2.1.2 Education and productivity
2.1.3 Educational investment and externalities
2.1.4 Education, equity and income distribution
2.1.5 Concluding remark
There has been a long standing debate about the contribution educational investment makes to economic growth. For a now familiar set of reasons there is no single answer to the question "how much does has education contribute to economic growth" and even less to the question "how much does education con tribute to development." It would be surprising if there were. The relationships between educational investment and economic growth are complicated by many intervening variables which interact in different ways in different national economies at different points in time. And, of course, definitions of the characteristics of development are not stable either. But this does not mean that in either case we cannot reach inferences from the large volume of studies that have been undertaken. Rather we have to recognise that what may be true under certain circumstances may not be true under others and that the role education plays in supporting growth and development is one which is constantly evolving.
The economic literature focuses on measurable returns to educational investment to the individual and to society as a whole. Historical and sociological perspectives emphasise more the interactive relationships between educational development and economic change. At the lowest levels some measure of economic development often appears as a pre-cursor to the development of school systems in recognisably modern forms - infrastructural investment has to have taken place and economic surpluses are needed to provide the resources to pay for a school system. As an education system is established it may begin to catalyse further economic development. Thus, as Foster has pointed out (Foster 1987:94), the significance of increased schooling as an instrument of economic development may be highly variable over time. Expansion may have substantial economic and developmental pay-off at some stages and not at others. Some types of educational provision (at different levels, of different orientations, of different qualities) may have much greater effects than others.
The early studies of Denison (1962, 1967, 1979), Harbison and Myers (1964) and Schultz (1961) and Becker (1964) are well known. Denison approached the problem of how much education contributes to economic growth by attributing a proportion of economic growth not explained by increases in capital, labour and productive land to improvements arising from increased educational levels in the labour force. This produced results suggesting that 23% of US economic growth was a result of educational investment between 1930 and 1960, and 15% for the period from 1950 to 1962, and 11% for 1948 to 1973. This kind of analysis claims to provide estimates of both the direct contribution of education and the indirect benefits that arise from advances in knowledge. The latter are argued to be responsible for about 29% of growth in Denison's last study thus attributing 40% (29%+ 11 %) to improvements in human capital and education broadly defined (Hicks 1987: 102). When the approach was applied to other countries the results varied widely - from 2% to 25% in a group of developed countries and from 1% to 16% in a group of developing countries (Psacharopoulos and Woodhall 1985:16). Bowman (1980) suggested that in over 22 countries where estimates could be made for the period 1950-62 education made a direct contribution to economic growth of more than 10% in only four. She also noted that the residual to be explained seemed to be greatest the higher the economic growth rate but that the contribution of education seemed to be smaller where growth rates were high. Others (e.g. Christensen and Jorgenson (1969)) have argued that if inputs and outputs are more completely specified than in the Denison model the residual to be explained is much more modest in size than suggested and, by implication, the contribution of education is over-estimated.
Harbison and Myers approach was to develop indicators of human resource endowments and compare these with indicators of economic development. Predictably choosing different indicators produces somewhat different results, but the overall correlation between greater human resource endowments and greater levels of economic development is robust. It leaves open the question of causality. Richer countries do indeed invest more in education and have higher endowments of human resources as a result. But this cannot lead to the simple conclusion that more investment in education in poor countries will lead to more economic development.
Schultz (1961) and Becker (1964) used an approach based on the rate of return to human capital. This assumes that individuals invest in education up to the point where the returns in extra income are equal to the costs of participating in education. Returns are both private (to the individual in the form of additional income) and public (to society in the form of greater productivity). Rates of return studies in developing countries have generally shown that returns at primary level are greater than for higher levels; private rates exceed social rates; social rates of return often exceed a 10% threshold- rates of return for education are higher in poorer countries (Psacharopoulos 1981, 1985); rates of return fall as economic development takes place; the greatest reductions occur at the lowest education levels as access becomes more universal (Haddad et al 1990:6). It has to be remembered that there are many well established methodological difficulties with rates of return analysis which include the problems of estimating incomes over time in changing labour market conditions and the validity of the assumption that the additional income received by the more educated is a result of additional education rather than other factors (Dore 1977, Carnoy 1980)). Indeed where modern sector salaries have declined, as they have in much of Sub Saharan Africa, and where expanded schooling has greatly increased the supply of graduates of a particular level, returns will have dropped, perhaps considerably, over the last decade. Rates of return will therefore be specific to countries, to particular levels and to particular periods in the development process.
