Students' expectations of debt in UK higher education

• 1. This article was downloaded by: [University of Westminster - ISLS] On: 05 December 2013, At: 13:11 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UKStudies in Higher Education Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/cshe20Students' expectations of debt in UK higher education Ray BachanaaBrighton Business School , University of Brighton , Brighton , UK Published online: 16 Jan 2013.To cite this article: Ray Bachan , Studies in Higher Education (2013): Students' expectations of debt in UK higher education, Studies in Higher Education, DOI: 10.1080/03075079.2012.754859 To link to this article: http://dx.doi.org/10.1080/03075079.2012.754859PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/termsand-conditions

2. Studies in Higher Education 2013, 126, iFirst ArticleStudents expectations of debt in UK higher education Ray Bachan*Downloaded by [University of Westminster - ISLS] at 13:11 05 December 2013Brighton Business School, University of Brighton, Brighton, UK The funding of students in UK higher education (HE) has undergone radical reform over the past two decades. Using a unique dataset, this paper investigates student expectations of debt. We nd that a students gender, ethnicity, and year of study play an important role in determining their expected debt. Students in receipt of nancial support from their parents and those with part-time jobs anticipate a lower level of debt. We also nd that the higher a student discounts future income, the greater their expected debt, and the more risk averse a student the lower the expected debt. Our ndings also suggest that a students expected earnings post-graduation has a positive impact on current debt. If the level of student debt which is expected to rise in the future compromises access to HE, then the close targeting of student nancial aid is warranted and an equitable repayment framework needs development. Keywords: student debt; debt expectations; higher education; tobitIntroduction The UK higher education sector has grown rapidly over the last 50 years. This expansion was particularly evident in the 1960s in the wake of the publication of the Robbins Report (1963), which created a rst wave of new universities, and in 1992 following the introduction of the Further and Higher Education Act, when former polytechnics were granted university status. A key feature of this expansion was to widen participation rates amongst young adults from disadvantaged socioeconomic backgrounds and to encourage participation of individuals from families with no previous participation in higher education. The policy was given major impetus in September 1997 when the then prime minister declared a desire to increase the higher education participation of young adults by the next century.1 The government set out to widen and increase higher education participation to 50% in the 1730 year age category by 2010 (Department for Education and Skills 2003a, 57). Recent gures suggest that initial participation rates for all 17- to 30-year-olds in UK higher education in 2010/ 2011 were about 47%; up from about 42% in 2006/2007 (Department for Innovation, Universities, and Skills 2012).2 The rapid expansion of the UK higher education sector and the desire to widen participation, particularly in the 1990s, led to a funding crisis and the adoption of student loans as an alternative method for funding student higher education (Barr and Crawford 1998). Under such a system it was envisaged that the limited public funds available for student support should be targeted on those whose personal nancial circumstances *Email: [email protected] ISSN 0307-5079 print/ISSN 1470-174X online 2013 Society for Research into Higher Education http://dx.doi.org/10.1080/03075079.2012.754859 http://www.tandfonline.com 3. Downloaded by [University of Westminster - ISLS] at 13:11 05 December 20132R. Bachancould hinder access to higher education. Student loans were introduced in 1990/1991 to reduce the public subsidy towards student living costs, with the remainder provided by a combination of educational grants, bursaries, and parental contributions. Over time these loans gradually replaced the existing maintenance grant and by the academic year 1996/1997 the maximum loan accounted for 50% of the grant. In 1998 mandatory grants were replaced by income-assessed loans and new entrants to UK higher education received all their maintenance allowance in the form of an income-assessed loan. At the same time means-tested tuition fees were also introduced. The 2004 Higher Education Act viewed student loans as a means by which students, after graduation, could repay their tuition fees. In the academic year 2006/2007 new students attending higher education institutions in England and Northern Ireland were charged a variable fee of up to GBP3000 as a contribution to the cost of their tuition. Students were given the option to take out a tuition-fee loan to cover these extra costs. Tuition fees were raised incrementally in line with ination and by 2010/ 2011 the standard fee for full-time UK-domiciled students was GBP3290 set to rise to a maximum of GBP9000 per annum in 2012/2013 (Browne 2010). These changes in student nancial support meant that graduate indebtedness was expected to rise considerably and particularly from 2006/2007 onwards. Nevertheless, advocates of student loans often argue that students are the main beneciaries of higher education (e.g. in terms of enhanced future earnings) and should contribute to its cost (Barr 2004; Friedman and Friedman 1980; Glennerster et al. 1968). This argument is grounded in human capital theory (Becker 1993) which assumes that students are able to rationally assess the cost and benets of investing in human capital (post-compulsory education). Such investment is expected to enhance future productivity which in turn raises the individuals future labour market earnings. If this assumption holds then one may expect students in general to have a positive attitude towards using loans as a means of nancing their higher education. This may be true for students who expect high future earnings with a low risk of unemployment. Thus the level of debt students anticipate by the end of their studies may reect the higher future earnings they expect in the graduate labour market. Several UK studies have found evidence of substantial returns to higher education qualications (Harkness and Machin 1999; Blundell et al. 2000; Bratti, Naylor, and Smith 2007), which is consistent with international evidence (Psacharopoulos and Patrinos 2004). However, the evidence is more mixed as to whether or not students can form realistic expectations of their future earnings. For the UK, Jerrim (2011) nds that rst-year full-time students over-estimate their expected starting salary by 20% and expectations become more realistic over time with nal-year students overestimating their starting salary by about 15%. Brunello, Lucifora, and Winter-Ebmer (2004) found that business and economics students in 10 European countries also tend to over-estimate their future earnings, but Dutch and US students in contrast made more accurate predictions (Webbink and Hartog 2004; Dominitz and Manski 1996; Betts 1996). There is another strand of literature which interprets a system of student loans, and the fear of indebtedness, as a potential barrier to access to higher education. This is particularly evident for potential students from lower socioeconomic groups, which may compromise the stated policy objective to widen participation (Connor et al. 2001; Knowles 2000; Callender 2003; Callender and Jackson 2005; Pennell and West 2005). Moreover, there is evidence that nancial support for students is inadequate (Callender and Wilkinson 2003), which can contribute to rising student 4. Studies in Higher Education3debt. Government concern over the perception of student debt and participation in higher education was highlighted in the lead up to the 2004 Higher Education Act (Department for Education and Skills 2003b). Indeed in 2001, the then Secretary of State for Education, Estelle Morris, highlighted the issue of student debt and its negative impact on widening participation in higher education:Downloaded by [University of Westminster - ISLS] at 13:11 05 December 2013I recognise that for many lower-income families the fear of debt is a real worry and could act as a bar to higher education. I want to make sure that our future reform tackles this problem. (Department for Education and Skills 2001)3Furthermore, the expectation of a high level of indebtedness can have further unintended consequences