University Training for "Gramsevaks" in India: An Example of Recurrent Education in a Low-Income Country

Recently it has become fashionable to discuss a concept of lifelong or recurrent education. Recurrent education may be viewed as encompassing two types of decisions. The first involves the postponement of educational investments, and the second, educational investments in older individuals.' This paper concentrates on the second type, using as the criteria of evaluation economic efficiency and equity. Unlike most of the discussion in the available literature, this paper employs both criteria to evaluate a scheme of recurrent education in a low-income country. In 1961 at the request of the government of India, G. B. Pant University of Agriculture and Technology initiated a special 2-year bachelor of science agriculture program for gramsevaks, or village-level workers.2 To qualify for admission, gramsevaks had to meet these minimum qualifications: first, a pass in both the high school and intermediate examinations;3 second, a pass in the 2-year diploma course for gramsevaks; and third, 5 years of experience in extension or its equivalent. The program represents a departure from the normal 3-year undergraduate program in agriculture: first, by lowering its admission standards with respect to previous academic work; and second, by decreasing the degree program by 1 year. The 1-year reduction in the degree program is tantamount to substituting 5 years of experience in agricultural extension for 1 year of university course work.


Introduction
Recently it has become fashionable to discuss a concept of lifelong or recurrent education. Recurrent education may be viewed as encompassing two types of decisions. The first involves the postponement of educational investments, and the second, educational investments in older individuals.' This paper concentrates on the second type, using as the criteria of evaluation economic efficiency and equity. Unlike most of the discussion in the available literature, this paper employs both criteria to evaluate a scheme of recurrent education in a low-income country.
In 1961 at the request of the government of India, G. B. Pant University of Agriculture and Technology initiated a special 2-year bachelor of science agriculture program for gramsevaks, or village-level workers.2 To qualify for admission, gramsevaks had to meet these minimum qualifications: first, a pass in both the high school and intermediate examinations;3 second, a pass in the 2-year diploma course for gramsevaks; and third, 5 years of experience in extension or its equivalent. The program represents a departure from the normal 3-year undergraduate program in agriculture: first, by lowering its admission standards with respect to previous academic work; and second, by decreasing the degree program by 1 year. The 1-year reduction in the degree program is tantamount to substituting 5 years of experience in agricultural extension for 1 year of university course work.
Socioeconomic characteristics of students expecting to graduate in July 1971 were gathered through a 10 percent random sample of seniors. This sample included 40 students.

