|Educational Cost-benefit Analysis - Education Research Paper No. 02 (DFID, 1993, 27 p.)|
As the preceding sections indicate, a large number of academics and others have been and are sufficiently in favour of educational cost-benefit analysis in Third World countries to devote a great deal of time, energy, and expense to undertaking such studies. Equally, other writers have been critical of various aspects of such work. Some of these criticisms have referred to a number of the problems set out in section 4 above, whilst others have raised questions particular to the circumstances in developing countries.
Such criticisms are not new. An early paper by Handa and Skolnik (1975) was very pessimistic regarding the contribution that rate-of-return analysis, which was termed inadequate and misleading, could make to educational policy decision-making. The authors referred to a number of the points outlined above in section 4, were particularly critical that distributional effects on different groups in society were usually ignored, and concluded "it is time the energy of researchers was directed to other allocation models". At about the same time, Griffin (1976) had found alpha-coefficient corrections to be unsatisfactory; his work showed that separate adjustments were necessary in respect of different groups, notably men differently from women, blacks differently from whites.
Perhaps the most comprehensive critique was given by Leslie (1990), who argued that rate-of-return studies were essentially flawed and were inappropriate policy devices for educational aims. He was particularly critical of the fact that calculated benefits almost always failed to take account of the consumption value of education and that calculations of costs failed to correct for the subsistence expenditure that the student would have incurred elsewhere.
Leslie argued that other private investment benefits should, but rarely do, include greater fringe benefits and superior working conditions (including paid vacations and holidays), better ability to select advantageous forms of savings, better health and longer life, lower unemployment and lower disability rates, fewer unwanted children and better health for offspring, more informed purchases, better, education-related, child-rearing practices leading to greater likelihood of future college attendance, and selection of spouse with higher earnings potential. Hence the findings by Becker, 1975, and Haveman and Wolfe, 1984, that true private rates-of-return are essentially double conventionally-calculated rates if nonmonetary benefits are included. "The reliability of social rates-of-return is even less" (Douglas, in Bowen, 1977). Calculations of costs, for example, have routinely failed to disaggregate different instructional costs for students at different levels and have neglected the incidence of cost subsidies from which particular groups of students may benefit. Wider social benefits than those encompassed by the students' future earnings are usually ignored: an example would be the future benefits to society from certain forms of research, such as into improved forms of agriculture.
The writer saw it as particularly important that in many Third World countries the ingredients that determine rates-of-return are impacted by quite different government policies, especially relating to government pay scales; as a consequence, it is likely that there is overproduction of graduates in social sciences and humanities alongside real shortages of technicians and engineers. Due to the effects of discounting, rates-of-return are far more sensitive to cost than to benefits differences, so that "such studies primarily are cost studies not benefit studies" (Leslie, 1990).
In summary, Leslie argues, the effects of all such adjustments would greatly increase the rates-of-return to education and thus there emerges a much strengthened case for increasing allocations to education but "generally rates-of-return do a poor job of identifying unmet and saturated manpower needs".
An important line of criticism of the standard cost-benefit approach has been developed by Behrman and Birdsall (1983, 1985, 1987) who argue that the cost-benefit model is seriously in default in concentrating on the quantity of education and neglecting the factor of educational quality. Quantity of education is almost always included via data for number of years of schooling but few rate-of-return studies have included any indicators of quality of schooling. Using data for Brazil, Behrman and Birdsall found that the standard approach may cause biases in the estimated returns to years of schooling, probably in the upward direction; that the standard approach tends to overstate regional and urban-rural differentials in the impact of schooling; and that most of the apparent differential returns to schooling in the standard estimates for migrants vs. nonmigrants, often attributed to migrant selectivity on personal characteristics, are due to variations in school quality.
The researchers conclude that rates-of-return have typically been overstated, perhaps by a factor of three and that once quality is taken into account the results:
"indicate that 'deepening' schooling by increasing quality has a higher social rate of return than 'broadening' schooling by increasing quantity" (Behrman and Birdsall, 1983).
The same writers subsequently commented that their conclusions point to a productivity/equity trade-off, since greater productivity gains would be possible if years of schooling and schooling quality were concentrated among fewer individuals rather than being spread broadly (Behrman and Birdsall, 1987). Behrman (1987), separately, produced detailed computations to support the above findings.
The above emphasis on quality of schooling is closely reflected in the development of the literature relating to educational effectiveness, which is referred to in section 11 below.
A study by Knight and Sabot (1987) was particularly concerned with the fact that the labour market may operate imperfectly and thus the marginal product of labour may not be accurately measured by the average wage, data for which tend to be more readily available than data for marginal wages. Using data for Kenya and Tanzania, the researchers found the marginal rate-of-return to secondary education to be lower - by between one and three percentage points - than the average return, thus suggesting a potential source of bias in standard returns calculations. They also suggest that over time such bias could become increasingly important as secondary school-leavers filter down into unskilled wage- or self-employment occupations in which their education has less value.
Tsang (1988) found five major methodological problems relating to rates-of-return studies. First, the results are based on past conditions and may not be reliable predictors of the future; second, most studies use cross-sectional data instead of longitudinal data; third, most studies use data for quantity of schooling and ignore educational quality; fourth, most studies ignore significant noneconomic benefits of education; fifth, the assumption that the labour market is perfectly competitive is unlikely to be true in developing countries where governments are major employers.
Tsang also points out that a number of previous writers have questioned the basic assumption underlying cost-benefit studies, namely that education raises future productivity; if it does not, if, e.g., education is merely a screening device or productivity is determined primarily by job structure or labour market characteristics such as segmentation, or if there is underutilization of education in production (leading, perhaps to lower work effort and lower productivity), then cost-benefit analysis loses its validity. Finally, Tsang notes that Bowles & Gintis (1976) found the focus on the productivity and earnings benefits of education too narrow, given their thesis that the central function of education is to reproduce the social relations of production in a capitalist economy.
McMahon (1988) found cost-benefit studies of vocational and technical education to be often unsatisfactory, partly because the "vocational" course content may not be up-to-date, and partly because there may be an imbalance between vocational and general curricula. In some circumstances, corrected rates-of-return would be negative.
Most recently of all, Bourguignon (1991) commented that during the recent period of major education transition, the very rapid development of education may lead to:
"a drastic change in the educational structure of the labour-supply, which in turn may induce changes in the structure of earnings by educational levels, and therefore changes in the observed returns to education".
Depending on a number of factors, this may mean that the standard "static" rate-of-return results may be over- or underestimated. Also, externalities relating to educational development - following, for example, significantly increased education levels among urban workers - may be overlooked. Bourguignon also stressed the potential importance of externalities typically excluded from cost-benefit calculations, notably the reduced fertility of more educated women, the ability to adapt quickly to a changing environment and make technical innovations, or the enhanced national cohesion and democratic sense of a more educated population.
Overall, these criticisms are so comprehensive that it may seem a matter of some surprise that so many researchers are still engaged in producing educational rate-of-return studies for Third World countries. The explanation must relate partly to the desire to constantly improve and refine the technique, partly to the fact that alternative techniques, such as manpower planning, are beset by at least as many problems.
Perhaps at times too much is expected of cost-benefit analysis. One recent study concludes with a salutary caution:
"Rates of return estimates are not precise results. Their policy purpose is to indicate desirable directions of policy changes. The composition of social or government investment should be shifted in directions where returns are highest" (McGavin, 1991).