may ('7G THE IMPACT OF AGRICULTURAL EXTENSION: THE TRAINING AND VISIT SYSTEM IN INDIA Gershon Feder and Roger Slade wo centuries ago, Malthus predicted a grim future for man- kind, with population growth outstripping the rise in food production. His projections have proved false because agricul- tural output has grown faster than he expected. It has done so not only because more land has been cultivated, with more irrigation and drainage; technology has also improved. In technical terms, each farmer has a production function that dictates how much output will come from the inputs he uses. While these inputs are subject to his discretion, the production function is affected by many other factors beyond his control. Some are environ- mental. But others can be affected by public investment, even though an individual farmer considers them as given: public irrigation and drainage, for example, and agricultural research organizations. These factors can increase output without requiring extra inputs from the farmer. Of course, such investments are not cheap. In many develop- ing countries, public investment in irrigation is a large part of the government's agricultural budget. As for research, worldwide spend- ing has been estimated at $7.4 billion in 1980. Between 1974 and 1984, the World Bank financed some 300 projects dealing partly or wholly with agricultural research, spending almost $1.5 billion. This paper summarizes the results of a study sponsored by the World Bank, "The Impact of Agricultural Extension: A Case Study of the Training and Visit System in Haryana, India" (RPo 672-29). More detailed discussions of these results can be found in other papers by the authors cited in the text. This paper has benefited from the editorial assistance of Joan S. Hazell. Copyright( 1986 by the International Bank for Reconstruction and Development/The World Bank 139 However, better technology alone may not increase farm productiv- ity. There is often a gap between available knowledge and what farmers actually do. They need to know about the costs and benefits of new technology before they adopt it. The public sector is often involved in spreading the knowledge. What is the rationale for its doing so? Information on new farming technology is often a public good, since the provider of information to one user cannot stop other users from getting it without charge, and the value of the information is not affected by the number of users. Information that is more specialized and specific is less of a public good, so could be provided by private entrepreneurs. Indeed, commercial agents dealing with cer- tain types of information such as pest control may operate alongside a public agency providing more general information (Hall 1977). And they may have an incentive to tell farmers about new technology that is embodied in the supplies they are selling. In general, however, farming information has enough public-good qualities to justify public sector involvement in its dissemination. Farmers obtain most of their information from each other. The time and resources they devote to acquiring it vary considerably. Also, a farmer does not consider the potential benefits to others when he decides how much information to obtain. It follows that, from a social point of view, farmers tend to underinvest in information ac- quisition. The availability of public sources of information tends to lower the cost of acquiring it for all farmers, and thereby increases social welfare if the cost of the public program is not too high (Feder and Slade 1984a). The main public channel for spreading agricultural knowledge is usually the extension service. Public spending on agricultural exten- sion is large. Judd, Boyce, and Evenson (1983) estimate that it totaled $3.5 billion worldwide in 1980. In 1974-84, investment in agricultural extension in projects financed by the World Bank was $2.3 billion. However, many analysts doubt that extension services can bring the big increases in agricultural productivity promised by the growing quantity of new or modified technology emerging from agricultural research institutions. Extension services in developing countries are often criticized for: (a) lack of staff training and incentives and channels for updating agents' knowledge; (b) inefficient organizational structures that pre- vent adequate supervision of field workers; (c) requirements for staff to perform tasks other than spreading information, such as collecting data; (d) staff shortages; (e) absence of organized feedback about farmer problems from fieldworkers to researchers. Accordingly, agencies have been paying more attention to improv- ing the management and efficiency of extension systems. One result has been the Training and Visit (T&v) extension system (described in 140 Research Observer 1, no. 2 (July 1986) Benor and Baxter 1984). T&V was originally tested in Turkey in the late 1960s. It has been introduced during the past ten years in more than forty developing countries, in many cases as nationwide systems, and often with the assistance of the World Bank. It has been most widely adopted since 1977 in India, replacing the system of multipur- pose village-level workers. With T&V, staff deal exclusively with extension work. They are organized in a single line of command, so at each rank an officer has few enough staff under him to allow effective and personal guidance, supervision, and training. At the bottom of the hierarchy are the village extension workers (VEWS). They usually cover areas containing 700-800 farming families, divided into about eight groups. In each group, about 10 percent are chosen as "contact farmers." The VEW visits each of the eight farmers' groups once every two weeks, on a specified day. These visits mostly take place in the fields of contact farmers, but other farmers are expected to participate. Occasionally, the VEW may organize a large group meeting, but most of his contacts are with small groups or individuals. The fixed schedule of visits helps the VEW's superiors to supervise his work and also encourages interest and confidence among farmers. To be effective, the T&V system aims to focus advice on the main crops grown and the most important farming methods. Simple meth- ods that do not need money are promoted first, so that all farmers can benefit. Contact farmers are expected to adopt (or at least try) recom- mended practices and transmit them to other farmers. Because T&V involves a high ratio of extension workers to farmers, it is relatively expensive. Much research has already been done on T&V (see, for example, Cernea 1981; Howell 1982a, 1982b, 1983, 1984; von Blanck- enburg 1982; Jaiswal 1983; Singh 1983; Moore 1984). Their opinions vary. Some observers argue that intensive personal contact between farmers and extension agents is unnecessarily expensive and that good results could be obtained by using written materials and the media. In many countries the mass media supplement direct contacts between farmers and extension workers (Perraton 1983). In some areas, they are the main form of diffusing public information. However, advo- cates of personal contact argue that the media do not give a fast feedback from farmers, nor can they adequately tailor their advice to local circumstances. Other skeptics of T&V argue that it is undesirable to separate advisory work from the supply of inputs. The counterar- gument is that the provision of inputs will tend to dominate the agents' activities; fieldworkers should, however, coordinate their