The Model
The efficiency criterion utilized is the internal social rate of return, a discount rate equating the present value of a stream of net benefits accruing to society from the investment to the present value of resources, with alternative uses, utilized in the investment. Thus, where PVVsB equals the present value of the stream of net social benefits, and P Vsc, the present value of a stream of social costs. Since both P VsB and P Vsc are incurred over time, equation ( (1 + SR) j, in which SR, the internal social rate of return, equates net present value of social benefits, SBE', to the present value of social costs, SC1. The stream of benefits to society involves (a) those that result in net direct increases in national income and (b) those that involve net increases in social welfare customarily omitted from national income accounts. Net increases in national income resulting from educational investments are measured by the increase in labor productivity and the contribution that education makes to the efficient use of technology.7 Indirect social benefits or externalities omitted from national income accounts include, to name only a few, ( reflected in factor price relationships, market prices will be poor indices of the real cost to society in undertaking the investment.
Net social benefits, SBi', are measured by the incremental increase in earnings which results from the investment in an additional unit of education. Since nonschool factors such as ability and socioeconomic characteristics of the family contribute to observed differences in earnings, the full amount of an earnings increment cannot justifiably be attributed to education as long as data are not standardized for these characteristics. It is appropriate to allocate only a proportion of the increment to schooling. This adjustment, A0, is the effect of nonschool factors, and 1 -AO measures the portion of net earnings accounted for by the difference in schooling.9 The net social benefit, SBi', in year i must also account for the probability of living to the ith year. Therefore, the stream of benefits should be adjusted by the probability of living from the initial period of the investment, year 0, to the ith year. This adjustment is, Loi, in equation ( where Ymi is the monthly earnings in rupees in the inth month for the ith individual since graduation, and X1i is the set of independent variables for the ith individual.
The model postulates that the independent variables comprising the set of XjK are multiplicative rather than additive. Furthermore, the set is divided into two subsets: X,L and XKi, where L + R = J.
The first subset, X,i, is composed of variables affecting the initial earnings of graduates. These are determinants of the earnings in the first 9 There is a growing number of empirical papers on the effect of nonschool factors on earnings. These suggest that job. The second set, XR,, may effect earnings at any given point in time, but primarily at times of promotion and job changes.
The first subset, X,L, includes: X1, = year of graduation, X2i = age at graduation, X3,i = overall grade point average, and X4, = initial period of unemployment in months."' The second subset, XR,, includes: Xs, = number of jobs previously held, X,i = location of the job in Uttar Pradesh state,11 and X,7 = number of months since graduation.
This function was selected because it is linear in logs and can be estimated by ordinary least squares (OLS), proved a reasonably good fit to the data without violating the underlying assumptions of OLS, was consistent with expected characteristics of age-earnings profiles, and produced reasonable projections of the life earnings of graduates.12 The earnings of graduates vary according to the type of firm in which they are employed. To measure these differences, employers are classified into five general categories which are introduced in the equation as dummy variables: D, = university research, extension, and/or teaching; D2 = government of India corporations or research institutions; D3 = military service; 13 D4 = state government; and D5 = farming and private business. Employment category is assumed to interact with the variable "months since graduation," X7, which gives the estimated age-earnings profile its characteristic shape. Incorporating category of employment, equation ( 12 Other functional forms were tested, including the omission of the exponential term, [exp (cl X7-1)], and the removal of the constraint which includes time as an inverse. In the former, the exponential term becomes [exp (c, X7)]. In both these cases, similar results were obtained regarding "goodness of fit," or explanatory power of the function. However, these functions give unreasonably high projections for the ageearnings profiles. 13 This category, although employed in the initial runs of the equation, was dropped in the final form presented here because of too few cases.

Results of the Regression
Equations were estimated (a) for the regular 3-year agriculture undergraduate and (b) for the special 2-year degree for gramsevaks. Results are presented in table 1. With 97 percent of the jobs held by gramsevaks in state government, the dummy employment variables were omitted in the estimation of their equation.  Year of graduation, InX1, is significant in both equations; the more recent the degree, the higher initial earnings. The variable primarily accounts for changes in pay associated with increases in dearness allowance in the public sector and bonuses in the private sector, since no significant change occurred over time in the initial employment pattern of graduates which would be reflected in higher starting salaries.14 Overall grade point average, InX3, is significant. A higher academic record at the university is associated with higher earnings. From the values of the coefficients in the two equations, the impact of grades is more pronounced on the earnings of the regular 3-year graduate. Regular 3year graduates employed in private business and in corporations sponsored by the government of India have higher grade point averages than those employed in state government, although the differences in mean grade point averages among these groups are not significant statistically. Higher grades are also associated with entry into the firm at a higher level. For gramsevaks, higher grades resulted in a decrease in the time lag between graduation and promotion. 15 The initial period of unemployment, In X4, is significant for both earnings functions. The anomalous behavior of the coefficient in the gramsevak equation is explained by reference to a small number of gramsevaks who did not return to state government service. The normal pattern for gramsevaks is a return to their previous post upon graduation before being promoted. For this group unemployment is zero. On the other hand, 3 percent of the gramsevaks failed to return to their original positions. Unemployment is observed among this group. Yet since they secured positions outside state government service at higher pay scales, unemployment has a positive effect on earnings. For the regular 3-year graduate, longer periods of initial unemployment are reflected in lower starting salaries. This indicates a willingness among graduates to lower their reservation wage as the number of months unemployed increases. Employment in Uttar Pradesh, In X6, results in significantly lower earnings for both groups of graduates. This is as anticipated on the basis of knowledge concerning state-wise annual per capita income in India.
Over the graduate's lifetime, employment in private business or farming means significantly higher earnings than employment in other areas. The coefficients for employment in university research, extension, and/or 14 Base pay is determined by pay commissions in India. These commissions are established by the central government for their employees, and independent pay commissions are established in the various states for state government employees. For central government employees, the last relevant pay commission was 1959. Recently the Third Pay Commission submitted its recommendations to the government of India, but at the time of this paper's writing no action has been taken to implement them. 15 Overall grade point average, In X3, and number of previous jobs held, In Xs, are not highly intercorrelated. The impact of grades on earnings is through a decrease in the time that elapses between being hired and the first promotion, not the number of promotions. teaching, government of India corporations or research institutions, and state government employment are not significantly different from the one estimated for In X7. Therefore, the incremented increase in earnings associated with months since graduation for graduates employed in these areas is not statistically different from the one associated with In X,.
Both time variables are significant. Earnings increase over time, but each successive month adds less to earnings than the previous month, because of the impact of clX7. Thus the function has the property of increasing at a decreasing rate. The inverse of time, cl/X,, may measure the deterioration in a unit of human capital with the passage of time. The greatest increment in earnings occurs in the immediate postgraduation period. Without further investments to upgrade the individual's human capital stock, previous investments will contribute increasingly less to an individual's earnings.
Number of jobs previously held, In X5, is significant only for gramsevaks. This variable measures the impact of promotion on earnings.
In both equations reasonably good fits are obtained. The R2 adjusted in the equation for gramsevaks is .661, and for regular bachelors of science in agriculture, .541.
Summary projected lifetime earnings are given in table 2. The average age at graduation is 31 years for gramsevaks compared to 21 for regular agriculture graduates.

