The Impact of Distance to Nearest Education Institution on the Post-compulsory Education Participation Decision

Abstract

This paper uses data sources from England with the unique capacity to measure distances between home addresses and education institutions, to investigate, for the first time, the effect that such distance has on an individual’s post-compulsory education participation decision. The results show that there is a small overall effect. However, when attention is focused on young people who are on the margin of participating in post-compulsory education (according to their prior attainment and family background) and when post-compulsory education is distinguished by whether it leads to academic or vocational qualifications, then greater distance to nearest education institution is seen to have a significant impact on the decision to continue in full–time post-compulsory education.

1. Introduction

This paper examines the decision to participate in post-compulsory education in England, which occurs after year 11, the year in which pupils turn 16. A considerable literature exists that examines the various factors that influence this decision (see, amongst others, Ashford et al., 1993; Gray et al., 1993; Lenton, 2005; McIntosh, 2001; McVicar and Rice, 2001; Payne, 1998; and Rice, 1999). Repeated studies have found that the biggest influences on the participation decision are prior attainment and family background. Over and above the effect of such individual characteristics, geography and local area effects can also be important determinants of post-compulsory participation. More generally, geography and educational attainment are linked through the effects of social composition in local areas. As Butler and Hamnett (2007) point out, if different groups of the population were evenly distributed across space, and if education provision was also evenly spaced and homogeneous across areas, then average educational attainment should not vary across areas, only within areas according to ability, family background, etc. Of course, this is not what we observe and individuals with similar characteristics tend to group together. Thus we observe separate areas of high and low socioeconomic congregation, while it may also be the case that school resources vary systematically across areas, so that lower attainment in more deprived areas is due to poorer resources. The outcome is significant spatial variation in educational outcomes (Gordon, 1996). Of the two sources of spatial variation, evidence suggests that it is individual characteristics rather than school resources that are the more important (though see Taylor, 2001, for a discussion of variation across regions in choice and diversity within secondary education in England). For example, Webber and Butler (2007) show how average GCSE¹ attainment varies positively with the geodemographic status of the area in which a pupil lives, so that individuals from wealthier areas tend to have higher school attainment. Furthermore, the geodemographic status of a school’s pupil intake is also positively related to GCSE performance, so that even if a pupil lives in a low-status area, she will achieve higher attainment if she attends a school containing many pupils who live in high-status areas.

Therefore, not only an individual’s own family background, but also the geodemographic characteristics of their neighbourhood and school help to determine educational attainment. For example, Garner and Raudenbush (1991) show that, for a sample of young people in a district of Scotland, there is a negative effect of neighbourhood deprivation on educational attainment, that is observed after controlling for the pupil’s own ability, family background and schooling. Similarly, Gibbons (2002) observes a neighbourhood effect over and above all the usual individual and family controls, although the estimated effect is small. Leventhal and Brooks-Gunn (2004) show through a randomised experiment in New York that when low-income ethnic minority teenage children are moved to low-poverty neighbourhoods, their educational achievement improves, relative to their peers who remain in high-poverty neighbourhoods, with the effect particularly large for adolescent boys. As a final example, Gordon and Monastiriotis (2007) show significant effects of neighbourhood geodemographic characteristics on educational attainment and, further, that the transmission mechanism works through schools’ absence rates and proportions with Special Educational Needs. The latter result suggests that the influence of other children on one pupil’s educational outcomes can also work through disruption effects within schools.

The relevance of such peer effects has become more of an issue in inner-city urban areas. The gentrification of previously poorer areas of the big cities, increasing professionalisation of work leading to more middle-class families and higher aspirations of the less well-off,² have led to the merging of individuals from different backgrounds in such areas, and so a greater heterogeneity within schools in the social background and academic attainment of the peers that a family’s children might experience at school. This in turn has led, for some, to a hunt for better educational opportunities, through geographical relocation to areas with higher proportions of more prosperous families, and hence with schools that have a larger proportion of their intake from such families. The ‘flight’ from inner cities to suburban and rural areas is a well-known example of this. This increases the demand for housing in such areas, thus pushing up house prices so that only the well-off can afford them. As evidence, Gibbons and Machin (2003) find that a one percentage point increase in neighbourhood attainment at primary school is associated with a 0.67 per cent increase in neighbourhood property prices.

The aim of this paper is to investigate a particular aspect of geography that has received less attention in the literature as far as its effect on educational attainment is concerned—namely, distance to nearest educational institution and its influence on the decision to participate in post-compulsory education in England. Given the flight from inner-city areas by middle-class families, towards more suburban or rural areas, distance to nearest institutions may have become an increasingly important consideration when making post-compulsory choices. Such suburban and rural areas may have been convenient for attendance at local village primary schools, but are not necessarily so for the large schools, sixth-form centres and Further Education colleges³ that provide post-compulsory education, which are fewer in number than primary schools and therefore more spatially spread. Rural settings on average will be further away from educational institutions,⁴ so if distance is important, moving to a rural setting can affect post-compulsory participation.

Why should distance affect participation? Greater travelling distances to the nearest institution will increase the cost of staying in education, in financial terms through travel costs, in temporal terms through the lost time spent commuting and also in psychological terms caused by the inconvenience and possible unpleasantness of a lengthy commute every day.⁵ If participation in post-compulsory education is an investment decision to be undertaken when the present value of the future benefits outweighs the current costs, as suggested by human capital theory, then an increase in such costs can reduce the likelihood of undertaking the investment.