especially for those students from disadvantaged socioeconomic groups and students with limited nancial support from their parents. For instance, it can potentially reduce the time these students devote to study through taking on paid part-time employment by necessity (Purcell and Elias 2010),4 which can adversely affect academic achievement (Ford et al. 1995; Humphrey 2006; Brennan et al. 2005; Callender 2008). Evidence also suggests that the fear of debt can affect a students desire to continue in higher education (Davies and Elias 2003; Yorke and Longden 2008) and future labour-market choices (Purcell and Elias 2010). Studies in the United States also nd that nancial pressure is positively associated with term-time working which can have a detrimental effect on academic performance and future choices (Strinebrickner and Strinebrickner 2003; Chapman and Lounkaewa 2010; Kalenkoski and Pabilonia 2010; Zhang 2010; Scott-Clayton 2012). The foregoing review suggests that changes in public nancial support for students in UK higher education have impacted student indebtedness, the student experience and performance, and their future career choices. It is not surprising that rising student debt has become a major issue of public concern. However, the determinants of student debt expectations are relatively under-researched in the UK, although several studies have focussed on student attitudes to debt (see for example, Davies and Lea 1995; Callender 2003; Brennan et al. 2005; Callender and Jackson 2005; and Johnson et al. 2009). A notable exception is the study by Purcell and Elias (2010) who explored the extent to which nal-year students worried about debt using OLS regression analysis. They nd evidence that male students are less fearful of indebtedness compared to their female counterparts and higher expected earnings on graduation lessened the extent of these worries. In addition, they found students were more fearful of debt in their nal year of study than they were in their rst. This study contributes to the literature in several important ways. First, it uses unique survey data to examine a variety of factors expected, a priori, to inuence student debt expectations. The identication of these factors is important from a policy perspective, particularly when considering the targeting of student nancial support and the design of debt repayment schemes. Second, the study examines the impact that students attitudes to debt and risk and their future wage expectations have on their expected debt during their period of study. Third, this study offers a novel contribution to the sparse literature that exists on the determinants of student debt expectations. This paper is organised as follows. The next section presents a brief background to the current state of student debt in UK higher education. This is followed by a description of the data, and then an outline of the statistical methodology employed. The penultimate section presents the results of the empirical analysis, and the nal section provides some conclusions. 5. Downloaded by [University of Westminster - ISLS] at 13:11 05 December 20134R. BachanStudent debt in the UK Student debt 1990/1991 to 2005/2006 Figure 1 shows that the average loan taken out by eligible students, domiciled in England and Wales, increased considerably over the period 1990/1991 to 2005/2006. We note that in the academic year 1990/1991, when maintenance loans were introduced, the average student loan was GBP390 in 1991 prices, and over the decade to1998/1999 increased to GBP1509 in real terms as loans replaced maintenance grants. Between 1998/1999 and 1999/2000 the maintenance allowance was replaced by a means-tested loan and the real average value of the loan increased by just over 33%. By the academic year 2005/2006, the average student loan stood at GBP2338. This increase in the size of the average loan was also accompanied by an increase in the number of students taking out loans. In the academic year 1990/1991, 180,200 students secured a loan, which represented a 28% take-up rate across all eligible students. By 1999/2000, 699,700 students obtained a loan, representing a 72% take-up rate which rose to 80% representing 880,700 students by 2005/2006 (Student Loans Company). These trends were anticipated given the changes in student funding arrangements in England and Wales. Surprisingly, given the growing public concern over student debt, there is a dearth of ofcial statistics on the level of actual student debt post-graduation between 1990/ 1991 and 2006/2007. The Barclays Bank Graduate Debt Survey (2005), covering the period 1994 to 2004, showed that student debt for new graduates increased in real terms by a factor of 3.7 from GBP2047 in 1994 to GBP9653 by 2004. The Natwest survey (2007) suggested that in real terms student debt increased by aFigure 1. Average student loan (GBP) 1990/19912005/2006 (1991 = 100). English- and Welsh-domiciled students. Source: Student Loans Company available at http:www.slc.co.uk/ and authors own calculations. 6. Studies in Higher Education5Downloaded by [University of Westminster - ISLS] at 13:11 05 December 2013factor of just over 3.5 between 2000 and 2006 in 2005 real student debt was GBP8789 and in 2006 it was GBP8929 (see Bolton [2009] for details). These surveys are not directly comparable but illustrate the increase in student debt between 1994 and 2006 (see Figure 2). Data from other surveys broadly support these trends. The Push survey (2007) showed average student debt remaining static or even falling in 2004 and 2005, and the Unite survey (2007) revealed a small decline in average debt among students in 2005. Callender and Wilkinson (2003) and Callender and Kemp (2000) report that average student debt amounted to GBP8666 in 2003 2.5 times higher than for those graduating in 1998 and 3.5 times higher than for those graduating in 1996. Student debt post 2006/2007 Figure 3 reveals that between the period 2006/2007 to 2010/2011 the average maintenance loan decreased by about 4.6% and tuition fee loans remained relatively constant in real terms, which is due in part to the increase in the maximum tuition fee institutions were permitted to charge. Overall, the average level of total student public borrowing decreased in real terms by about 3%. It is also worth noting that the number of student borrowers increased from 2.9 million in 2008/2009 to 3.2 million in 2009/2010 (Student Loans Company 2010). There is very little ofcial UK data on the level of debt experienced by graduates who enrolled in the academic year 2006/2007 or later. The 2007/2008 Student Income and Expenditure Survey, covering only English-domiciled students, found that student loans made up 88% of all student borrowing and average debt for nal-yearFigure 2. Average student debt (GBP) 19942006 (1991 = 100). English-domiciled students. Note: It should be noted that the gures reported above do not apply to students who were liable for variable tuition fees. The data for both series were derived from a survey of graduates carried out 618 months after graduation. Source: P. Bolton (2009) and authors own calculations. 7. Downloaded by [University of Westminster - ISLS] at 13:11 05 December 20136R. BachanFigure 3. Average maintenance loan and tuition fee loan (GBP) (1991 = 100). England 2006/ 20072010/2011. Note: These gures represent awards to English-domiciled students who entered university in November of the relevant academic year irrespective of where they study. Source: Students Loans Company, statistical rst release (November, various years)full-time students graduating in 2008, who were not part of the new student funding regime, was estimated to be GBP7783 (Johnson et al. 2009). Purcell and Elias (2010), using more recent data, report that full-time UK students graduating in 2009 anticipate an average debt of about GBP15700 on graduation. However, the Push survey (2010) estimated that students graduating in 2009 expected a greater amount of debt totalling GBP22,000 on average, and those graduating in 2010 or 2011 reckoned to be in debt by an average amount of GBP23,000 and GBP24,700 respectively. The portrait that emerges from this evidence is a signicant increase in student expected indebtedness from 2009 onwards. It is worth noting that international evidence on the debt students expect on graduation is rare and studies often focus on the repayment of debt post-graduation (see for example, Shen and Zilderman 2009; Chapman and Lounkaewa, 2010; Chapman and Sinning, 2011). Usher (2005), however, reports a wide variation in student debt burden post-graduation across eight countries including the UK. The study nds that UK students have the fourth-highest debt at graduation and only Swedish, Canadian, and US students having higher debts. Recent evidence for the United States suggests that students graduating in 2009 from private not-for-prot higher education institutions who relied on student loans to fund their undergraduate studies had an average debt of USD26,200 (GBP16,725), and those graduating in 2010 an average debt of USD28,100 (GBP18,175) (College Board 2011).5 It is not easy to compare these gures with student debt in the UK, but if student debt expectations are realised then these gures may suggest that UK graduates will be at least as indebted as their US counterparts in the near future. 