Estimation of Social Costs
Social costs include the total value of resources allocated to a particular educational activity. For Pant University, the relevant costs are, (a) annual recurring expenditure per student, (b) annual rent per student for the fixed capital invested in buildings and other durable assets used in teaching, (c) student's annual expenditure on books and stationery, (d) net annual cost for food and lodging while at the university,16 and (e) loss in productivity while in school.
The annual recurring expenditure per student and the annual rent per student were computed from the official university records. First, the annual expenditures for staff and contingencies were assumed to be the major items of recurring expenditure. For each year a per student cost was estimated. Second, the per student rent was calculated by summing the university's investment in its physical plant for the 1960s, depreciating these costs using a straight-line method over a 60-year period, and dividing by the number of students enrolled. In addition, the cost of annual 16 An individual has food and lodging costs in his alternative activity. Therefore, the costs associated with schooling are those that are a direct result of going to school. If the alternative is indeed remaining at home, the costs to society are any additional food and lodging costs associated with sending an individual to school.    H ostel ............................ complex including library were estimated following the procedure outlined above. Instead of partitioning the costs to teaching, the assumption was made to allocate to teaching the full recurring annual expenditures plus depreciation. This was done for two reasons. First, no reliable means was found for assigning administrative costs according to time spent in teaching.

Second, it was felt that the administrative complex was largely an adjunct of the teaching activities of the university, although it peripherally serves research and extension facilities.19
Based on a survey of 10 percent of the students expecting to graduate in July 1971, the annual expenditure on books for a student in the College of Agriculture was RS 52. It was assumed that both gramsevaks and regular B.Sc.Ag. students spent equal amounts on books.
From the records available in the comptroller's office the per student annual expenditures on food and hostel were calculated. The average gramsevak spent RS 467 the first year for food and lodging and RS 601 the second. The average regular bachelor-of-science-in-agriculture student spent RS 481 the first year, RS 563 the second, and RS 645 the third. These are averages for the decade of the 1960s. To arrive at a net cost per student required an estimation of the cost of living at home. The National Sample Survey's thirteenth, sixteenth, and eighteenth rounds contain information on per capita monthly food expenditure by consumption class for Uttar Pradesh state.20 Based on the survey of seniors, which 19 The administrative complex is largely taken up with offices directly related to teaching, such as the registrar, student welfare, and the comptroller. Foregone earnings for graduate gramsevaks were calculated from the data supplied on their returned questionnaires. A large share of the graduates listed their earnings before attending the university. These were compared with their earnings immediately after graduation before promotion. In this comparison, earnings failed to reflect an increase due to university training. Gramsevaks received higher earnings through the promotional advantages and opportunities available to them after receiving the college degree. Using the reported earnings prior to promotion, it was possible to estimate an alternative earnings stream.