In a somewhat different analysis Hicks (1980) has compared literacy levels (a proxy for educational levels) with historic rates of economic growth in 83 countries. He concludes that the twelve developing countries with the fastest growth rates also had levels of literacy above the average (68% compared to 38% in 1960). These countries had higher income levels and, since income is correlated to levels of literacy, this result might have been expected. However when income level is controlled, literacy rates were still 12 % greater in the fastest growing countries, suggesting that faster growth rates were coincident with more developed human resources.
Wheeler's (1980) study of data on 88 countries tries to take into account interactions between economic growth and investment in human resources over time and give some insight into the direction of causality. His findings imply that literacy does have a strong effect on output levels and that greater literacy influences fertility downwards. This study suggests that increases in average literacy rates from 20% to 30% are associated with increases in GDP of 8% to 16%, with the strongest relationships in African countries. Marris (1982) uses data from 66 countries to argue that the cost benefit ratio of educational investment in human resources (based on primary enrolment ratios) ranges between 3.4 and 7.4 compared to ratios of 0.4 to 1.0 for investments in other types of capital. He also suggests that general investment has less effect on growth rates when it is not accompanied by educational investment. Psacharopoulos and Woodhall (1985:22) also suggest that investments in human capital have higher rates of return than those in physical capital in many developing countries, whilst the reverse tends to be true in developed countries.
At the meso and micro level there is evidence of the effects of educational investment on productivity and this has also been widely studied. These analyses can be conveniently separated into those that relate to agriculture, the modern sector, and the urban informal sector.
Agricultural productivity does seem to have a positive relationship to the education of farmers. After reviewing 18 studies containing 31 data sets which bear on agricultural productivity Lockheed, Jamison and Lau (1980) concluded that four years of primary education increased productivity by 8.7% with a standard deviation of 9%. When weighting was introduced to account for variations in standard errors associated with the various studies, the result was a 7.4% gain with a standard deviation of 6.8%. Though there were some studies that did not show gains the overall effect is clear. As might be expected there are reservations. Output was measured in terms of crop value in most cases - this is dependent on price structures that vary widely between crops and countries. Different studies measured educational inputs in different ways - e.g. number of years, highest grade completed, dichotomous achievement of literacy. They also associated the educational variables with different individuals or groups - head of household, an aggregate for all family members, or for all farm workers. In addition other input factors were measured in a wide variety of ways - by quantity or value or time input, by type of capital available, by technological characteristics of farming (irrigation, new seed varieties, fertilizers etc.).
This review also indicated that agricultural productivity was more influenced by education in modernising than in traditional environments as Shultz (1975) had earlier suggested was plausible. Traditional environments were defined in terms of primitive technology, traditional farming practices and crops, and minimal reported levels of innovation. Modernising environments include access to new varieties of seeds, innovative farming practices, the control of erosion, the availability of pesticides, fertilisers and farm machinery, access to extension services and the existence of market orientated production. When the studies were simply classified into traditional and modernising environments the result was to suggest that four years of primary education increased productivity by a mean value of 1.3% in traditional environments and 9.5% in modernising ones. When regressions were undertaken the average gain in modernising environments was consistently 10% greater than in traditional environments. A recent update [Jamison, Lau, Lockheed and Evanson 1992] reaffirms this general picture.
Findings of five studies on education and agricultural productivity are reported in Haddad et al (1991:5) which include those by Jamison and Lau (1982) in Korea, Malaysia and Thailand and Jamison and Moock (1984) in Nepal. Four of these show positive and significant effects of education under different conditions. They support the view that the effects are greater in modernising environments. Though Thai farmers physical productivity and choice of technology was related to their educational level, it was not the case that they achieved higher prices for their outputs or lower prices for inputs, suggesting perhaps that they were no better at exploiting comparative advantages and using market information than others with less education. Market efficiency may therefore depend more on factors other than increased educational levels. The evidence from Nepal suggests that education does have an effect independent of family background and that increased productivity is related to improvements in farmers numerical skills giving some clue as to why the observed correlations exist.
