STATISTICAL DISCRIMINATION AND EMPLOYERS' RECRUITMENT: Practices for low-skilled workers

Theories of employers' recruitment behaviour emphasise the widespread use of statistical discrimination techniques in order to select job applicants in a situation that is characterised by uncertainty and asymmetric information. Employers tend to use ‘signals’ to identify potentially problematic employees among prospective candidates.

In this paper, we are interested in the recruitment strategies of employers in the low-skilled segment of the labour market. In all developed countries, jobs with low formal skill requirements are on the decline, and have shifted from the agricultural and industrial sectors to parts of the service economy. We focus on low-skilled workers for three reasons. First, they are overrepresented among jobless people and constitute the bulk of the clientele included in various activation and labour market programmes that have mushroomed over the last two decades in Europe. A better understanding of the constraints and opportunities of interventions in this labour market segment may help improve their quality and effectiveness.

Second, in today's Europe, the bulk of actual and prospective workers in the low-skill sector tend to display some of the signals that have traditionally been associated with labour market disadvantages, such as lack of basic skills or command of the language, higher age, immigrant status, or experience of unemployment. Rejecting applicants with these characteristics, as implied by the traditional version of statistical discrimination theory, would dramatically reduce the pool from which employers can choose their employees. As a result, if statistical discrimination techniques are to be applied in this labour market segment, they need to be adapted.

Third, there are reasons to believe that statistical discrimination will play a particular role for this group of jobseekers. They have no certified occupational skills and, normally, no proof of formal qualifications is necessary for a successful application. Depending on the type of job, what matters are individual qualities such as motivation, social competencies, physical fitness, or outside appearance. In addition, given that low-skilled jobs can be performed without much training, employers may easily assess applicants’ productivity on the job through tests. The availability of this option, which would not be available for skilled positions that require extensive on-the-job training, suggests that in this segment of the labour market employers may prefer tests to statistical discrimination.

In this paper, we try to shed some light on employers' recruitment strategies concerning workers with low or no formal occupational qualifications. Qualitative evidence is based on 41 semi-structured interviews with employers largely relying on low-skilled workers. The interviews were carried out in 2009 in six European countries (see section 3 for more details). The results are presented in sections 4–6. First, we deal with the impact of the signals that are known to play a role in the recruitment process: higher age, immigrant status and being unemployed (section 4). Then, on the basis of the empirical evidence collected, we identify a number of other signals that our sample of employers use in order to detect motivation, a key quality in this segment of the labour market (section 5). Finally, we turn to alternatives to statistical discrimination, essentially testing and social networks (section 6).

Statistical discrimination as a tool to screen job applicants in the recruitment process has been identified by neoclassical economists in the 1970s (e.g., Phelps 1972; Arrow 1973; Spence 1973), though theoretically modelled somewhat earlier (Becker 1971, first 1957). Every hiring entails fixed costs for the enterprise. Among them are the costs of screening applicants, providing some training or giving time to familiarise with the job before the newly recruited worker becomes fully involved in production. Hiring is thus a risky process for employers due to information asymmetry (Akerlof 1970): The worker may quit before the investment pays off, and whether he or she performs according to the expected productivity will be known to the employer only some time after appointment. Moreover, uncertainty on the part of the employer is permanent due to the indeterminacy of the labour contract: a certain quantity and quality of work to be performed by the employee cannot be completely defined in advance. Rather, to some extent, it remains within the discretion of the workers to determine their output – whether they ‘cooperate’ or somehow show less commitment (‘shirking’) and withhold full effort (Offe and Hinrichs 1985).

In order to minimise risks and costs, employers may decide to avoid applicants belonging to groups of workers who, because of certain average characteristics, are believed to deliver less than the expected job performance or are more likely to create other problems in the firm. Typically, employers may decide to exclude applicants with immigrant background, older workers, long-term unemployed people, women, or persons who combine any of these attributes. Employers' assessment can be based on various sources, such as previous experience but also societal prejudice. Economists tend to regard statistical discrimination as an efficient tool for screening applicants, at least to the extent that employers' assessment of group characteristics is roughly accurate. At virtually no cost, statistical discrimination allows employers to limit the number applicants that need to be examined more closely, thus, reducing the cost of the recruitment process.

