The rural pay penalty: youth earnings and social capital in Britain

Young people in rural areas are under-researched, and there is a particular paucity of studies on rural youth in the labour market. This paper addresses that dearth. I pose the research question: how does rural location affect the earnings of young people in full-time employment in Britain? I consider the background of rural disadvantage, and its specific effects on young people and outline the relevance of social capital to this topic, identifying norms and networks as the two constituent elements of the concept. Using data from the British Household Panel Survey Wave 17 (2007/8), I find that rural youth are paid less than urban counterparts. When coupled with higher living costs, this amounts to a double disadvantage for rural youth. Concerning social capital, I find that norms – in terms of trust in individuals, community and institutions – exert a significant effect on net pay. Conversely, despite prior research positing the importance of informal contacts for rural jobseekers, networks do not exhibit any significant effect on wages. This corroborates accounts of social capital as a protean concept, illustrating how one facet alters the likelihood of finding work, while the other facet determines outcomes once in employment.


Introduction Á rural Britain and disadvantage
England is densely populated (Scotland and Wales less so) and heavily urbanised, yet a significant minority live outside of the larger towns and cities. Rural Britain has an ageing population (Lowe and Speakman 2006), so young people in such areas are a minority within a minority and have largely been neglected by research. In this paper, I address this oversight by comparing labour market outcomes for rural and urban youth. I pose the research question: how does rural location affect the earnings of young people in full-time employment in Britain? I argue that coupled with higher living costs, lower pay in rural areas amounts to a double disadvantage for young people. Informal networks, highlighted in the existing literature as crucial in rural job searches, do not affect earnings significantly, while the other dimension of social capital, norms, does exert a significant effect.
Today, Britain extends the same benefits to rural residents as to others Á the NHS, education and other services of the welfare state. A drastically different identity and social outlook are not features of British rural life in an era of universal compulsory education and media proliferation (although there have been some suggestions to the contrary Á see Neal and Walters 2006). The countryside in Britain has mostly shifted to a post-productivist economy, and those employed in rural districts generally work in the same sectors as urban counterparts (Cherry and Rogers 1996, 110;Taylor 2008, 123). The difference between urban and rural life in Britain is less striking than elsewhere, including Western nations with vastly different geographies, such as Canada, Australia and the USA. Britain may be unique in this respect.
Despite population density and welfare provision rendering the rural/urban disparity less pronounced than in other countries, it remains a concern that location can be a determinant of life-chances. The lack of existing research into how rural youth are disadvantaged compared with urban peers leaves this issue relatively unexplored. Concentrated deprivation is far more prevalent in urban Britain (Milbourne 2010, 164) yet solely studying such areas ignores other experiences of disadvantage, which can be exacerbated by remoteness (Noble and Wright 2000). As much as 928,000 rural households live below the poverty line, with this deprivation camouflaged by illusions of rural idyll and 'hidden urban biases in policy and delivery' (Burgess 2008a, 3). Income is clearly a major contributor to rural disadvantage. There have long been concerns that rural residents 'suffer the additional burden of the relative affluence of their surroundings', their hardship concealed in statistical overviews by the wealthier majority living in the vicinity (Cloke et al. 1995, 360). This paper considers how rural disadvantage is reflected in youth labour market outcomes. The next section discusses young people in rural areas.

Rural youth
Britain's rural population is older (Lowe and Speakman 2006;Hardill and Dwyer 2011), yet the rural advocate also notes a decline in net out-migration of young people from rural areas (Burgess 2008b, 2). These young people have difficulties obtaining affordable housing (Wilcox 2006, 19;Taylor 2008, 86). They often find that rural employers, usually small businesses (Spielhofer, Golden, and Evans 2011, 7;CRC 2012, 40), struggle to supply well-coordinated external training; both of these factors impact upon future prospects (OECD 2008, 98). Post offices and primary schools are closing and public transport has become less frequent and more expensive (Milbourne 2004, 569Á570;CRC 2008;Burgess 2008b).