We would not expect this distance effect to dominate the other key drivers of post-compulsory participation discussed earlier—namely, prior attainment and family background. However, at the margin, when a young person is just undecided (or indifferent) between participating or not, distance to travel may just tip the balance one way or the other. This in turn could mean that distance has a differential effect on the decision to participate in academic or vocational education, if the decision to participate in vocational education is a more marginal decision relative to the choice of employment than the decision to participate in academic education, which may have been a longer-term aim. Few studies in the literature have considered how the likelihood of post-compulsory participation varies by type of education, with Conlon (2005) and Lenton (2005) being exceptions. Furthermore, the effect of distance may vary by other characteristics. For example, whether a location is urban or rural may make a difference. Being several miles distant from an educational institution may have a small impact in urban settings served by good public transport, but may be an impossible commute in a rural setting without private transport. Similarly, the impact of distance may vary by gender or ethnicity, if such characteristics affect parents’ willingness to allow their children to commute.

Another reason for studying the impact of distance on the post-compulsory education participation decision is that it is more amenable to immediate policy influence than the main determinants of participation, prior attainment and family background.⁶ For example, improved public transport or subsidised travel could both cut down on the costs of travelling to a place of learning, while the provision of distance learning is an alternative way to negate the impact of distance.

A final motivation for studying the effect of distance on education participation is a methodological one. An issue in the literature that studies the wage returns to education is the endogeneity of education, so that the higher wages of the better-qualified may be due to the fact that they were better motivated or more able in the first place (see, for example, Ashenfelter et al., 1999; Card, 1994; and Harmon et al., 2003). One possible solution to this econometric problem is to find an exogenous variable that affects individuals’ education choices, but has no effect on the wages that they eventually earn, which can then act as an instrument for education. Distance is one such potential instrument, if it can be shown to affect enrolment in further education. Such an argument is made by Card (1993), who shows that men who grew up in areas with a nearby college acquire more education on average; he then successfully uses the distance measure as an instrument in his wage equation.

There are a limited number of papers in the literature that have considered the effect of distance to nearest education institution on post-compulsory education participation. Most papers that do so focus on university participation. Research across a number of countries has shown that distance to the nearest university is negatively associated with the likelihood of attending university—for example by Frenette (2002) for Canada (particularly for women and individuals from low-income families), by Sa et al. (2004) for the Netherlands and by Spiess and Wrohlich (2010) for Germany. One issue with such research in the context of England, however, is that a large majority of English students do not go to the nearest university to their home and so the distance to their nearest university does not affect their decision-making. Consistent with this, Gibbons and Vignoles (2009) show a negligible effect of distance to nearest university on the likelihood of participating in higher education, although they do show that, conditional on participation, the probability of students attending a particular institution declines with the distance of that institution from their parental home (Kelchtermans and Verboven, 2006, find similar results for Flanders in Belgium). On the other hand, since most young people in England who attend further education between the ages of 16 and 18 do tend to remain in the family home, we would argue that distance to nearest institution is likely to be a more important determinant of participation in further education between the ages of 16 and 18, rather than of participation in higher education at university after the age of 18. We are not aware of any studies, however, which assess the impact of distance on immediate post-compulsory participation at age 16; hence the need for the present study.

The remainder of the paper is structured as follows. The following section describes the two datasets that we use to analyse the impact of distance on post-compulsory education participation, and the methodology used. The results of the analysis are presented in section 3, while a final section concludes.

2. Data and Methodology

In order to analyse the impact of distance on the post-compulsory participation decision, we use data from two recent cohorts of young people in England—namely, cohort 12 of the Youth Cohort Study (YCS), and the Longitudinal Study of Young People in England (LSYPE). Both of the surveys were developed and supplied by the UK government’s Department for Education.

2.1 The Youth Cohort Study

The YCS was designed primarily to provide information on young people’s transitions from compulsory education to further and higher education, and/or the labour market. Repeated cohorts have been surveyed. Here, we use the latest available full dataset, which is cohort 12. Survey participants were interviewed annually in 2004, 2005, 2006 and 2007. We focus on sweep 1 which was carried out in spring 2004 for individuals who completed their compulsory education 8 months earlier (i.e. were eligible to leave school for the first time in summer 2003). The YCS collects data on education and labour market activity, qualifications gained and sought, details on current employment, as well some background socioeconomic information about families and their attitudes. Coverage is England and Wales, although we use data for England only, to be consistent with the LSYPE.

2.2 The Longitudinal Study of Young People in England

The LSYPE is also focused on young people’s transitions from education into the world of work. It is a single cohort study, tracking a sample of young people from age 13/14 (year 9) in spring or summer 2004 in order to reach a better understanding of their development from their early teens while still in education (as compared with the YCS which only starts post-16). Interviews (known as ‘waves’) take place annually. We primarily focus on wave 3 conducted in 2006 to obtain data on the explanatory variables about the young people and their families. At this point, the respondents were (mostly) aged 16 and were coming towards the end of their compulsory education or, in some cases, had just completed it. The post-compulsory education participation variable in the LSYPE was derived from information on respondents’ current activity in wave 4 of the survey, conducted in 2007, one year after the end of their compulsory schooling.

