Stepping stones or trapdoors? Paid and unpaid graduate internships in the creative sector

Abstract Policy discourse on graduate internships rests on the assumption that, paid or unpaid, they improve the employability of interns. Employing data from a survey of UK creative and mass communications graduates, surveyed two to six years after graduation, this article examines the impact of graduate internships on subsequent job prospects. While paid internships are associated with better pay and increased chances of having a creative or graduate-level job, unpaid internships are not, and are associated with lower pay in the short to medium term. Findings contribute to theory by challenging the ‘stepping stone’ view of unpaid internships and much policy discussion about these. Results reinforce, with enhanced specificity about the role internships play in the graduate labour market, theories that characterise the graduate labour market as a ‘positional’ hierarchy of opportunities rather than a labour market that is essentially meritocratic in nature.


Introduction
Graduate internships sit squarely at the intersection of debates around graduate employability, transitions into employment, social mobility and socio-economic reproduction. Policy discourse contends that, paid or unpaid, graduate internships help to develop work-related skills and abilities not easily learned in the classroom and get a foot on the career ladder (Milburn 2009(Milburn , 2014. That such experiences provide an 'invaluable opportunity' to further one's career was repeatedly invoked by those opposing a UK parliamentary private member's bill seeking to outlaw unpaid internships (HC Deb, 4 Nov 2016, vol 616 cols 1156-1226. This 'stepping stone' view is evident among many interns, who see internships as a necessary step or 'paying your dues' (Shade and Jacobson 2015;Siebert and Wilson 2013). Academic literature on the topic is more critical, raising concerns about potential exploitation of young aspirants and that unpaid internships act to exclude those without the right economic, cultural and social capital from accessing key industries (Bathmaker et al. 2016, Wright and Mulvey 2021, Leonard, Halford, and Bruce 2016. However, such concerns implicitly assume that: • unpaid internships are nevertheless beneficial to individuals; and • it is access to unpaid internships, and not paid ones, that creates particular problems for social mobility (e.g. Shorthouse 2010).
This paper focuses on a UK sector in which the use of graduate internships has been particularly marked, the creative industries (broadly defined), to address the first of these two questions and contributes to theory by challenging these above assumptions. The paper contributes to the understanding of graduate internships in three ways. First, the findings contribute to theory on the operation of the graduate labour market by providing tentative evidence of how graduate internships contribute to positioning, based on external markers. Second, the findings make an empirical contribution by providing quantitative data that challenges prevailing policy assumptions about the value of paid and unpaid internships for individuals. Third, the findings contribute to policy by showing that not all internships are 'invaluable' and questioning the logic of not outlawing the practice of unpaid internships.
The findings suggest that, while paid internships improve labour market outcomes, unpaid ones do not and instead appear to have a negative 'scarring' impact in the short to medium term, at least for creative and mass communications graduates. The findings also challenge theory about the role internships play in socio-economic reproduction: they suggest that unpaid internships may not be so instrumental in accessing certain jobs.
To avoid doubt, we define internships here as specifically post-graduation employment experiences of a finite duration, aimed at improving one's career chances. As argued elsewhere, there are good reasons for seeing these 'open market' internships as distinct from other work experiences such as placements carried out as part of an educational course and voluntary work (Hunt 2020;O'Higgins and Pinedo 2018), although some studies referring to 'internships' use less focused definitions. In the UK, educational work placements are: supported by the students' institution, provide academic credit, and are often covered by student finance (Lawton and Potter 2010). Such 'governed' work experiences are more likely to be structured and developmental and may have quite different consequences for employability (Lain et al. 2014;Pollard et al. 2015). Furthermore, while graduate internships are commonplace in the voluntary sector (Leonard, Halford, and Bruce 2016), we distinguish between these experiences and volunteering carried out for altruistic rather than career reasons.
The paper proceeds as follows. First, we outline the context in which the practice of internships has emerged. Second, we highlight contradictions and inconsistencies in the literature around the value of unpaid internships for individuals. Third, we present findings from a survey of creative and mass communications graduates that suggest that unpaid internships may be less valuable to individuals than policy discourse assumes and may position graduates lower down the graduate labour market.