Several empirical studies have shown that statistical discrimination techniques are used by employers in many different settings. Most studies are concerned with wage differences that cannot be explained by observable characteristics and systematically put at a disadvantage a given group, e.g., women or immigrants. Statistical discrimination is also used to explain the obstacles that women encounter in reaching top level jobs in organisations.

Studies focusing most specifically on selection procedures have tended to use a technique known as ‘paired resume (or CV) audit’. In this type of studies, pairs of fictional CVs are submitted to employers. The paired CVs are identical and differ only in relation to the feature that is supposed to be used for discriminating against a given group, for instance females, immigrants or older workers. By comparing response rates, it is possible to estimate the impact of the discrimination factor on the likelihood of receiving a positive reply (e.g., an invitation to an interview). In general, as will be seen below, studies of this type find ample evidence of statistical discrimination in the way employers select candidates.

Some studies, like ours, have focused on interviews with employers, who are asked about the criteria they use in screening applicants. This approach has the advantage of providing detailed and in depth information on employers motivation and strategies. It may, however, underestimate the extent to which statistical discrimination techniques are actually used. This unsurprising result is confirmed at least in one US study which combined the two methodologies (paired CVs and interviews with employers). Six months after a paired CVs audit, the same employers were approached again and asked about their use of statistical discrimination as a selection tool. The extent of discrimination emerging from employers' interviews was considerably lower than what came out of the paired CVs part of the study (Pager and Quillian 2005). We will keep this result in mind when interpreting the responses in our interviews.

Only a few studies are concerned with the population of interest to us, namely, low-skilled workers. American studies of recruitment processes for ‘entry-level jobs’ have found a clear effect of discrimination against African Americans and, in particular, males (Neckerman and Kirschenman 1991; Moss and Tilly 2001; Pager and Quillian 2005; Zamudio and Lichter 2008). Other studies have identified further features associated with a reduced chance of positive reply to an application. These include former offenders (Pager and Quillian 2005) and older women (Lahey 2008).

For our purposes, these findings have to be reconsidered in the context of the low-skilled labour market segment in Europe. Accordingly, we assume statistical discrimination techniques to play an important role, albeit in particular ways.

First, we can expect people who do not display any of the typical negative signals to avoid many of the low-skilled jobs that are available in Europe's labour markets, particularly, when working conditions are harsh, wages low, working hours unfavourable, the job grants low prestige, or any combinations of these characteristics apply. The necessity to match low-skilled low-wage jobs and low-ambition applicants opens up (re-)employment chances for workers showing typically negative signals. As the owner of a German industrial laundry service put it: ‘If someone doesn't have labour market problems, he or she doesn't want to work in a laundry’ (GE3 LAUNDRY). Thus, some employers can simply not make blanket decisions leading to the avoidance of groups displaying certain negative signals. We therefore expect that firms employing low-skilled personnel will tend to resort to more sophisticated types of statistical discrimination than those mentioned in the literature.

Second, low-skilled jobs, by definition, do not require elaborated occupational skills. What matters instead are ‘soft skills’. This is an ill-defined concept which involves a fair degree of subjectivity. However, it is possible to get a better grip on the concept by differentiating two dimensions that are usually implied in the notion of soft skills. The first dimension consists of interpersonal skills. These include abilities to get on with clients, co-workers or supervisors. They are more important for jobs in the service sector although not irrelevant in the industrial one. The second dimension refers to motivation, or commitment, a positive work attitude, and so-forth (Moss and Tilly 2001). Employers will try to ascertain whether features that are connected with these skills can be attributed to the individual applicant. For example, as we will see below, an unsolicited application may be taken as a sign of positive motivation whereas a referral by the public employment service may emit a negative signal (see also Hieming et al. 2005: 122).