The LSYPE questionnaires cover a broader range of topics than the (shorter) YCS questionnaire.⁷ Thus, in addition to the information collected in the YCS, the LSYPE also includes: attitudes to school and involvement in education; parental expectations and aspirations; risk factors (absences, truancy, police contact, bullying) and a range of parental questions.

2.3 Methodology

A multinomial logit model is estimated, since the dependent variable takes one of three values, indicating academic study, vocational study and no post-compulsory education. The latter formed the reference category. Only those in full-time education are considered.⁸

A range of explanatory variables is available in both datasets used. These include measures of respondents’ prior attainment which is indicated by whether they have achieved 5 or more GCSEs at grade C or above at the completion of compulsory schooling, and also whether these 5 or more successes included Maths and English. Standard demographic controls are available in both datasets, as well as indicators of family background such as type of housing tenure, parental occupation and parental education. Attitudes to schooling and education are measured by whether respondents report a history of truancy or have ever been suspended or expelled from a school.⁹ The only school characteristic observed in the YCS dataset is whether the young person attended an independent or grammar school in their last year of compulsory schooling. The LSYPE dataset, with matched-in administrative data, contains considerably more information about the school(s) attended and the respondents’ peer groups, such as the proportion of pupils in the school who achieve 5 or more good GCSEs, the proportion of pupils in receipt of free school meals, as an indicator of socioeconomic background of the school’s intake, and the unauthorised absence rate. There are other variables available in LSYPE but not in the YCS, in particular derived from a parental questionnaire asking about their attitudes to their children’s education. The post-compulsory participation rate is 75 per cent in the YCS and 72 per cent in LSYPE.¹⁰ As for background characteristics, the average values amongst the two cohorts of young people are very similar.¹¹

The main contribution of this paper is the analysis of the impact of distance to the nearest education institution on the post-compulsory participation decision at age 16. Distance variables were created by using information on individuals’ home locations based on their full postcodes. This information was combined with postcode information for all institutions providing post-compulsory education in England. The postcode information for individuals and institutions was then converted to grid references based on the centre of each postcode using GeoConvert.¹² Finally, ArcGIS¹³ was used to calculate the distance ‘as the crow flies’ for each individual from their home to their nearest education institution.

Two variables were created, one measuring the distance to the nearest institution offering ‘academic’ qualifications and the other measuring the distance to the nearest institution offering ‘vocational’ qualifications.¹⁴ Note that all types of institution offer academic qualifications. In classifying education institutions, it was assumed that only further education colleges and sixth-form centres offer vocational education, in addition to academic qualifications. Thus, for every individual, the distance to the nearest institution offering vocational education must be greater than or equal to the distance to the nearest institution offering academic education. As a consequence of this, on average, individuals live further from an institution offering vocational qualifications than from an institution offering academic qualifications. The average distance from an academic institution is 2.29 km in both datasets (with a standard deviation of 2.5 km in both datasets), while the average distance from a vocational institution is 5.96 km in the YCS and 5.63 km in LSYPE (with a standard deviation of 5.8 km in the YCS and 5.2 km in the LSYPE).

3. Results

3.1 Participation in Post-compulsory Education: The YCS

The results of the multinomial logit analysis of the YCS data are displayed in Table 1. The first row shows that the further individuals live from an institution offering academic qualifications, the less likely they are to undertake post-compulsory academic study, with each kilometre in distance reducing the likelihood of participating in academic education by 1.5 percentage points. A one standard deviation increase in distance to institutions offering academic qualifications is therefore associated with a 3.75 percentage point reduction in the probability of participation. By contrast, the further individuals live from an institution offering academic qualifications only, the more likely they are to undertake study of vocational subjects, by 0.9 percentage points per kilometre distance (or by 2.25 percentage points for a one standard deviation increase in distance). Given the overall participation rates (approximately one-half of 16/17-year-olds engaged in study for academic qualifications,one-quarter in study for vocational qualifications and one-quarter not participating in post-compulsory education) and the average distances involved to institutions, then the marginal effects per kilometre distance in Table 1 are not large. At the margin, the absence of a local institution providing academic qualifications will persuade a small number of individuals to switch to vocational study. The idea of the ‘marginal’ individual being influenced by distance will be developed further in the following sub-section. Note that the distance to an institution offering vocational qualifications does not seem to have any effect on the decision to participate in any form of post-compulsory education.

Table 1.

Multinomial logit for participation in post-compulsory education, by type of education: the YCS (N = 11 381)