ad- vice with the suppliers of inputs. Other criticisms of T&V are (a) that it is hard to implement in areas of extensive agriculture, because it requires well-organized farmers' groups and frequent personal contact; and (b) that T&V iS biased Gershon Feder and Roger Slade 141 toward wealthier farmers. These arguments are seldom informed by suitable empirical evidence. Few authors, therefore, fail to mention the need for objective empirical information with which the effects of T&V can be assessed. This paper assesses those effects by using farm survey data from India. First, it reviews the sources of data, then goes on to compare the importance of different channels for informing farmers about new or improved farming methods. The next section examines the charac- teristics of contact farmers, and the biases inherent in their selection. The following section analyzes the frequency with which contact and noncontact farmers are reached by extension and some of the factors that determine these frequencies. The paper then looks at the effect of T&V on farmers' knowledge. Subsequently, an econometric estimation of the impact of T&V on farm productivity is undertaken. Conclusions follow a cost-benefit analysis of T&V. Empirical This paper draws on two main sources of data. The first is a series Sources of sample surveys from India, spanning four consecutive crop seasons in two districts in the state of Haryana (Jind and Karnal) served by a T&V extension system, and two crop seasons for Kairana Tehsil (part of Muzaffarnagar district) in the neighboring state of Uttar Pradesh. In Jind and Karnal districts, two random samples, consisting of roughly the same numbers of contact and noncontact farmers, were chosen. The sample from Kairana comprised only noncontact farmers as the T&V system had not yet been introduced there. Respondents were interviewed twice in each season and repeatedly assured that the surveys were unconnected with the extension system or any other government department. Because Jind is much drier than Karnal and Kairana, comparative analysis is mainly confined to the latter two areas. It covers both the rainy and dry growing seasons of 1982-83, for which comparable data are available. The second source of data is monitoring and evaluation (M&E) reports from seven states in India over a number of years. The data are comparable across states because all M&E units use the same sampling design, definitions, and questionnaire (Slade and Feder 1985). The data in this study do not allow conclusive assessment of the relative merits of T&V extension and those methods based on the mass media. Neither can they illustrate the potential of T&V in areas of extensive agriculture, because Indian conditions are different. T&V To assess the various channels of spreading information, it is useful Operations to review how farmers rank them. Table 1 shows that contact farmers consider extension to be their primary source of information. A signi- 142 Research Observer 1, no. 2 (July 1986) Table 1. Sources of Information forFarmers in Northwest India, by Ranked Preference, 1983 (percent) Karnal District, Haryana, Kairana Tebsil Ranked Contact Noncontact Uttar Pradesh' Source preference farmers farmers (allfarmers) Individual advice First 87 19 2 from VEW Second 1 9 1 Third 1 4 1 Advice from contact First 1 16 1 farmer Second 3 4 0 Third 1 3 0 Advice from other First 9 47 82 farmersb Second 36 33 9 Third 21 10 3 Demonstration/field First 0 0 1 days Second 10 2 28 Third 3 1 8 Agricultural radio First 1 10 9 programs Second 28 27 38 Third 39 34 32 Salesmen and agency First 0 8 3 officials Second 15 17 17 Third 18 21 17 Research personnel First 1 0 0 Second 1 2 0 Third 9 7 0 Journals First 1 0 0 Second 1 3 7 Third 5 7 13 Other First 0 1 1 Second 4 3 1 Third 5 13 26 Note: For each source of information, farmers were asked to say whether they regarded it as a primary, secondary, or tertiary source. a. Karnal District is serviced by T&V extension, Kairana Tehsil by a different and older extension system. b. Other farmers could be contact farmers. Not all contact farmers are known as such to noncontact farmers. Source: Feder and Slade 1985. ficant proportion of other farmers, 19 percent in Karnal, shares that view. However, about three-quarters of the farmers in Karnal, and rather more in Kairana, say that other farmers are their main source of advice. Radio programs tend to be a secondary or tertiary source, probably because farmers prefer to rely on direct and personal con- tacts, who can clarify points and answer additional questions. The minor role played by written materials is compatible with the area's Gershon Feder and Roger Slade 143 Table 2. Relative Importance of Extension Workers and Other Farmers as Sources of Information for Wheat Farmers in Kairana and Karnal, Rabi 1982 Type of farmer and practice Smallfarms Large farms Total Non-T&V farmer Less expensive practices 0.00 0.07 0.04 More expensive practices 0.00 0.16 0.09 T&V noncontactfarmer Less expensive practices 0.20 0.36 0.27 More expensive practices 0.28 0.68 0.47 T&V contact farmer Less expensive practices 3.51 4.27 3.98 More expensive practices 4.43 7.50 5.14 Note: Data are ratios of the number of times a VEW is cited as the main source of informa- tion to the number of times that "other farmers" is cited. Less expensive practices include choice of seed variety, seeding rate, and spacing; more expensive practices include use of phosphate, potash, and zinc, seed treatment against termites and disease, and timing of nitro- gen fertilizer application. Source: Feder, Slade, and Sundaram 1986. low rate of literacy (35 percent). However, even in an area of high literacy (and very thin extension services) in Thailand, farmers used the radio much more than written sources (Hutanuwatr and others 1982). These findings highlight the importance of communication among farmers. They also explain why several extension systems concentrate on a relatively small number of farmers, who are expected to transmit information to the rest. The choice of information source tends to be affected by the complexity or cost of the farming methods being learned. Many farmers regard firsthand or specialized information sources as more accurate and reliable (Feder and Slade 1986, Howell 1984). Table 2 compares the ratio of farmers who have learned com- plex or expensive practices from extension agents (firsthand source) to that of farmers who have learned from other farmers (secondhand source). The message is clear: the easier it is to consult extension agents, the more likely they are to be the source of information. And irrespective of extension availability, the more complex or risky prac- tices tend to be learned more often from extension. When spreading the more complex methods of farming, therefore, agents should try to maximize their direct contact with farmers. Characteristics Contact farmers receive a direct and regular flow of information of Contact from extension agents, so their selection is a matter of considerable Farmers importance. Although their potential for leadership is the main criter- ion, other considerations should be kept in mind. For instance: 144 Research Observer 1, no. 2 (July 1986) Contact farmers must be willing to try out practices recommended by the extension workers and be prepared to have other farmers visit their fields. But they should not be the community's more progressive farmers who are usually regarded as exceptional and their neighbors tend not to follow them. On the other hand, very