Rural youth are more reliant on temporary work, the use of private vehicles and jobs without promotion opportunities (Hodge et al. 2002;Midgley and Bradshaw 2006), yet there are very few studies into rural youth employment prospects. Cartmel and Furlong (2000) provide the most comprehensive contribution to this topic. They generalise about rural labour markets, identifying three common features: 'restricted opportunities, the need for private transport, and the use of local contacts for recruitment' (Cartmel and Furlong 2000, 27). The claim that rural areas exhibit these characteristics is plausible, but this plausibility would be increased by showing that these traits did not extend beyond rural areas, thus proving their distinctiveness. Cartmel and Furlong fail to do this, despite implicit claims of familiarity with urban markets: 'the chances of finding work in depressed urban areas are much poorer and long-term unemployment is much more common' (35). This putative distinction is not explained in greater depth or supported by comparative empirical evidence, so the claims here remain unwarranted.
Cartmel and Furlong bemoan how 'a lack of incentive to embark on education or training routes may leave rural youth in a more or less permanent cycle of unemployment followed by short term employment' (12). They conclude that whilst unemployment is more deeply entrenched in deprived urban areas, different yet significant problems exist in rural locations, namely, limited opportunities and poor public transport (35). The study is laudable for investigating this under-researched area, yet suffers from setbacks, such as the failure to support some claims with appropriate comparators. Nevertheless, Cartmel and Furlong opened the door for further enquiries into this area, enquiries which are, for the most part, yet to materialise. This is surprising, as their data is from 1997 to 1999, and rising house prices, increases in NEET youth, falling youth out-migration from rural areas and declining public services have been witnessed during the intervening decade, indicating that the problems facing rural youth have increased.
There have since been modest advances around this topic, such as Phimister, Theodossiou and Upward's (2006) investigation of low-paid work in urban and rural areas, using British Household Panel Survey (BHPS) Waves 1Á8 (1991Á8). They find that 'urban low-pay durations are somewhat shorter on average, with a higher probability of movement to a higher paid job ' (2006, 693) and that young people in rural areas are likelier to leave low-paid jobs for unemployment than urban peers, although the differences are modest (2006,708). Their attempt to systematically compare urban and rural employment represents progress from Cartmel and Furlong (2000) by adding further detail regarding pay and duration.
The evidence suggests that rural Britain is now home to more young people, who struggle to find permanent employment with good prospects and pay and are disadvantaged by the lack of affordable housing and public transport. Research into this topic, particularly in Britain, has been almost non-existent since Cartmel and Furlong's study in 2000, despite the fact rural youth are seen to face difficulties which are both significant and distinct. The next section introduces social capital, often seen as pivotal for employment in rural areas (Cartmel and Furlong 2000;Hodge et al. 2002;Mathews, Pendakur, and Young 2009;Spielhofer, Golden, and Evans 2011, 6Á12). I start with general definitions of the concept followed by its application to labour market outcomes, youth employment and rural areas.

Social capital and rural disadvantage
In this section, I begin by outlining the concept of social capital, with specific reference to the work of Coleman and Putnam, before expounding its relevance to the study of rural youth labour market prospects. In essence, 'the central idea of social capital is that social networks are a valuable asset' (Field 2003, 22). As such, Coleman portrays social capital as a positive, productive resource, 'making possible the achievement of certain ends that would not be attainable in its absence ' (1990, 302). He argues that social capital in terms of community and family support can compensate for a lack of public resources. Much research posits social capital as a significant predictor of positive outcomes, such as educational attainment and employment prospects, for young people (Coleman 1988;Putnam 2000;Porfeli et al. 2009, 72). However, Coleman's depiction of social capital as wholly positive renders his contribution rather one-dimensional. Even his admission that social capital facilitates some actions whilst constraining others (Coleman 1990, 311) falls short of explicit reference to negative aspects.
Putnam also views social capital as an asset to communities and societies, proclaiming that 'the core idea of social capital theory is that social networks have value . . . social contacts affect the productivity of individuals and groups ' (2000, 18Á19). His principal argument is that the USA has suffered endemic social decapitalisation characterised by declining civic participation (1995). Putnam believes civic disengagement is a national problem, revealing that he sees social capital, in terms of shared norms, trust and networking as a valuable social force (91). Crucially, however, he also looks at the other side of the coin.