	Academic study		Vocational study
Covariates	M.E.	(S.E.)	M.E.	(S.E.)
Distance from nearest academic institution in km	−0.015	(0.003)***	0.009	(0.002)***
Distance from nearest vocational institution in km	−0.000	(0.001)	−0.001	(0.001)
Lives in an urban area	−0.056	(0.020)***	0.028	(0.013)**
Female	0.043	(0.012)***	0.004	(0.008)
Belongs to ethnic minority	0.224	(0.020)***	−0.038	(0.012)***
5+ A*−C GCSEs or equivalent	0.452	(0.020)***	−0.215	(0.013)***
5+ A*−C GCSEs including Maths and English	0.298	(0.015)***	−0.168	(0.011)***
Has health problem or disability	-0.013	(0.033)	0.015	(0.020)
Played truant in year 11	−0.090	(0.013)***	0.009	(0.009)
Expelled or suspended in years 10 or 11	−0.127	(0.027)***	0.018	(0.017)
At grammar/independent school in year 11	0.279	(0.026)***	−0.161	(0.021)***
Lives in owner-occupied house	0.027	(0.025)	−0.009	(0.016)
Lives in a council house	−0.033	(0.031)	0.022	(0.018)
Lives with father only	−0.002	(0.028	−0.004	(0.018)
Lives with mother only	0.017	(0.020)	−0.005	(0.013)
Lives with neither parent	−0.057	(0.036)	−0.013	(0.023)
Only mother employed	−0.023	(0.027)	−0.010	(0.017)
Only father employed	0.019	(0.026)	−0.021	(0.016)
Both parents employed	−0.006	(0.023)	−0.017	(0.014)
Father’s occupation in SOC 1–3	0.068	(0.014)***	−0.040	(0.009)***
Mother’s occupation in SOC 1–3	0.057	(0.015)***	−0.019	(0.010)*
At least one parent with degree	0.127	(0.016)***	−0.048	(0.011)***
At least one parent with A level	0.062	(0.015)***	−0.018	(0.010)
Index of local deprivation	−0.097	(0.047)**	−0.013	(0.029)
Constant	−0.266	(0.043)	0.146	(0.027)

Briefly considering the other marginal effects in Table 1, post-compulsory participation is strongly related to prior attainment (GCSE scores) and family background, consistent with previous research. Over and above any distance effect, young people in urban areas are less likely to participate in academic post-compulsory education and more likely to participate in vocational post-compulsory education, relative to similar individuals in rural areas.

The results in Table 2 consider interaction effects between the distance variable and some of these other explanatory variables. None of the interaction coefficients are statistically significant at the 5 per cent level. This is still the case when the interactions are entered one at a time, rather than all at once as reported in Table 2, which might have caused multicollinearity leading to high standard errors. A couple of the interaction coefficients are statistically significant at the 10 per cent level. One such coefficient is that on the interaction between living in an urban area and distance to an academic institution, in the academic participation equation (coefficient of -0.011). The negative sign on this coefficient suggests that distance reduces participation in urban areas more than in rural areas. This is contrary to the hypothesis that distance might have less effect in urban areas due to more widespread public transport. Perhaps individuals in rural areas are simply more used to having to travel distances to access any amenities and so are less deterred by having to commute, or city-centre congestion is making commuting time greater than a simple measure of distance would suggest. This issue is one that would benefit from further research.

Table 2.

Multinomial logit for participation in post-compulsory education, by type of education, interactions effects: the YCS

	Academic study		Vocational study
Covariates	M.E.	(S.E.)	M.E.	(S.E.)
Distance from nearest academic institution in km	−0.010	(0.005)**	0.005	(0.003)
Distance from nearest vocational institution in km	−0.001	(0.002)	−0.000	(0.002)
Academic distance × urban	−0.011	(0.006)*	0.003	(0.004)
Vocational distance × urban	−0.001	(0.002)	0.001	(0.002)
Academic distance × female	−0.002	(0.005)	0.006	(0.003)*
Vocational distance × female	0.003	(0.002)	−0.003	(0.002)*
Academic distance × ethnic minority	0.013	(0.016)	−0.006	(0.009)
Vocational distance × ethnic minority	−0.009	(0.007)	0.000	(0.004)

Notes: The table presents marginal effects (M.E.) on the probability of participation. Standard errors (S.E.) in parentheses. *significant at 10 per cent; **significant at 5 per cent. All control variables shown in Table 1 are also included.

The other interaction coefficients to be statistically significant at the 10 per cent level are on the interaction terms between gender and distance in the vocational participation equation. These coefficients suggest that women are slightly more affected by distance than men, being less likely than men to undertake vocational post-compulsory education when the nearest vocational institution is further away and more likely than men to undertake vocational post-compulsory education when the nearest academic institution is further away.

3.2 Participation in Post-compulsory Education by Marginal Students: The YCS

It was suggested in both the introduction and in the previous sub-section that distance might have a greater impact on the marginal individual, who is indifferent between participating in post-compulsory education or not, or between undertaking academic or vocational study. The question remains, however, how to identify the marginal individuals in our analysis.

One possibility is to use the number of GCSEs achieved at the completion of compulsory schooling. The vast majority of those with 7 or more good GCSEs automatically follow the academic route, whilst this route appears closed to all but a few young people with 3 or fewer good GCSEs. However, those with 4–6 good GCSEs appear quite evenly split between undertaking academic post-compulsory education or not.¹⁵ It would appear that our marginal individual will therefore be found in this group. The hypothesis to be tested is therefore that the distance effect will be greater for young people in the 4–6 GCSE group than in any other GCSE group.