A provocative study by Mingat and Tan (1988) suggests that Project Related Training (PRT) yields high rates of return in both agricultural and non-agricultural development. This study is based on an analysis of 115 World Bank projects taking the success of the project, rather than direct measures of earnings, as a criteria. However, high returns are concentrated heavily in countries where the general educational base is well established. Where illiteracy rates are high and educational participation rates are low PRT does not appear to be an effective investment. This may arise both because individuals with low levels of formal education are handicapped in absorbing training inputs and because countries where educational infra-structure is weak may also be those where management capacity is least developed and organisational capabilities are most limited. In countries where more than half the population are literate rates of return for PRT are more strongly positive for agricultural rather than non-agricultural projects. This may be the result of diminishing returns to training (nonagricultural projects tended to have more than four times as much training associated with them) and because agricultural projects tend to have greater dependence on people and skills and less on capital than many nonagricultural projects (Mingat and Tan 1988:238). The conclusions of this work argue that changes are need in PRT where its effectiveness appear very low for reasons associated with poor infrastructure and low educational endowments in the population as a whole. In these conditions institution building and support for improvements in basic education are a priority. Under other circumstances PRT appears very cost effective.
Studies of productivity in urban areas and in industry are much more common in developed countries than in developing countries. Much of this literature has addressed the debate between human capital proponents (who argue that education increases productivity which is rewarded by higher earnings) and screening theorists (who attribute the higher earnings of the more educated to factors other than the cognitive changes which are associated with studying to higher educational levels). The evidence does not conclusively favour one or the other approach (Winkler 1987:287). Part of the reason lies in the difficulty of measuring the dependent variable - productivity. If simple output measures are not available e.g. piecework production under standardised conditions, comparison is difficult between workers with different educational levels. Supervisor and peer group ratings can be used though these may not have high reliability. Comparison between jobs with different characteristics is problematic - the relative productivity of lawyers and plumbers cannot simply be assumed to be reflected in their earnings for a long list of reasons. And in any case the occupational mobility of urban workers is often high, resulting in situations where additional educational inputs may be reflected in increased productivity in subsequent not current jobs.
Global syntheses of the evidence on urban and industrial productivity and education are not very meaningful for the reasons given above. There are certainly studies which show positive effects on productivity of education amongst urban workers in developing countries (Fuller 1970, Berry 1980). Equally there are those that question the strength and nature of such relationships and which show how widely such correlations can vary across different types of job, from strongly positive to strongly negative (Little 1984). Selectively citing those studies that support a particular viewpoint would therefore be misleading. It should be noted however, despite the mixed evidence, employers in many countries adhere to a set of beliefs which does value explicitly educational attainment in the selection of employees. The research does not suggest this is an unqualified judgement independent of job or job level. Moreover there is evidence that employers often conceive of the problem in terms of minimum levels of education suitable for different types of employment, above which other factors may become more important in the selection of employees (Oxenham (ed) 1984:66). To the extent that this is generally the case it introduces the possibility of curvilinear relationships between educational level and productivity, where it is possible to be over educated and those with most education may produce less than those with more modest attainments in particular jobs.
Knight and Sabot's (1990) work based on the "natural experiment" of comparing samples of about 2000 employees in Kenya and Tanzania provides detailed insights into education, employment and income relations in those two countries at the beginning of the 1980's. Their findings support the human capital view that there is a positive rate of return to investment in secondary schooling in both countries. More specifically they argue that whereas the labour market returns to reasoning ability (as measured by Raven's Coloured Progressive Matrices) are small, and to years of schooling modest, the returns to cognitive achievement (as measured by numeracy and literacy tests) acquired through schooling are large (ibid: 17). In Kenya cognitive skill accounts for three times more variance in earnings than do ability and years of schooling combined; in Tanzania the ratio is two to one. In neither country is being amongst the brightest of one's peers a sufficient condition for performance in the labour market - the predicted earnings of the most able primary completers are less than those of less able secondary completers. In both countries how much is learned in primary or secondary schools has a substantial influence on performance and income at work. Moreover the evidence suggests a complementary relationship between cognitive skill levels and returns on experience; the greater the former the greater the benefit from experience and training over working lifetimes. Intriguingly, cognitive achievement gains for secondary schooling for students of mean ability are 17 % higher in Kenya than they are in Tanzania, despite the fact that per student expenditures are greater in Tanzania suggesting there may be differences in the effectiveness with which resources are allocated to teaching and learning (ibid: 23). Educational expansion also appears to have led to compression of wage differentials over time contributing to a reduction in income inequality (ibid: 30) as jobs for secondary educated students have expanded in number.