Putnam distinguishes between bonding and bridging social capital, with the former defined as inward looking, typified by denser networks of homogeneous individuals, tightly connected by virtue of their commonalities. This is contrasted with bridging social capital, which creates connections between people who are likely to be more diverse (2000,411). He suggests that some bonding social capital may discourage the formation of bridges to other groups, and vice versa (362), indicating that emphasis on certain kinds of connections can be detrimental to other types of relations. The explicit admission that social capital 'can be directed towards malevolent, antisocial purposes, just like any other form of capital' (22) demonstrates awareness that not all norms and networks are good for individuals and society.
Although Putnam clearly sees social capital as predominantly positive, the acknowledgement of its complexity is a major strength of his argument, which has been supported by numerous empirical studies (Granovetter 1973;Lin 2001;Mathews, Pendakur, and Young 2009). The purported prevalence of personal networks for rural jobseekers (Cartmel and Furlong 2000, 27) suggests social capital is a concept with empirical purchase on the relationship between location and youth employment outcomes. Whilst access to networks is crucial, shared norms and values are deemed an equally important facet of the concept, manifested in Putnam's emphasis on trust and civic participation. Stone and Hughes (2002, 5) state that the 'key measures of social capital are norms and networks' and argue for social capital to be operationalised in a way which allows for its different components to be analysed distinctly. Later, I find this to be an effective approach to studying social capital and labour market outcomes.
It has been suggested that rural locations enjoy a comparatively strong community ethos (Glendinning et al. 2003, 151) and that access to local networks is vital for gaining employment in rural areas (Mathews, Pendakur, and Young 2009). Migration to rural areas has increased during recent years, and it is claimed that rural areas enjoy higher levels of community activity such as 'volunteering' and involvement in 'political action' (Burgess 2008a, 63).
However, there is also evidence that rural areas are not so superior to urban locations in terms of social capital stocks. For example, as rural areas are behind in terms of broadband access (Burgess 2008a, 30), and as the positive effect of such technology has been postulated by research (Lin 2001, 215), this suggests that they are disadvantaged where this element of networking is concerned. Furthermore, as transport in remote areas is widely regarded as inadequate (Burgess 2008b, 19), and young people in rural areas are arguably more vulnerable to isolation than urban counterparts (Valentine et al. 2008, 29), remote locations appear punitive to young people who might otherwise benefit from access to networking opportunities taken for granted by urban peers. Glendinning et al. found that rural residents overcame the lack of services via the safety and security of family and community, although one caveat is that females feel more affected by social claustrophobia and gossip than males in such locations (2003,151). However, the concerns highlighted by girls here indicate that the networks available to some can be inaccessible to others, effectively acting as barriers. Watkin and Jones also discuss the importance of rural networks but contend that only a small minority can gain entry. Without well-established local reputations and 'untarnished family backgrounds' people can be excluded from the crucial informal channels offering job opportunities (2008,230).
This idea was explored more explicitly by Mathews, Pendakur, and Young, who claim that jobseekers in rural areas use different means to find employment, arguing that 'rural job-finding is strongly influenced by constraints on the labour market and on social capital and networks that do not exist in cities ' (2009, 308). They qualify this statement by adding that good connections in such places are mitigated by the shortage of job opportunities compared with urban areas (310) and find higher rates of self-employment and unemployment in more remote areas (314). Additionally, their findings dispute Granovetter's (1973) argument that weak ties, such as professional contacts or college acquaintances, are more useful than stronger connections, such as family members or long-term neighbours, for finding work in urban labour markets (Mathews, Pendakur, and Young 2009, 317), contending that urban jobseekers are likelier to rely on informal means (321).