Table 3 reports the coefficients on the distance measures for four sub-groups of the cohort defined by GCSE performance. There is clear evidence in support of the hypothesis. Amongst young people with either 0 or 1–3 good GCSEs, distance to nearest academic institution has no impact on their academic participation, while there is only a small negative effect of distance to nearest vocational institution on vocational participation for the latter group. For these groups, their participation is low, particularly for academic qualifications which are more affected by distance effects in general as shown in Table 1. In most cases, they have already decided not to participate in academic post-compulsory study and the presence of a nearby academic institution is not going to change their minds. For individuals with 4–6 or 7+ good GCSEs, distance to nearest institution offering academic qualifications does affect their likelihood of participating in post-compulsory education. The marginal effects are statistically significantly larger for individuals in the marginal group with 4–6 good GCSEs, however. Thus, for each kilometre that this group live distant from an institution offering academic qualifications, their probability of undertaking post-compulsory academic study falls by 2.1 percentage points (or a 5.25 percentage point fall in participation for a one standard deviation increase in distance), while their probability of undertaking post-compulsory vocational study increases by 1.4 percentage points (or by 3.5 percentage points for a one standard deviation increase in distance). Thus, for these ‘marginal’ participants, distance is an important determinant of their participation decision.

Table 3.

Multinomial logit for participation in post-compulsory education, by type of education and number of GCSEs: the YCS

	Academic study		Vocational study
Covariates	M.E.	(S.E.)	M.E.	(S.E.)
0 GCSEs at grade A*–C
Distance from nearest academic institution in km	−0.000	(0.003)	0.012	(0.008)
Distance from nearest vocational institution in km	−0.000	(0.001)	−0.003	(0.003)
1–3 GCSEs at grade A*−C
Distance from nearest academic institution in km	−0.005	(0.005)	0.004	(0.007)
Distance from nearest vocational institution in km	0.000	(0.002)	−0.006	(0.003)**
4–6 GCSEs at grade A*−C
Distance from nearest academic institution in km	−0.021	(0.007)***	0.014	(0.006)**
Distance from nearest vocational institution in km	0.000	(0.002)	0.000	(0.002)
7+ GCSEs at grade A*−C
Distance from nearest academic institution in km	−0.005	(0.001)***	0.003	(0.001)***
Distance from nearest vocational institution in km	−0.000	(0.001)	0.000	(0.000)

Notes: The table presents marginal effects (M.E.) on the probability of participation. Standard errors (S.E.) in parentheses. **significant at 5 per cent; ***significant at 1 per cent. All control variables shown in Table 1 are also included.

We also considered alternative definitions of ‘marginal’ participants in post-compulsory participation defined in terms of family background. Assuming that marginal pupils are more likely to be found in lower-social-status families, our results found that the negative impact of distance on academic post-compulsory participation is indeed greater amongst young people from less socioeconomically advantaged families, when defined as low-education versus graduate parents, and also when comparing council-owned tenure with other forms of homeownership.¹⁶

3.3 Participation in Post-compulsory Education: The LSYPE

Table 4 reports the results from similar multinomial logit equations as before, but now using the LSYPE rather than the YCS data. The LSYPE contains considerably more information with which to control for other factors that might influence post-compulsory participation. In particular, there is much more information on the schools attended by respondents whilst in compulsory education, as well as more information on the attitudes to education of their parents. The results presented in this section are therefore a robustness check on the key results obtained with the YCS, to determine whether the same results are obtained when these additional controls for school quality and parental attitudes are included. Only the distance coefficients are reported, for reasons of space.¹⁷

Table 4.

Multinomial logit for participation in post-compulsory education: the LSYPE

	Academic study		Vocational study
Covariates	M.E.	(S.E.)	M.E.	(S.E.)
Panel A: full sample (N = 8708)
Distance from nearest academic institution in km	−0.010	(0.004)**	0.009	(0.003)***
Distance from nearest vocational institution in km	0.005	(0.002)**	−0.005	(0.002)***
Panel B: by GCSE attainment
0 GCSEs at grade A*–C
Distance from nearest academic institution in km	−0.000	(0.001)	−0.004	(0.010)
Distance from nearest vocational institution in km	−0.001	(0.001)	−0.003	(0.005)
1–3 GCSEs at grade A*–C
Distance from nearest academic institution in km	0.000	(0.005)	0.012	(0.009)
Distance from nearest vocational institution in km	0.004	(0.002)*	−0.010	(0.004)**
4–6 GCSEs at grade A*–C
Distance from nearest academic institution in km	−0.038	(0.012)***	0.024	(0.009)***
Distance from nearest vocational institution in km	0.006	(0.004)	−0.008	(0.004)**
7+ GCSEs at grade A*–C
Distance from nearest academic institution in km	−0.001	(0.002)	0.002	(0.002)
Distance from nearest vocational institution in km	0.001	(0.001)	−0.000	(0.001)

Notes: The table presents marginal effects (M.E.) on the probability of participation. Standard errors (S.E.) in parentheses. *significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent. Control variables for individual, school, family, parental attitude and area characteristics also included.

The results in Panel A of Table 4 show that this greatly expanded list of control variables does not significantly affect the size of the distance impact on participation. In the LSYPE, each additional kilometre distance from an institution providing academic qualifications reduces respondents’ likelihood of participating in academic post-compulsory education by 1 percentage point (a 2.5 percentage point fall for a one standard deviation increase in distance), while increasing their likelihood of participating in vocational post-compulsory education by 0.9 percentage points (a 2.25 percentage point increase for a one standard deviation increase in distance). These effects are similar in magnitude to the results reported in Table 1 for the YCS data. In addition, there is a disincentive effect of living at a greater distance from vocational education provision on vocational education participation in the LSYPE that was not observed in the YCS, of 0.5 percentage points per kilometre distance (or 2.6 percentage points for a one standard deviation increase in distance).

