Beyond Networks: ‘Social Cohesion’ and Unemployment Exit Rates*
Carmel Hannan Institute for Labour Research University of Essex
Contact address: Carmel Hannan, Institute for Social and Economic Research,
(incorporating the ESRC Research Centre on Micro-social Change), University of Essex, Wivenhoe Park, Colchester C04 3SQ. UK. E-mail: [email protected]
, Fax: 01206 873151, Phone: 01206 872588 Acknowledgements
The support of the Leverhulme Trust, the Economic and Social Research Council and the University of Essex is gratefully acknowledged. The author would especially like to thank Marco francesconi for his numerous helpful comments. An earlier version of this paper was presented at the European Social Fund conference “European Societies or European Society? Inequality and Social Exclusion in Europe: The Role of Family and Social Networks”, Castelvecchio di Pascoli, Italy, 3-7 April 1998 and the Work, Employment and Society conference, Cambridge, UK, 14-16 September 1998. All comments appreciated.
Beyond Networks: ‘Social Cohesion’ and Unemployment Exit Rates
This paper provides convincing new evidence on the role of social resource patterns in shaping an individual’s chances of entry to the labour market. It links movements out of unemployment into employment to constructed indicators of ‘social cohesion’. These are social participation, social support and the social network. It was found that the current duration in a state has an influence on the probability of exit from that state. However, even after controlling for this and many other demographic and economic factors, the social network measure remained a significance influence on whether the unemployed found a job. Respondents who have close employed friends are significantly more likely than those who do not to exit unemployment. Why is this the case? Previous research has shown that the more socially integrated individuals have greater access to useful job information flows. In addition, this study has found that the unemployed who have close employed friends are significantly less likely to suffer psychological distress. In this sense, policies which isolate the unemployed into ghettos (for example, council housing schemes) do much harm and may play a large role in keeping the unemployed, unemployed.
Flagged by C. Jenck (1972) as dealing with luck and later by Granovetter (1973) as having the right contact in the right place at the right time, unemployment exits rates have been explained from a number of different standpoints. Economic explanation is built around notions of state and duration dependencies or the effects of previous spells of unemployment on the probability of exit from that state. Sociology relies upon notions of social exclusion and social networking or linkages between actors as channels for the transfer of resources. Using data from the first six waves of British Household Panel Study (BHPS), a nationally representative random sample, it is possible to construct measures for both these type of explanation. It is therefore possible to examine the role social resource patterns in shaping an individual’s chances of entry to the labour market after controlling for the traditional, and relatively easy to measure, economic factors. Three limited measures of social cohesion were constructed; a social participation scale, a social support index, and most importantly, a social network measure.
Of course, it was found that the current duration in a state has an influence on the probability of exit from that state. However, even after controlling for this and many other demographic and economic factors, the social network measure remained a significance influence on whether the unemployed find a job. Respondents who have close employed friends are significantly more likely than those who do not to exit unemployment. It was also found that these people display less psychological distress than those whose close friends were all unemployed. Future work will
address the linkages between psychological health and social integration. For the moment, it seems clear that social resource patterns play a crucial role in the determination of who finds a job. In particular, the segregation of the unemployed population in any way renders the maintenance and development of links (formal and informal) with the working world impossible, thereby worsening the psychological strain placed on these people and statistically rendering them much less likely to find a job
Beyond Networks: ‘Social Cohesion’ and Unemployment Exit Rates.
For many years the analysis of unemployment was solely the domain of Economics. However, in recent decades it has become apparent that who exits unemployment is fundamentally determined by a social process. Over twenty years ago Mark Granovetter (1974) found that over 60 percent of the professional, technical and managerial workers he interviewed reported obtaining their jobs through personal contacts. A recent British Department of Social Security report (1997) reported that around 38 percent of job seekers had contacted friends and family as a means of job search, and numerous US studies have estimated that over 50 percent of the unemployed who found jobs did so through ‘word of mouth’ (see Montgomery 1992 for a summary). Yet, large scale quantitative analysis of unemployment exit rates rarely even alludes to this process. It seems that ultimately who finds a job has much to do with ones social contacts. Who people know, how they know them and the consequences of differing social relationships on peoples’ lives are important research questions.