Implicitly supporting research which points to less time spent on lower wages in urban areas compared with rural locations (Phimister, Theodossiou, and Upward 2006, see above), Mathews, Pendakur, and Young find weak ties likelier to result in low earnings in rural communities (2009, 320Á321). This corroborates the studies stating the importance of familiarity with local markets (Cartmel and Furlong 2000;Hodge et al. 2002). They observe that whilst weak ties might prove useful in finding work, insecurity, modest pay and poor career prospects characterise the positions filled by using informal connections (Mathews, Pendakur, and Young 2009, 326). Moreover, they discover that rural communities contain dual labour markets, with one for established local residents, and another for those who are newer to the area. Those living there for six years and longer are likelier to use strong ties in finding work (322). This demonstrates how rural employment shows significant favour to those able to access local networks and that a lack of social capital in this regard makes entry into the labour market much more difficult. Mathews, Pendakur, and Young (2009) explore an area where explicit comparisons are scarce, but their study has limitations. They state clearly that the two datasets used represent vastly different areas in terms of economic buoyancy (310). Thus, it is unsurprising that rural labour markets present more hostile conditions in general, let alone to those who are unable to utilise personal networks for jobsearches. Also the average age of rural respondents is higher (54.8 years) than in the urban regions (43.3). It is expected that older people have greater difficulty accessing jobs in rural labour markets as professional occupations have declined in rural Canada, where the study is based (Mathews, Pendakur, and Young 2009, 310). Finally, although this is not a criticism of the study itself, Canada is unique geographically Á as is any country Á so the findings are not necessarily applicable elsewhere. Despite this, Mathews, Pendakur, and Young warrant praise for systematically comparing rural and urban regions and incorporating social capital into the analysis.
Social capital is clearly a collective concept, given that it is defined as the ability for people to achieve shared goals through working together, microdata can only reflect the experiences of trust, norms and networks as perceived by individual survey respondents. However, measuring social capital in individuals remains useful. For example, organisation membership or frequency of interaction with friends, family or neighbours are all indicative of someone's integration into different social networks. The principal tenet of social capital theory is that these networks are more than the sum of their parts, and how individuals engage with them represents their personal stock of social capital. Having articulated the salience of this concept to rural youth employment, I now introduce the data and methods used.

BHPS Wave 17 (2007/8) data
The BHPS includes respondents in England, Wales, Scotland and Northern Ireland. The total sample size for Wave 17 is 14,910, with 2242 aged 16Á24. Regional identifiers are not included in the standard data-set and are obtained through special licence permission, which have been merged with the individual respondent data file to create the sample used in this paper. Despite the different number of rural/urban categories in each country, classificatory guidelines include advice as to how these can be collapsed into a dichotomous schema. The criteria for rural/urban status vary according to nation, and this paper analyses respondents from the entire UK. Rather than imposing a single cut-off point for population size or density that may be insensitive to the different geographies in question, separate classifications for each country are used. These are based on definitions created by the Department for Environment, Food and Rural Affairs in England, the Scottish Executive in Scotland and the Department for the Environment in Northern Ireland. All define rural areas according to settlement size, population density and distance from larger conurbations, although each country adopts slightly different measures (for full details, see Institute for Social and Economic Research 2008). Therefore, all respondents are allocated a rural/ urban status that fits the national context rather than categorised according to potentially inappropriate universal benchmarks. Thirty-three per cent of the full BHPS individual sample is rural, compared to 30% for respondents aged under 25. Table 1 shows that a slightly higher number of rural youth are unemployed, although the difference is not statistically significant. A marginally higher proportion of urban youths are employed (this group also includes those who report being selfemployed), while there are more young people in rural areas in full-time education. While youth training is important, there are not enough respondents on government training programmes to warrant a separate category here (n04), so this group is added to the 'other' category. This also includes family carers, those on maternity leave and with long-term illness or disabilities. This exemplifies how youth unemployment in rural areas is overall very similar or possibly higher than in urban locations, which is curious given the lack of attention paid to rural youth in prior research. With roughly equal proportions of rural and urban youth in employment, I now assess the inequality of labour market outcomes, using net monthly pay as the measure.

Operationalising social capital
BHPS contains 19 variables on potential social capital indicators. Entering these into a regression model would prove unwieldy, so factor analysis is an appropriate method to reduce this number. There is a significant theoretical rationale for believing that some relationship exists between many of the 19 variables, strengthening the case for this approach. I have argued above that each variable used here is included for its relevance as an indicator of the overarching concept social capital, so it is expected that the factors here are related. Therefore, the results that follow are from oblique Á specifically, oblimin Á rotation. In practice, the rotation method makes little difference to the results here. This is to be expected with the analyses incorporating enough variables and stable factors with relatively high loadings (Tabachnick and Fidell 1983, 404).