Internships in context
Following the introduction of, and subsequent increases to, student contributions to university tuition fees in the UK there has been growing policy concern about graduate 'employability' (Tholen and Brown 2018). While, initially during this period, the term was used to refer to an individual's chances of gaining and maintaining employment (Hillage and Pollard 1998), drawing on market individualism and human capital theory, policy discourses often view employability as the possession of skills and attributes needed to perform work (Holmes 2010;Tholen 2013). In this view, internships enable individuals to accumulate the skills, human capital and contacts to succeed in the graduate labour market (GPCF (Gateways to the Professions Collaborative Forum) 2013; CIPD (Chartered Institute of Personnel and Development) 2015). An alternative view emphasises the relational nature of the graduate labour market as a positional competition (Brown and Hesketh 2004). From this perspective, internships act as a 'hard currency' of employability, a line on one's CV, and provide a chance to gain 'soft currencies' , such as skills, adaptability and self-confidence (Brown et al. 2014).
Consequently, during the last fifteen years internships have emerged as a key route into certain professions and industries (Milburn 2009(Milburn , 2014, as a strategy for graduates to develop employability and for employers to 'try out' new recruits (Roberts 2017). This can be viewed in the context of two main developments.
First, the rapid expansion of higher education (HE) in the UK, combined with the polarisation of jobs, has led to an increasingly congested labour market (Tholen and Brown 2018). In this 'positional conflict' graduates must distinguish themselves from their peers by accumulating more, better or higher status credentials or other signifiers of employability (Brown and Hesketh 2004). Extracurricular activities, work placements and internships are part of this landscape (Bathmaker, Ingram, and Waller 2013;Tomlinson 2008).
Second, increasing financialisation and competition within the global economy means that employers are increasingly concerned with profitability and flexibility, and are cautious about committing to and investing in labour market entrants (Thompson 2013). Consequently, it increasingly falls to individuals, supported by labour market institutions like the educational system, to develop and display markers of employability (Brown and Souto-Otero 2020). Internships are emblematic of this, where the responsibility for developing work skills has shifted from the employer to the individual (Smith 2010;Thompson 2013). The short-term nature of such assignments, and ambiguous legal status of unpaid interns, means that an intern who does not measure up can be easily jettisoned (Hunt and Tzanakou 2020;Roberts 2017).
In light of these developments, some have raised concerns that access to the right kinds of internships act as a barrier to social mobility (Bathmaker, Ingram, and Waller 2013;Brown et al. 2014), particularly in industries and professions that lack institutionalised pathways (Leonard, Halford, and Bruce 2016;Shade and Jacobson 2015). Some theorists argue that, in this congested graduate labour market, middle class graduates are better able to 'play the game' and deploy their superior capitals, entrenching socio-economic reproduction (Bathmaker, Ingram, and Waller 2013;Brown et al. 2014;Roberts 2009). Those lacking the contacts and funds may be less able to access and afford opportunities that are thought to help further build valuable social and human capital, which is key in sectors dominated by fixed-term projects and freelance work (Grugulis and Stoyanova 2012).
The creative sector represents a critical case, for our purposes, as an industry where internships are increasingly considered an important route towards a career . The sector lacks institutionalised pathways from education to employment where an oversupply of graduates, and the promise of enjoyable work for glamorous firms, puts aspirants at risk of exploitative remuneration and conditions; and is a sector where short-term contracts, freelancing and non-standard forms of employment are commonplace Lawton and Potter 2010;Shorthouse 2010). It is perhaps unsurprising that the creative industries have high rates of participation in internships, both paid and unpaid (Hunt and Scott 2018;Hunt 2020).