Panel B of Table 4 splits the LSYPE sample by prior attainment at GCSE level. The results are even more striking than the YCS results (compare Table 3) in that it is clearly only the marginal group with 4–6 good GCSEs where distance has any role in explaining post-compulsory participation. The marginal effects for this group are larger than those observed in Table 3 with the YCS. Each kilometre increase in the distance an individual with 4–6 good GCSEs has to travel to an institution offering academic qualifications reduces their likelihood of participating in academic post-compulsory education by 3.8 percentage points (or by 9.5 percentage points for a one standard deviation increase in distance), while raising their likelihood of participating in vocational post-compulsory education by 2.4 percentage points (or by 6 percentage points for a one standard deviation increase in distance). Table 4 also shows a smaller, but statistically significant, negative effect of distance to nearest institution offering vocational education on the likelihood of undertaking vocational post-compulsory learning for this marginal GCSE group.

3.4 Non-linear Effects of Distance

Thus far, the effect of distance on post-compulsory participation has been assumed to be linear. However, the effect may be non-linear, such that the negative influence of distance may have an increasingly larger effect the greater the distances involved, or may only have any impact at all for large distances. To investigate whether there are such non-linear effects, the continuous distance measures were replaced with dummy variables indicating 2–8 km from the nearest education institution and more than 8 km from the nearest education institution (with the base category being individuals who live less than 2 km away), separately for institutions providing academic and vocational qualifications.¹⁸ The marginal effects for the new distance variables are shown in Table 5 for the YCS and the LSYPE specifications that continue to control for the same covariates as used earlier. The full sample results in Panel A show significantly lower academic participation rates amongst those who live more than 8 km distance from an academic institution relative to those who live less than 2 km distance, by around 10 percentage points in both datasets. When the sample is split by level of GCSE attainment as in Panel B, the statistically significant effects are again clearly clustered amongst the marginal participation group with 4–6 good GCSEs. The effects are also very large in absolute size. Those young people who live more than 8 km distance from an academic institution are 18 percentage points (YCS) to 27 percentage points (LSYPE) less likely to participate in academic post-compulsory education, compared with those who live less than 2 km distance. Given that the academic post-compulsory participation rate of this group is 47 per cent (YCS) or 38 per cent (LSYPE), these marginal effects are clearly very large. In addition, in the LSYPE results only, there is a negative effect of living more than 8 km from an institution offering vocational qualifications on the likelihood of post-compulsory vocational participation amongst the marginal (4–6 good GCSE) group, by 15 percentage points relative to those who live less than 2 km distance.

Table 5.

Multinomial logit for participation in post-compulsory education, by type of education: distance categories

	YCS				LSYPE
	Academic		Vocational		Academic		Vocational
Covariates	M.E.	(S.E.)	M.E.	(S.E.)	M.E.	(S.E.)	M.E.	(S.E.)
Panel A: full sample
Distance to academic institution 2–8km	−0.067	(0.015)***	0.041	(0.010)***	−0.038	(0.021)*	0.042	(0.015)***
Distance to academic institution >8km	−0.099	(0.034)***	0.062	(0.022)***	−0.103	(0.046)**	0.086	(0.035)**
Distance to vocational institution 2–8km	0.021	(0.016)	−0.014	(0.010)	0.020	(0.019)	−0.030	(0.014)**
Distance to vocational institution >8km	0.012	(0.020)	−0.021	(0.013)	0.056	(0.028)**	−0.051	(0.021)**
Panel B: by GCSE attainment
0 GCSEs at grade A*−C
Distance to academic institution 2–8km	−0.017	(0.014)	0.086	(0.036)**	0.001	(0.021)	0.055	(0.044)
Distance to academic institution >8km	−0.011	(0.025)	0.055	(0.081)	−0.000	(0.065)	−0.004	(0.118)
Distance to vocational institution 2–8km	−0.004	(0.011)	0.009	(0.036)	−0.001	(0.005)	−0.044	(0.041)
Distance to vocational institution >8km	0.001	(0.015)	−0.047	(0.049)	−0.005	(0.009)	−0.001	(0.064)
1–3 GCSEs at grade A*−C
Distance to academic institution 2–8km	−0.022	(0.023)	−0.020	(0.034)	−0.007	(0.023)	0.055	(0.044)
Distance to academic institution >8km	0.010	(0.056)	−0.007	(0.079)	−0.030	(0.060)	0.130	(0.101)
Distance to vocational institution 2–8km	0.027	(0.022)	−0.014	(0.034)	−0.010	(0.019)	−0.024	(0.041)
Distance to vocational institution >8km	0.019	(0.029)	−0.045	(0.044)	0.034	(0.030)	−0.092	(0.060)
4–6 GCSEs at grade A*−C
Distance to academic institution 2–8km	−0.100	(0.033)***	0.071	(0.028)***	−0.126	(0.045)***	0.064	(0.039)
Distance to academic institution >8km	−0.183	(0.082)**	0.111	(0.067)*	−0.269	(0.115)**	0.198	(0.096)**
Distance to vocational institution 2–8km	0.040	(0.034)	−0.029	(0.029)	0.032	(0.040)	−0.073	(0.035)**
Distance to vocational institution >8km	0.026	(0.045)	−0.042	(0.038)	0.111	(0.060)*	−0.152	(0.054)***
7+ GCSEs at grade A*−C
Distance to academic institution 2–8km	−0.021	(0.007)***	0.012	(0.005)**	0.000	(0.011)	0.010	(0.008)
Distance to academic institution >8km	−0.035	(0.014)**	0.025	(0.010)**	−0.016	(0.022)	0.021	(0.017)
Distance to vocational institution 2–8km	0.004	(0.008)	−0.005	(0.006)	0.003	(0.011)	−0.003	(0.009)
Distance to vocational institution >8km	0.002	(0.010)	−0.006	(0.007)	0.009	(0.014)	−0.000	(0.011)

Notes: The table presents marginal effects (M.E.) on the probability of participation. Standard errors (S.E.) in parentheses. *significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent. All control variables previously used are also included.