weak farmers tend to be slow in adopting new methods. Further- more, the contact farmers must be of good standing in their community so that their views on new practices will be respected by other farmers (Benor and Harrison 1977, pp. 13, 14). Evidence suggests that there is a large overlap between the qualities of "opinion leaders" and those of "innovators," or fast adopters (Kilvin and others 1971). Traditionally, extension workers have tended to concentrate on the well-to-do farmers, because their efforts were more likely to produce an immediate and visible impact and because wealthier farmers could offer them personal benefits (meals, accom- modation, produce). This bias has often been noted in the literature (Chambers 1976, Cernea 1981, and von Blanckenburg 1982). Some studies have argued that, on efficiency grounds, scarce extension re- Table 3. Factors Affecting the Selection of Contact Farmers Coejficient Jind Karnal Pooled Factor District District sample Farm size .0275 .0339* .0317 (3.471) (3.410) (5.154) Education (years) .0397 .0573' .0453* (1.637) (2.227) (2.593) Dummy variables Ownership of tubewell .0570* .8987- .3956* (3.016) (1.949) (2.250) Membership in village .3416 .4374# .4090' institutions (1.443) (1.524) (2.260) Previous participation 1.042* .8262'¢ .8172 - in agriculture training (2.180) (2.213) (2.851) Likelihood ratio statistic 60.55 44.96 99.23 Number of observations 599 361 960 Significant at 5 percent (one-tailed) level of significance. Significant at 7 percent (one-tailed) level of significance. Note: The results in this table were obtained from a logit analysis in which the probability of a farmer's being selected as a contact farmer is related to socioeconomic explanatory vari- ables. Numbers in parentheses are asymptotic t values. a. No estimate of the constant is presented; the sample is not random with respect to the proportions of contact and noncontact farmers, and the estimated constant is therefore biased. The other coefficients are not affected, however (McFadden, personal communication). Source: Feder and Slade 1984b. Gershon Feder and Roger Slade 145 sources should be used to greatest effect-implying a focus on larger, and usually richer, farmers (Welch 1979). However, this bias has disadvantages as well. Smaller and poorer farmers may not be con- vinced that the new practices are better: to them, a better performance by contact farmers could simply reflect the latter's greater wealth. There are other complications as well. Extension workers are re- sponsible for choosing contact farmers in their area, and personal preferences cannot be eliminated. As in many extension systems, agents tend to favor the wealthy and influential, even if they are not suitable as communicators of innovations. It is sometimes necessary to avoid antagonizing such powerful individuals, who may otherwise make extension work difficult to carry out. Furthermore, the choice of contact farmers will vary according to how far agents understand the extension system and how well they have been trained. Hoeper (1983) has shown that the selection of contact farmers varies considerably. In a study on India, Feder and Slade (1984b) found that the caste com- position of two groups of contact and noncontact farmers was almost identical, but that the farmers chosen as contacts tended to be weal- thier, more educated, with better irrigation facilities, and of higher social status (see Table 3). However, although farmers owning less than two acres were underrepresented in the contact group their share was not negligible-12 percent-compared with 30 percent in the general population. Relations The links between extension workers and contact farmers can be with Farmers regarded as the "supply" of extension services, because the T&V sys- tem requires these services to be provided regularly to contact farm- ers. Contact farmers, in turn, are expected to pass on this information to noncontact farmers. Moreover, it is not imperative that noncontact farmers have regular visits from extension workers; thus, the interac- tion between noncontact farmers and extension agents is likely to be determined by "demand" from noncontact farmers. However, the extension agents are expected to accommodate requests for informa- tion from all farmers. It is also expected that noncontact farmers will attend some meetings between extension agents and contact farmers. Some idea of the frequency of agents' visits to contact and noncon- tact farmers can be gleaned from Table 4, drawn from the reports of monitoring and evaluation surveys from seven states in India over several seasons. The reference period for visit frequencies is one month. But it is possible that noncontact farmers, since they do not receive regular visits, had a longer time horizon in mind, so the data may overstate the frequency of visits. The critical indicator is the percentage of farmers who report not seeing an extension agent. For contact farmers, this ranges from 1.2 to 146 Research Observer 1, no. 2 (July 1986) 34.7 percent; for noncontact farmers, from 21.4 to 59.2 percent. Across all seven states, the average percentage of "no-visits" reported by contact farmers is 15.4 percent (that is, about 85 percent of contact farmers were visited at least once in the reference month); 34.5 per- cent of the noncontact farmers reported no visits. Considering that some share of no-visits must be due to "normal friction" (Feder and Slade 1984a)-illness of extension workers, vacant posts, and unavail- ability of contact farmers-the actual supply of T&V services seems adequate relative to the potential supply. The demand for extension services (measured by noncontact farm- ers' interaction with agents) is significantly lower than the supply (measured by agents' visits to contact farmers). However, there is not necessarily much unused capacity, since the actual supply available to noncontact farmers must be less than that to contact farmers. Further- more, the demand for extension services is far higher in a T&V area: data for the Kairana area, which is not covered by T&V, show that between 89 and 97 percent of the farmers were not visited by (or did not seek out) the extension worker during the reference period (Feder and Slade 1986). In non-T&V areas, no distinction can be drawn between contact and noncontact farmers, so the Kairana figures could be the result of either low demand or low supply. It is known, however, that the ratio of extension workers to farmers is lower in non-T&v areas and that agents there have many duties other than extension. The small supply of extension in non-T&V areas could therefore increase the cost to the farmer of acquiring information from extension, thus reducing the contact between farmers and exten- sion workers. Table 5 summarizes data on visits and no-visits by farm size. Among both contact and noncontact farmers, there is a remarkable similarity between large and small farms. Among contact farmers, 15.9 percent of the small farms and 14.5 percent of the large farms were not visited, a difference of 1.4 percentage points. For noncontact farmers, the difference was only 3.2 points. While these differences are statistically significant at the 99 percent level (see Table 5), their size indicates that the bias toward large farmers is not enough to war- rant serious concern. Further evidence comes from Feder and Slade (1984b), showing that "area of land owned" has a positive but not significant effect on the probability that a contact farmer would be visited. Since