Functions of internships
While there have been insightful qualitative studies of graduate internships in the UK and beyond (e.g. From the extant literature, three main themes can be discerned. First, graduate internships are thought to improve the employability of interns -particularly in sectors like the creative industries that lack established pathways into employment -by providing valuable work experience and developing networks, confidence, and industry-specific skills and knowledge (Milburn 2009; Mellors-Bourne and Day 2011; OCL and CRAC (Oakleigh Consulting Ltd. and CRAC) 2011). Internships and work placements, particularly for wellknown or prestigious organisations, are also thought to provide a valuable addition to a CV and help get a foothold in creative careers (Wright and Mulvey 2021;). This view is reflected in government and wider guidance on internships (GPCF (Gateways to the Professions Collaborative Forum) 2013; CIPD (Chartered Institute of Personnel and Development) 2015), and in the views of creative graduates and interns themselves who often see internships as a 'rite of passage' or a 'necessary evil' (Shade and Jacobson 2015;Siebert and Wilson 2013;Jacobson and Shade 2018).
Second, alongside the above view, there are concerns that employers exploit an oversupply of graduates as a surplus pool of free labour (Shade and Jacobson 2015;Frenette 2013;Perlin 2012). The allure of work in glamorous sectors, such as advertising, the media and the creative industries, makes aspirants vulnerable to exploitative practices in order to 'get noticed' (Siebert and Wilson 2013;Shade and Jacobson 2015;McLeod, O'Donohoe, and Townley 2011). The opportunity to 'try out' potential recruits at little cost or risk is seen as a benefit to employers but leaves interns in a precarious position (Smith 2010;Roberts 2017). Finally, a recurring theme, especially in the more policy-oriented literature (Milburn 2009(Milburn , 2014Lawton and Potter 2010;Roberts 2017;Montacute 2018), relates to the role internships play in terms of social mobility and socio-economic reproduction. Lack of the social contacts and information needed to access opportunities, and/or the financial means to work for low or no pay, are thought to exclude those from some backgrounds from accessing key industries and professions Waller 2013, Bathmaker et al. 2016). In these debates, paid internships, on the other hand, are assumed to be unproblematic due to more formal recruitment practices and removal of financial barriers (Lawton and Potter 2010).
Thus, internships, particularly unpaid ones, might be viewed as having a dualistic nature comprising both advantages and disadvantages, with some of the latter contingent on the former. As noted above, concerns about social mobility linked to the practice of unpaid internships rely on the assumption that a) these sorts of opportunities really do lead to better jobs -the 'stepping stones' or 'pipeline' view (Frenette 2013;Shade and Jacobson 2015;Siebert and Wilson 2013); and b) it is access to unpaid internships, and not so much to paid ones, that is problematic.
Overall, the extant literature provides a patchy and sometimes contradictory picture of the practice. First, qualitative research contains some inconsistencies. While unpaid internships are seen as unfair and exploitative, symptomatic of a changing labour market (Shade and Jacobson 2015;Frenette 2013), the argument that they exclude less privileged graduates from key industries -as stepping stones or a necessary evil -assumes they help get better jobs (Siebert and Wilson 2013;Jacobson and Shade 2018). Second, the picture painted by quantitative research is inconclusive. While some quantitative studies show positive effects for unpaid graduate internships others find the opposite and all have methodological limitations.
Evaluations of government backed schemes in the UK indicated that unpaid internships improved interns' employability but the analysis failed to control for confounding factors (Mellors-Bourne and Day 2011; OCL and CRAC (Oakleigh Consulting Ltd. and CRAC) 2011). Research using a YouGov survey of young people that did control for age, gender, location and social background indicated that engaging in one unpaid internship after graduation improved subsequent earnings but multiple unpaid internships did not (Cullinane and Montacute 2018). Full results of this analysis were not presented, making the findings hard to verify. Interestingly, this finding contradicts qualitative research that suggested such 'opportunity stacking' is beneficial to interns (Wright and Mulvey 2021), underlining contradictions in the literature. Analysis of a large-scale graduate survey in the UK showed that while unpaid work prior to graduation increased the chances of having a graduate job 1-2 years after leaving university, unpaid work after graduation did not ). However, the analysis did not distinguish between unpaid internships and other forms of unpaid work such as volunteering and unpaid family work. Also, analysis of Destinations of Leavers from Higher Education data that found that unpaid work six months after graduation led to less favourable outcomes three years later (Holford 2017), did not capture paid internships or unpaid internships undertaken after six months after graduation. Finally, analysis using panel data from a German graduate survey (Cerulli-Harms 2017) that found graduate interns subsequently earned less than their non-intern peers did not distinguish between paid and unpaid internships. Moreover, the German educational system and labour market differ substantially from the UK context. Thus, important questions remain unanswered.
In summary, the ways in which graduate internships operate in the labour market, as 'stepping stones' or as a mechanism of socio-economic reproduction, are contingent on two, as yet untested, propositions that we investigate below: • That graduate internships (paid or unpaid) really do improve employability; • That it is access to unpaid internships, and not paid ones, that creates problems for social mobility.