Knight and Sabot question some of the methodological assumptions that lie behind conventional rate of return analysis and suggest that some results may be misleading. They draw attention to the dynamics of labour markets fed by expanding numbers of school graduates as others have done. In particular false conclusions are likely to be drawn if the starting wages of new school leavers stand in a changing relationship to average wages for previous cohorts who entered the labour market when wage differentials were different. In the case of Kenya they suggest that this leads to an exaggeration of rates of return for primary school leavers sufficient to bring these below the level for secondary leavers when calculated on a marginal rather than average basis. This can be taken to suggest that greater emphasis on primary investment would have an efficiency penalty from a purely economic point of view. It should be noted that the data used in this very comprehensive study dates from 1980 since when labour market conditions will have changed. Like other studies of its type it appears to give little attention to unpacking educational quality variables - e.g. taking account of the school attended (since many employers are known to value the school attended in making recruitment decisions) and separating years of schooling from information on qualifications and grades achieved (qualification levels and grades are likely to be used by bureaucratic employing organisations for recruitment and promotion).
The impact of education on productivity is also a matter of concern in the urban informal sector. Here the problems of measurement are even more challenging than in relation to agriculture and modern sector employment, and the evidence is equally mixed. Hallak and Caillods (1981) were unable to establish a clear relationship when reviewing studies by the World Bank on entrepreneurship, by PREALC on the informal sector in Paraguay, Ecuador, the Dominican Republic, by Nihan on Mauritania and Togo, by Aryee on Ghana, and by Van Dijk on Senegal. The Aryee study did suggest that gross output and earnings of heads of enterprise did increase with educational level, but only up to middle school level. In the PREALC studies the level of education that made a difference to income varied by employment activity being least in basic services and greater for repair and maintenance work. King's review of studies of training for the informal sector (King 1991 98-113) lists a large number of initiatives that have been taken to explore the interfaces between education, training systems and informal employment. As might be expected from the wide diversity of situations that they investigate these do not appear to lead to any singular conclusions. The impact of education on productivity within the areas reviewed is therefore variable and likely to appear at many different levels of significance. The conclusions that can be drawn have to be surrounded with caveats. There are likely to be circumstances where education has a strong effect on productivity and cases where this seems improbable. Which are which has to be the subject of specific research projects.
The existence of externalities associated with investment in education, which extend beyond the benefits to individuals, has already been highlighted in the earlier discussion of the benefits of education for girls and women. These are a subset of a range of externalities that have been identified by McMahon (1987:133) and many others. This is an extensive field and one at an early stage where much of the force of the arguments exists at a qualitative level. McMahon's classification draws attention to external benefits to society at large, the local community, and spill overs to other communities, though it must be remembered that its perspective is essentially that of a developed country with substantial welfare benefits and educational services.
The first category includes a range of effects all of which are difficult to demonstrate but most of which are widely recognised. Education is attributed with benefits in creating more efficient markets with more sophisticated producers and consumers better able to process information and adapt to technological change. That this should be theoretically so is almost self evident if education does enhance cognitive skills. Whether the effects are more or less than the other factors that constrain the development of markets and limit access to information and new technologies is more difficult to judge.
Education is also argued to have benefits for civil society and public service. More educated citizens may be more likely to demand and exercise a democratic franchise for the collective benefit, they may be more likely to take part in public service activities and voluntary work, be more employable and they may be less likely to display criminality. To the extent that more educated citizens maintain their health status at a higher level, the costs of publicly financed health care may be reduced over what they would otherwise be.
