4  Discussion

This paper looks for evidence of the wider benefits of education in a number of different areas of life. As a whole, the literature on the wider benefits of education appears considerably less advanced than the literature on earnings-related benefits. Some areas, most notably the links between education and health, are reasonably well-researched, but it is probably true to say that there are more high-quality studies on the effects of education on wages than there are on all the other ‘wider’ areas put together. However, it does appear that the wider benefits of education is becoming a more popular topic in the literature, as evidenced, for example, by the number of recent NBER studies cited in this paper.

The research discussed in this paper suggests that greater education, as measured by either the time a person spends in formal education or the qualifications they attain, may cause a reduction in cigarette smoking, anxiety disorders, anti-social disorders, suicide, crime, teenage pregnancies, unemployment and reliance on welfare benefits. The evidence in these particular areas is largely based on the results of the Christchurch HDS and Dunedin MHDS, whose strengths are that they study young New Zealanders (and so are relevant to this country), have been recently conducted (the cohorts were at school in the early 1990s) and control for a relatively large number of early life factors. These studies measure outcomes between the ages of 18 and 21. It is less clear whether the benefits of education in terms of cigarette smoking, etc., persist into adult life, although some evidence from instrumental variables studies suggests a causal link with adult outcomes in these areas.

The evidence for later adult outcomes in terms of physical health, social connectedness and political participation is less convincing than it is for the adolescent outcomes covered by the New Zealand birth-cohort studies. However, the large number and variety of studies which show a positive association between education and health does suggest that education has a causal influence on a person’s health. Studies which look at whether a person’s education benefits the next generation, in terms of their children’s achievement at school, yield conflicting results, although most find some evidence of a causal relationship between the generations.

Some of the studies suggesting a causal relationship between education and later outcomes do so by comparing the outcomes of people who have School Certificate, for example, or a high school diploma, with the outcomes of people who do not. Since these two groups are very broadly defined, it is not possible to estimate how much benefit a marginal increase in education would confer. In particular, these studies may appear to overstate the benefits of an individual gaining School Certificate or a high school diploma. Other studies compare groups of people with progressively higher qualifications or progressively longer education. These studies typically show that the more qualifications a person holds, or the longer they are in formal education, the greater the probability of good outcomes in later life. Even here, though, it is difficult to settle on the marginal benefits of increased education. Studies present their results in a variety of ways and, in any case, give differing results which are hard to reconcile. In general, however, it does seem that the marginal benefits of an additional year of education, or of an additional qualification, are relatively modest. In their review of the literature, Behrman and Stacey (1997) claim that the percentage of the variation in social outcomes explained by education is much smaller than the percentage of the variation in earnings explained by education.

Ideally, a comprehensive study of the wider benefits of education would not only provide estimates of the marginal benefits in each area of life, but would also go on to express these benefits in a common unit, so that the total wider benefits of an increased year of education, or of an additional qualification, could be measured. This would be extremely difficult to achieve, although Haveman and Wolfe (1984) do attempt it. They present a method for quantifying some of the wider benefits of education using information on the costs of alternative ways of obtaining better health, lower levels of crime, children’s attainments, and so on. In this way they come to their estimate that the total annual value of the wider benefits of education is of the same order of magnitude as the earnings-related benefits. Haveman and Wolfe admit, though, that theirs is a very rough and speculative estimate of the value of additional education.

To heap ideal upon ideal, the perfect study would also take into account the interactions between benefits in different areas: that unemployment might increase mental health problems; that smoking might contribute to health problems; that crime might decrease social connectedness; that anti-social disorders might increase crime; and so on. Double counting of benefits would then be avoided, although it would be extremely difficult to control for all these interactions in an empirical study. It is possible to cut through this difficulty, however, by measuring people’s self-assessed well-being – assuming that it captures the value to them of all the possible benefits of education – and then comparing the well-being of individuals (or the average well-being within countries) with different levels of education. As discussed in this paper, Helliwell (2002) shows a positive association between education and self-assessed well-being, but his study includes only minimal controls for early life factors and cannot therefore establish a convincing causal association.

The literature on the wider benefits of education does not give a good indication of whether benefits are greater at particular times, for particular people or for particular skills. Following Mincer (1974), wage models typically assume that the marginal return to education is constant, so that an additional year of education has the same percentage benefit in graduate school as it does in primary school. It is not clear, however, that such an assumption is justified when considering the wider benefits of education. If the payoffs to increased education diminish as the level of education gets higher, then an additional year of schooling at age 16 will be more valuable than an additional year of university study, and the attainment of School Certificate will be more valuable than the attainment of a PhD. There might also be different benefits accruing from different courses of study, for example from attending vocational training rather than academic courses. The literature gives little guidance here, although some studies suggest that completing a year of education which results in a significant qualification is more valuable than completing a year which doesn’t.