Adjustments The source of the mortality adjustment, Loi, was the life survivorship table for India in the United Nations Organization's Demographic Yearbook, 1966. This table resembled the model life table given in the UNO's Age and Sex Patterns of Mortality: Model Life Tables for Underdeveloped
Countries for a population with a life expectancy of 40 years. This overestimates the incidence of age specific mortality for Plant University graduates for two reasons. First, college-educated individuals come from a higher socioeconomic stratum, whose class-specific mortality should be lower than those from lower strata, because of differences in the standard of living. India's life tables are dominated by the poor and the rural. Second, the UNO's estimate of life expectancy in South Asia, which included Nepal, India, Pakistan, and Bangladesh for the period 1965-70, was 48 years. Therefore, the Demographic Yearbook's 1966 estimate based on 1951 and 1961 census data fails to account for more recent changes in life expectancy.22 The ability or nonschool factor is assumed to be 0.40. Forty percent of the observed differential in earnings is related to nonschool characteristics and 60 percent to schooling. The use of this proportion, although arbitrary, conforms to Blaug's use of 0.35 and 0.50 as well as Gounden's 0.50.23 Evidence cited earlier from United States data suggests that the effect of nonschool factors may be substantially less than 40 percent. Similarly, in a study of education in Kenya, Thias and Carnoy find that the effect of nonschool factors decreases as the level of schooling increases, the most pronounced effect being in primary school, with little or no impact at the highest level.24 Without the existence of a body of comparable evidence for India, the decision was made to use 40 percent to maintain conformity with the Blaug and Gounden studies.
The adjustment for nonschool factors was employed for the regular agriculture graduate and not gramsevaks. Since longitudinal earnings data existed for gramsevaks in the pre-and postgraduation periods, the observed differential in earnings may be assumed to reflect only schoolrelated characteristics, since all other factors are constant. On the other hand, the observed differential between the earnings of regular agriculture graduates and matriculates is a function of both school and nonschool . 25 The use of the adjustment for the regular undergraduate and not gramsevaks hinges on the measure of the alternative stream of earnings. For 3-year graduates the earnings differential is arrived at by comparing two data sources-the earnings of matriculates from the Urban Income Survey and the earnings of Pant University graduates. The Urban Income Survey represents a broader cross section of the educated in India. A comparison of their socioeconomic characteristics with those of Pant University students indicates that the latter come from a significantly higher social stratum. If earnings are positively related to socioeconomic characteristics, one may postulate that without a college education, the Pant University student would earn a higher wage than the average matriculate. Therefore, to use the average matriculate's earnings as a measure of the alternative earnings stream overestimates the rate of return due to schooling. One may argue that the use of a nonschool factor adjustment of 0.40 goes to the other extreme of underestimation. However, if Ao is assumed to be 0.10, based on recent U.S. evidence, the conclusion of this paper, that both programs are equally as efficient, is not altered. The crucial factor in the analysis is not the use of Ao in one case and its absence in the other but the reduction in the degree program by 1 year for gramsevaks.