8. Downloaded by [University of Westminster - ISLS] at 13:11 05 December 2013Studies in Higher Education7Data The data for this study were collected through a questionnaire administered to 425 students in lectures and seminars who were enrolled full-time on three-year business undergraduate degree programmes at a UK university during January/February 2009. This means that the nal year students sampled commenced their university career during the autumn of 2006 and were among the rst to face a charge of up to GBP3000 to cover the cost of their tuition. Information was collected on a students socioeconomic characteristics, their programme of study, expected debt at the end of their programme of study, earnings expectations, risk attitudes, and time preference. After allowance was made for missing values a sample of 308 useable observations was obtained. The descriptive statistics for the variables used in the empirical analysis are presented in Table 1. Column 1 reports the summary statistics for the sample as a whole and columns 2 and 3 report the sample means and proportions by gender. We rst note that the proportion of undergraduate students expecting to be in debt at the end of their time at university is just over 81%. The full sample of students report an average expected debt of GBP14,022 and those who anticipate to be in debt by the end of their undergraduate studies reckon for an average level of debt of GBP17,276. This latter gure is within the range of expected debt levels reported for students graduating in 2009 (Purcell and Elias 2010; Push 2010). Although not reported in the table it is interesting to note that expected debt declines by cohort. First-year students expect an average debt of GBP14,556 by the end of their degree programme and the corresponding gures for second and third/nal year students are GBP14,211 and GBP11,950 respectively. The level of debt reported for third/nal year students, however, is lower than that reported by Purcell and Elias (2010). There is a dominance of male students (59%). A large proportion of the sample are classied as white British (61%) with 14% classied as white non-British and 25% from other ethnic groups (see notes to Table 1). The sample average age is just over 20, about 29% hold a part-time job during term time which is lower than proportions generally reported in national surveys, 85% are from families that own their own home, and the majority are in their rst year of study. In terms of student nance about 51% of students report they are in receipt of a student grant, scholarship or bursary, and just over 46% receive parental contributions averaging about GBP152 per month. In terms of future graduate earnings students expect to earn on average GBP25,108 per annum in their rst post-graduation job, which is higher than that reported in other studies (see, Johnson et al. 2009; Purcell and Elias 2010). By the time the individual reaches the age of 30 they expect to be earning, on average, GBP50,397 per annum. It is also interesting to note that for these students they would expect to be earning about GBP5900 per annum less had they not studied for a degree by the time they reach 30 years of age. This gure suggests substantial anticipated returns to their investment in human capital if such expectations are actually realised. We measure the personal discount rate by presenting students with ve different scenarios. In each scenario students were asked to consider if they felt better off, worse off or the same by comparing a given sum of money received by a friend in a years time compared to GBP1000 received by the student today.6 These sums were: GBP950, GBP1000, GBP1050, GBP1100, GBP1200. The discount factor was elicited on the basis of when students selected the option the same. Thus, they were implicitly given one of the following discount factors: -0.05, 0, 0.05, 0.1 or 0.2. The average discount rate was a plausible value of 0.084. Although simple, this 9. Summary statistics. All students 1Female 2Male 3t-stat/z-scorea/ chi-squared testb0.8150.7780.8411.4014022.76 (9543.25) 17276.04 (7473.83) 0.591 20.5 (3.195)12499.88 (9460.82) 16236.96 (7432.02) n/a 20.222 (3.091)15077.05 (9482.959) 17934.8 (7449.39) n/a 20.692 (3.258)2.35Ethnicity: White non-Britishc Other ethnic groupd White British x2 20.139 0.250 0.611 n/a0.183 0.238 0.5790.110 0.258 0.6321.81 0.40 0.93 0.137Cohort: First-year student Second-year student Third-year student x2 20.623 0.198 0.179 n/a0.667 0.182 0.1510.593 0.209 0.1981.30 0.57 1.06 0.043Proportion of students expected to be in debt by the end of their studies Students individual characteristics Expected debt (GBP) all students. Expected debt (GBP) students who expect to be in debt. Male students Age (years)1.76 n/a 1.27(Continued .)R. BachanDownloaded by [University of Westminster - ISLS] at 13:11 05 December 20138Table 1. 10. (Continued .) All students 1Has a grant/scholarship Has part-time job during term time Students family characteristics Estimated family annual income (GBP) Father went to university Mother went to university Receives parental contribution Monthly parental contribution (GBP) Family home owners Students future income expectations Expected earnings in rst job after graduation (GBP) Expected earnings at age 30 with a degree (GBP) Expected earnings at age 30 without a degree (GBP) Students attitudes to debt and risk Discount rate Risk attitude Debt aversionFemale 2Male 3t-stat/z-scorea/ chi-squared testb0.516 0.2920.548 0.2460.494 0.3240.92 1.4853897.73 (36292.16) 0.314 0.256 0.464 152.15 (240.47) 0.85452563.49 (35537.36) 0.349 0.310 0.444 151.49 (227.97) 0.84954821.43 (36874.69) 0.291 0.220 0.478 152.60 (249.37) 0.8570.5425108.77 (8227.349) 50397.73 (22652.85) 44561.69 (44109.64)23821.43 (8119.72) 46190.48 (18447.64) 39984.13 (11131.92)26000.00 (8205.24) 53310.44 (24789.02) 47730.77 (56478.61)0.084 (0.077) 6.166 (2.105) 3.406 (1.289)0.082 (0.077) 5.540 (2.226) 3.810 (1.237)0.086 (0.076) 6.598 (1.906) 3.126 (1.253)1.08 1.97 0.58 0.04 0.19 2.30 2.74 1.52 0.51 4.48 4.739(Continued .)Studies in Higher EducationDownloaded by [University of Westminster - ISLS] at 13:11 05 December 2013Table 1. 11. R. BachanDownloaded by [University of Westminster - ISLS] at 13:11 05 December 201310Table 1.(Continued .) All students 1Uncertainty aversion N aFemale 2Male 32.990 (1.072) 3082.770 (1.104) 1263.143 (1.025) 182t-stat/z-scorea/ chi-squared testb 3.042t-tests are used to test differences in means between male and female students, and z-scores are used to test difference in proportions. The appropriate critical value at 5% level of signicance is 1.96. b Chi-squared values are used to test the assumption of independence in the sets of categorical variables between male and females. The appropriate critical value at the 5% level of signicance is 5.99. c White non-British includes EU and other overseas students. d Other ethnic groups include British Asian, British Afro-Caribbean, British Chinese and other overseas students. e Standard deviations are reported in parentheses for continuous variables. 