the links between noncontact farmers and extension workers are probably demand-driven, the difference between large and small farmers may merely indicate-as predicted by theory (Feder and Slade 1984a)-the tendency of larger farmers to invest more in gathering information. Extension workers may visit farmers more often in the dry season. Data for both contact and noncontact farmers show that the inci- Gershon Feder and Roger Slade 147 Table 4. Frequency of VEW Visits to Contact and NoncontactFarmers in India, by State and Farm Size Contact farmers Noncontact farmers Monito ring or Percentage with visits Percentage with visits monitoring and in the past four weeks in the past four weeks evaluation Sample Sample State and farm size survey size 0 1 2 or more size 0 1 2 or more Haryana Rabi 1981-82 M Small farms, 251 17.1 14.0 68.9 333 48.6 27.9 23.3 Large farms 202 15.9 12.8 71.3 138 45.7 29.7 24.6 Kharif 1982-83 M Small farms 232 16.4* 17.2 66.4 309 40.4 35.9 23.7 Large farms 219 10.0 20.1 69.9 138 45.6 21.0 33.3' Kharif 1982-83 M&E Small farms 64 7.8 21.9 70.3 658 52.3 24.5 23.2 Large farms 64 6.2 17.2 76.6 267 49.8 21.3 28.9 Karnataka Rabi 1981-82 M Small farms 2,024 15.4' 17.3 67.3 1,482 30.2 21.8 48.0* Largefarms 1,143 12.3 20.6 67.1 530 32.6 26.4 41.0 Rabil981-82 M&E Small farms 86 17.4 16.3 66.3 395 40.3 21.8 37.9 Large farms 159 11.4 17.6 71.0 493 40.2 25.8 34.0 Kharif 1982-83 M Small farms 1,499 10.1 10.8 79.1 1,869 28.0 16.2 55.8 - Large farms 1,133 10.9 14.7 74.4 760 25.8 25.8 48.4 Kharif 1982-83 M&E Small farms 307 13.0 24.1 62.9 2,065 50.7 19.7 29.6 Large farms 235 8.9 27.2 63.8 712 50.6 20.5 28.9 Rabil982-83 M Small farms 1,280 13.0 21.6 65.4 1,621 28.2 24.6 47.2 Large farms 988 14.0 15.0 71.0' 698 27.8 22.7 49.5 Rabi 1982-83 M&E Small farms 69 18.8 23.2 58.0 606 57.9- 18.0 24.1 Largefarms 168 11.3 17.3 71.4' 544 48.7 20.8 30.5 Kharif 1983-84 M Small farms 1,157 20.4S 13.9 65.7 1,593 36.4:' 18.8 44.8 Large farms 944 15.4 18.6 66.0 648 31.3 19.4 49.3 Gujarat Kharif 1981-82 M Small farms 503 33.2 19.9 46.9 .. Large farms 328 28.6 21.6 49.7 .. Kharif 1981-82 M&E Small farms 298 8.0 19.5 72.5 .. Large farms 237 8.0 20.7 71.3 .. Rabil981-82 M Small farms 527 23.9 17.5 58.6 .. Large farms 308 20.1 14.0 65.9* .. Rabil981-82 M&E Small farms 208 8.6 14.9 76.5 .. Large farms 184 12.0 20.1 67.9 .. .. Kharif 1982-83 M Small farms 498 24.3 14.3 61.4 .. .. Large farms 337 19.3 14.2 66.5 .. .. 148 Research Observer 1, no. 2 (July 1986) Table 4 (continued) Contactfarmers Noncontact farmers Monitoring or Percentage witb visits Percentage with visits monitoring and evaatorion Sample in the pastfour weeks Sml in the pastfour weeks evaluation Sample Sample State and farm size survey size 0 1 2 or more size 0 1 2 or more Gujarat (continued) Kharif 1982-83 M&E Small farms 394 12.4 15.0 72.6 .. Large farms 375 13.1 20.3 66.6 .. Rabi 1982-83 M Small farms 481 21.9 18.8 59.3 .. Large farms 354 20.2 22.1 57.7 .. Assam Kharif 1982-83 M&E Small farms 394 27.3k 9.6 63.0 423 48.0* 22.7 29.3 Largefarms 332 11.2 18.0 70.8' 308 21.4 9.1 69.5S Maharashtra Rabil983-84 M Small farms 1,200 14.0* 12.2 73.8 .. Large farms 735 10.2 11.4 78.4r .. Bihar Kharif 1983-84 M Small farms 734 32.2 17.5 50.3 854 59.2 15.2 25.6 Large farms 352 34.7 16.8 48.5 249 55.0 16.1 28.9 Tamil Nadu Summer 1982 M Small farms 347 2.0 4.6 93.1 .. Large farms 83 1.2 8.4 90.4 .. Kbarif 1982-83 M Small farms 1,317 3.4 2.8 93.8 .. Large farms 248 2.8 3.6 93.6 .. ..= not available Significant at 5 percent level of significance. Note: Small farms were defined as less than 5.1 hectares in Haryana and Gujarat; 4.1 hectares in Tamil Nadu and Karnataka; 3.1 hectares in Madhya Pradesh and Bihar; and 2.1 hectares in Assam. Source: Feder, Slade, and Sundaram 1986. dence of no-visits during the dry season is significantly lower than in the rainy season, although the absolute difference is small. This result is consistent with Feder and Slade (1986), who show that the rate at which knowledge spreads tends to be higher for dry-season crops. These findings support the proposition that extension workers play a greater role in the dry season, although the cause may have more to do with the available technology and the riskiness of rainfed agricul- ture than with the efficiency of the extension system. As experience with the T&V system increases, so the pattern of extension visits changes. The proportion of contact farmers not visit- Gershon Feder and Roger Slade 149 Table 5. Summary of VEW Visits to Contact and Noncontact Farmers in India, by Farm Size (percent) Contact farmers Noncontact farmers One or more One or more Farm size No visits visits No visits visits Small 15.9 84.1 35.4 64.6 Large 14.5 85.5 32.2 67.8 Note: Data are based on evidence from large sample surveys conducted by monitoring and evaluation units; states, years, and seasons as in Table 4. Source: Feder, Slade, and Sundaram 1986. ed goes up significantly: among projects that are four or more years old, nearly one in five contact farmers was not visited (see Table 6 and Figure 1). This may be partly due to extension workers' replacing some contact farmers with others but not telling the former of the change. Conversely, the proportion of noncontact farmers not visited declines, from about 48 to 36 Figure 1 percent. This may be due to the fact that, over time, more Percentage 80 noncontact farmers come to of farmers not visited hear of T&V and take advan- 60 - tage of the service; T&V work- ers are expected to respond to 40 Noncontact queries from all farmers (Benor _ _ _ _ _ ~farmers and Baxter 1984). Because con- 20 - Contact tact farmers form only about farmers 10 percent of the farming com- _____________________ _--------- ,_, munity, the most important 1 2 3 4 finding is that the proportion Project life (years) of all farmers visited increases as projects mature. In short, although the T&V system is not without flaws, it does reach the majority of contact farmers regularly and, less regularly, a substantial proportion of noncontact farmers as well. Furthermore, the data do not support the contention that T&V has atrophied, leaving an empty structure and no change in extension operations (Jaiswal 1983). In the state of Haryana, even in unsettled times,' T&V still seemed to work better than the traditional system in neighboring Uttar Pradesh. Moore (1984), Jaiswal (1983) and other commentators have claimed that, in many parts of India covered by T&V, farmers see little benefit in the reformed system; that workers are not known to their clients; and that contact farmers fail to pass on information to others (and 150 Research Observer 1, no. 2 (July 1986) Table 6. Frequency of VEW Visits in India, by Project Life and Farm Size (percent) Contactfarmers Noncontactfarmers Project life One or more One or more andfarm size No visits visits No visits visits One year or less Small farms 5.0 95.0 48.7 51.3 Large farms 7.5 92.5 45.7 54.3 All farms 5.7 94.3 47.8 52.2 Two years Small farms 17.1 82.9 32.3 67.7 Large farms 13.9 86.1 36.3 63.7 All farms 16.0 84.0 33.8 66.2 Three years Small farms 14.3 85.7 28.1 71.9 Large farms 14.1 86.0 26.7 73.3 All farms 14.2 85.8 27.7 72.3 Four years or more Small farms 22.2 77.8 38.8 61.2 Large farms 14.3 85.7 28.1 71.9 All farms 19.4 80.6 36.1 63.9 Note: Data are based on monitoring and evaluation reports of state governments in India; states, years, and seasons as in Table 4. Source: Feder, Slade, and Sundaram 1986. may not even know that they are contact farmers). The data for the study area in Haryana do not