The importance of this ‘social’ being has recently received increased attention due to the popularity of a rather confused debate on the collapse of community or the decline of social cohesion in modern society. This concept of ‘social cohesion’ is mostly misleading and ambiguous. It is assumed to be some natural ‘good’ yet strong cohesion can be disabling and counter productive. Take, for example, a highly
organised cohesive criminal gang. However, this debate has focused attention on the differing ways in which individuals are linked to each other and to society. Social network analysis has chosen to plot such linkages in a mathematical way, as some sort of aid for getting things done, and the debate on social cohesion seems to view them as some sort of glue to provide order and meaning to social life (Pahl 1996). The crux of the issue rests on ones view of social relationships. An increasingly popular view is to argue that through such relationships people build up a stock of ‘social capital’. This theory presents numerous reasons for why social contacts are important. What follows is a critical review of the implications of adopting such an approach.
It is argued that this sociological theory differs from its economic counterpart in that the ability to obtain social capital does not inhere in the individual, as the possession of money (material capital) or education (human capital) does. It is instead a product of the individual’s set of relationships with others; a product of ‘embeddedness’2. Such a point has been invaluable in highlighting that individuals are not isolated beings. However, research has focused on what Granovetter (1985) called ‘structural embeddedness’, that is, the ways in which an actor’s mutual contacts are connected to one another. This, Granovetter argues, is the most important type of structure in which economic transactions are embedded. This is because it easily allows for the transfer of resources (instrumental aid) within groups. Therefore, research on networks has been limited to firms, interorganisational ties and the environment of organisations because the linkages between actors can be easily plotted. Of course,
concern is often expressed over the generalisability of findings. Yet, people are connected to and influenced by each other even when these connections are scattered diffusely through out the population. The intention of this paper is to move social network concerns beyond instrumental aid and hence organisational ties and, in this way, test the wider relevance of such concerns.
It seems logical to suppose that we are influenced by our social contacts in a variety of ways but the concept of social capital itself is misleading and, at best, vague. “It is a potentially powerful concept which is so useful that it is given many different meanings by many different people who use it in many different ways to explain many different things” (Newton 1996: 1). Let me explain this further. The definition and measurement of social capital remain unresolved in the literature. The first thing to note is how this concept echoes the economic version of human capital. The emphasis being on capital as an investment one makes expecting a future return. Yet again highlighting the instrumental aspect of human activity. For some social capital is indeed simply a broader version of human capital but one which includes a social and cultural dimension and highlights the importance of informal learning (see Morrow 1998). For Wall, Ferrazzi and Schryer (1998) the confusion surrounding this term may be clarified into three separate approaches. Firstly, Colemans (1992) view with an emphasis on the positive outcomes of social capital transformation. The approach adopted here is probably closest to this view as the focus is on differing (both positive and negative) outcomes of, what I prefer to call, social resource patterns.
Secondly, Bourdieu’s view (1990) of human activity as primarily aimed at controlling and accumulating different kinds of capital, with economic capital as the prime form. This approach is commendable in that his definition of social capital is part of a more complex and intricate typology rooting social capital in the practises of everyday life (the habitus). Yet, in practise, this concept of social capital is vague and immeasurable. Morrow (1998) on this point argues that when researching such a view, it seems best not to conceptualise social capital as a measurable ‘thing’, but rather as a set of processes and practises that are integral to various aspects of our lives.
Finally, there is Putnam’s definition of social capital (1993) which is fundamentally about civic involvement and revolves around the notion of social trust, the norms of reciprocity and networks of civic engagement and successful co-operation. This community based approach highlights the centrality of the concepts of reciprocity and trust or the subjective dimension of social capital. His views are also open to criticism. One interesting critique attacks Putnam’s belief that people who join are people who trust. When in truth, little is known about this causality (see Newton 1996: 13 for an in-depth analysis).
The notion of trust must take central place in any discussion on why and how people become friends. For Bourdieu (1990) actors use ‘cultural capital’, prestigious forms of knowledge and style or distinctive speech forms, as bases for interpreting one anothers character and intentions. Cultural style therefore signals a probability that
trusting relations can be constructed. This idea is also evident in Homan’s (1951) Homophily Principle where individuals become friends with persons similar to themselves. For social network theory, the most important characteristic of the network is the degree to which it is composed of individuals with differing social status’ (esp. higher ones). In the case of employment, this is to facilitate movements up the earnings ladder and for the unemployed, so as to gain useful job information. This implies that certain people have more culturally diverse and therefore, according to this approach, more ‘useful’ relationships than others. But is this really the case (in particular here, for the unemployed) and if it is so, why so?