4. Conclusions

The geography of education matters. A number of papers in the literature have shown that the geodemographics of the area in which families live will have an effect on the educational participation and attainment of their children, over and above the combined influence of the family’s own background and the ability of their children (for example, Gibbons, 2002; Leventhal and Brooks-Gunn, 2004; and Webber and Butler, 2007). While these family characteristics are the dominant influences, at the margin, the smaller effects of local area characteristics can make a difference between success and failure, and families are willing to seek out areas with desirable geodemographic characteristics, if that even slightly improves their children’s chances of success. Usually such location decisions are made in relation to the primary schools and secondary schools that their children will attend, with the availability of post-compulsory institutions not of immediate concern. If that means that families, having moved to suburban or rural areas, find themselves some distance from post-compulsory education provision, does this mean that such distance will affect the post-compulsory choices of their children?

This paper has used the unique provision of distance information to investigate, for the first time in the literature, the impact of distance to one’s nearest education institution, on the probability of participating in post-compulsory education, specifically at age 16. For many individuals, distance is not an issue, as they live within easy (walking or cycling) distance of their nearest education institution. In addition, many young people have their post-16 lives already clearly mapped out, in terms of further academic study if they have achieved good results in their GCSE examinations, or leaving full-time education if they have obtained few or no GCSEs. There is, however, a group of young people who are on the margin between participating in post-compulsory education or not. The results presented here show that, whether we define the marginal participant in terms of having a mid-level of prior attainment (4–6 good GCSEs), or in terms of coming from a relatively socioeconomically disadvantaged background, we consistently find that distance to nearest institution can affect the decision to partake in post-compulsory education. These results are robust across two separate datasets. Some of the effects of distance become very large, with the biggest effect suggesting that those who live more than 8 kilometres distance from an academic institution are 27 percentage points less likely to participate in academic post-compulsory education, compared with those who live less than 2 kilometres distance (Table 5, the LSYPE). Since only a small minority of young people live at such distances from their nearest education institution, and the result only applies to marginal students living at such distances, this result is clearly not widely applicable. However, for those individuals affected, distance to the nearest education institution is clearly an important aspect to take into consideration when deciding whether to participate in post-compulsory education.

Therefore, while distance is not the dominant influence on post-compulsory education choices, it can make an important difference at the margin for those with a close decision to make about whether to participate or not. Thus, moving to and living in an area with more distant post-compulsory education provision may have a negative effect on participation in such education. There is no evidence, however, that distance is more of a problem in rural areas than in urban areas. Indeed, if anything, the effect of distance on participation is greater in urban areas than in rural areas (the difference in the effect of distance between rural and urban areas being statistically significant at the 10 per cent level). This result is not a function of the characteristics of the rural and urban populations, since the analyses control for family background and local area deprivation scores.

It therefore would not appear that more limited or less widespread provision of post-compulsory education in rural areas is a major issue, given the result just mentioned, and given the fact that the marginal students with only average GCSE results or from less well-off families are, on average, less likely to be found in rural areas. It is therefore in urban areas that the issue of distance is in greater need of being addressed, given the greater impact of distance there and the higher likelihood of finding marginal students there.

As far as policy implications are concerned, the analysis cannot identify the cause of the distance effect. Thus, it is not known whether greater distances reduce the incentive to participate because of financial travel costs, or time and psychological (boredom) costs of frequent long commutes. If the former turns out to be the cause, then subsidised travel for those participating in post-compulsory education would be an effective policy. If the latter is the cause, then one policy response would be to improve the frequency of travel services and reduce travel time perhaps through priority for public transport on congested city-centre roads. Since distance affected only marginal students, making any such policies universally applicable would involve considerable deadweight. Targeted policies on those individuals whose decisions are influenced by distance would therefore be the most effective. Clearly, it would be easier to target financial travel subsidies than general travel services. It would also be difficult to justify targeting subsidised travel specifically on the group with 4–6 GCSEs. The other definitions of the marginal participant, in terms of family background will therefore be more useful, with subsidised travel for students from less well-off families being an appropriate policy response.

Alternatively, if distance is discouraging post-compulsory participation amongst some young people, then distance learning policies need to be considered. The technology is increasingly available to make distance learning a more engaging possibility, from remote access to libraries, through teaching websites where course material can be posted, to the use of podcasts and videos of lectures and classes. Thus, providing resources to education institutions to make distance learning possible, as well as helping less well-off families with the facilities to access remotely provided material, is another policy response to the issue of distance. There is another issue to raise here, however, in that distance learning may be more appropriate for academic learning than for vocational education where there is often a requirement for physical presence for at least part of the syllabus, when manual skills are learned and practised. Given that vocational education rather than academic education will be considered more by the marginal student than on average in the population as a whole, distance learning may not be the solution to distance effects in all cases.