While the literature gives little guidance in this area, it is tempting to think that significant benefits could be realised by increasing the educational attainment of the very lowest achievers, since a basic level of literacy and communication would seem to be necessary in order to find something in a newspaper, read to children, understand what is written on a medicine bottle, use a phone book, interpret a nutrition label, and so on. In other words, a minimum level of educational attainment may be necessary to avoid social exclusion. Also, since people with the poorest attainment at school tend to have the worst outcomes in adulthood, focusing on improving the achievement of the lowest achievers might help to reduce disparities in adult outcomes.

Why should greater education cause good outcomes later in life? This review has not attempted to systematically uncover the factors which might mediate between educational performance and early adult outcomes, although it is worth considering what some of these factors might be. The association between educational achievement and unemployment should be relatively straightforward to explain, since employers want workers with a proven level of basic skills and attitudes. It is also possible that the association between school achievement and poor outcomes in early adulthood springs from this: that is, that adolescent substance use, crime, sexual risk-taking, etc., are a consequence of unemployment or, more generally, a consequence of not being gainfully occupied either at school or at work. The juvenile crime literature, for example, suggests that school attendance, rather than achievement of qualifications, might provide a constraint on criminal behaviour by keeping young people gainfully occupied. For the most part, however, the Christchurch HDS and Dunedin MHDS studies do not test for these mediating factors. We also know that studies showing causal relationships between school achievement, unemployment and early adult outcomes cannot always be interpreted transitively, for whatever reason.[20]

Alternatively, it might be the case that the associations between education and poor outcomes between 18 and 21 are mediated by income. What we consider to be the wider benefits of education for young adults may in fact be due simply to higher earnings. The New Zealand birth-cohort studies did not test for this. It is unlikely, however, that better-educated people would have earned significantly more than less-educated people by ages 18 to 21, especially since people who continue with formal education are usually not in full-time paid employment.

Other explanations rely more on psychosocial factors. Miech et al (1999) mention a number of possible mediating factors which might explain the effect of education on mental health. These factors are stress, poor social or psychological coping resources, and lack of occupational direction, control and planning. Ross and van Willigen (1997) highlight the importance of a sense of personal control and the ability to form close relationships. They consider that these factors are affected by access to work and in particular to non-alienating work. For outcomes later in life, it is probably true to say that education affects outcomes though a whole range of mechanisms, which differ for different people in different circumstances. The effect of increased education at age 16 on increased health at age 66, for example, will necessarily be complex, involving many intervening factors across this 50-year period.

Causality and coincidence

Section 2 of this paper discussed the issues of causation and coincidence, and of individual and social benefits. These are key issues for policy-makers and so are worth revisiting.

The discussion above refers to studies ‘suggesting’ a causal association between education and later outcomes. This might be considered an unnecessarily guarded way of describing these relationships, especially where high-quality, New Zealand-relevant studies have demonstrated a persistent link between educational attainment, or participation, and subsequent outcomes. However, while much of this appears to be good evidence it is as well to be cautious in ascribing causality. Even in the richest of longitudinal studies, for example, controls inevitably remain incomplete and critics always have the opportunity of suggesting potentially confounding factors that are not included in the analysis and perhaps could never feasibly be included. Fuchs (1982), for example, suggests that the observed relationship between education and health could reflect personality factors differing between individuals and in particular their rate of time preference. People with a low rate of time preference will tend to invest both in education and in health for the longer-term payoffs that these provide; people who by their nature live for the now will not be prepared to sacrifice current pleasures in such a way.

The instrumental variables studies reviewed in this paper avoid this problem of omitted variables. However, they do raise questions about generalisability to contemporary New Zealand. All the IV studies cited here are from overseas, and a number look at the effects of changes in compulsory schooling laws from as long ago as before the Second World War. In all cases, the people affected were those who otherwise would have left school, but who were forced to stay for an extra year or two by the changes in schooling laws. Therefore, while the estimates from IV studies are unbiased by omitted variables, how much weight should we place, for example, on a study of pre-war students in the United States who were compelled to stay at school until they were 13 instead of leaving to work on a farm at age 11? Twin and adoption studies also avoid some of these problems of omitted variables, but have demanding sample requirements and are difficult to execute well, as section 3.8 on children’s education illustrates. Therefore, while some study designs are better than others, none are perfect. In any case, even the best study of the link between education and outcomes over people’s lives would effectively be 50 or 60 years out of date when it reached its conclusions, since formal education mostly occurs in a person’s youth and several important outcomes, such as mortality, occur (usually) after a long intervening period.