12. Downloaded by [University of Westminster - ISLS] at 13:11 05 December 2013Studies in Higher Education11procedure is a standard method of eliciting discount rates from individuals, has generally found to be a signicant determinant of student debt (Oosterbeek and van den Broek 2009; Booij, Leuven, and Oosterbeek 2012), and has been adopted in many contexts (see, Harrison et al. 2005; Harrison, Lau, and Williams 2002; Meier and Sprenger 2012). Frederick, Loewenstein, and ODonoghue (2002) provide a critical review of the studies and techniques used to elicit discount rates since the 1970s. Risk attitudes were elicited by presenting students with the following question: How do you see yourself? Are you a person who is fully prepared to take risks, or do you try to avoid taking risks? Students were then invited to select a value on an 11-point risk scale ranging from 0 (not prepared to take risks) to 10 (fully prepared to take risks). This question has been used to gauge individual global risk attitudes in the 2004 wave of the German Socioeconomic Panel (SOEP)7 and has proved to be a reliable predictor of actual risk-taking behaviour (see Dohmen et al. 2005, 2011; Ding, Hartog, and Sun 2010; Booth and Nolan 2012). The average for the sample of students is 6.16 suggesting that, on average, students are prepared to undertake some risky behaviour. This is broadly similar to that reported by Oosterbeek and van den Broek (2009). Figure A1a contained in the Appendix depicts the distribution of risk attitudes across the sample of students. Each bar represents the percentage of individuals selecting a particular value on the risk scale. There is some degree of heterogeneity across the sample. In the extremes a small percentage of students (1.6%) are not willing to take any risk and have selected zero on the risk scale, and about 3.6% of students report a high degree of willingness to take a risk by selecting 10 on the scale. The modal response is 7, which was selected by just over 26% of students. Information on student debt aversion was obtained from the question I am not scared of being in debt. Responses were recorded on a ve-point scale strongly agree (1), slightly agree (2), neither (3), slightly disagree (4) and strongly disagree (5). Thus, the higher the score the more is the students aversion to debt. The mean score is 3.4 and the median 3 suggesting that on average students are neutral in respect of their aversion to debt. Similar questions have been used in previous studies attempting to elicit the degree to which students are averse to debt (Purcell and Elias 2010; Oosterbeek and van den Broek 2009). Students were also asked to indicate their dislike of uncertainty and to indicate their strength of agreement with the following question: I do not handle uncertainty well. Responses were recorded on the same ve-point scale as described above. Thus, the higher the score the better the student copes with an uncertain situation. The mean score is 2.99 suggesting that students are generally neutral in respect of their concerns about uncertainty. There are gender differences in the responses to the questions asked. Relatively fewer females expect to be in debt by the end of their undergraduate studies compared to their male counterparts 78% and 84% respectively. Male students report a higher level of expected debt by the end of their studies and expect to be more indebted than females by GBP2577 on average. However, there is no signicant difference between the level of debt expected between male and females who anticipate being indebted by the time they complete their studies (t = -1.76). In terms of ethnicity and age, the female and male sub-samples are broadly comparable, although white non-British are more represented in the female than the male sub-samples. In terms of cohort issues, the majority of females are in their rst year and proportionately more males are in their second and third years, but these differences are not statistically signicant at conventional levels. The sub-samples are also broadly 13. Downloaded by [University of Westminster - ISLS] at 13:11 05 December 201312R. Bachancomparable in terms of the proportion of males and females in part-time work, in receipt of a grant/scholarship and receiving nancial support from parents. The monthly amount of parental nancial support received on average is also comparable. Male graduates expect to earn GBP2179 more than their female counterparts in their rst job and by the time they are 30 years of age they expect to be earning GBP7120 more. The nding that males expect higher earnings in their rst job after graduation is also generally found in previous surveys (see Purcell and Elias 2010; Johnson et al. 2009). However, there is no statistically signicant difference between what male and female students expect to be earning by the time they are 30 years of age had they not studied for a degree. The mean discount rate reported by males and females is very similar and the distribution of the discount rate by gender is also broadly comparable. However there are signicant gender differences in their reported risk attitudes, their aversion to debt and their dislike of uncertainty. In terms of risk attitudes females are relatively less willing to take risks compared to their male counterparts their scores being 5.54, and 6.59 respectively. Figure A1b in the Appendix depicts the distribution of risk attitudes by gender. We rst note that the modal response is 7, which is the same across gender groups. The gure also reveals that very few male and female students, approximately 1.6%, report an unwillingness to take risks. However, at the other end of the distribution, a larger percentage of males are more willing to take risks than females. For instance about 60% of males report a risk attitude in the range of 7 to 10, and 5.5% have selected 10. In contrast about 45% of female students report a risk attitude in the rage of 7 to 10, and less than 1% have selected 10. In terms of debt aversion female students are on average more averse to debt than their male counterparts, which is a standard nding in the literature. We note that about 73% of females disagreed with the statement I am not scared of being in debt with the comparable gure for males about 45%. Similarly, females in the sample appear to dislike uncertainty more than their male counterparts.Methodology The data employed in this study can be described as censored. In other words, the dependent variable, expected debt, records a zero for students who expect not to be in debt by the end of their studies or a positive non-zero value otherwise. As a signicant proportion of the sample of students report zero expected debt, application of OLS to these data will potentially lead to biased and inconsistent coefcient estimates. We therefore model student expectation of debt using a censored tobit model which gives consistent coefcient estimates when the dependent variable is censored.8 We dene the latent debt equation as: y = xi b + ui i(1)where y is a partial latent dependent variable that captures the ith individuals propeni sity to be in debt, xi is a vector of debt determining variables for individual i, is a vector of xed unknown coefcients to be estimated, and ui N(0, 2). Thus: yi = y if xi b + ui . 0 and i yi = 0 if xi b + ui 0 14. Studies in Higher Education13where yi represents the actual expected debt by the ith individual. Thus yi is either positive (yi > 0) or zero (yi = 0). Using this information, the log-likelihood function (L) may be expressed as follows: nL=Zi lnDownloaded by [University of Westminster - ISLS] at 13:11 05 December 2013i=1F[(yi x b) 4 s] +(1 Zi ) ln[1 F(x b 4 s)] s(2)where Zi = 1 if yi >0, and Zi = 0 if yi = 0, is the standard normal density function, is the cumulative distribution function of the standard normal, and ln() is the natural log operator. The parameter values , and are chosen to maximise L (the log-likelihood function) using non-linear iterative methods (e.g. the Newton-Raphson method). The resultant estimates are known to be consistent and asymptotically normal. The estimated coefcients are not readily interpretable as the underlying stochastic index, expression (1), is not observed when students report zero expected debt. To aid interpretation the marginal effects for each estimated coefcient are computed (see Greene 1999). The marginal effects for the tobit model can be expressed: E[yi |xi ] x i b = bF xi s(3)Thus, the tobit coefcients have to be adjusted by a factor equal to (x'i/) to nd the effect on expected debt for small changes in the independent variables. This scaling factor is constructed in the current application using the sample average values for xi variables. Expression (3) can also be used to approximate the impact effects on expected student debt associated with the discrete rather than continuous independent variables.Empirical results Five alternative specications are estimated using a tobit model, and the corresponding marginal/impact effects are also estimated for each specication. Each model includes variables that capture a students individual and socioeconomic characteristics. The rst specication can be described as an uncertainty model that includes variables that capture students expected (uncertain) earnings post-graduation. Specications 2 and 3 augment specication 1 to allow for student discount rates and attitudes to risk to enter the analysis. Specications 4 and 5 can be described as behavioural models that augment the uncertainty model to include students reported debt