Rates of Return to Degrees across British Regions

Abstract

O'Leary N. C. and Sloane P. J. Rates of return to degrees across British regions, Regional Studies. Earlier papers have found considerable heterogeneity in the returns to degrees in relation to subjects of study, degree classification and higher education institution. This paper examines heterogeneity of returns across British regions using the Labour Force Survey. It finds substantial regional variations in the financial rewards available to graduates, with much higher returns in London and the South East than elsewhere. Adjusting for regional differences in the cost-of-living narrows such differences considerably. Decompositional analysis suggests that coefficient effects dominate composition effects, consistent with agglomeration and productivity spillover effects being important.

O'Leary N. C. et Sloane P. J. Le taux de rendement des diplômes universitaires à travers les régions de la Grande-Bretagne, Regional Studies. Des articles antérieures ont trouvé que le rendement des diplômes universitaires par rapport à la matière, à la mention et à l'université est dans une large mesure hétérogène. A partir de l'enquête sur l'emploi, cet article cherche à examiner l'hétérogénéité du rendement à travers les régions de la Grande-Bretagne. Il s'avère d'importantes variations régionales des récompenses financières offertes aux diplômés universitaires, dont des récompenses nettement plus élevées à Londres et dans le Sud-est qu'ailleurs. Une fois ajustées pour tenir compte des écarts des coûts de la vie, les différences se réduisent fortement. Une analyse par composantes laisse supposer que les effets des coefficients l'emportent sur les effets des composantes, ce qui correspond à l'importance des retombées d'agglomération et de productivité.

Education Diplômes universitaires Taux de rendement Régions

O'Leary N. C. und Sloane P. J. Die Höhe der Einkommenssteigerung durch Hochschulabschlüsse in verschiedenen britischen Regionen, Regional Studies. In früheren Beiträgen wurde eine ausgeprägte Heterogenität hinsichtlich der Einkommenssteigerung durch Hochschulabschlüsse festgestellt, was das Studienfach, die Art des Abschlusses und die Art der Hochschule anbelangt. In diesem Artikel untersuchten wir mit Hilfe des Labour Force Survey die Heterogenität der Einkommenssteigerung zwischen verschiedenen britischen Regionen. Wir stellten erhebliche regionale Schwankungen hinsichtlich der finanziellen Vorteile für Hochschulabsolventen fest; insbesondere fällt die Einkommenssteigerung in London und im Südosten Großbritanniens deutlich höher aus als anderswo. Diese Unterschiede verringern sich jedoch beträchtlich, wenn man die regionalen Unterschiede hinsichtlich der Lebenshaltungskosten berücksichtigt. Aus einer Dekompositionsanalyse geht hervor, dass die Koeffizienzeffekte die Kompositionseffekte dominieren, was zur Tatsache passt, dass die Effekte von Agglomeration und Produktivitätsübertragung eine wichtige Rolle spielen.

Bildung Hochschulabschluss Höhe der Einkommenssteigerung Regionen

O'Leary N. C. y Sloane P. J. Índices de los ingresos en función de los diplomas en las regiones británicas, Regional Studies. Se ha observado en otros documentos previos una heterogeneidad considerable en los ingresos en función de los diplomas con relación a las asignaturas, la clasificación de licenciaturas y la universidad. En este ensayo examinamos la heterogeneidad de las devoluciones en las regiones británicas mediante la Encuesta de Población Activa. Observamos variaciones regionales significativas en las recompensas financieras disponibles para licenciados que son muy superiores en Londres y el sureste que en otras regiones. Estas diferencias se limitan de manera significativa si se tienen en cuenta las diferencias regionales en cuanto al índice del coste de la vida. Los análisis de descomposición indican que los efectos de coeficiente dominan los efectos de composición de acuerdo con los efectos de aglomeración y desbordamientos de productividad que desempeñan un papel importante.

Educación Diploma Tasas de devoluciones Regiones

Keywords:

• Education
• Degree
• Rates of return
• Regions

Acknowledgements

The authors are grateful to Andy Dickerson and two anonymous referees for helpful comments. Material from the Labour Force Survey is Crown Copyright: it has been made available by the Office for National Statistics (ONS) through The Data Archive and has been used by permission. All views expressed in this work are entirely those of the authors and not those of either the ONS or the UK Data Archive.

Notes

1. Payments after graduation would be through the tax system, linked to ability to pay, with the threshold at which graduates would have to start repaying their fee contribution and maintenance loan fixed at £15 000.

2. More recent research has suggested that this figure of £400 000 somewhat overestimated the additional career earnings a graduate could be expected to earn. Based on regression analysis, O'Leary and Sloane (2005) estimate that a representative male graduate would enjoy a £141 539 lifetime earnings advantage and a female graduate would enjoy a £157 982 advantage. These figures are more in line with a figure of £120 000 referred to in a recent Hansard Report of 8 June 2005 by the Minister of Higher Education in response to a parliamentary question arising from the above study.