Specifying a two-factor solution produces the results seen in Table 2 (for a fuller account of the factor analysis process, see Appendix A). First, there is a factor relating to networks. Personal trust does not load significantly on this; nor do frequency of meeting people and talking to neighbours. The variables relating to contacts outside of the household who can lend money or help find a job also do not load heavily on this factor. Instead, the factor extracted here relates more to personal support networks. Whilst such connections emerge from the social capital literature as critical for finding work, there is no evidence of networks or any aspect of social capital, determining outcomes once in employment. Hence, this paper examines whether these networks are as crucial for those who are in employment as for jobseekers.
The second factor extracted concerns community norms and trust. If loitering youths, vandalism, fear of crime and racism are rife, this reflects negatively on the norms of the community and is indicative of a decline in social capital. This is understandably linked to perceived trustworthiness of other people. The norms and trust which an individual sees as characteristic of their community could feasibly affect employment outcomes. This is consistent with extant social capital literature which correlates strong family and community bonds with positive outcomes in education and employment (see above). Thus, this factor is worth entering into the model. I now present findings attesting to the rural pay penalty for young people in full-time employment and revealing the complex effect of social capital on earnings.

The rural pay penalty
Having reduced 19 social capital variables to two factors, these can now be entered into a multiple linear regression model. The outcome variable is pay, a reliable proxy  Pouliakas and Theodossiou [2010] argue that while low pay equates to less job security and satisfaction in some European nations, this does not apply to the UK), and indicative of the skill level required for the work in question. Many of the sample are not earning, owing to unemployment or remaining in full-time education (which also includes many low earners, 95% earned under £500 in the previous month Á see Table 3). Table 4 compares rural and urban respondents' earnings by gender, age and qualifications. Urban males enjoy the highest median earnings. Rural females earn the least, perhaps a reflection of greater service sector opportunities, typically staffed  by females, in urban locations. The earnings gap between females in rural and urban areas is £103 per month, far greater than the rural/urban difference in male pay, which stands at £12 per month. Furthermore, the gap between male and female earnings in rural areas, £171 per month, is more than double the gendered pay disparity among urban respondents. Whilst addressing unequal labour market outcomes according to gender is not a priority for this paper, it is worth highlighting this finding, which could warrant further investigation. The urban premium is also slightly greater for 20-to 24-year-olds, probably due to bigger companies offering better prospects in larger conurbations (OECD 2008, 98;Spielhofer, Golden, and Evans 2011) and graduates staying in urban areas after university. Consequently, urban graduates report the highest mean and median earnings. Urban youth with no qualifications also earn far more than rural counterparts at the same level of attainment. However, one caveat worth noting is that median pay for urban respondents with A levels or equivalent qualifications is slightly lower than for rural youth in the same attainment category and also lower than pay reported by urban youth with GCSEs or equivalent, or no qualifications, although there is no clear explanation for this in the data.