A final issue addressed by this paper is an econometric point. The results presented here have relevance for any empirical field where an instrument is required for education or qualifications acquired—for example, in the returns to education literature. Given the effect of distance on participation, and assuming family household location is judged to be exogenous as far as the future earnings of the children are concerned, then distance to nearest institution at age 16 appears to be a good instrument for education participation.

Footnotes

Notes

Funding

This research was funded in part by DEFRA under their Rural Research Programme (grant number CTX 0808 002).

References

Ashenfelter

Harmon

Oosterbeek

(1999) A review of estimates of the schooling/earnings relationship, with tests for publication bias, Labour Economics, 6(4), pp. 453–470.

Ashford

Gray

Tranmer

(1993) The introduction of GCSE exams and changes in post-16 participation. Youth Cohort Report No. 23, Employment Department Research Series, London.

Butler

Hamnett

(2007) The geography of education: introduction, Urban Studies, 44(7), pp. 1161–1174.

Card

(1993) Using geographic variation in college proximity to estimate the return to schooling. Working Paper No. 4483, National Bureau of Economic Research, Cambridge, MA.

Card

(1994) Earnings, schooling and ability revisited. Working Paper No. 4832, National Bureau of Economic Research, Cambridge, MA.

Conlon

(2005) The determinants of undertaking academic and vocational qualifications in the United Kingdom, Education Economics, 13(3), pp. 299–313.

Frenette

(2002) Too far to go on? Distance to school and university participation. Analytical Studies Research Paper No. 191, Statistics Canada.

Garner

Raudenbush

(1991) Neighborhood effects on educational attainment: a multilevel analysis, Sociology of Education, 64(4), pp. 251–262.

Gibbons

(2002) Neighbourhood effects on educational achievement: evidence from the census and National Child Development Study. Discussion Paper No. 18, Centre for the Economics of Education, London School of Economics.

10.

Gibbons

Machin

(2003) Valuing English primary schools, Journal of Urban Economics, 53(2), pp. 197–219.

11.

Gibbons

Vignoles

(2009) Access, choice and participation in higher education. Discussion Paper No. 101, Centre for the Economics of Education, London School of Economics.

12.

Gordon

(1996) Family structure, educational achievement and the inner city, Urban Studies, 33(3), pp. 407–424.

13.

Gordon

Monastiriotis

(2007) Education, location, education: a spatial analysis of English secondary school public examination results, Urban Studies, 44(7), pp. 1203–1228.

14.

Gray

Jesson

Tranmer

(1993) Boosting post-16 participation in full-time education: a study of some key factors. Youth Cohort Report No. 20, Employment Department Research Series, London.

15.

Harmon

Oosterbeek

Walker

(2003) The returns to education: microeconomics, Journal of Economic Surveys, 17(2), pp. 115–155.

16.

Kelchtermans

Verboven

(2006) Participation and schooling in a public system of higher education. Discussion Paper No. 5690, Centre for Economic Policy Research, London.

17.

Lenton

(2005) The school-to-work transition in England and Wales, Journal of Economic Studies, 32(2), pp. 88–113.

18.

Leventhal

Brooks-Gunn

(2004) A randomized study of neighborhood effects on low-income children’s educational outcomes, Developmental Psychology, 40(4), pp. 488–507.

19.

Lucas

Heady

(2002) Flexitime commuters and their driver stress, feelings of time urgency, and commute satisfaction, Journal of Business and Psychology, 16(4), pp. 565–571.

20.

McIntosh

(2001) The demand for post-compulsory education in four European countries, Education Economics, 9(1), pp. 69–90.

21.

McVicar

Rice

(2001) Participation in further education in England and Wales: an analysis of post-war trends. Oxford Economic Papers, 53(1), pp. 47–66.

22.

Novaco

Stokols

Campbell

Stokols

(1979) Transportation, stress and community psychology, American Journal of Community Psychology, 7(4), pp. 361–380.

23.

Payne

(1998) Routes at sixteen: trends and choices in the nineties. Research Report No. 55, Department for Education and Employment, London.

24.

Reay

David

Ball

(2005) Social Class, Gender and Higher Education. Stoke-on-Trent: Trentham Books.

25.

Rice

(1999) The impact of local labour markets on investments in further education: evidence from the England and Wales Youth Cohort Studies, Journal of Population Economics, 12(2), pp. 287–312.

26.

Florax

Rietveld

(2004) Determinants of the regional demand for higher education in the Netherlands: a gravity model approach, Regional Studies, 38(4), pp. 375–392.

27.

Spiess

C. K.

Wrohlich

(2010) Does distance determine who attends a university in Germany?, Economics of Education Review, 29(3), pp. 470–479.

28.

Stutzer

Frey

(2008) Stress that doesn’t pay: the commuting paradox, Scandinavian Journal of Economics, 110(2), pp. 339–366.

29.

Taylor

(2001) The geography of choice and diversity in the ‘new’ secondary education market of England, Area, 33(4), pp. 368–381.

30.

Webber

Butler

(2007) Classifying pupils by where they live: how well does this predict variations in their GCSE results?, Urban Studies, 44(7), pp. 1229–1253.