Individual and social benefits

As discussed above, this paper draws together evidence suggesting that people who choose to further their education are likely to experience (some) better outcomes in later life, and are therefore likely to gain a benefit from this greater education. Following the framework set out in section 2.1, we can also consider whether adding up the benefits to individuals of increasing their own education would under-estimate or over-estimate the benefits of additional education to society as whole.

If the benefits resulting from increased education have significant spill-overs then the sum of individual benefits might under-estimate the benefits to society as a whole. Spill-over, or ‘external’, benefits occur when an individual’s increased education provides benefits to other people, and there is no market for these benefits. What might some of these spill-over benefits be? If education causes a decrease in cigarette smoking then the spill-over benefits might include a reduction in the harm done to their babies by pregnant women who smoke, or the harm done to others through passive smoking. Similarly, spill-overs from a decrease in criminal activities will include the reduction in resource costs to other members of society. Spill-overs from a decrease in mental health problems, suicide, teenage pregnancy, and other outcomes are likely to include a reduction in the distress suffered by family and friends. If increased education has benefits for an individual’s children then this would seem to be an intergenerational spill-over. By definition, social capital is a property of a community, so if an individual’s education increases their membership of groups, trust, volunteering, etc., then a resulting increase in social capital is a benefit shared by the rest of the individual’s community.

Pinning down what exactly constitutes a spill-over, however, is difficult, and some of the examples given above are disputed. For example, it is not clear whether benefits which accrue to members of an individual’s own family should be considered as spill-overs (Behrman 1997). Grossman and Kaestner (1997) consider that spill-overs include costs or benefits that the decision maker imposes on himself or herself but neglects due to imperfect information. Manning, Keeler, Newhouse, Sloss and Wasserman (1991) question whether maternal and passive smoking are in fact spill-overs. In general, the difficulty is that spill-over (external) benefits and costs are defined in a precise way in microeconomic models (e.g. Buchanan and Stubblebine 1962) but in real-world situations are much less clearly identified. It does seem, however, that some good outcomes arising from increased education, such as a reduction in crime, plainly have some spill-over benefits and therefore that the benefit of education to society as a whole might be greater than the sum of individual benefits.

On the other hand, the benefit to society as a whole might be less than the sum of individual benefits if, at least to some extent, education promotes good outcomes through social stratification, rather than solely through the skills, knowledge and behaviour that are developed. In other words, if the lasting effect of education is the status it confers on people, such as being known as a university graduate, then it may confer relative benefits (being better off than the next person) but not absolute benefits to society as a whole (since the other person is now worse off than you). Even early in adulthood, people who do not have educational qualifications may be being selected into lower social strata. Being low in the social hierarchy may lead to psychosocial stress, which could have physical or mental health effects. Better-educated people might also be able to work ‘the system’, for example by knowing how to get preferential treatment from health providers. This paper provides some evidence in favour of this social stratification hypothesis, from United States studies showing that the risks of mortality, or of being in prison, decline markedly upon the completion of a high school diploma or a college degree (Backlund et al 1999; Lochner and Moretti 2001). Obtaining a key qualification may confer an additional benefit, over and above the skills, knowledge, etc., that are developed during an additional year of education.

Conclusion

In summary, the literature points to a number of areas where greater education might improve outcomes, over and above the benefits that accrue from being better paid. Knowledge of these wider benefits of education should complement discussions on the link between education and individual earnings, and the link between education and economic growth. The wider benefits of education are difficult to quantify, however, and the degree of uncertainty around them is considerable. It would be unwise, with our present knowledge, to consider education as the springboard for widespread social change, especially since New Zealand already mandates compulsory schooling for 10 or 11 years.[21] Greater education seems likely to improve outcomes such as health and crime, but may do so only modestly. If this is the case then there are two policy implications: firstly, the wider benefits of education may not be great enough, on their own, to justify large increases in public spending on education and secondly, greater education may not be the most cost-effective way to improve outcomes in the areas of health, crime, etc. These implications need to be considered in more specific research.

Furthermore, the literature gives little guidance on where improvements in education would be most effective (let alone most cost-effective) in promoting wider benefits. It may be that the greatest benefits could be realised by increasing the educational attainment of the very lowest achievers, although this is more a product of speculation than of empirical evidence. The literature also cannot tell us whether the benefits of education to society as a whole are more or less than the sum of benefits to individual students. While this seems a vague and unsatisfying conclusion, it is worth noting that the evidence on whether the earnings-related benefits to society as a whole are more or less than the sum of individual benefits is also unclear, despite being more closely studied (Temple 2001). Additional education, on top of current levels of provision, probably does make us healthier, wealthier and wiser but the evidence for health impacts, as well as for other wider benefits of education, leaves a lot of questions yet to be answered.