support these contentions. From an original sample of 192 contact farmers selected at random from the extension lists in Karnal district, 175 turned out, on their own admis- sion, to be contact farmers. Most of the others said they had been contact farmers in the recent past. All farmers in the sample surveys in Karnal and Kairana were asked in 1983 if they had seen changes in the style of the extension system during the previous few seasons. Predictably, farmers in the Kairana area of Uttar Pradesh saw no change, as none had taken place.2 In Karnal (where T&V was introduced in 1979) almost all contact farmers were aware of a change in extension operations, and they thought the change beneficial. Among noncontact farmers, less than half had noticed a change, but most of those who did were favorably impressed. This relative lack of awareness of the change among noncontact farmers, particularly on small farms, suggests that earlier efforts to publicize the availability of extension advice had been inadequate. Almost all contact farmers and about half of the others knew the extension worker who visited their group. The comparable proportion Gershon Feder and Roger Slade 151 for the non-T&V area (Kairana) was little more than one-tenth. Simi- larly, 60 percent of noncontact farmers reported knowing at least one contact farmer in their area. More than half of all contact farmers claimed to have discussed extension advice with other farmers, while more than 30 percent of those noncontact farmers who had talked to extension workers also claimed to have passed on information ob- tained from them. T&V Effects Extension aims to increase farmers' knowledge about crops and on Farmers' cropping practices, obviously in the hope that additional knowledge Knowledge will lead to improved husbandry and thence to increased agricultural productivity. However, many other factors affect the adoption of tech- nology and output, and they cannot easily be disentangled from ex- tension (as will be discussed in the next section). This section com- pares the levels of knowledge (thus largely avoiding such complica- tions) among different groups of farmers, drawing on data from the sample surveys in Karnal (T&v area) and Kairana (non-T&v) during the rainy and dry seasons of 1982-83. The data show that for most practices not involving specialized technical knowledge or major expense, contact farmers under the T&V system learned mostly from the extension service. Noncontact farmers learned mostly from other farmers, including contact farmers. Where specialized technical knowledge was involved, all farmers tended to learn from knowledgeable primary sources, such as extension agents. This pattern suggests that the spread of knowledge about the more demanding practices is likely to be much faster in an area such as Karnal, which has ample extension staff, than in a less well-endowed area such as Kairana. During the sample surveys, farmers were also questioned about their knowledge of specific practices and when they first learned them. Knowledge is difficult to measure without a thorough examination of a respondent's understanding. For some practices this was possible; for others, detailed testing was beyond the time and resources avail- able. In such cases, however, it was possible to establish the farmers' awareness of a particular practice; a farmer who is unaware of a practice cannot, by definition, be familiar with its detail. The resulting data show the growth in the number of farmers who were aware of different technologies in 1978, the year before T&V extension was introduced in Haryana, and four years later. To increase the validity of comparisons between Karnal and Kairana, contact farmers in Kar- nal have been excluded from the analysis because they receive a disproportionate amount of extension advice and may also be differ- ent in other ways (as discussed earlier). The analysis employed two alternative standard specifications (lo- 152 Research Observer 1, no. 2 (July 1986) gistic and negative exponential) of the time path of growth in the spread of knowledge.3 The results showed that among ten practices for high yielding varieties (HYV) of paddy,4 the rate of growth in farmer knowledge was clearly faster in Karnal for only three of them. For two paddy practices, knowledge spread faster in Kairana. It is noteworthy, however, that two of the three practices that spread faster in Karnal involved considerable technical content and needed cash inputs. These results are consistent with the argument that such prac- tices are most commonly learned directly from extension agents. Where HYV wheat was concerned, knowledge spread faster in Karnal for nine out of the ten practices examined. These are interesting results, but they must be qualified. First, they are based on sample surveys, and all such surveys have a margin of error. Second, even when knowledge about a practice has increased, it may not be useful or profitable to farmers. Consequently, these results do not prove whether gains in yields result from the observed increas- es in knowledge or whether such gains outweigh the increased costs of T&V extension. These issues are discussed later. Nevertheless, the results suggest that T&V extension in Karnal speeded up the spread of knowledge for almost all recommended practices for HYV wheat and several important practices for HYV rice. Such results are consistent with Karnal's significantly greater exten- sion activity. They are also consistent with other survey findings that knowledge spreads faster among contact than noncontact farmers. As this article has already shown, contact farmers have more direct links with extension workers and should therefore (other things being equal) be more knowledgeable. However, as the contact group is not necessarily representative of all farmers, its superior knowledge may be the result of other factors. The process by which extension affects crop yields involves many Information variables. If extension efforts are successful, however, this success Sources and must eventually result in increased output per unit of input, reduced Farm costs per unit of output, or both. .o *i. Since the contact point between the extension system and the farmer is the village extension worker, it is essential that the VEW is "better" than other sources of information. A testable hypothesis is therefore implied: other things being equal, farmers whose main information source is the extension worker will have higher productivity or yields than those who rely on other sources. Of course, there may be some systematic relationship between farmers who use extension as a main source of information and their other inherent attributes (such as intel- ligence) which make them better farmers who obtain higher yields. Drawing again on the state M&E reports in India, we use data on Gershon Feder and Roger Slade 153 crop yields in the rainy and dry seasons classified by information source. For the rainy season, we use paddy yields; for the dry season, wheat (under both irrigated and unirrigated conditions). State average yields were calculated by applying weights based on the sample sizes for irrigated and unirrigated farms and for contact and noncontact farmers. The resulting overall average for each state was set equal to 100. Each subset of yields was then expressed as an index number relative to the state average. This conversion permits paddy and wheat yields to be compared (since they differ in absolute magnitudes) and minimizes differences in agroclimatic