3. Notable exceptions to this are Bell and Sarajevs (2004) and Taylor and Wright (2005) who focus upon the position of graduates in Scotland.

4. A conventional approach to measuring the returns to degrees is to base estimates relative to those who could have pursued further education but chose not to do so. This comparator role is filled by those individuals who have gained two or more A-Levels.

5. Estimation of the earnings functions is by ordinary least squares (OLS). There is no doubt, however, that those who choose to undertake a degree are a self-selected group and likewise that the decisions over which subject to study and which institution to attend are not random. More generally, Dearden et al. (2002) highlight a number of potential sources of bias that may arise in OLS estimation. While a number of econometric techniques have been suggested to address these issues, the emerging view in the literature is that ability bias and measurement error more or less cancel each other out in OLS estimation (see, for example, Bonjour et al., 2003). Furthermore, sample selection does not appear in practice greatly to affect estimated returns to education (Dearden, 1999; Chevalier and Walker, 2001). Again, given that there is no evidence to suggest that such biases may have a regional impact, it is unlikely that the regional relativities presented in this work will be affected significantly by such issues.

6. Weale (1993) provides a good discussion of the likely biases that may arise in the calculation of both private and social rates of return. However, given the comparison group of 2 + A-Level holders used in this analysis and the fact that the focus is upon private returns, the only issue that may be relevant arises because of our assumption of continuous employment. Thus, if employment prospects are enhanced by a degree the true benefit of gaining such a qualification may be understated.

7. Since spring 1997, respondents to the LFS are asked about their earnings during their first (wave 1) and final (wave 5) interviews. In the analysis conducted here, we select only wave 1 respondents.

8. Graduates are defined as those who possess a university degree and will include those whose highest educational qualification is either a first degree or a higher degree. This classification is maintained in the rate of return calculations that follow and is adopted to account for any possible ability bias that might occur by concentrating upon undergraduate degree holders only. The reasoning behind this is that if the more able students go on to pursue postgraduate studies, excluding them would truncate the ability distribution and provide biased results. To counter this, all degree holders are retained and a dummy variable included denoting the possession of a higher degree. As it turns out, there is no evidence to support the hypothesis of the ability distribution being truncated and the results are unaffected by whether higher degree holders are included or not. Meanwhile, region is taken to denote region of work as opposed to the region where a graduate lives. Again, such a definition is maintained in the rate of return calculations that follow.

9. All returns for graduates are calculated from the same region so as to provide a common benchmark. The West Midlands was chosen as it fits the criterion of a representative region on a number of grounds: average earnings in the region are at the median of the distribution of earnings across all regions; the magnitude of its under-representation of graduates in its workforce is similar to that in many other regions; the size of the local labour market is reasonably large; the cost-of-living in the region is representative of the cost-of-living in the majority of other regions outside of London and the South East. It should be remembered, though, that the choice of the West Midlands is still arbitrary and any other region could have been selected. These returns are calculated for a representative individual, details of which are given in the notes to Table 2, and are based upon a ‘typical’ 3-year degree course. For this reason, holder of degrees in medicine, dentistry and languages are all excluded as such degree programmes will typically have a duration in excess of three years. Similarly, holders of degree-equivalent qualifications (including HNC/HNDs, teaching and nursing diplomas, and NVQs above level 3) have also been excluded as such courses will rarely fit the typical 3-year course offered by universities. It should be stressed, though, that these assumptions and those listed under Table 2 are not pivotal to the nature of the results and the implied regional relativities. The variables used in the rate of return calculations are described in Appendix Table A2.

10. The rates of return calculated in Table 2 are based upon the assumption that students do not work during the course of their studies. As such, if students have some sort of paid employment at the same time as they are studying, the implied annual rate of return to their increased career earnings will increase as the costs of obtaining these (in terms of foregone earnings) will have fallen. For example, assuming that a student works for 16 hours per week at the national minimum wage for 18–21 year olds of £4.25 (October 2005) for 42 weeks of the year, the calculated annual returns to a degree for men in Wales and London would increase to 5.2% and 29.9% respectively. It is likely that the returns reported in Table 2 will represent a lower bound to the returns available to graduates as some degree of part-time employment is becoming increasingly more important for university students. Furthermore, the question of whether such increasing distractions from studies will affect educational outcomes and subsequently returns is an interesting issue, but it is one that is ultimately beyond the scope of this current analysis.

11. Although annual rates of return are derived from additional career earnings, such earnings are not presented in Table 2 as it would be misleading to make a direct comparison between additional earnings in different regions as these are allowed to accrue to graduates at different rates across regions. As already shown in equations (1) and (2), the age-earnings profiles of workers (both graduates and non-graduates) are estimated separately for each region. Thus, it is not only how much more a graduate earns that is important, but also at what time in their career they earn it. It should be borne in mind that the estimated return of 5.3% for male graduates in Scotland is calculated on the basis of a four year degree course and not the three assumed for the regions of England and Wales. Some students in Scotland will, however, complete an ordinary degree in three years.