Data
The data used in this research comes from the Creative Graduates' Careers Survey: a bespoke survey of graduates from twelve UK higher education institutions (HEIs) that broadly reflect UK provision in creative arts and design (CAD) and mass communications and documentation (MCD) (e.g. publicity studies, media studies and journalism) subjects and account for 14 per cent of provision in these subjects. The survey employed a systematic probability sampling method covering UK and EU domiciled full-and part-time first degree graduates. These are two subject areas with a high incidence of graduate internships (Hunt 2017;Hunt 2020) that reflect industries lacking institutionalised pathways where concerns have been raised about social closure (Milburn 2009). The survey was carried out in autumn/winter 2014 and surveyed graduates two, four and six years after graduation (2007/08, 2009/10 and 2011/12 graduating cohorts) allowing investigation of labour market outcomes over time. The data collection period thus covers a time when participants had experienced the impact of financial crisis, which remains relevant backdrop, given the contemporary economic shocks of the pandemic and Ukraine war. Importantly, there is no evidence that participation in internships has declined significantly in subsequent years (Hunt 2020;Cullinane and Montacute 2018); concerns about unequal opportunities and exploitation in this sector remain current (SMC (Social Mobility Commission) 2021). If anything, the Covid-19 pandemic has made the labour market tighter for the present cohort of labour market entrants (Davis 2021). The questionnaire covered: course details; current and previous employment activities; perceptions about internships and other forms of employment; career satisfaction; and personal characteristics. Care was taken in question wordings to distinguish internships from voluntary work for a third-sector or public organisation, between paid and unpaid internships, and between work placements carried out while studying and after graduation. Development interviews and a semantic pilot were carried out on a small sample of stakeholders and previous interns, and a full pilot was also carried out. Full details of the methodology can be found in Hunt (2017, 125-145).
Graduates were invited to take part by their institution via email and completed the survey online. A total of 616 eligible responses were received, representing a response rate of 7.4 per cent after adjusting for any known undeliverable emails. While the response rate is low, as has been shown elsewhere, a low response rate does not necessarily lead to bias and increasing the response rate does not necessarily improve sample representativeness (Curtin, Presser, and Singer 2000;Keeter et al. 2000;Meterko et al. 2015). Of greater importance is that the achieved sample can be considered as broadly representative of the target population in terms of age, gender, ethnicity and broad subject area (See Appendix, Table A1), and while the sample is slightly skewed towards the more recent cohort (2011/12) and those achieving a first class degree these factors are controlled for in the multivariate analysis.

Outcome measures
If internships improve employability, those with internship experience should have more favourable labour market outcomes and will be more likely to secure 'better' jobs. In order to test this proposition the analysis looked at three measures of labour-market outcomes: professional/graduate-level jobs, creative jobs and income.

Graduate-level jobs
While there is considerable debate about what constitute a graduate job (James et al. 2013) and a number of definitions exist (e.g. Elias and Purcell 2012;Green and Henseke 2014), 'graduate-level' occupations were defined as Standard Occupational Coding system (SOC 2010) major groups 1-3: 'managers, directors and senior officials' , 'professional occupations' and 'associate professional and technical occupations' . 1 Although an imperfect yardstick, this measure gives an indication of whether the jobs graduates were doing were of a level generally requiring a degree and largely occupied by graduates. The measure is also commonly used by HEIs in internal analysis of graduate outcomes and in the statutory Key Information Sets (KIS) in the UK. 2 Furthermore, closer inspection of the occupations held by respondents in the current survey revealed a high level of agreement (95 per cent) between the measure used here and the definition proposed by Elias and Purcell (2012).
Graduates were considered to have a graduate-level job if their 'main' job (i.e. the one they spent the most time doing) was in SOC major groups 1-3. 'Jobs' were defined as any work activity described as: 'permanent employment (with wage/salary)'; 'self-employed/ freelance (including own business and commission work)'; and 'temporary employment/ fixed-term contract (with wage/salary)' . Self-employed, freelance and fixed-term employment were included in this definition of graduate jobs for the following reasons. First, such work is commonplace in the creative industries and a conscious choice for many creative graduates for whom the pursuit of creative practice is important for developing personal identity and meaning (Higdon 2018). A major study of creative graduates found a third (33%) were working in self-employed or freelance work 4-6 years after graduation and 62% anticipated some self-employed work during the following five years . Second, personal development, doing relevant work and the use of knowledge and skills are relatively more important for creative graduates than formal markers of success like earnings and status . While standard employment with a well-known employer may be an ideal for some, this model has declined in recent years in the creative industries, as outsourcing and short-term contracts have become increasingly the norm. As such, self-employed or freelance work that is either creative or at an associate professional level is likely to be considered an achievement for many creative graduates at this early career stage.

Creative jobs
Graduates from creative and mass communications disciplines are often motivated by a desire to be creative and to use the skills they have developed during their course Hesmondhalgh 2010). Therefore, a measure of creative jobs was derived, based on a definition developed in collaboration with a wide range of HEI, policy and creative sector stakeholders during the Creative Graduates, Creative Futures project . It includes traditional creative occupations such as graphic, clothing and product designers and artists as well as broader media and communications related occupations such as journalists, public relations and marketing professionals and associate professionals. A detailed list of the four digit SOC codes included in the definition of creative jobs can be found in Hunt (2017). In many cases, occupations defined as 'creative' here are also in SOC major groups 1-3; there is thus a high level of correlation between the definitions of creative and graduate-level jobs.