and socio- economic factors between states. The net result is a series of index numbers that are comparable across states, crops, and cropping sea- sons. Table 7 summarizes the data for irrigated, unirrigated, and all farms, classified by main source of farmer information. Farmers whose main source of information is the VEW have the highest yield index of 114.5. This is followed by those whose source is other farmers; their yield index is close to the average. Other sources (such as radio, demonstration days, sales personnel) have a lower yield index of 95.77. Farmers who receive "no advice" had an index of only 86.11. Those using extension workers as the main source of information seem to have yields that differ substantially from all other sources, but the difference between those using other farmers and other sources is much smaller. All three, however, have higher yields Table 7. Yield Indexes, by Type of Farm and Main Source of Information Other Other No Type offarm VEW farmers sources advice Irrigated farms 116.25 89.52 92.64 93.22 (13) (13) (13) (13) Unirrigated farms 114.75 101.90 103.29 79.88 (13) (13) (13) (13) Allfarms 114.50 99.08 95.77 86.11 (15) (15) (15) (15) Note: Data are based on monitoring and evaluation reports from seven state governments in India; states, years, and seasons as in Table 4. Figures in parentheses indicate sample size. Sample sizes differ because, for two states, data classified by irrigated and unirrigated farms were not available. The actual number of farmers in the sample is more than 1,500; the sample sizes in the table refer to the number of average yield index figures and hence represent a mean of means. a. One state, in one cropping season, had an unduly high yield figure, and the sample base was extremely low in relation to the rest and hence was significant in the computation of weighted average yields for all farms. However, in computing the average across unirrigated farms in all states, this number receives equal weighting. Hence, this particular figure should be considered an overestimate. Source: Feder, Slade, and Sundaram 1986. 154 Research Observer 1, no. 2 (July 1986) than those receiving no advice. The figures shown in Table 7 were rigorously tested with econometric techniques, and the conclusions confirmed these results (Feder, Slade, and Sundaram, 1986). One difficulty with yield comparisons is that no allowance is made for differences in, for example, soil types, farmer attributes, or extent of irrigation, which may also contribute to the variability of yields. We therefore made a deeper analysis, using the farm-level data from the Karnal and Kairana sample surveys. Although this analysis takes into account certain differences be- tween the two areas, Karnal and Kairana are similar in many respects. They lie on opposite banks of the Jamuna river, are flat, and have light alluvial soils. Average annual rainfall in Karnal is 803 millime- ters, and in Kairana 794 millimeters. Both districts are heavily irrigat- ed: in Karnal 74 percent of the net cropped area, in Kairana 84 percent. Linguistically and ethnically, the two are similar. In the dry season, wheat is the dominant crop in both areas. In the rainy season, however, paddy is the main crop in Karnal, sugarcane being less important; in Kairana, it is the other way round. In the state of Uttar Pradesh, of which Kairana is one of the most western parts, the extension system at the time of the study consisted of the traditional network of village-level workers (VLWs) administered by the Community Development Department. These workers are re- sponsible not only for providing extension advice but also for regulat- ing the supply of inputs and credit and the administration of other subsidy and incentive schemes. They are also the link between the rural population and several other government agencies. In 1981, there were some 140 VLWS in Kairana: one worker for every 6.1 villages. In Karnal the ratio was 4.7. In terms of numbers of people, there was one village worker for every 11,500 rural dwellers in Kair- ana, and one for every 7,400 in Karnal. In Kairana VLWs are supple- mented by staff from the Department of Agriculture, working mainly under the aegis of special crop programs. Karnal's extension system was reformed in late 1979. The reorgani- zation reduced the ratio of villages to VEWS from an initial 6.05 (similar to the ratio in Kairana at the time of the study) to 4.7. It created new senior positions-for example, supervisors and special- ists. VEWS were relieved of nonextension duties. By March 1983, 99 percent of VEW positions and 88 percent of specialist positions (techni- cal specialists of intermediate rank) were filled. However, in May 1982, 25 percent of the positions for agricultural extension officers (supervisors of VEWS) were still vacant.5 The study included only high yielding varieties of wheat and paddy; traditional varieties are rarely grown in Karnal. Initially, yield differ- ences between the two regions for the two main crops, wheat and paddy, were estimated. These estimates took account of differences in Gershon Feder and Roger Slade 155 the quantities of variable and fixed inputs, the types of soils, human capital, and irrigation (both quantity and quality).6 It is thus tempting to assume that any yield differences between Karnal and Kairana in 1983 were wholly attributable to differences in the extension system. However, this need not be true if other systematic (but unobserved) factors differentiated the two areas, or if yield differences had been significant in 1979. To minimize the possibility of misinterpreting the results, the control sample (the non-T&V case) was from that part of Muzaffarnagar district (Kairana Tehsil) next door to Karnal district. Thus, the villages in Kairana's sample were no more than 30 miles from the center of Karnal. Spillover effects were minimal because farm- ers from the two regions were not regularly in touch with each other. These precautions could nonetheless fail to account for some fixed and systematic differences between the areas, or possibly for the fact that knowledge spread more rapidly in one of the areas before T&V began. In such cases, even the adjusted yields in 1979 would not be equal, so further adjustments would have to be made. Ideally, what would be needed would be a complete sample for 1979, so that it and the 1982-83 sample could have been subject to detailed econometric analysis. It would then have been possible to test the hypothesis that the 1982-83 residual yield difference was larger than the 1979 differ- ence. Any difference between these two levels would then have been due to T&V extension. Unfortunately, no such detailed sample from Karnal and Kairana was available for 1979. However, there were some data derived from the seasonal crop-cutting estimates. These data have several deficiencies: (a) the sample sizes for subdistricts are small; (b) they do not differentiate between irrigated and unirrigated conditions or between high yielding and traditional varieties, whereas the 1982-83 data focus only on high yielding varieties under irrigated conditions; (c) they provide no informa- tion on inputs or other similar variables that might explain the differ- ences in output; (d) in any one year they include random elements that fluctuate over time, such as disease or bad weather. To overcome these deficiencies, the data were adjusted to derive mean yields for 1979 that were comparable to the sample used in the detailed analysis of the 1982-83 