The regression model below (Table 5) only includes those respondents who report being in full-time work (n0835), defined as 30 hours per week or more (following Eurostat 2005). For these, some have missing data for pay and others for gender, while a number of cases are also invalid for social capital factors, which have been produced using list-wise deletion. Thus, the final sample for this model (including in the regression only respondents with valid responses for all variables) is 680. Whilst this is a substantial decrease from the full youth sample of 2242, focus on those in full-time employment produces a more meaningful model as comparing the monthly pay of someone working 20 hours per week with another doing 40 hours heightens the risk of unreliable conclusions. This measurement difficulty is noted by Salverda and Mayhew (2010, 128), despite their insistence that excluding part-time workers 'is necessarily incomplete and misleading' (131). However, the specific concern here is with youth, so including only full-time employees is logical given that many of the  16Á24 BHPS sample remain in full-time education and have not settled into permanent jobs. Despite the importance of part-time work in helping young people to gain experience, build confidence and earn money (Spielhofer, Golden, and Evans 2011, 6Á7), this is more of an important factor in long-term development and is, therefore, somewhat peripheral to synchronic analyses of labour market outcomes. The sample is, therefore, restricted to those working 30 hours or more per week. Proportionately, there is no difference between rural and urban respondents here and the full under-25 sample (30% rural and 70% urban), suggesting that there is no location effect on working part-time or full-time. Fifty-one per cent are female, and unsurprisingly, more (70%) of the final sample are aged 20 or above, a reflection of higher numbers of the under 20s remaining in education and thus being predominantly (although not exclusively) restricted to part-time work. That said, 1.3% of youth working at least 30 hours per week also report being in full-time education. The upper limit of earnings reported has been capped at £2000 to reduce the influence of outliers on the model. In any case, this applies to only 1.5% of fulltime workers aged below 25 (this model still includes 4 outliers with standardised residuals at 93. The model without these outliers shows a slightly higher rural pay penalty Á see Appendix B1).
Model 1 explains 9.2% of variance and shows that rural youth working full-time can expect £90 less per month than urban counterparts (p B0.001). Whilst this may be considered a small effect, it further demonstrates that earnings are lower for young people in rural areas. None of the social class variables have a significant effect, except for the no-data category. This is accounted for by the large amount of missing data. Once the sample is reduced to those working full-time, the numbers with valid data for each class are too small for a significant effect to emerge. This model also includes the two social capital factors, of only 'norms' exerts a significant effect (B 080.13, pB0.001). This provides a clear link between the impression respondents have of their local area and their earnings. Also noteworthy is the absence of a significant effect for 'networks', suggesting that personal connections, while important in finding work according to previous studies, have little bearing on outcomes once in the labour market. Longitudinal analysis may produce different findings, but the effect of networks in this cross-sectional analysis reveals no significant effect. This further illustrates that social capital should not be treated as a monolithic concept and that its constituent elements can produce distinct effects.
Model 2 adds age and gender into the analysis. Unsurprisingly, 20-to 24-yearolds earn significantly more than those under 20, owing to obvious factors such as greater chance of finishing education and finding work, more time to gain promotion, and so on. Also female full-time workers earn less than males (B 0( 126.46, p B0.001). Again, 'norms' have a significant effect, while 'networks' do not. The rural pay penalty is reduced to £80 (net monthly earnings) once age and gender are controlled for. In particular, this reflects the gendered dimension to rural/urban pay disparities discussed above. Despite this, the overall rural/urban earnings gap remains evident, significant at the p B0.01 level.
Model 3 again displays significant effects for rural location and norms and also adds the highest academic qualification achieved. Holding a degree predicts an increase in the earnings of £118 per month (pB0.05), but other levels of educational attainment have no significant effect, and the model explains 25% of total variation, suggesting other variables are also important. Potential predictors of earnings, such as innate ability, the effort applied whilst at work and personal life constraints, beyond what is captured in the social capital indicators analysed are unobserved in the data-set. There is also an argument that qualifications should be considered outcomes rather than predictors. This requires a more longitudinal approach and should be addressed by future research. The third model also shows that 'parental class unknown' is not statistically significant, despite being significant at the pB0.01 level in model 1, and at pB0.05 in model 2. This suggests that while some pattern concerning parents' occupation affects earnings, this is not as powerful as the predictors which remain significant throughout (rural location, norms and trust, age, gender and holding a degree). It is also possible that class origin exerts greater effects on earnings for those aged 25 and above by which time career trajectories are more developed.
For now, it is clear that young rural workers receive lower net earnings than urban peers, earning £76 less per month even when controlling for age, gender, qualifications and social capital indicators. The wage penalty is a problem in itself but must be considered alongside recent evidence that rural living is more expensive, with those living in more remote areas required to earn more to afford the same standard of living as urban residents. Single working age people face the biggest relative discrepancy in rural/urban affordability, over £40 per week for those in the most isolated locations (Smith, Davis, and Hirsch 2010, 37). Rural youth, therefore, face the double disadvantage of lower wages and higher living costs. That they earn less is explained to some extent by the difference in occupational status of jobs held by young people according to location, as seen in Table 6.
