households' milk herd may also be an indicator of experience with dairying as well as a measure of the importance of the dairy enterprise to the household.  -45- Table 4.2: HOLDERS OF CROSS-BRED CATTLE CLASSIFIED BY MILK ANIMAL OWNERSHIP a/ No. of Milk Holders Animals Owned Non-holders Holders Total in Class -----------------