Income
Although income may not always be the primary concern of creative graduates , it provides another indicator of labour-market success. Furthermore, because occupational codes generally provide little indication of the relative level of employment within an occupational code -a junior graphic designer would have the same SOC code as a more senior one -income provides a supplementary measure of graduate outcomes. If graduate internships are thought to enhance the employability of interns it might be reasonable to expect this to be reflected in their remuneration.
Self-reported annual income was collected using banded response categories, increasing in £5,000 bands up to £50,000+. Two measures of income were used in the analysis: (1) a banded version with three levels (£15,000 or less; £15,001-£25,000; £25,001 or more) and (2) an interval income variable using the mid-point of each £5,000 band with £50,000 assigned as a conservative value for those reporting income in the top band. The former measure was used in the Ordinal Logistic Regression (OLR) analysis and the latter was used in a supplementary Ordinary Least Squares (OLS) regression analysis, both of which are described in the following section.

Approach to analysis
Multivariate analysis was used to examine the proposition that internships lead to enhanced labour market positions. In all analyses, the approach taken was to estimate a model containing all controls expected to predict labour market outcomes from previous research and then test whether the introduction of terms for internship participation improved the predictive power of the model. Effectively, this tested the hypothesis that graduate internships improve labour market outcomes.
Binary logistic regression tested whether experience of paid or unpaid internships increased the chances of having a graduate-level or creative job while controlling other factors shown to be related to graduate outcomes Saniter and Siedler 2014;Walker and Zhu 2011;De Vries 2014). The control variables used were as follows: • Age at graduation; 3 • Gender (Male or Female); • Ethnicity (White or Black, Asian and minority ethnic); • Parental experience of HE (a proxy for social class -Roberts 2010); • Current residence (London, elsewhere in the UK, or outside the UK); • Graduating cohort; • Broad subject area (CAD or MCD); • Classification of degree (first class, upper second, or lower second and below); • League table score of HEI (a proxy for institutional reputation); 4 • And whether or not graduates had undertaken a work placement whilst at university.
Ordinal Logistic Regression was then used to test whether internship participation was associated with propensity to be in the top (and bottom) income bands, using the same approach as for the logistic regressions outlined above. As noted previously, the creative sector is characterised by a high incidence of project work, self-employment/freelance work and part-time working (Grugulis and Stoyanova 2012;Ball et al. 2010). As a consequence, we include additional controls for hours worked (full-time or part-time), and if they had multiple jobs (so called portfolio working). Previous studies of creative graduates (e.g. Ball et al. 2010), suggest that those who have multiple work activities tend to earn less overall compared to those with one main job.
OLS regression, using the interval measure of income described in the previous section, was conducted as an indicative analysis to aid interpretation and as a robustness check. While the measure used in the OLS regression cannot truly be considered an interval variable, this analysis was considered justifiable for the following reasons. First, because the coefficients in OLS regression represent the estimated effect of the independent variable on the outcome in monetary terms this analysis is more intuitive to interpret than ordinal logistic regression (where odds ratios represent the change in odds of being in one income band versus another). Second, the distance between bands was uniform up to the maximum and the income measure used is broadly normally distributed (see Appendix, Figure A1). Third, this approach follows previous studies investigating graduate incomes that used a similar measure (e.g. De Vries 2014; BIS (Department for Business and Innovation and Skills) 2013).
While the above approach cannot prove causality it can examine whether participation in paid or unpaid internships is associated with improved labour market outcomes.

Results
While bivariate analysis appears to indicate that both paid and unpaid internships do improve labour market chances, deeper inspection tells a different story. While paid internships do improve employability, contrary to prevailing discourse unpaid internships do not convey the same benefits.