aversion and dislike of uncertainty (see Oosterbeek and van den Broek 2009). Table A1 in the Appendix reports the coefcient estimates for the tobit index functions. We rst note that the goodness-of-t measures, reported at the bottom of the Table A1, are satisfactory for models of this kind. Furthermore, the majority of the estimated coefcients reported in the table are well determined at a conventional level of statistical signicance. Each specication was also tested to determine if separate male and female equations tted the data better than a single pooled equation using likelihood ratio tests. The null of a pooled regression was upheld by the data in all cases. Table 2 reports the estimated maximum likelihood tobit marginal/impact effects derived from expression (3) for the ve expected debt equations. Again the majority of the estimated coefcients, in all ve specications, are well determined at a 15. Tobit maximum likelihood estimates: marginal/impact effects. Specication 1Gender (male) White non-British Other ethnic group White British Age (years) Grant/scholarship Has part-time job Monthly contribution (GBP) First-year student Second-year student Third-year student Family home owners Mother and father university educated Expected earnings >GBP30,000 after graduationSpecication 2Specication 3Specication 4Specication 52558.898** (1045.40) 2492.986 (1567.60) 4098.849*** (1228.81) f 97.137 (170.55) 1265.626 (1144.21) 8590.715*** (1208.96) 11.709*** (2.56) 1091.396 (1471.92) 4258.316** (1717.30) f 3098.619** (1420.32) 3281.140** (1381.42) 4050.037*** (1580.87)2490.071** (1039.27) 2629.599* (1559.04) 4365.615*** (1227.65) f 79.313 (168.93) 1048.938 (1141.36) 8536.019*** (1201.71) 11.628*** (2.55) 1457.336 (1471.88) 4257.002** (1707.18) f 3160.541** (1411.47) 3352.706** (1371.92) 3973.433** (1570.70)1799.670* (1059.86) 2294.468 (1553.61) 3499.635*** (1254.01) f 62.956 (167.92) 753.023 (1135.78) 8441.873*** (1191.73) 11.909*** (2.54) 1565.639 (1459.05) 4430.030*** (1693.76) f 2880.196 ** (1401.89) 3474.237*** (1362.91) 2568.635* (1533.15)2036.378* (1078.06) 2322.105 (1351.12) 3483.319*** (1251.82 ) f 58.074 (167.98) 850.813 (1137.06) 8457.571*** (1189.84) 12.029*** (2.54) 1494.338 (1457.77) 4447.743*** (1689.76) f 2944.440** (1400.45) 3459.429** (1360.58) 2773.774* (1640.40)1589.142 (1068.32) 2219.983 (1549.70) 3313.197** (1257.58) f 59.870 (167.44) 672.601 (1133.78) 8610.544*** (1194.48) 12.082*** (2.53) 1652.151 (1455.85) 4504.362*** (1689.31) f 3083.756** (1405.45) 3505.761*** (1358.57) 2622.472* (1629.20) (Continued .)R. BachanDownloaded by [University of Westminster - ISLS] at 13:11 05 December 201314Table 2. 16. (Continued .) Specication 1Expected earnings >GBP50,000 at 30 Discount rateSpecication 2Specication 3Specication 4Specication 52740.183*** (1023.38) 2638.543*** (1008.89) 13628.944** (6441.97) 797.997*** (271.66) 457.355 (400.48) 2711.933*** (1010.96) 13643.780** (6432.91) 707.777*** (269.71) Risk attitude2745.483*** (1016.79) 14199.556** (6507.81) Debt aversion2590.176*** (1009.78) 13985.731** (6449.61) 750.374*** (268.89) Uncertainty aversionLog-Likelihood Scale factore Observations a2686.349 0.9209 3082683.987 0.9227 308Asymptotic standard errors are reported in parentheses beneath coefcient estimates. * denotes signicant at 10%; ** signicant at 5%; *** signicant at 1%. c f denotes base category in estimation. d denotes variable not used in estimation. e Scale factor used in the computation of the marginal/impact effects. f All estimations reported were undertaken using NLOGIT 3.0 (2003). b2680.104 0.9249 3082679.454 0.9255 308635.735 (471.56) 2679.199 0.9258 308Studies in Higher EducationDownloaded by [University of Westminster - ISLS] at 13:11 05 December 2013Table 2.15 17. Downloaded by [University of Westminster - ISLS] at 13:11 05 December 201316R. Bachanconventional level of statistical signicance. It should be noted that there is no comparable method or routine for adjusting the standard errors for the presence of heteroscedasticity in the tobit model as there is in the OLS estimated linear regression model (e.g. through use of a robust estimator). Greene (2008) is critical of the use of such robust procedures in qualitative response models like the tobit because the asymptotic properties of such estimates are unknown. He suggests modelling heteroscedasticity directly, but this is not feasible in the current application due to the small sample size available. Moreover, it is not always transparent if such a procedure actually deals with the problem or merely controls for some underlying non-linear mis-specication that is unrelated to heteroscedasticity. We initially focus on a students individual personal and socioeconomic characteristics and examine their impact on expected debt. Several of the estimated coefcients relating to a students individual characteristics are statistically signicant at conventional levels and the results remain robust across all specications reported. There is a signicant gender effect on expected debt. Male students expect to have more debt than their female counterparts by the end of their undergraduate studies. For instance, the impact effect in specication 1 suggests that male students, on average and ceteris paribus, expect a level of debt that is about GBP2559 greater than that anticipated by female students at the end of their studies. This effect remains after controlling for time preference, though the impact diminishes to GBP1799. However, the estimated gender effect is statistically insignicant when dislike of uncertainty is included in the specication possibly suggesting a high intercorrelation between gender and this uncertainty measure. We nd some evidence that student ethnicity impacts on expected debt. In the basic uncertainty model, students classied as being non-white (e.g. British Asian, black and Chinese) expect to be about GBP4100 less in debt by the end of their undergraduate studies than their white counterparts, on average and ceteris paribus. This effect remains robust across all specications but the magnitude of this effect diminishes with movement across the ve regression models reported in Table 2. Students in their second year of study expect a higher debt than their nal year counterparts, but there is no statistically signicant difference between debt expectations of students in their rst and nal years of study. This may be suggestive of students trying to aggressively reduce their debt in the nal year of their study. A large and signicant negative effect on expected debt is found for students who have a part-time job during term time. The estimated coefcient suggests that holding such a job reduces expected debt, on average and ceteris paribus, by about GBP8500 in all specications compared to those students who are not in part-time employment. This seems to be a plausible estimate assuming a student earns on average about GBP81 per week for 35 weeks of the academic year. After some experimentation three variables that capture the socioeconomic characteristics of the students family enter the empirical models: parental contributions, parental educational background and home ownership. First, students whose parents contribute to their living expenses anticipate their debt to be reduced by just under GBP12 for every GBP worth of nancial help received per month according to specication 1. This represents a rather large multiplier effect and is broadly similar across all ve empirical specications reported here. Second, on the basis of specication 1, a student whose mother and father were university educated expects to be in debt by GBP3281 less than students with either one or no parent who was university educated, on average and ceteris paribus. Again this effect remains reasonably stable across all specications. Third, students from families that own their own home anticipate to 18. Downloaded by [University of Westminster - ISLS] at 13:11 05 December 2013Studies in Higher Education17be about GBP3000 less in debt than students who do not possess this socioeconomic attribute. There is no evidence that the age of the student or, rather surprisingly, the receipt of a grant or scholarship impacts signicantly on expected debt. However, student expected debt may be inuenced by the earnings they anticipate in the future, which may also reect a students expected set of employment opportunities. For instance, using the results from the rst specication, students who anticipate earning more than GBP30,000 in their rst post-graduation job expect to be about GBP4000 more in debt than students who report a more modest future earnings level. This effect is attenuated as we control for