Third, statistical discrimination may produce the biggest gains for employers when recruitment concerns positions where the worker's true productivity can only be observed after a relatively long period on the job – for example, because extensive on-the-job or formal training is needed. Under such circumstances, there is no simple way to ‘test’ candidates, and employers may decide to use statistical discrimination as a selection tool. However, most low-skilled jobs we addressed in our interviews did not require lengthy training periods, as a result of which the alternative of testing is not considerably more costly than applying statistical discrimination techniques although its applicability may be dependent on the rigidity of the labour market (foremost employment protection legislation).

In fact, there are good reasons to assume that testing will be more efficient than statistical discrimination. By making blanket decisions which exclude entire groups of applicants from entering the recruitment process regardless of individual worker's abilities, employers will miss some productive workers who just happen to belong to the wrong group. Testing may be beneficial to the employment chances of workers belonging to groups who are discriminated against. A study of hiring practices in the Chicago area, showed that employers who used tests as a screening device had a slightly higher proportion of blacks in their workforce, after controlling for numerous other factors (Neckerman and Kirschenman 1991).

Our objective in this study was to investigate issues pertaining to the specificities of the low-skill segment of the labour market. In view of scant research focusing on this specific topic, we decided to use qualitative face-to-face interviews rather than a paired-CV audit methodology. The latter methodology, in fact, requires a fairly precise idea of the factors that are likely to be used to discriminate against certain candidates, something we did not have prior to this study. Due to the focus of our study on the low-skill sector of the labour market, we elected to cover employers in different countries, because it allowed us to highlight sector-specific characteristics rather than a country's distinctive features. Countries were selected so as to maximise variation on a range of relevant dimensions: type of welfare state, system of labour market regime, and most importantly, level of protection against dismissal. This research design implies that we focus on similarities across countries, which are more likely to be sector specific, rather than on differences.

We interviewed 41 employers¹ based in six different European countries, three of them with rather strict employment protection (Germany, Italy and Slovenia) and three with a lower degree of job protection (Denmark, Hungary and Switzerland). Within these countries, employers were selected among those who rely heavily on low-skilled labour, defined as having completed less than 12 months vocational training. For employers with a more mixed labour force, we focused our questions on this labour market segment. The sampling of employers in each country was carried out by the national team of researchers (see acknowledgment for names) on the basis of convenience. In other words, we selected companies that we thought would be easily accessible, for example because of their geographical location. The national teams were instructed to include in each country sample both industrial and service firms as well as companies of different size. As shown in Table 1, national samples do fulfil this requirement. Interviews were conducted during the first semester of 2009 on the basis of semi-structured questionnaires (available from the authors). Within companies, we interviewed the managing director, the owner or, for larger companies, a manager from the human resources department. Interviews lasted about 60 to 90 minutes. They were recorded, transcribed in full and translated into English.

The first objective of the interviews was to check whether employers relied on the typical traits that are known to be used as statistical discrimination devices: age, immigrant status and (long-term) unemployment. Regularly, also ‘gender’ figures as a criterion of statistical discrimination. In the low-skilled segment of the labour market we looked at, however, most jobs tend to be exclusively or predominantly ‘male’ (metalworking, meat processing, construction, etc.) or ‘female’ (supermarket, cleaning, clothing outlet, etc.) occupations and no real competition between applicants of different sex arises. Sometimes gender is mentioned as signal, but in conjunction with age. Younger women are regarded as more ‘risky’ because in case of maternity they may need to be replaced.

The application of statistical discrimination as a screening device is considered as an important causal factor for specific labour market problems of workers who exhibit those traits: they have an above-average risk of becoming or remaining unemployed, are at risk of leaving the labour market altogether (‘discouraged workers’, enforced early exit); they have to make more concessions than others in order to be hired (with regard to wages, working conditions – like shift work, dead-end jobs –, or regional mobility) and have fewer chances to utilise their once acquired formal skills.

Interviewees were asked whether (a) being unemployed and older than 50, (b) being unemployed and of immigrant origin, and (c) being long-term unemployed (for over 12 months) were considered to be signals of a potentially problematic candidate. Unsurprisingly, most employers (between 24 and 30 out of 41) denied using these features for screening candidates. We are aware that this may simply be a result of the desire of some interviewees to appear politically correct. Our interview data are nonetheless trustworthy. First, we were interested in finding out the motives given by those who admitted to rely on statistical discrimination of the type we suggested. Second, employers made reference to a range of other signals that they use to identify those they regard as ‘good’ candidates. Unlike direct questions about discrimination, we do not expect the responses to these questions to be strongly biased by the wish to appear politically correct.