data. Econometrically estimated relations were used to calculate the residual yield difference between Karnal and Kair- ana in 1979. This difference was subtracted from the one estimated from the 1982-83 sample, and the increased yield (if any) attributable to T&V extension was calculated on several different assumptions. The results suggest that in 1982-83, after three years of T&V exten- sion and holding all inputs constant, HYV wheat yields in Karnal were 8.9 percent higher than in Kairana. However, this estimate excludes any difference that existed before the more intensive T&V extension system was introduced.7 The productivity difference between Karnal 156 Research Observer 1, no. 2 (July 1986) and Kairana in 1979 (before the T&V system) was between 1.6 and 3.0 percent; it must be subtracted from the yield difference in 1982-83. The difference in yields, about 7 percent over the first three years of implementation, is attributable to T&V extension. Since Karnal and Kairana were both quite advanced even before T&V-almost all farmers used HYVs and nitrogenous fertilizers-what was the source of this gain in productivity? The study suggests that the gains came partly from the spread of improved production meth- ods, such as the timing of various farm operations. However, farmers and extension agents in the study areas also laid stress on the ability of extension workers to spot local problems, seek help from special- ists, and then give farmers the right advice on what to do to minimize yield losses. In areas without a link to expert advice, unforeseen and localized production problems cause bigger losses. The value of any increase in farm output attributable to T&V Cost-Benefit extension must be set against the additional costs incurred to make Analysis the extra output possible. Although the cost-benefit analysis of T&V in Karnal was made after the fact, a complete series of either costs or benefits for the whole of the project was not available. We were therefore obliged to make several as- sumptions. To estimate costs, we used Figure 2 data on actual costs incurred during In y the first four years of the T&V system, as well as projections made at the time of project appraisal. These were ad- With T& justed to constant 1979 prices. Where l the project's life was assumed to be shorter than that of physical structures and equipment, appropriate residual Without T&V values were calculated and deducted from cost. To estimate benefits, we 1979-80 1982-83 T' drew on the figures (presented earlier) a = productivity gain due to extension in 1982-83 on the extra yield attributable to T&V Y = yield t=time extension in the third year of the proj- In = natural logarithm ect. However, it is reasonable to ex- pect gains to continue beyond the third year. As there were no data with which to estimate such additional gains, we constructed a dy- namic model to simulate later changes in productivity, both with and without T&V (see Figure 2). In the absence of T&V extension, the average yield is assumed to grow at a constant rate, while the introduction of T&V initially accel- erates that growth by informing farmers of better farming methods and how to use them. Once that phase is over, however, productivity Gershon Feder and Roger Slade 157 growth will slow down. After a certain number of years (T-'), the average yield will be the same whether or not T&V was implemented. If a T&V project can be stopped as soon as marginal benefits are equal to marginal costs, that would maximize its efficiency. The model was used to estimate the project's benefits for varying periods. The results show, with a high degree of statistical confidence, that the internal rate of return on a project lasting for T- years exceeds 15 percent; on a project lasting for the most efficient period, it would exceed 20 percent. These calculations, it should be emphasized, refer to incremental costs and benefits. Thus, they reflect the returns to intensifying and improving the extension system, but not its overall return. Theoretically, the overall return could be low while that on incremental investment was high. The data, however, cannot be used to infer the overall return; that would need information on productiv- ity without any extension, which was not available. Knowledge of the overall return would be essential if there was an option to disband the extension system altogether, including those parts that existed before the introduction of T&V. In practice such an option rarely exists, because of bureaucratic rigidities. Conclusions This paper has analyzed some key hypotheses about the effects of i&v extension. The results, based on data from India, show that T&V greatly increases the number of contacts between farmers and exten- sion workers, and the proportion of farmers reached increases the longer the T&V system operates. Extension agents were found to be an important source of knowledge about new farming practices, particu- larly when these practices are complex and expensive. The paper shows that T&V led to significant increases in yields of a major crop in one area covered by a detailed study. Even after allowing for many other factors that help explain differences between farmers' perfor- mance, yield differences of about 7 percent over three years remain. When the costs of T&V are set against the value of the increased production, the project produced internal rates of return of at least 15 percent. These benefits seem to be due to improvements in overall farm management rather than to the induced use of more (or new) inputs. More specifically, the results suggest that the greater availabili- ty of extension workers and, through them, the advice of specialists substantially improved the ability of farmers to respond to local prob- lems. Moreover, these results pertain to an area where most farmers were already using high yielding varieties and fertilizers before the extension system was reformed. In less advanced areas several studies cited by Herdt and Capule (1983) show that the quality of extension affects farmers' adoption of modern varieties and inputs. As the results concerning productivity gains come from one of 158 Research Observer 1, no. 2 (July 1986) India's more advanced agricultural regions, it might be argued that they do not apply to less advanced areas. Moreover, it has been observed, for example, that profitable innovations spread fast during the green revolution in northwest India without intensive extension work. Nonetheless, the results of this study show that if extension produces even a small gain in one major crop, the extra cost is justified. The review by Herdt and Capule (1983) also cites several studies showing that extension can accelerate the spread of innova- tions such as high yielding varieties. Thus it seems that the basic elements of new and profitable technology may spread fast naturally, but the spread of more complicated methods and the adaptation of technology to local circumstances will be significantly improved if farmers have access to specific and up-to-date advice. Where there are not enough good and well-organized advisers, extension is likely to be much less effective. Moreover, in areas where appropriate technology is not yet available, it may be inadvisable to invest in expanding extension services. This article reviews the rationale for public sector involvement in the dissemination Abstract of technological information to farmers, concluding that free markets do not fully satisfy farmers' information needs, and that government support is justified. Agricul- tural extension is a principal way that governments can disseminate information, and the World Bank is financing many extension projects throughout the developing world. One specific approach to extension adopted in many Bank extension projects is the Training and Visit (T&V) system. Data from a Bank-sponsored survey in northwest India and from monitoring and evaluation reports issued by several Indian states are used in this article to evaluate T&v extension operations and their impact. Extension agents' interaction with farmers is found to be more intensive and more significant as a source of information in areas covered by T&v extension than in areas with a different extension system. The yield levels of farmers whose main source of informa- tion was the T&V extension agent are also shown to be higher. In one case study, the incremental investment in T&V extension is shown to be likely to generate at least a 15 to 20 percent rate of return. 