risk attitudes, aversion to debt and dislike of uncertainty. A similar, but considerably smaller, effect is found for those students who expect to earn more than GBP50,000 by the time they reach 30. In terms of time preference we note that in specication 2 a one percentage-point increase in the discount rate increases expected debt by about GBP142.9 A higher discount rate implies that an individual prefers current consumption to future consumption and thus we would expect a greater level of expected debt. The third specication adds the attitude to risk as an additional covariate. The estimated coefcient is well determined and statistically signicant and suggests that the less risk averse the student the higher expected debt.10 Specication 4 introduces the variable designed to capture the students attitude to debt. The estimated coefcient is counter-intuitively signed but is statistically insignificant at a conventional level. Thus, it can be inferred from this sample that a students attitude to debt exerts no independent effect on anticipated debt, on average and ceteris paribus. The nal specication replaces attitude to debt with an alternative that measures student feelings in regard to uncertainty. Again the estimated effect for this variable fails to attain statistical signicance at an acceptable level of statistical signicance. Conclusion This paper examined the determinants of student expected debt in UK higher education using a sample of business and nance students within a single university. We argue that such an approach using data for a single institution has the merit of removing the effects of inter-institution heterogeneity. However, it is acknowledged that some of the results presented may not be generalisable to the broader UK student population. Nevertheless, the quality and richness of the data permits the inclusion of variables that would otherwise be unavailable in national datasets. We believe that the analysis presented offers some stylised facts that are likely to be reected in most comparable UK institutions, and we take the view that the content of our ndings potentially has broader implications. The results presented here suggest that there is a gender dimension to student debt expectations with males having a greater expected level of debt on the completion of their undergraduate studies than females; supporting the ndings from the empirical analysis of Purcell and Elias (2010). This particular result may also reect the role of greater risk-taking behaviour on the part of male students and conrms the earlier ndings of Davies and Lea (1995) and Callender (2003) for pre-university students. Nonwhite students expect a lower level of debt compared to their white British counterparts. This may be reective of differing cultural and religious attitudes to debt (Callender 2003). There also appears to be some change in the level of debt a student expects 19. Downloaded by [University of Westminster - ISLS] at 13:11 05 December 201318R. Bachanover time with second-year students anticipating more in debt than their rst- and thirdyear counterparts. This could be explained by changing attitudes to debt over time with students becoming more tolerant of debt between their rst and second years, but less tolerant to debt in their nal third year (see Davies and Lea 1995). Students who work part-time during term time expect to have a lower level of debt. If there is a trade-off between being a full-time student and part-time work then raising tuition fees in the future may compromise the time students will devote to study and may result in an increase in the non-continuation rate and a reduction in student academic achievement, particularly among students from the lower socioeconomic groups. The importance of a students socioeconomic background as proxied by parental home ownership, parental experience of higher education and parental nancial support all were found to exert a signicant role in reducing student expected debt, and this conrms the ndings of previous studies (see for example Davies and Lea 1995). In particular, we nd a sizeable multiplier effect on debt expectations emerging through parental contributions perhaps suggesting that such contributions enhances the students sense of nancial security. Whether or not such contributions are affordable particularly amongst the lower-income groups merits a close targeting of student support through the use of bursaries and other forms of student subsidy. The empirical analysis suggested that the receipt of either a grant or scholarship or bursary has no signicant effect on reducing student expected indebtedness. This may be indicative of the fact that current levels of nancial support are inadequate and ineffective in reducing anticipated debt and resonates with the ndings of many recent studies and surveys on student nance and debt. This issue, and the issues raised above, have clear policy implications and suggest closer targeting of nancial support to those students from lower socioeconomic groups, which may ultimately impact favourably on widening participation in UK higher education. This is particularly important at a time when the majority of UK universities are set to increase their tuition fees to GBP9000 per annum in the academic year 2012/2013. There is evidence that students anticipate a high return to their university education in terms of high expected earnings, which in turn contributes to higher expected debt. Whether or not these expectations are to be realised in the future is difcult to discern and may indeed be an overestimate (Jerrim 2011). Moreover, if wage expectations are unfullled, for instance due to unforeseen shocks to the graduate labour market or simply because students are ill-informed on graduate wages and/or job opportunities, then their life choices could be severely compromised (Jerrim 2011; Callender and Kemp 2000; Purcell and Elias 2010). Improving the quality of information on graduate wages and jobs may not only help to match graduates to jobs but may also help to moderate student wage expectations and attenuate graduate debt. Finally, we found that students who have a high discount rate and those with a high-risk attitude are prone to higher expected debt. This is a novel nding for UK students and supports the ndings of Oosterbeck and van de Broek (2009) in their study of the borrowing behaviour of Dutch students. This study looked at debt expectations of business and nance students in a single UK university. It would be interesting to see if these results generalise to other universities and for students on different degree programmes across the UK higher education sector. It would also be interesting to see if these results extend to A-level students contemplating higher education and whether or not their expectation of debt and the debt they actually incur in higher education are correlated. These provide fruitful avenues for further research particularly in the wake of the 2010 Browne Report. 20. Studies in Higher Education19Acknowledgments The author would like to thank Barry Reilly, Mike Barrow, and Yvonne Hillier for constructive comments on earlier drafts of this paper. The author would also like to thank the editor of this journal for comments provided and is particularly indebted to two anonymous referees of this journal for their constructive and insightful comments. The revised paper has benetted immensely from their inputs. However, the usual disclaimer applies.Notes 1.Downloaded by [University of Westminster - ISLS] at 13:11 05 December 20132. 3.4.5. 6.7. 8.9. 