Although somewhat varying between EU member states, labour market statistics consistently show higher unemployment rates of workers in the upper age brackets than for those of prime age. Moreover, they face a higher risk of becoming long-term unemployed and, partly as a consequence, people with low formal education (involuntarily) exit the labour market much earlier than their counterparts with higher qualifications (OECD 2006: 35–45). The statements of our interviewees, however, largely contradict well-known attitudes of employers towards elderly applicants leading to relatively worse (re-)employment chances, namely, outdated skills, impaired physical condition, not adapting and learning as well and as fast as younger workers (cf. Brauer et al. 2009; OECD 2006: 63–5, 103–10).

Across countries, older workers were generally considered as more reliable and showing a stronger work ethic than their younger colleagues, and those firms which employ workers above age 50 indeed value them as an ‘asset’ and HR managers develop a more positive image of ‘age’.² There were few mentions (in 10 out of 41 interviews) of potential risks or problems when hiring older workers. These concerned practical obstacles (hard physical work) but also behavioural ones like those mentioned before. Image problems for companies wishing to appeal to young people were also raised. The director of a chain of cafés in Switzerland summed up the problem for his enterprise:

Quite the opposite position is held by the manager of a small supermarket, mainly employing women. He even prefers a workforce above age 40. ‘Older people have life experience. They know how to deal with other people. That's very important for us in the service sector’ (GE5 SUPERSMALL).

Less strict employment protection can be another reason for not discriminating against older applicants, but rather, to prefer them. In Germany, older unemployed people, if hired, are less well protected against dismissal than prime-aged worker. This, according to the manager of a cleaning company, may turn out as an advantage for their reemployment chances:

All in all, it seems that (higher) age only plays a secondary role in recruitment and is not per se an excluding attribute although it remains unclear to what extent the employers we interviewed actually do hire older unemployed. Indeed, most of the positive comments referred to older workers who had grown old ‘on the’ job rather than older unemployed applicants.

Immigrants are rarely mentioned as a potentially problematic group, associated with higher risks. Beside political correctness, this result may also be due to the strong reliance of the low-skill labour market segment on immigrant workers in most EU countries. In several Danish, Swiss and German companies the interviewees pointed out that they employ immigrant workers to a large extent, often over 50 percent of the workforce. As a result, it would be unreasonable and counterproductive to apply immigrant status as an exclusion criterion in the selection process. However, basic command of the language is often considered an essential requirement to ensure safety at work and being able to follow the instructions of supervisors.

Contrary to common assumptions, in some cases, immigrant status was in fact mentioned as a positive signal. According to the manager of a Danish cleaning company:

A different reason for explicitly hiring immigrants is presented by the manager of a large supermarket in Germany, located in an area with a high share of Turkish migrants. He explained that he is developing an ‘Ethno-Marketing’ with products especially for Muslims: ‘For this, I even needemployees with a migration background who can pick out this assortment. I can't judge that at all by myself’ (GE4 SUPERLARGE). Moreover, he stressed the advantage in customer interaction: ‘So employees with a migration background help us to bridge problems of understanding and for example, to better understand purchasing habits.’

In some cases, immigrant status is probably used as a proxy for tractability (uncritical acceptance of management's decisions). A US study pointed out that preference is given to Latinos over African Americans in parts of the low-skill sector, precisely because belonging to the former group is considered to be a signal of tractability (Zamudio and Lichter 2008). We found similar results with regard to some groups of immigrants, such as Asians in Denmark. The manager of a meat processing plant in Denmark found that: ‘if we need people to work overtime, they [immigrants of Asian origin] always say yes, and we sometimes get a few remarks about that, but we don't get any problems as a result of it’ (DK5 MEAT).