Graduate-level and creative jobs
At a bivariate level, graduates who had completed a paid internship were more likely to have a graduate-level or creative job than those without internship experience (Figure 1) (respectively -χ 2 = 9.37, p = .009, column p< .05; χ 2 = 18.087, p< .0005, column p< .05). However, while those who only had experience of unpaid internships appear to be more likely than those without internship experience to have a graduate-level or creative job the differences in column proportions were not statistically significant at the α= .05 level. So while on the face of it both paid and unpaid graduate internships appear to improve employability the figures presented here fall short of supporting that picture. The findings of the multivariate analysis further challenge the stepping stone view for unpaid internships.
Binary logistic regression was used to test whether graduate internship experience helped predict outcomes while controlling for other variables including grades, HEI reputation and previous placement experience (Table 1, Appendix Table A2). We estimated a model including just the control variables (Model 1) and then tested whether the addition of terms for internship experience significantly improved the model by looking at the change in −2 log-likelihood (Model 2). In the case of both graduate-level and creative jobs, the addition of terms for internship experience did significantly improve the model (P grad = .036; P creative .041), and the Nagelkereke Pseudo R 2 indicated that both models were a good fit of the data (Nag grad = .187; Nag creative = .182).
Consistent with previous research on graduate jobs, coefficients for classification of degree, gender and geographical location were found to be statistically significant, with higher grades, being male and living in London associated with an increase in the odds of having a graduate or creative job. Parental experience of HE was associated with an increase in the odds of having a creative job (significant at the .1 level), but not a graduate job. The lack of a strong association between social class and propensity to have a graduate job could be due to the fact that class is associated with league table score (t(613) = −5.302, p<.001) and participation in paid internships (χ 2 = 8.584, df = 1, p = .004), both of which help predict graduate jobs. As such some of the variance between social class and chances of having a graduate job may be accounted for by university attended and participation in paid internships. The fact that parental experience of HE was still associated with chances of having a creative job may reflect the importance of networks and social capital in the creative sector (Grugulis and Stoyanova 2012).
Similarly, participation in a university work placement was associated with an increase in the odds of having a graduate but not a creative job. That coefficients for league table score in both models and university work placement experience in the model for creative jobs change from statistically significant to non-significant may be due to both of these variables being strong predictors of internship participation in Base: graduates in work. ***P < 0.001, **P < 0.01, *P < 0.05, + P < 0.1. Base: graduates in work (n = 521). ***P < 0.001, **P < 0.01, *P < 0.05, +P < 0.1.
previous analysis carried out using the same data (Hunt and Scott 2018), thus some of the variance associated with these variables may be accounted for in the internship variables. Cohort and age group at graduation were not found to be statistically significant in either model. Perhaps because the occupational measure used, based on SOC codes, is unable to pick up on differences in progression based on experience or that occur over time. However, in the income analysis in the following section, both variables were found to be statistically significant. The lack of statistically significant coefficients for subject area is likely to reflect that both subject areas tend to be associated with similarly low labour market outcomes relative to other subjects, such as STEM (Belfield et al. 2018;De Vries 2014). While non-significant coefficients for ethnicity could either indicate that outcomes are not generally worse for graduates from a BAME background or that disparities in outcomes between graduates from different ethnic groups are masked by the broad category used. The wide confidence intervals for ethnicity may support this latter explanation.
Crucially, however, for both outcome variables while experience of a paid internship was associated with an increase in the odds of having a graduate-level or creative job relative to not having completed an internship (by more than a factor of two -ceteris paribus), the unpaid internship coefficients were not statistically significant (p = .110 for graduate-level and p = .536 for creative jobs). While paid internships do appear to help graduates secure a graduate-level or creative job as a main job, the idea that unpaid internships confer similar advantages is not confirmed. In other words, we can say that paid internships do help CAD and MCD graduates get better jobs, but we find no evidence that unpaid internships do.

Income
When it comes to income, those who had completed an unpaid internship tended to earn less on average than those without prior internship experience ( Figure 2) and those with paid internship experience earned more (F(2,540) = 3.455, p = .032), although only the difference between those with paid internships and those with unpaid internships was statistically significant (p = .031). However, when controlling for other factors, such as grades and institutional reputation, the differences in the value of paid and unpaid internships were more pronounced.
Using OLR (Table 2, Appendix Table A3), the addition of terms for graduate internship experience (Model 2) helped improve the prediction of income over a base model containing only controls (Model 1) (Δ-2Ll = 8.168, df = 15.825, p< .0005), with the Nagalkerke pseudo R 2 indicating that the model was a good fit of the data (.408).
The following controls were associated with a statistically significant increase in the odds of being in a higher income band compared to being in one band lower: • Age at graduation; • Classification of degree; • Living in London; and • Cohort of study with greater time in the labour market associated with increased odds of being in a higher income band.
As expected, having multiple jobs or having only part-time jobs was associated with a statistically significant decrease in the odds of being in a higher income band.
Coefficients for league table score of HEI, ethnicity, gender, broad subject group and parental experience of HE were not found to be statistically significant. As with the models for graduate and creative jobs, it may be that league table score of HE and parental experience of HE are non-significant in this analysis because they are strongly linked to internship participation and so some of the variance associated with these variables on income are accounted for in the internship variables. For ethnicity and gender it is plausible that differences in earnings have not had long enough to emerge at this stage or are masked by income disparities between ethnic groups within the BAME category, while the two broad subject areas are both relatively low earning subjects compared to other subjects such as medicine, economics and engineering (Belfield et al. 2018).
Crucially for our analysis, participation in a paid graduate internship was found to be associated with a statistically significant increase in the odds of being in a higher income band by a factor of 2.28. Participation in unpaid internships, however, was associated with a statistically significant decrease in the odds of being in a higher income band of by a factor of 2.44 (the inverse of 0.41, Model 2). Moreover, because of the proportional odds assumption in OLR, when reversing the income bands (Model 3) we can see that participation in unpaid internships is associated with a statistically significant increase in the odds of being in the lowest income band by the same factor (2.44).
To put this into context using OLS regression (Model 4), participation in a paid graduate internship is associated with an increase in income of around £2,258 while participation in an unpaid internship is associated with a decrease in income of around £3,963. While these estimates should be treated with some caution -as the outcome measure is based on £5,000 bands and not strictly interval -the positive effect of having engaged in a paid internship is of a similar order to two years in the labour market (as reflected in the coefficient for the 2009/10 cohort) and the negative effect of having engaged in an unpaid internship is considerably more than this. This latter finding contradicts prevailing beliefs that unpaid internships are a stepping stone to successful careers, but reflects the analysis for graduate-level and creative jobs, outlined above, and reinforces the findings of other studies that have found that unpaid work after graduation can have a negative effect on labour market outcomes Holford 2017). Additionally, while this finding contradicts Cullinane and Montacute (2018) finding of a salary premium for those who had engaged in a single unpaid internship, it is not clear whether their analysis controlled for grades, institution or other factors that may determine wages.