12. As a consistent series is not available for all years in our pooled sample, figures for 2004 are used to deflate nominal earnings in all years. For information on how the regional price series is constructed see Wingfield et al. (2005). As respondents do not necessarily work and live in the same region, the deflator used to calculate real earnings is based upon region of residence and not region of work.

13. Even excluding London, there is a fall in the spread of returns across regions from 6.5 percentage points to 4.0 percentage points.

14. Other than providing information on region of residence and region of work, the LFS does not contain any further details on commuting, such as commute time, which could be used as a control in the analysis.

15. As an alternative estimation strategy, a full set of regional dummies was included in the earnings functions modelled in equations (1) and (2) to capture regional fixed effects. For example, one might surmise that migration and commuting exert some particular influence around London and south east England that will not be picked up elsewhere in the analysis. However, the effect upon estimated rates of return of this alternative strategy was negligible and the pattern of returns identified above was unaltered (results available on request).

16. This is borne out of a priori expectations and inspection of the data and is also vindicated by the research of Blackaby and Manning (1990). In their examination of nominal earnings in the UK, regional cost-of-living differences and differences in the industrial and occupational base were found to be the major determinants of regional wage differentials between the South East of England and the rest of Great Britain. Thus, our examination of real earnings will concentrate upon regional differences in the mix of jobs across industrial and occupational dimensions. It should be noted, though, that such information is not included in the calculation of rates of return, as where graduates find themselves employed will be part of the return to education in itself.

17. The Duncan index of dissimilarity (Duncan and Duncan, 1955) is based upon absolute deviations in the percentage employed in each occupation and indicates the percentage of graduates in any region that would have to move between occupations to achieve distributional equality with London.

18. To calculate average earnings, a random sample of graduates in the North East was selected such that there was an exact match in the proportion in each of the five broad occupations used in Table 3(a) with that found in London. To avoid any potential bias, the sample was redrawn 1000 times and earnings averaged across all draws. The same procedure was repeated in each of the standard regions.

19. Given the distinct patterns of industrial clustering across regions, it comes as no surprise to note that the Duncan index for all regions (measured relative to London) is higher than that calculated along the lines of occupation.

20. Given the wage premiums afforded to public sector workers in the UK (Blackaby et al., 1996), public sector employment is seen as an important institution for holding up wage levels in many regions. Thus, the process of adjustment used above would reduce the number of relatively highly paid public sector workers and so have a detrimental effect upon average earnings levels.

21. The decomposition results that are presented have been derived from the following model:

where E denotes gross hourly earnings, Z represents a vector of characteristics that determine earnings (described in Appendix Table A2), ρ denotes a conformable vector of estimated rewards to these characteristics, a bar denotes a mean value, a hat denotes a coefficient estimate, and the superscripts L and O denote the London region and another comparison region respectively. The first term on the right-hand side will measure the extent to which differences in mean characteristics between regions explain the difference in average earnings and is labelled the composition effect. Meanwhile, the second term on the right-hand side will measure the extent to which differences in the way that these characteristics are rewarded across regions explain the difference in average earnings. This second term is labelled the coefficient effect.

22. While Table 5 presents only summary results of the decomposition analysis, full results, including a list of control variables that were included, are given in Appendix Tables A3(a) (males) and A3(b) (females).

23. The approach of Yun (2004) allows the identification of the individual components on the coefficient effect by imposing a linear restriction of zero upon any set of dummy variables (or any single category dummy variable). While this provides a tractable way of circumventing the baseline problem highlighted by Jones (1983), it does not solve the scale problem associated with the inclusion of continuous variables. For this reason, the underlying specification used in the estimation of the regional wage equations was composed entirely of dummy variables.

24. The constant term is the dominant element in all of the regional decompositions with the exception of London versus the East Midlands for male graduates. From Appendix Table A3(a) we can see that the difference in returns to the employment status dummy variable in this decomposition is substantially greater than for any other region. Why this should be the case is not clear. However, when the sample of graduates was restricted to full-time workers only, this anomalous result disappeared and the constant term emerged as the dominant element in all decompositions. Indeed, the magnitude of the influence of the constant term and its relative effect increased markedly when the analysis excluded part-time workers.

25. While we have raised the possibility of agglomeration and spillover effects being important, data limitations prevent an assessment of their relative contributions. To pursue this matter further would require data that could match workers within workplaces in addition to identifying individuals at a finer regional level than is publicly available within the LFS.

26. Consistent findings are presented by Dickerson (2006), where returns to National Qualifications Framework (NQF) at Level 4 + are lower in London as compared with the other regions of the UK.