10.This commitment was made in a speech delivered at the Labour Partys conference in 1997. Parry (2006) discusses the merits and potential pitfalls in moving from an elite higher education system in the UK to a mass and more varied higher education system in the1980s and 1990s. It is instructive to note that participation rates for students under 21 years of age from the lower social classes have increased from 30.6% in 2006/2007 to 39.8% by 2010/2011. Also participation rates for students from low-participation areas and students from state schools have also increased between these years from 9% to 10.5% and 87.4% to 88.7% respectively (Higher Education Statistical Agency 2012). Evidence from UK national surveys of students income and expenditure surveys report that the proportion of full-time students working term time and the hours worked has generally increased over the last decade, but fell slightly towards the end of the decade (Callender and Kemp 2000; Callender and Wilkinson 2003; Johnson et al. 2009; Purcell and Elias 2010). The Bank of England annual spot exchange rates were used to convert the USD value debt to its sterling equivalent these were 1.5665 for 2009 and 1.546 for 2010. This was undertaken to check for consistency in responses. For instance, if a student considered GBP1050 received by the friend the same as GBP1000 today then for lesser amounts they should feel better off and for greater worse off. Only students who responded consistently were included in the sample. Dohmen et al. (2005) provides summary details for the 2004 survey. The questionnaire is available at http://www.diw.de/en/diw_02.c.222729.en/questionnaires.html. The tobit estimator has been used in studies on the determinants of student loan default (Greene 1989), the determinants of student part-time work hours (Kalenkoski and Pabilonia 2010), and household and individual debt expectations (Brown et al. 2005). Amemiya (1984) provides an early but extensive survey of tobit models. Given that the discount rate is expressed in proportional terms, multiplying the coefcient estimate by 0.01 yields the effect of a one percentage-point rise in the discount factor. In this specication we test the assumption that risk attitude is exogenous using the approach suggested by Smith and Blundell (1986). In conducting this test risk attitude is regressed on a set of four exogenous identifying dummy variables (instruments) that include information on whether or not the student: gambles (i.e. regularly partakes in lotteries); played a fruit machine in the week in which the questionnaire was administered; had a personal savings account; and/or ever participated in a dangerous sport. The residuals from this model are inserted into specication 3 and the statistical signicance of the estimated coefcient is tested using a t-test. These variables were found to be individually and jointly signicant in inuencing risk attitudes with a F(4, 303) = 11.67 [prob-value = 0.000] and the t-test suggests that risk attitude is exogenous (t-statistic= 1.48 [prob value = 0.141).References Amemiya, T. 1984. Tobit models: A survey. Journal of Econometrics 24: 361. Barclays. 2005. Annual graduate survey. Barclays Bank Press Release, April 14 2005. Barr, N. 2004. Higher education funding. Oxford Review of Economic Policy 20, no. 2: 26483. Barr, N., and I. Crawford. 1998. Funding higher education in an age of expansion. Education Economics 6, no. 1: 4570. 21. Downloaded by [University of Westminster - ISLS] at 13:11 05 December 201320R. BachanBecker, G.S. 1993. Human capital: A theoretical and empirical analysis, with special reference to education. 3rd ed. Chicago, IL: University of Chicago Press. Betts, J.R. 1996. What do students know about wages? Evidence from a survey of undergraduates. Journal of Human Resources 31, no. 1: 2756. Blundell, R., L. Dearden, A. Goodman, and H. Reed. 2000. The returns to higher education in Britain: Evidence from a British cohort. Economic Journal 110: F82F89. Bolton, P. 2009. Student and graduate debt statistics. SN/SG/1327, House of Commons Library, London. Booji, A., E. Leuven, and H. Oosterbeek. 2012. The role of information in the take-up of student loans. Economics of Education Review 31, no. 1: 3344. Booth, A., and P. Nolan. 2012. Gender differences in risk behaviour: Does nurture matter? Economic Journal 122: F56F78. Bratti, M., R. Naylor, and J. Smith. 2007. Different returns to different degrees? Evidence from the British cohort study 1970. Warwick Economic Research Paper 783. Warwick: Department of Economics, University of Warwick. Brennan, J., A. Duaso, B. Little, C. Callender, and R. Van Dyke. 2005. Survey of higher education students attitudes to debt and term-time working and their impact on attainment. Bristol: HEFCE, http://webarchive.nationalarchives.gov.uk/20100202100434/http://www. hefce.ac.uk/pubs/rdreports/2005//rd15_05/rd15_05.pdf. Brown, S., G. Garino, K. Taylor, and S. Weatley Price. 2005. Debt and nancial expectations: An individual- and household-level analysis. Economic Enquiry 43, no. 1: 100120. Browne, J. 2010. Securing a sustainable future for higher education: An independent review of higher education funding and student nance. London: Department for Business Innovation and Skills. Brunello, G., C. Lucifora, and R. Winter-Ebmer. 2004. The wage expectations of European business and economics students. Journal of Human Resources 39: 111642. Callender, C. 2003. Attitudes to debt: School leavers and further education students attitudes to debt and their impact on participation in higher education. London: Universities UK. Callender, C. 2008. The impact of term-time employment on higher education students academic attainment and achievement. Journal of Education Policy 23, no. 4: 35977. Callender, C., and J. Jackson. 2005. Does fear of debt deter students from higher education? Journal of Social Policy 34, no. 4: 50940. Callender, C., and M. Kemp. 2000. Changing student nances: Income, expenditure and takeup of student loans among full- and part-time higher education students in 1998/9. Research Report No. 213. London: Department for Education and Employment. Callender, C., and D. Wilkinson. 2003. 2002/03 student income and expenditure survey: Students income, expenditure and debt and changes since 1998. Research Report No. 487. Nottingham: Department of Education and Skills. Chapman, B., and K. Lounkaewa. 2010. Repayment burdens with US college loans, Discussion Paper No. 647. Canberra: Centre for Economic Policy Research, Australian National University. Chapman, B., and M. Sinning. 2011. Student loan reforms for German higher education: Financing tuition fees, IZA Discussion Paper No. 5532. Bonn: IZA. College Board. 2011. Trends in student aid. New York: College Board. http://trends. collegeboard.org/downloads/Student_Aid_2011.pdf. Connor, H., S. Dawson, C. Tyers, J. Eccles, J. Regan, and J. Aston. 2001. Social class and higher education: Issues affecting decisions on participation by lower social class groups. Research Report No. 267. London: Department for Education and Employment. Davies, E., and S. Lea. 1995. Student attitudes to student debt. Journal of Economic Psychology 16: 66379. Davies, R., and P. Elias. 2003. Dropping out: A study of early leavers from higher education. Research Report No. 369. London: Department for Education and Skills. Department for Education and Skills. 2001. Key challenges of the next decade. Speech by Estelle Morris, given on October 22, at London Guildhall University. Department for Education and Skills. 2003a. The future of higher education. London: HMSO. Department for Education and Skills. 2003b. Student loans and the question of debt. London: Department for Education and Skills. http://www.dfes.gov.uk/hegateway/hereform/index. cfm. Accessed 20/02/2010. 22. Downloaded by [University of Westminster - ISLS] at 13:11 05 December 2013Studies in Higher Education21Department for Innovation, Universities, and Skills. 2012. Participation rates in higher education: Academic years 2006/072010/2011. http://www.bis.gov.uk/assets/biscore/ statistics/docs/h/12-p140-participation-rates-in-he-2010-11. Ding, X., J. Hartog, and Y. Sun. 2010. Can we measure individual risk attitudes in a survey? IZA Discussion Paper No. 4807. Bonn: IZA. Dohmen, T., A. Falk, D. Huffman, U. Sunde, J. Schupp, and G.G. Wagner. 2005. Individual risk attitudes: New evidence from a large, representative, experimentally-validated survey. IZA Discussion Paper No. 1730. Bonn: IZA. Dohmen, T., A. Falk, D. Huffman, U. Sunde, J. Schupp, and G.G. Wagner. 2011. Individual risk attitudes: Measurement, determinants and behavioural consequences. Journal of the European Economic Association 9, no. 3: 52250. Dominitz, J., and C. Manski. 1996. Eliciting student expectations of the returns to schooling. Journal of Human Resources 31, no. 1: 126. Ford, J., D. Bosworth, and R. Wilson. 1995. Part-time work and full-time higher education. Studies in Higher Education 20, no. 2: 187202. Frederick, S., G. Loewenstein, and T. ODonoghue. 2002. Time discounting and time preference: A critical review. Journal of Economic Literature 40, no. 2: 350401. Friedman, M., and R. Friedman. 1980. Free to choose. London: Penguin. Glennerster, H., S. Merrett, and G. Wilson. 1968. A graduate tax. Higher Education Review 1, no. 1: 2638. Greene, L. 1989. An economic analysis of student loan default. Education Evaluation and Policy Analysis 11, no. 1: 6168. Greene, W. 1999. Marginal effects in the censored regression model. Economics Letters 64, no. 1: 4349. Greene, W. 2008. Econometric analysis. New Jersey: Pearson Education. Harkness, S., and S. Machin. 1999. Graduate earnings in Britain, 197495. Research Report No. 95. London: Department for Education and Employment. Harrison, G., M. Lau, E. Ruststrm, and M. Sullivan. 