Discussion and conclusion
The idea of graduate internships as 'stepping stones' , albeit with redeemable flaws, pervades much policy discussion. Policy discourse on internships assumes that, paid or unpaid, they help improve employability and get a 'foot in the door' . Similarly, debates about the role internships play in social mobility and socio-economic reproduction are also contingent on this assumption, as concerns that unpaid internships exclude disadvantaged aspirants implicitly assumes these opportunities are useful. We have presented quantitative evidence to interrogate such claims and better incorporate the nature of internships into theories of the education -employment transition. The study has certain limitations: caution must be exercised in extrapolating results from a small survey in one -albeit significant -sector to all sectors, especially in relation to income data; response rates are inevitably relatively low in places; the survey did not enable measurement of variables such as internship duration or perceived prestige of employers. Previous research indicates that unpaid internships tend to be shorter and less developmental than paid internships (Mellors-Bourne and Day 2011; Scott 2018, 2020) and it may be that these and other factors like prestige (Wright and Mulvey 2021; influence outcomes. Future research would do well to disentangle these factors. However, the key point here is that the type of internships that tend to be unpaid appear to be, by and large, less valuable in the graduate labour market. While our analyses cannot prove a causal link between internship experience and labour market outcomes, the findings challenge the 'stepping stone' view and provide further evidence of the differences between paid and unpaid internships. The findings suggest that, for CAD and MCD graduates at least, while paid internships do appear to help graduates in the graduate labour market, there is no evidence that unpaid ones do, and may have negative consequences as longer-term career markers -especially in the case of serial unpaid internships. When controlling for other factors such as grades and institutional reputation (among others), there was no evidence to suggest that unpaid interns were any more likely than non-interns to secure a graduate-level or creative job (all else being equal), and actually tend to earn less.
On its own, the lack of any evidence to reject the null hypothesis that unpaid internships make no difference to individuals' labour market chances might be considered a non-finding. Given a larger sample, for example, might a significant association between unpaid internships and positive labour market outcomes be found? Our data cannot refute such a prospect entirely, but closer analysis of the results renders this unlikely, for two reasons. First, in the analyses for creative and graduate-level jobs, although the coefficients for unpaid internship experience were non-significant, the odds ratio for graduate-level jobs is in the negative direction. While the coefficient was in a positive direction for creative jobs, the confidence interval is sizeable. Second, the results of the analysis for income, which show that engaging in an unpaid internship is negatively associated with pay in the short to medium term, make the prospect of finding a positive effect for unpaid internships on occupational outcomes seem less likely. Furthermore, the findings chime with previous research, employing different methodologies, that found that unpaid work carried out after graduation led to less favourable labour-market outcomes Holford 2017), and that paid internships are more selective than unpaid internships and perceived as more developmental (Hunt 2020;Scott 2018, 2020).
On the issue of pay, at least, there may be less reason to question the finding that unpaid internships appear to have a negative impact in the short to medium term. The vast majority of our interns (81 per cent of unpaid and 82 per cent of paid) ended up working in the same ten out of 24 industries (e.g. design, fashion, advertising/publicity and education), suggesting that industry does not account for the observed differences in earnings. While it is hard to say from the current research exactly why unpaid interns tend to earn less, even after several years in the labour market, there are several possible explanations. First, engaging in unpaid work for up to a year, or longer, may simply put them behind their peers on the graduate labour market. However, given the relative size of the negative effect on income -much larger than the estimated effect of two years in the labour market and of similar magnitude to that found elsewhere (Holford 2017) -this is unlikely to explain the entire effect. Second, engaging in an unpaid internship may put graduates in a weaker bargaining position when accepting their first paid job, with lower reservation wages (Cerulli-Harms 2017). However, this is impossible to confirm from our data and many employers may have standard starting salaries for labour market entrants. Finally, and perhaps most intriguing, it is possible that unpaid interns are simply 'positioned' below those with relevant paid experience and those snapped up before leaving university (e.g. through work placements: Pollard et al. 2015). Unpaid internships may therefore have a negative signalling effect (Cerulli-Harms 2017). Even within unpaid internships, there is evidence to suggest a hierarchy of opportunities with more privileged graduates seemingly able to access the better opportunities (Holford 2017;Wright and Mulvey 2021). Further research is needed on such questions.
