2  Concepts and methods

2.1  Individual and social benefits

The individual benefits of additional education are those which are realised by the person being educated. Individuals consider what their own benefits (and costs) are likely to be when they make decisions about whether to continue with their education. Social benefits, on the other hand, are those which fall on all members of society, not just on the individuals who receive additional education. Policy-makers consider what the social benefits (and costs) are likely to be when they make decisions about government investment in education.[2]

A simple thought experiment may be used to illustrate the nature of social benefits. If 1,000 students stayed at school for an additional year and each received 5 units of benefit as a result, then society as a whole might be expected to benefit by 5,000 units. However, if an individual’s additional education benefits other people, for which the individual is not compensated, then it is likely that more than 1,000 people would be better off from this additional year of education and the benefit to society as a whole would be greater than 5,000 units. This might occur if, for example, additional schooling reduced crime rates or increased social capital. The existence of these ‘spill-over’ benefits to other people constitutes one argument for government involvement in education: individuals under-invest in education from society’s point of view because they don’t take into account spill-over benefits, and the government could therefore use subsidies to encourage greater investment. Identifying spill-over benefits in a precise way, however, can be extremely difficult.

On the other hand, if education is, at least in part, a signalling or credentialing process, then the benefit to society as a whole might be less than 5,000 units. Signalling and credentialing are effects which have been hypothesised in explaining the earnings-related benefits of education (Weiss 1995; Collins 1979). Collins, for example, argues that education establishes a pecking order for jobs rather than teaching job-relevant skills, which are mostly acquired on the job. This need not concern individual students, since increased education still leads them to higher-paying jobs. However, from the point of view of society as a whole, more resources may be going into education than are necessary to turn out productive workers. There may also be an analogous effect with the wider benefits of education. If the attainment of qualifications is simply a social ranking mechanism, for example, rather than a means of gaining life-relevant skills, then education may benefit individuals far more than it benefits society as a whole.

Table 1 below shows the four combinations of education benefits which result from considering wider benefits as distinct from earnings-related benefits and also by considering social benefits as distinct from individual benefits.

This paper is mostly concerned with the shaded box in Table 1, that is, with the wider benefits of additional education to society as a whole. The paper does this by considering individual benefits across a number of domains and then considering whether simply adding up these benefits, within each domain as well as across all of them, would under-estimate or over-estimate the social benefits of additional education.[3]

2.2  Causality and coincidence

In this review, the term ‘causal’ is used as in the economics literature.[4] The issue that this review explores is whether, if more New Zealanders achieved a particular qualification or stayed in the formal education system longer, we could reasonably expect to see improvements in health, crime rates, and so on. If greater education merely coincided with good outcomes in adulthood then we would not expect this to be the case.

As stated in the introduction above, studies consistently find that more-educated New Zealanders tend to do better on a range of outcomes. The question remains, however, as to whether these associations are causal, or whether educational attainment and other good outcomes simply coincide. It may in fact be the case that a third factor, such as a person’s family upbringing, influences both their schooling decisions and their later life outcomes, thereby giving the (false) impression that education has an effect on later outcomes.

For example, New Zealand children from disadvantaged families are more likely to do poorly at school (Ministry of Education 2002a) and there is evidence from overseas that children from disadvantaged families are also more likely to have poor health as adults (van de Mheen, Stronks, Looman and Mackenbach 1998). Coming from a disadvantaged background might therefore explain both poor school performance and poor adult health. Also, as in studies of education and earnings, there is the potential for confounding by natural ability. Naturally clever, perceptive or determined people might do well both at school and in other areas of life.[5]

Empirical studies commonly involve regressions of the outcome being measured, such as people’s health status, against their level of education, with the inclusion of other background variables (such as family background and child health status) to control for potentially confounding factors. However, most researchers have only a limited number of control variables available to them, and suspicions usually remain of residual confounding from unmeasured, or poorly-measured, influences of family backgrounds, inherited characteristics or childhood experiences.

Three main methodological approaches are used to address this problem, and to test causality. These are the use of longitudinal studies, the use of instrumental variables (IV) models, and the use of information on twins and siblings. Longitudinal studies usually have rich data sets, with information taken from the same participants at intervals over a period of time. Birth-cohort studies, in particular, tend to have good information on the early lives of participants. As a result it is possible to examine the ways in which variations in family and social conditions in childhood are related to an individual’s longer-term adjustment and well-being.

In IV studies, researchers try to identify naturally-occurring experiments, where the duration of education varies across individuals for reasons which are not likely to be related to later outcomes in any other way. A number of studies reviewed in this paper use changes over previous decades in state schooling laws in the United States as an instrument for predicting the length of schooling. The reasoning here is that compulsory schooling laws affect the number of years of schooling that people undertake but are not related to outcomes in any other way. Therefore if people who spent their childhoods in states that required them to go to school for longer are more healthy or live longer or commit fewer crimes or are less likely to take drugs, it can be concluded that education influences health or crime or drug use. In particular, education influences health or crime or drug use amongst those people most affected by the instrument, namely those people who would have dropped out of school earlier but were prevented from doing so by the compulsory schooling law.[6]

Twin and sibling studies are a way of controlling for common family backgrounds and genetically inherited endowments. Identical twins who have been brought up together, for example, share 100% of their genes and have essentially the same family background. Adopted children, on the other hand, share the same upbringing as their adoptive siblings, but none of their genes. For the most part, this review discusses twin and adoption studies in the context of determining whether increasing a person’s education has a causal effect on his or her children’s education.