2005. Eliciting risk and time preferences using eld experiments: Some methodological issues. In Field experiments in economics, Vol. 10. ed. J. Carpenter, G.W. Harrison and J.A. Lists, 125218. Greenwich, CT: JAI Press. Harrison, G., M. Lau, and M. Williams. 2002. Estimating individual discount rates in Denmark: A eld experiment. American Economic Review 92, no. 5: 16061617. Higher Educational Statistical Agency. 2012. Performance indicators: Summary tables and charts. http://www.hesa.ac.uk/index.php?option=com_content&task=view&id=2397&Itemid=141. Humphrey, R. 2006. Pulling structured inequality into higher education: The impact of part-time working on English university students. Higher Education Quarterly 60, no. 3: 27086. Jerrim, J. 2011. Do UK higher education students overestimate their starting salary? Fiscal Studies 32, no. 4: 483509. Johnson, C., E. Pollard, W. Hunt, M. Munro, J. Hillage, J. Parfement, and N. Lowe. 2009. Student income and expenditure survey 2007/08 English-domiciled students. Research Report No. 0905. London: Department for Innovation, Universities, and Skills. Kalenkoski, C., and S. Pabilonia. 2010. Parental transfers, student achievement, and the labor supply of college students. Journal of Population Economics 23: 46996. Knowles, J. 2000. Access for few? Student funding and its impact on aspirations to enter higher education. Widening Participation and Lifelong Learning 2, no. 1: 1423. Meier, S., and C. Sprenger. 2012. Discounting nancial literacy: Time preferences in nancial education programs. Journal of Economic Behaviour and Organisations. DOI:10.1016/j. jebo.2012.02.024. Natwest Bank. 2007. Student money matters. Natwest Bank Press Release, August 14. Oosterbeek, H., and A. van den Broek. 2009. An empirical analysis of borrowing behaviour of higher education students in the Netherlands. Economics of Education Review 28: 17077. Parry, G. 2006. Policy-participation trajectories in English higher education. Higher Education Quarterly 60, no. 4: 392412. Pennell, H., and A. West. 2005. The impact of increased fees on participation in higher education in England. Higher Education Quarterly 59, no. 2: 127237. Psacharopoulos, G., and H. Patrinos. 2004. Returns to investment in education: A further update. Education Economics 12, no. 2: 11134. 23. Downloaded by [University of Westminster - ISLS] at 13:11 05 December 201322R. BachanPurcell, K., and P. Elias. 2010. The impact of paid and unpaid work and of student debt on experience of higher education. Futuretrack Working Paper 3. Warwick: Institute for Employment Research, University of Warwick. Push. 2007. The Push student debt survey. http://www.push.co.uk/. Push. 2010. The Push student debt survey. http://www.push.co.uk/Debt-Survey-2010Summary. Robbins, Lord. 1963. Report of the committee appointed by the Prime Minister under the Chairmanship of Lord Robbins. London: HMSO. Scott-Clayton, J. 2012. What explains trends in labor supply among US undergraduates? National Tax Journal 65, no. 1: 181210. Shen, H., and A. Ziderman. 2009. Student loans repayment and recovery: International comparisons. Higher Education 57, no. 3: 31533. Smith, R., and R. Blundell. 1986. An exogeneity test for a simultaneous equation tobit model with an application to labour supply. Econometrica 54: 67985. Strinebrickner, R., and T. Strinebrickner. 2003. Working during school and academic performance. Journal of Labor Economics 21, no. 2: 47391. Student Loan Company. Various years. Facts and gures. http://www.slc.co.uk/statistics/facts% 20and%20%20gures/take_up_stats_9105.html. Student Loan Company. 2010. Student loans for higher education in England, nancial year 200910. http://www.slc.co.uk/media/77939/slcsfr032010.pdf. Statistical rst release 03/ 2010. Unite. 2007. The student experience report 2007. http://www.unite-group.co.uk/Attachments/ 000171/Student%20Experience%20Report%202007.pdf. Usher, A. 2005. Global debt patterns: An international comparison of student loan burdens and repayment conditions. Toronto: Educational Policy Institute. Webbink, D., and J. Hartog. 2004. Can students predict starting salaries? Yes!. Economics of Education Review 23, no. 2: 103113. Yorke, M., and B. Longden. 2008. The rst-year experience of higher education in the UK. York: Higher Education Academy. Zhang, L. 2010. Effects of college educational debt on graduate school attendance and early career and life choices. http://ssrn.com/abstract=942793 or http://dx.doi.org/10.2139/ssrn. 942793. 24. Studies in Higher EducationDownloaded by [University of Westminster - ISLS] at 13:11 05 December 2013AppendixFigure A1a.Student risk attitudes: all students.Figure A1b.Student risk attitudes by gender.23 25. Tobit maximum likelihood estimates: index function. Specication 1Gender (male) White non-British Other ethnic group White British Age (years) Grant/scholarship Has part-time job Monthly contribution (GBP) First-year student Second-year student Third-year student Family Home Owners Mother and father university educated Expected earnings >GBP30,000 after graduationSpecication 2Specication 3Specication 4Specication 52778.829** (1135.057) -2707.252 (1703.659) -4451.135*** (1338.231) -105.485 (185.223) 1374.403 (1242.840) -9329.065*** (1314.503) -12.71511*** (2.785663) 1185.199 (1598.316) 4624.307** (1865.06) f -3364.937** (1542.454) -3563.146** (1502.076) 4398.127** (1718.200)2698.596** (1126.108) -2849.808* (1691.023) -4731.201*** (1334.493) -85.95528 (183.083) 1136.778 (1237.174) -9250.845*** (1303.963) -12.60206*** (2.772099) 1579.376 (1595.03) 4613.493** (1850.392) f -3425.212** (1529.731) -3633.47** (1488.673) 4306.177** (1703.67)1945.72* (1145.521) -2480.673 (1680.933) -3783.644*** (1358.557) -68.06565 (181.5533) 814.1341 (1228.101) -9126.963*** (1290.228) -12.87514*** (2.754683) 1692.697 (1577.31) 4789.544*** (1831.436) f -3113.935 ** (1515.613) -3756.185** (1475.511) 2777.089* (1663.05)2200.331* (1164.468) -2509.062 (1677.272) -3763.769 *** (1355.364) -62.74967 (181.5119) 919.3138 (1228.78) -9138.506*** (1287.42) -12.99704*** (2.751435) 1614.65 (1574.97) 4805.839*** (1826.018) f -3181.503** (1513.154) -3737.955** (1472.095) 2997.096* (1773.093)1716.42 (1153.611) -2397.787 (1675.025) -3578.558*** (1360.942) -64.66597 (180.8673) 726.4714 (1224.727) -9300.18*** (1291.812) -13.05032*** (2.747557) 1784.475 (1572.276) 4865.126*** (1824.783) f -3330.741** (1517.94) -3786.544** (1469.363) 2832.511* (1760.28) (Continued .)R. BachanDownloaded by [University of Westminster - ISLS] at 13:11 05 December 201324Table A1. 26. (Continued .) Specication 1Expected earnings >GBP50,000 at 30 Discount rate2975.695*** (1112.742) Specication 2 2975.397*** (1103.299) 15388.66** (7053.512) Risk attitudeDebt aversionUncertainty aversion9113.382 (425.2541)Goodness of t stats: R2 - ANOVA R2 - Decomposition Log-Likelihood Likelihood Ratio Test -0.263 0.272 -2686.349 138.95 [0.000]9034.38 (421.4977) 0.275 0.284 -2683.987 143.68 [0.000]Specication 3Specication 4Specication 52800.379*** (1092.935) 15023.42** (6973.531) 811.2701*** (291.0654) 2850.978*** (1091.354) 14726.24** (6961.176) 862.2457*** (293.8944) 494.1774 (432.7096) 2929.138*** (1093.169) 14736.54** (6948.69) 764.4649*** (291.6456) 8923.831 (493.4554) 0.277 0.287 -2680.104 151.44 [0.000]8901.029 (490.5441) 0.281 0.291 -2679.454 152.75 [0.000]686.6526 (509.2374) 8888.974 (492.6188) 0.277 0.290 -2679.199 153.25 [0.000]x2 k(Continued .)Studies in Higher EducationDownloaded by [University of Westminster - ISLS] at 13:11 05 December 2013Table A1.25 27. R. BachanDownloaded by [University of Westminster - ISLS] at 13:11 05 December 201326Table A1.(Continued .) Specication 1Observations a308Specication 2 308Specication 3 308Specication 4 308Specication 5 308All estimations reported were undertaken using NLOGIT 3.0 (2003). Asymptotic standard errors are reported in parentheses. c * denotes signicant at 10%; ** signicant at 5%; *** signicant at 1% d f denotes base category in estimation. e denotes variable not used in estimation. f 2 R - ANOVA = variance in predicted conditional mean over variance in dependent variable. g 2 R Decomposition = variance in predicted mean over variance in predicted mean plus model residual variation. h The likelihood ratio test is dened: -2 (Log-likelihood value (constant only) - Log-likelihood value (full model). The test statistic is a chi-squared statistic with the degrees of freedom determined by the number of independent variables (k) in the relevant specication. The null tests the joint restriction that all the estimated coefcients from a specic specication are simultaneously equal to zero. b