Finally, belonging to a certain group of immigrants may be an advantage because it is easier to integrate them in an already existing team. The manager of a German food processing firm emphasizes the merits of assembling a homogeneous group:

Taken together, being a job applicant with immigrant background does not necessarily imply a negative signal to potential employers. Dependent on the actual reliance of employers on this reservoir of workers or on immigrants as customers or as co-workers they may experience no disadvantages or are even be preferred over nationals. This is especially true for certain groups of immigrants whose ethnic background is seen as a signal of tractability.

It is generally assumed that the longer an unemployment spell lasts, the more difficult it will be for the jobless person to find a new job. Unemployment status is thus a self-enhancing negative signal. Although a large majority of interviewed employers (30) denied using long-term unemployed status as a selection criterion, we find more consistency in the reasons given by those interviewees who explicitly acknowledged their reluctance to hire from long-term unemployed applicants because of their assumed lack of motivation or other behavioural defects. The director of a Swiss café chain stated: ‘… if someone has not found anything for a long period of time there has to be a problem in addition to job loss – most frequently lack of motivation’ (CH6 CAFÉ). And the human resources manager of a Hungarian industrial company ponders about long-term unemployed applicants:

Not only Hungarian managers are prepared to hire long-term unemployed if the applicant can produce an acceptable justification for failing to find a job earlier. However, other interviewees are not very concrete about what would be a ‘good reason’. Interestingly, in a period of rising unemployment the stigma carried by (long-term) unemployed may decline. According to one Swiss employer:

Our findings confirm the commonly held assumption that being out of work for a longer period increasingly impedes re-employment, unless the unemployed person is able to bring forward a convincing justification for not finding a job earlier, or the state of the labour market is such that there is a substantial excess of labour supply.

We hypothesised that in the low-skill segment of the labour market, soft skills and in particular motivation is likely to be the key feature employers are going to look for when examining candidates. Of course, motivation is likely to be important in every job, but it may be even more crucial in the low-skill segment. First, these jobs tend to be badly paid, offer few opportunities for advancement and might have a negative image. Second, while for most other job openings professional qualifications will count as much as motivation, here, given the simplicity of the tasks involved, differences in ability are unlikely to become a discriminating factor in the hiring process. In fact, the employers we spoke to have developed more sophisticated strategies to spot motivation.

It came out in a number of interviews and across countries, that the very fact of being sent by the public employment service (PES) is considered to be a negative signal. Employers tended to regard candidates coming from the PES as persons who apply in order to comply with the job search requirements imposed by unemployment benefit rules. Not taking the initiative themselves to get back to a job qualifies them as less motivated candidates. According to the manager of a Danish meat-processing firm:

Quite similar is the assessment of a Hungarian manager:

Along the same lines, the manager of a Slovenian supermarket clearly states:

The owner of a German laundry also takes a reserved stance towards registered unemployed:

However, she cannot be very ‘choosy’:

The negative attitude towards the unemployed who apply at the instigation of the PES and the actual experience that they ‘were not so reliable’ (GE2 MEAT) is also reflected in Figure 1. The recruitment channels an employer uses – regularly more than one – are determined by past experiences with the various channels, the (sectoral) labour market situation and the type of labour looked for. Different channels mobilise varying numbers of applicants and certain categories of jobseekers (Behrenz 2001: 263). Recruitments for low-skilled jobs via the PES channel rank only fourth, far behind direct applications and the informal route of utilising recommendations from already employed workers.

Figure 1 shows that the preferred recruitment channel by the employers we spoke to is unsolicited applications (i.e., applications that are not made in response to a vacancy announcement). Employers are keen to obtain signals which demonstrate motivation and social skills. The fact of making the effort to apply is taken as a signal of motivation. Several employers suggested this line of reasoning.

A German employer is quite convinced that the commitment a person puts in the job search process is a good indicator of his or her subsequent job performance:

Likewise, a Danish cleaning company prefers applicants who come forward unsolicited:

Direct or unsolicited applications are widely perceived as a signal of ‘true’ interest in getting a job and of motivation. However, it has been pointed out that the multiplication of such unsolicited applications as a result of public authorities’ intensified activation efforts may reduce the usefulness of such activity as a positive signal (cf. Hieming et al. 2005: 220) because the employers can hardly distinguish between ‘genuine’ and ‘fake’ applications, foremost sent to comply with the rules of benefit receipt.