The higher numbers of service class workers in urban areas are noteworthy, as is the disproportionately large proportion of rural youth employed in manual jobs. This finding is an interesting extension to the model outlined above, yet the shortcomings of the model demand one particular solution. As the earnings of 16-to 24-year-olds may not accurately reflect the success they enjoy in the labour market during later life, a longitudinal analysis of the effect of location on outcomes in employment is needed. This would make it possible to analyse more fully the effect of educational attainment, which made little difference in the model presented here.
There are further limitations of the data which should also be noted here. The high amount of missing data on parental occupation prevents firm conclusions around the social class. It is plausible that parental class can affect a young person's success in the labour market and their ability to cope with the challenges of living in a rural area. Additionally, while the analysis has produced clear findings based on different facets of social capital, the data do not permit investigation of how the bonding and bridging dimensions conceptualised by Putnam interact with rural/ urban location to determine labour market outcomes. Future research should pursue these lines of enquiry to build on the findings of this paper. Although rural youth are no likelier to be unemployed than urban peers, once in the labour market they face lower earnings alongside higher living costs, amounting to a double disadvantage. The impact of this on future prospects should be explored, but for now policy solutions such as helping young people in rural areas to find affordable housing and assisting with transport are more pressing priorities. For example, the Government's 2012 Rural Statement (Department for Environment, Food and Rural Affairs 2012) pledged that a 'national Wheels to Work coordinator' would be appointed, with the aim of extending access to cheap moped lease schemes such as those already operating in some rural locations. This is a welcome advance, but further support for both public transport and similar schemes promoting independent travel would help to make young workers and jobseekers more mobile, increasing their labour market potential. Given the lower wages and higher living costs faced by young people in rural areas, tax relief for transport or housing would benefit those in remote locations with low-paid jobs.

Conclusion
Perceptions of rural idyll and biases towards urban areas in terms of research may obscure the fact that rural youth can face difficulties different to those confronting urban counterparts, namely, limited opportunities and, as seen here, lower labour market returns. Rural areas now have broadly similar sectoral structures to urban regions, but lower pay and a higher proportion of manual occupations remain barriers for rural youth in terms of employment prospects. Overall, unemployment is generally higher in urban areas, where intergenerational deprivation is more deeply entrenched, but for young full-time workers urban pay is higher even when London is excluded. Other variables affecting earnings are age, qualifications and gender.
I postulated that social capital is an important determinant of labour market outcomes, informed by the literature surrounding the concept. The image of rural areas as harmonious and tranquil suggests that community is central in such locations, and there is evidence that accessing local networks can be crucial for rural jobseekers. The two key components of social capital, norms and networks, were, therefore, entered into the regression model of net pay for young people in full-time employment. I found that while norms, defined as trust in individuals and institutions along with perception of community safety and cohesion, exerted a significant effect on earnings, networks did not. Personal and professional contacts may improve the chances of finding work, as indicated by previous research. However, this facet of social capital does not appear to affect outcomes once someone is in employment. This reinforces the view that social capital is a protean concept, with one facet more important when seeking work, while the other affects outcomes once in a job.
The rural pay penalty must be considered alongside recent evidence that living costs are higher in rural areas, effectively doubling the disadvantage for rural youth. These are challenging times for all young people, but this paper has shown that location makes a difference regarding youth employment outcomes, with the rising cost and declining availability of transport posing major problems. Youth unemployment is lower in rural areas, but rural youth still face labour market disadvantages regarding earnings and living costs. Further research should consider how where one lives or grows up shapes the opportunities and obstacles to youth employment, with a view to informing policy which takes into account the effect of location. Table B1 shows that the rural effect on youth pay is increased slightly by removing outliers with standardised residuals 93, but the same variables remain significant, suggesting that the overall findings remain the same. Table B2 shows that the effects are very similar when London-based respondents are excluded from the analysis. Variable 1 (13.0) 2 (11.0) 3 (7.9) 4 (7.0) 5 (6.0) 6 (5.7) Anyone you can count on to offer comfort