The contrary is of course possible. More education may increase the demand for public health services despite reductions in morbidity. Since prolonged unemployment is often negatively correlated with higher levels of education it may also be that the public costs of unemployment could reduce with higher levels of education. Equally though it is possible that more education would result in a weakening of the employment-educational level link resulting in higher public costs for unemployment. In most societies the more educated display less criminality in terms of offences warranting imprisonment. Though more educated criminals may be more adept at evading prosecution and specialise in different types of crime, only a confirmed pessimist would argue that this was a dominant effect. Spill over benefits occur when groups other than those paying the costs obtain benefits from educational investment. This may occur regionally within countries where, for example, educated youths migrate and take with them skills and capabilities that are lost to their region but not to their country. It is also relevant to international migration which is increasingly an issue related to migration between developing countries as well as from developing to developed.
Income inequality may be affected by educational investment in a number of ways. These include the ability of more educational provision to raise income levels in general and remove groups from absolute poverty - richer countries tend to have lower levels of income inequality; the ability of education to raise incomes disproportionately amongst the poorest and provide avenues for social mobility; the financing and organisation of education in ways which generally favour poorer rather than richer families in terms of participation and which thereby diminish income inequality arising from the higher earnings of the more educated; and the interaction of educational level with other variables - fertility, mortality, health which have a bearing on income distribution at the family and individual level.
The studies which have attempted to link increased educational investment and participation with income inequality do not show that strong relationships usually exist. Jallade (1974) argued that income inequality in Brazil had not diminished as a result of increased educational provision. Field's (1980) review of five developing countries concluded that though individual incomes were determined more than anything else by educational level achieved, relationships between distribution and performance on educational indicators at the country level were weaker than those with aggregate economic growth. Carnoy et al (1979) suggests that the explanation for this apparent paradox lies in the fact that earnings are influenced more by government incomes policies and by organisational features of employment than by educational levels of employees. Leonor and Richards (1980) argue in a similar vein using data from the Philippines and Shri Lanka. In both cases however both educational opportunity and income distribution appear to have improved over time.
Knight and Sabot (1982) argue that educational expansion in Kenya has reduced the relative earnings differential associated with secondary graduates by roughly 20% and has thereby reduced income inequality. This they suggest has been a more effective policy than government intervention on wages in the public sector, the favoured strategy in Tanzania. They thus argue the contrary point of view to Carnoy. Intergenerational mobility appears to have become more dependent on the educational background of parents and employers have become more likely to discriminate between potential employees on the basis of family backgrounds in Kenya (ibid: 37). Income inequality reductions as a result of wage compression may therefore have been accompanied by increases in the intergenerational transmission of social status.
It is therefore difficult to find a consensus on the relationships between education and income distribution. Expanded access does appear to reduce income inequality, at least where a substantial modern sector exists as in Kenya. The relative impact of government policy on progressive taxation, incomes and subsidies will depend on how draconian these are and how effectively they are implemented. Educational expansion is frequently more attractive politically than direct interventions to transfer wealth and income earning opportunities away from the relatively privileged (Blaug 1978).
What then can we conclude from the literature on the relationships between education and economic growth? First, that there is no single answer to the question some wish to pose - there are many answers depending on circumstance, developmental status and the specifications of the variables.
Second, the direct policy implications of macro level research are very limited. They are constrained by dependence on historical relationships which may or may not persist, the level of aggregation is often so high that effective and ineffective years of schooling are treated as similar, and the application of findings from individual countries or groups to other countries is analytically hazardous.
Third, far more studies imply, suggest and demonstrate plausible and positive links between educational investment and economic growth than suggest that the effects are nonexistent. Even fewer studies suggest a negative relationship. It would be pessimistic in the extreme to suggest that the widespread faith in educational investment as a component of economic development was an aberration that could persist so extensively for so long if it did not contain elements of truth no matter how difficult these are to demonstrate.
Fourth, there is evidence in many studies of productivity benefits derived from educational investment. The most policy relevant ones appear to be those based on recent data which relate to circumstances in particular countries which can give some guidance on the most worthwhile types of educational interventions. Placing them in context is a necessary pre-condition for confidence in conclusions that can be drawn.
Fifth, educational effects are associated with various externalities that may have economic consequences. They may also extend to influencing income distribution and wider social inequalities through dynamic processes that need careful unravelling.
Sixth, there are many methodological questions in the analysis of relationships between education and economic development which have only partial resolutions. These are extensively debated in the literature (e.g. Psacharopoulos et al 1983, Little 1986, Hough 1992) and need no repetition here. The results of the various studies have to be understood in the light of these.