Participation in a labour market programme may also demonstrate the motivation to find a job. A majority of employers (29 out of 40) considers this as a positive signal, though sometimes without much enthusiasm, just believing that it is better than inactivity. A German study dealing with recruitment for low-skilled jobs in the service sector found that employers prefer applicants with certified formal training (although not needed for the job) because it signals traits such as staying power, punctuality and reliability as well as the transferability of certain pertinent skills and social competencies in particular (Hieming et al. 2005: 225).

At first sight, statistical discrimination may appear as an efficient technique to single out among candidates. The tool, however, is likely to lack precision, as blanket judgements will exclude highly productive individuals who just happen to display the ‘wrong’ signals. For this reason, we may expect employers to be interested in alternatives to statistical discrimination when selecting candidates.

We assume two alternative strategies to play a particularly important role: testing and relying on networks. As hypothesised above, testing can be particularly attractive in this labour market segment, because the productivity of prospective employees can be assessed rapidly, without providing extensive initial training, as may instead be the case in higher skilled jobs. Alternatively, networks are known to be extremely important in job matching, and constitute a source of reliable information on applicants’ productivity (Granovetter 1995). The employers we interviewed like both techniques. Finally, employers have also a third alternative to statistical discrimination, consisting simply in outsourcing the selection process to a private recruitment company. This alternative, however, comes at considerable costs.

The information provided in Figure 1 shows that the employers in our sample tend to prefer private employment agencies to the PES. These agencies are used in two different ways. First, in some cases, the complete selection process is simply outsourced to a private employment company, which knows well the employer's needs and, accordingly, proposes suitable candidates. Second, many employers use private employment firms to recruit temporary workers. The use of ‘temp agencies’ provides employers the opportunity to ‘test’ several workers. When a vacancy arises, they can recruit the worker from a pool of temporary workers they already know well. However, workers who do not fulfil expectations can very easily be refused and sent back to the agency.

Unsurprisingly, employers having utilised this recruitment channel look upon it positively:

A similar positive statement on the ‘try and hire’ strategy is given by the HR manager of a German food-processing company which, due to rationalisation of production, has not recruited new workers for a while:

While this recruitment channel reduces the risks for the hiring company, it is questionable whether the low-skilled workers displaying negative signals benefit to the same extent. Their problem in finding a job is shifted from employers to the temp agency that decides about the engagement in the first place. The agency, most likely, applies similar selection criteria in order to reduce the risk that the worker it leases is refused due to poor performance and has to be replaced. If this happens too often, the employer may quit the cooperation with the agency. Therefore, the group of job seekers we focus on have to go through the needle eye of the recruitment process in any case.

In many countries employers, willing to hire hard-to-employ jobless people, are offered wage subsides, limited to a certain time. From the point of view of employers, these programmes provide an opportunity to test a (long-term) unemployed person for an extended period (typically, a few months) at a reduced wage costs.

Overall, the employers we interviewed appreciate those subsidies, although with some reservations. A subsidised internship and/or a ‘trial period’ is seen as an incentive to give people with labour market problems a better chance in the recruitment process as they reduce employers' costs and risks. For example, it can improve the job prospects of older unemployed applicants: ‘We have also hired employees over the age of 50. There's public support for that’ (GE2 MEAT).

During such a ‘trial’ or ‘probation period’ employers have the opportunity to observe whether the former unemployed performs well on the job and proves his/her reliability. In the end, initially emitted negative signals may not play a significant role anymore if the employee shows positive properties (like punctuality, politeness, productivity, reliability). According to an Italian manager:

Some employers, however, are less enthusiastic about financial support. They wonder why someone is eligible for a subsidy and fear that this may indeed represent another negative signal. In addition, to make use of this instrument, employers must cooperate with the public employment services, and many are reluctant to do so.