In policy discourse, informed by market individualism and human capital theory, unpaid internships should still provide interns with the opportunity to develop the human capital (skills, networks and experience) that should be valued in the labour market, particularly in professions and sectors where relevant paid opportunities are hard to come by (Milburn 2014;Montacute 2018). However, the finding that unpaid interns appear to do less well than non-interns questions this logic. Theorists, critical of the dominant discourse, have argued that in an increasingly positional labour market employers use a range of 'hard' and 'soft' currencies to identify the most 'talented' candidates, including, educational credentials, work experiences and extracurricular activities, as well as less tangible qualities (Brown and Hesketh 2004;Bathmaker, Ingram, and Waller 2013). While there is some evidence that paid internships may be more developmental than unpaid internships and at a level requiring greater level of qualification Scott 2018, 2020), doing an unpaid internship should still provide opportunities to develop industry-specific skills and networks relative to not doing one or doing less relevant paid work. Admittedly, while some non-interns in our sample will have managed to secure relevant work straight after graduation unpaid work is often considered 'inevitable' for creative sector aspirants and 'often the only route' (Siebert and Wilson 2013, p712;Brook, O'Brien, and Taylor 2020). It is easy to see why paid interns may be seen as a safer bet relative to unpaid interns as they might assume that payment shows that interns have been more valued by the host organisation and have been given a greater level of training and responsibility. However, the finding that unpaid interns appear to fare less well than non-interns, even though the stepping stone discourse claims they are a way to develop skills and networks and 'get ahead' (GPCF (Gateways to the Professions Collaborative Forum) 2013; Siebert and Wilson 2013), lends further weight to the idea of a positional labour market based on a hierarchy of opportunities and external markers, rather than a meritocratic matching of skills and human capital possessed to jobs (Brown 2013;Tholen 2013).
In a post-Covid economy where the creative industries have been particularly hard hit, the fact that unpaid internships do not appear to help CAD and MCD graduates in the graduate labour market is particularly concerning for two main reasons. First, while there is some evidence that unpaid internships have declined in recent years (Hunt 2020), resurgent recessionary pressures may put new graduates at greater risk of embarking on a course of action with high personal costs, in terms of hardship and loss of income, but little chance of producing the expected employability benefits. The findings make it hard for employers to justify the exploitation of free labour as a quid pro quo for employability benefits and challenge the justification for not bringing unpaid internships unambiguously within minimum wage legislation. Unpaid internships may be more about free labour and offsetting risk than a stepping stone, supporting views of a less patient, financialised capitalism that expects labour market entrants to 'hit the ground running' (Thompson 2013;Brown and Souto-Otero 2020). Second, analysis from the same survey shows that those from more privileged backgrounds are more able to secure the more beneficial paid internships (Hunt and Scott 2018), indicating that class (dis)advantage plays a significant role in dictating who is able to access the best opportunities. It is not altogether clear why this is the case, but previous studies have shown that middle-class graduates are more able to recognise and 'play the game' , and to package their 'self ' in ways that are more attractive to employers, as well as being able to draw on their frequently superior social and cultural capital (Bathmaker, Ingram, and Waller 2013;Tholen et al. 2013). Reproduction theorists have already noted how in the education system middle-class cultural capital and ways of being are rewarded while working-class cultural capital is disparaged (Roberts 2009;Bourdieu 1984). In this sense, internships can be seen as an extension of the same mechanisms that reproduce patterns of advantage and disadvantage that are evident in the education system (Boliver 2011).
Our findings reveal that graduate internships do appear to operate as a mechanism of socio-economic reproduction, but not necessarily in the way predicted by prevailing wisdom. While unpaid internships are certainly problematic from an ethical viewpoint, ability to access the better, paid, internships is a greater concern for social mobility, as it is these opportunities that hold greater currency in the graduate labour market. Issues of social mobility go much deeper than questions of ability and affordability.
Notes assumption. This was not found to be a problem in either of the income analyses and so the continuous variable was preferred as it contains more information. 4. Based on the most recent league table scores available for the relevant subject areas from the Guardian and Good University Guide league tables.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Data availability statement
At the time of the survey participants of this study did not agree for their data to be shared to third parties, so supporting data is not available.

Data deposition
The data supporting this research has not been deposited.