Social Rates of Return
Two relevant pairs of social rates of return are computed. The first assumes no delay in the promotion of gramsevaks.26 In this case the social rate of return for gramsevaks is compared with the return to regular 3-year agriculture graduates employed in state government service. The second assumes that the real measure of the social benefit from investment in gramsevaks is the earnings of the regular 3-year agriculture graduate. The support for the latter assumption is threefold: first, the similarity in university training programs; second, the significantly higher grade point average for gramsevaks while at the university; and third, the experience of gramsevaks who elected not to return to their posts in state government service.27 A comparison of the academic performance of gramsevaks and regular 3-year agriculture undergraduates indicates that gramsevaks had significantly higher grade point averages. Objection may be raised that gramsevaks take academically less demanding courses. No evidence of this appears in the university's course outline for each program. However, if theory courses are considered more difficult than practical courses and it could be shown that gramsevaks take less theory, this would not alter the argument that the earnings profile of the average 3-year graduate is a measure of the social benefit for gramsevaks with university training. It can be demonstrated that the practical agricultural courses offered by the university are one of its distinguishing characteristics, which differentiates Pant University agricultural graduates in the labor market. The evidence for this is twofold. First, in discussions with private businessmen, the outstanding practical training of students at Pant University was emphasized as important in their hiring policies. Also students generally felt that 26 Graduate gramsevaks realized the benefits of university training through promotion. Therefore, a long delay between graduation and promotion results in a net loss in social benefit as measured by earned income. The average promotional delay during the 1960s was 19.4 months. During this period the benefit to society from university-trained gramsevaks is not reflected in their earnings profile. For this reason, the earnings profile is computed assuming no promotional delay. 27 It should be noted that the purpose of this paper is to compare the social rates of return of two degree programs offered by the university. Therefore, the conclusion depends on the relative relationship of the two rates of return and not on their absolute levels. For example, the crucial question is which is higher or lower. Government subsidy on the cost side affects the private rate of return, not the social, ceteris paribus. Government subsidy on the demand side through the establishment of pay scales does affect the absolute level of the rate of return, but it does so for any graduate employed in government service. The comparison of the rates of return to both gramsevaks and regular agriculture graduates employed in state government service is made to compensate for observed differences in their employment pattern. The use of the earnings profile of regular undergraduates as a measure of the social benefits derived from training gramsevaks is addressed to the absolute level of the social rate of return. Pay scales in state government service are lower than those in the large private corporations hiring regular agriculture graduates. Therefore, the lower pay of gramsevaks compared to regular agriculture graduates underestimates the social benefit. employers placed more emphasis on practical training. For companies serving the agricultural sector, this form of training is highly valued. Second, a large proportion of agricultural undergraduates is initially employed in university research and extension for up to 2 years after graduation. From these jobs, graduates move to higher-paying positions in private business. Therefore, employment in university research and extension may be characterized as appendant practical training to the formal educational program of the university and hence a stepping-stone to private-sector employment.
In the first case, the average social rate of return for gramsevaks is 8.3 percent compared to 9.9. percent for the regular agriculture undergraduate employed in state government service. In the second case, the social rate of return to gramsevaks is 13.5 percent compared to an average of 10.3 percent for the regular agriculture graduate.
Given the similarity in training experience, the higher academic performance of gramsevaks, and the ability of a few gramsevaks to compete effectively in the same employment market, the claim might be made that the program is more efficient than the regular 3-year one. However, given the lack of evidence concerning the actual alternative stream of earnings for regular agriculture graduates, a more reasonable conclusion is that the gramsevak program is at least as efficient as the regular undergraduate one. This is attributable to the 1-year reduction in the degree program for gramsevaks, which counters their shortened work life. Gramsevaks, who are older than regular undergraduates, have approximately 10 fewer years in the labor force. The gramsevak program has allowed the participation of a lower socioeconomic group in the Pant University's agricultural program. Although gramsevaks by no means come from the lowest income groups in rural India, their inclusion in the university's agricultural program is a significant step toward expanding the university's participation base. Without such a program of recurrent education, these groups would find it difficult to compete for admission to the regular 3-year program, given their socioeconomic backgrounds.

Importance for Rural Development
The efficiency and equity criteria are satisfied in the university training program for gramsevaks. It may be useful to speculate about the importance of the program for rural development. Extension plays an important role in the advance of new technology. Often there is a lack of information about the specifics of new agricultural practices. The traditional forms of cultivation no longer suffice. Having been shown that an innovation is profitable, the farmer's information requirements become increasingly more technical and specific. He needs information on disease control, planting times, fertilizer utilization, and water control. The response to these needs requires competent agriculturalists working at the village, block, and district levels. Therefore, the university training of gramsevaks with experience working in rural areas equips them to meet the emerging needs of rural development in India. By promotion to district agricultural extension officers, the college-trained gramsevaks become an integral link in the development process by providing at a higher administrative level individuals who are aware of the practical side of agriculture in rural India, who have empathy with gramsevaks working under them, and who have the technical skills to handle the flow of agricultural information downward to the village.

Conclusions
The gramsevak program demonstrates the effective use of a program of recurrent education from the standpoint of both efficiency and equity. The social rate of return was shown to be equal to or greater than the comparable return for the regular agricultural graduate. The higher return is dependent on the reduction of the program for gramsevaks by 1 year. The better academic performance of gramsevaks despite weaker educational backgrounds and being out of school for 10 or more years indicates that no deterioration has occurred in their ability to perform in an academic environment. It may well be that gramsevaks are more motivated and aware of career objectives. These two factors may have compensated for any deterioration in learning ability.