Other employers have expressed the view that productive workers do not need subsidies, and since employers are after productive workers it makes no sense to rely on subsidies: ‘Either a person is ok and works ok, and I hire him, or he is not ok, and I don't hire him’ (CH7 BUILD). Danish and German managers express a similar attitude:

Taken together, our evidence on the role of subsidised trial periods is mixed. Future research should identify under what conditions this instrument can be most effective in opening access to employment for hard-to-employ jobless people.

As shown in Figure 1 above, numerous employers tend to use informal contacts as a recruitment channel, either utilising recommendations from their current employees (29 out of 38 employers) or other contacts in the sector (11 out of 38). This finding is consistent with previous research on the important role played by social networks for attaining a job (e.g., Granovetter 1995; Korpi 2001; Brandt 2006; Larsen 2008).

There are several reasons why many employers like this way of identifying potential candidates. Applications recommended by already employed workers reduce firms’ search and selection costs and show a high probability to deliver the desired results if the ‘sponsor’ provides trustworthy information to both the employer and the applicant. Hiring a person who is acquainted with an already employed worker (a friend, relative, neighbour) is likely to result in better job performance and work discipline as well as a lower fluctuation rate. The new employee, in fact, is probably better informed about job demands and practices, shop morale and so-forth. No less important, the employee who recommends an applicant for a position, to some extent, also takes responsibility for his/her aptitude, motivation and smooth integration into the company. Virtually, the employee vouches for the person who is recommended. His or her own reputation may be at stake in the event of discontent with the newly hired worker, which may result in some pressure being exerted on the newcomer so as to ensure good performance (Windolf and Hohn 1984: 139–47; Hieming et al. 2005: 121, 220–1). For that reason also applicants displaying negative signals and who would not be selected otherwise may get a chance.

These deliberations clearly come out in three statements which all relate to trust, either in the employee who recommends someone else or the person who is recommended:

Other employers are less explicit with regard to the precise motivation beside using networks, but clearly as much satisfied: ‘When we are looking for staff, we find someone relatively fast through word of mouth. We've had the best experiences this way’ (GE1 CLEAN).

A large quantitative study carried out in Germany (Rebien 2010) confirms that social networks are a key recruitment channel particularly for small firms, when jobs with low (or very high) skill demands have to be filled and when working conditions are rather unfavourable (shift work, noise). In our interviews we also found that employees with a migration background often learn about vacancies from already employed relatives or friends who bring them along or encourage them to apply for the job. If they are appointed then the ‘sponsor’ assists them in getting used to the new work environment (GE2 MEAT).

This extension of an internal labour market (Manwaring 1984) is beneficial for hard-to-employ people who otherwise would be hardly taken into consideration by an employer. To learn about a vacancy this way and to have an intercessor supporting an application, however, is a matter of social capital that gives the applicant a comparative advantage to obtain a job. However, unemployed people regularly dispose of less social capital, and with increasing length of unemployment the (shrinking) social network counts fewer and fewer people holding a job themselves and, thus, are able to inform a friend, relative, neighbour about a vacancy (Korpi 2001; Lindsay 2010). Moreover, when firms can rely on this informal channel to fill a vacancy for low-skilled jobs, they will not report the vacancies to the public employment service or use other formal channels (announcements in newspapers or in the internet), among others, to limit the number of applications they have to deal with. Then, the transparency of the labour market declines further, and unemployed people have less chances to learn about vacancies as such. For these two reasons only a limited number of hard-to-employ jobseekers may actually benefit from social networks as a device for finding a job.

The evidence presented above shows that employers in the low-skill segment of the labour market do use statistical discrimination techniques. However, as suggested in our first hypothesis, the signals that are used most frequently differ from those that are typically found in the literature. Immigrant status, older age and long-term unemployment do not seem to systematically disadvantage applicants showing these characteristics. Due to the often unpleasant working conditions, in the low-skill segment, employers cannot be too ‘choosy’ with regard to their selection criteria. Very often, those signals which are generally assumed to be the cause of discrimination turn out to be assets of the applicants. Being older or having an immigrant status can be an advantage when it comes to certain jobs. For example, older applicants may be preferred because of their assumed or experienced reliability, or certain immigrant groups are regarded as more tractable, provide an advantage for the employer due to their acquaintance with customers, or are easier to integrate in an already existing team of immigrant workers. Long-term unemployment, however, clearly remains a negative signal.