1. During 1982 and 1983 when the surveys in Karnal and Jind districts of Haryana Notes were conducted, there were widespread and disruptive transfers of field staff and a high rate of turnover among senior management. 2. The extension system in Uttar Pradesh is expected to be reformed along T&V lines starting in 1986. 3. Details are to be found in Feder and Slade (1986). 4. The term "paddy" is used to describe rice, whether growing or harvested, before the milling process. (Farmers grow paddy; millers produce rice.) 5. There were also other significant difficulties during 1982 and 1983. For example, state and regional committees charged with defining and programming technical recommendations were either not convened or worked only erratically. 6. A more extensive discussion of the analysis and results reported in the rest of this paper can be found in Feder, Lau, and Slade (1985). 7. The results for HYV paddy were not statistically significant. Gershon Feder and Roger Slade 159 References Benor, Daniel, and James Q. Harrison. 1977. Agricultural Extension: The Training and Visit System. Washington, D.C.: World Bank. Benor, Daniel, and Michael Baxter. 1984. Training and Visit Extension. Washington, D.C.: World Bank. Cernea, Michael. 1981. "Sociological Dimensions of Extension Organization: The Introduction of the T&V System in India." In Bruce R. Crouch and Shankarian Chamala, eds. Extension Education and Rural Development, vol. 2: International Experience in Strategies for Planned Change. Chichester: Wiley. Chambers, Robert. 1976. Two Frontiers in Rural Management: Agricultural Extension and Managing the Exploitation of Communal Natural Resources. Institute of Devel- opment Studies Communications Series 113. Brighton, England: Institute of Develop- ment Studies, University of Sussex. Feder, Gershon, and Roger H. Slade. 1984a. "The Acquisition of Information and the Adoption of New Technology." American Journal of Agricultural Economics 66, no. 3 (August): 312-20. _ 1984b. "Contact Farmer Selection and Extension Visits: The Training and Visit Extension System in Haryana, India." Quarterly Journal of International Agriculture 23, no. 1 (January/March): 6-21. -_____ 1985. "The Role of Public Policy in the Diffusion of New Agricultural Technology." American Journal of Agricultural Economics 67, no. 2 (May): 423-28. _____. 1986. "A Comparative Analysis of Some Aspects of the Training and Visit System of Agricultural Extension in India." Journal of Development Studies 22, no. 2 (April): 407-28. Feder, Gershon, Lawrence J. Lau, and Roger H. Slade. 1985. The Impact of Agricultural Extension: A Case Study of the Training and Visit System in Haryana, India. World Bank Staff Working Paper 756. Washington, D.C. Feder, Gershon, Roger H. Slade, and A.K. Sundaram. 1986. "The Training and Visit Extension System: An Analysis of Operations and Effects." Agricultural Administra- tion 22, no. 2: 407-28. Hall, D. C. 1977. An Economic and Institutional Evaluation of Integrated Pest Manage- ment. Environmental Protection Agency Report 68-01-2982. Washington, D.C.: U.S. Government Printing Office. Herdt, R. W., and C. Capule. 1983. Adoption, Spread and Production Impact of Modern Rice Varieties in Asia. Los Banos, Philippines: International Rice Research Institute. Hoeper, Bernhard. 1983. "Selected Results of the Agriculture Development Officers and Village Extension Workers Survey in Jind, Karnal and Mahendragarh Districts, Haryana, India." Organizational and Methodological Variables of the Training and Visit System of Extension, Working and Discussion Note 1. Berlin: Institute of Socio-Economics of Agricultural Development. Howell, John. 1982a. Managing Agricultural Extension: The T&V System in Practice. Discussion Paper 8. London: Overseas Development Institute, Agricultural Adminis- tration Network. _____. 1982b. Responses to Discussion Paper No. 8, Managing Agricultural Extension: The T&V System in Practice. Newsletter 9. London: Overseas Development Institute, Agricultural Administration Network. . 1983. Strategy and Practice in the T&V System of Agricultural Extension. Discussion Paper 10. London: Overseas Development Institute, Agricultural Admin- istration Network. -_____ 1984. Small Farmer Services in India. Working Paper 13. London: Overseas Development Institute, Agricultural Administration Network. 160 Research Observer 1, no. 2 (July 1986) Hutanuwatr, Narong, and others. 1982. "Socio-economic Constraints in Rain-fed Agricultural Production in the Lower North-East Thailand." Bangkok: Kohn Kaen University, Faculty of Agriculture. Jaiswal, N. K. 1983. "Transfer of TechnoLogy under T&V - Problem Identification." In Background Papers: Workshop on Managemenzt of Transfer of Farm Technology under the Training and Visit System. Hyderabad: National Institute for Rural Development. Judd, M. A., J. K. Boyce, and R. E. Evenson. 1983. Investing in Agricultural Supply. Economic Growth Center Discussion Paper 442. New Haven, Conn.: Yale University. Kilvin, Joseph, and others. 1971. Innovation in Rural India. Bowling Green, Ohio: Bowling Green State University Press. Moore, Michael. 1984. "Institutional Development, The World Bank, and India's New Agricultural Extension Program." Journal of Development Studies 20, no. 4 (July): 303-17. Orivel, F. 1983. "The Impact of Agricultural Extension: A Review of the Literature." In H. Perraton and others, eds. Basic Education and Agricultural Extension. World Bank Staff Working Paper 564. Washington, D.C. Perraton, H. 1983. "Mass Media, Basic Education and Agricultural Extension." In H. Perraton and others, eds. Basic Education and Agriculture Extension. World Bank Staff Working Paper 564. Washington, D.C. Singh, R. N. 1983. "T&V in Chambal Command Area (Kota District): Some Observa- tions." In Background Papers: Workshop on Management of Transfer of Farm Technology under the Training and Visit System. Hyderabad: National Institute for Rural Development. Slade, Roger H., and Gershon Feder. 1985. "The Monitoring and Evaluation of Training and Visit Extension in India: A Manual of Instruction." Washington, D.C.: World Bank. Processed. von Blanckenburg, Peter. 1982. "The Training and Visit System in Agricultural Exten- sion: A Review of First Experiences." Quarterly Journal of International Agriculture 21, no. 1 (January/March): 6-25. Welch. Finis. 1979. "The Role of Investments in Human Capital in Agriculture." In Theodore W. Schultz, ed. Distortion of Agricultural Incentives. Bloomington: Indiana University Press. Gershon Feder and Roger Slade 161