Second, we hypothesised that in this segment of the labour market, formal qualifications count less, but soft skills (social competencies, willingness to perform) are more important. Therefore, employers are keen to spot among applicants, who almost all exhibit certain negative signals, those who are most promising to possess the desired soft skills. The issue of motivation was indeed often highlighted by employers. Interestingly, we found that the channel through which the applicant establishes the first contact constitutes a positive or negative signal. Unsolicited applications are taken as an indicator for strong motivation, and recommendations from already employed workers are seen as a sort of guarantee for a low-risk candidate. In contrast, in all six countries the fact of being referred by the public employment service is considered as a signal of weak motivation. More in general, PES have a poor reputation and, reinforced by negative experience, are rarely used.

According to the third hypothesis, ‘testing’ is a more effective strategy to spot the least risky applicants than relying on statistical discrimination. In this segment of the labour market ‘testing’ is more feasible than for high-skilled jobs. The recourse to private employment (temporary work) agencies is indeed a widely used attempt to gain information on applicants’ future job performance. Hiring the ‘good’ ones after they showed their qualities in the work process (and ‘returning’ the others) is a reliable strategy also because these workers have passed through a ‘double selection process.’ The ‘try and hire’ strategy is also possible without the intermediation of a private employment agency. A German study (Hieming et al. 2005) reports that, wherever possible, firms test low-skilled applicants from a few hours up to 2 days in order to ascertain their competencies for the concrete job (power of concentration, exact handling of tasks, working speed, sleight of hand, etc.).

Our research design implied that we looked for results that are similar across countries rather than differences. It is nonetheless striking that we found very little cross-country variation in the statements made. Differences, if any, are mainly due to the size of the company, the sector they are operating in, or working conditions (harsh or not). Little variation also means that the market for low-skilled labour is a special one, very different from recruitments for (highly) qualified workers. The absence of cross-national variation suggests that recruitment practices in the low-skill segment are mostly governed by incentives and less by institutions. For example, it was a surprising result that large cross-national differences in terms of employment protection legislation do not seem to impact on recruitment practices. Everywhere, employers dislike dismissals – meaning ‘problems’ of various kind – and are eager to use all available alternatives that minimise the risk of hiring the ‘wrong person’, regardless of how difficult it is to lay him or her off later.

These findings have policy relevance. In particular, the fact that employers tend to regard a referral by an employment office as a negative signal may be a consequence of the ‘activation’ turn adopted in most of the countries we covered. PES put increasing pressure on jobless people to actively engage in job search. As a result, many applications are motivated by the wish to avoid a sanction and not by the desire to work in a given workplace. Rather paradoxically, compulsory job search reinforces the negative image employers have of the PES, and makes the placement of jobless people more difficult.

The research on which this paper is based has been carried out in the context of the network of excellence RECWOWE (Reconciling work and welfare in Europe), funded under the EU framework 6 programme. We would like to thank project participants Christian Albrekt Larsen, Michel Berclaz, Nevenka Černigoj Sadar, Miroljub Ignjatović, Vera Messing, Jacob J. Pedersen, Katalin Tardos, Patrik Vesan, Valeria Sparano, A. Caroline Warfelmann, and Sabine Wichmann for making their interview data available and for comments on previous versions of this article.

Giuliano Bonoli is Professor of social policy at the Swiss Graduate School of Public Administration (IDHEAP), in Lausanne. He has published extensively on the process of welfare state transformation, particularly on old age pensions and active labour market policy. His most recent book ‘The origins of active social policy’, is forthcoming with Oxford University Press.

Karl Hinrichs is Senior Research Associate at Bremen University's Centre for Social Policy Research (since 1990) and Professor of Political Science at Humboldt University in Berlin. His main research focus is on comparative welfare state analysis, the development of social policy in Germany, and the study of old-age security policies and politics in ageing societies. In 2012 he co-edited (together with Matteo Jessoula) Labour Market Flexibility and Pension Reforms. Flexible Today, Secure Tomorrow? (Basingstoke: Palgrave Macmillan).