All relationships, however, do involve a degree of uncertainty, risk, and vulnerability and hence, require trust in others (the social cohesion argument). It is argued that this is because the utilities derived from friendships are not constitutive as in market relations but rely more on a generalised form of reciprocity where good turns will be repaid at some unspecified time in the future (Newton 1996). However, yet again, this raises some important research questions about the degree to which this is actually the case. Given the complexities of the social process as outlined in this section of the paper, quantitative research can only provide limited measures of the effects of differing friendship patterns. It can not provide deep insights into the importance of reciprocity and trust in the development of friendships.
The preceding review of the social capital literature has provided important insights into how people become friends and why this process is crucial for their future life
chances. When combined with social network theory, it provides a useful theoretical understanding of the importance of social contact for the life of the unemployed. It can be seen that all three types of definitions of social capital are linked; trust and norms at the subjective level, networks at the objective level and efficacy/ social cohesion as a collective good. Hence, social capital refers to both the relations, networks, and obligations existing in social situations and the product of these interactions (Wall et al. 1998). That is, social capital is confusingly being viewed as an individual phenomenon and a community based resource all at the same time. In addition, politically the term is ill advised in that it focuses blame onto the community for numerous social problems (in place of the usual ‘blame the victim’ strategies). It is for these reasons that I have chosen to avoid the use of this term. Instead, the research focus is on social resource patterns, in particular measures of sociability, the social networks, and social support. What follows is built upon this basic premise, that an unemployed person’s sense of self-efficacy in relation to friends, support structures etc., and to the corresponding feelings of alienation or engagement (psychological effects), will have some influence on their life chances. With the dawn of high tech communications, many social relationships know no geographical boundaries and it has become almost impossible to collect detailed information on every social tie. But by adopting a broader view on the issue of social networks and, hence, focusing on a more general hypothesis, this paper demonstrates why all social relationships are important.
2. The instrumental approach.
As previously noted, most research in this area has emphasised the instrumental way in which social relations influence unemployment exit rates. Traditional economic accounts of this process assumed a simple matching process between labour supply and demand. At the qualitative levels, sociologists argue that there are multiple criteria by which employers select workers (for example Tilly and Tilly 1992). For employers to recruit through their existing workforce seems logical as it provides a cheap and useful screening devise and a clever method for reducing uncertainty and avoiding risk. One idea is that workers will tend to refer to others who are similar to themselves, so that employers will solicit referrals from high-ability employees (inbreeding bias). This policy is reflected in the employment strategies of many companies which now award bonus’ to employees who fill vacancies within the firm.
From the employee side, giving a referral is not a costless exercise, for if the recruit proves unsatisfactory the sponsors own reputation will be endangered (Greico 1987). So, when employees refer friends to their employer they have made an informed judgement about this persons employment ability. It is evident that among job seekers asking friends and family is indeed a popular job search method. In wave 6 of the British Household Panel Study respondents were asked about their job search methods, 69 percent of the unemployed men in this sample reported asking friends and contacts. Holzer (1987) suggests that job seekers prefer this referral procedure
because contacting friends and family generates a job offer with relatively high probability by inexpensive means. Of course, this implies that individuals who are cut off from the social networks in which job information is diffused may have a reduced probability of finding a job.
3. Unemployment and Psychological Well-being.
Individuals are connected to each other and society in numerous ways and to numerous extents with differing implications. Psychological research has emphasised the implications on happiness and stress levels. All aspects of social participation are seen as important means of avoiding loneliness (e.g. Knipscheer 1992). The ‘integration theory’ (Gove and Hughes 1980) suggests that people need satisfying, intimate relationships that give them affection, identity, and care. Yet little is known about the implications of differing levels of social involvement on the well-being of unemployed individuals.
What is known is that unemployed people have much lower levels of psychological well-being than those in work and that the long term unemployed show less distress than those who have recently lost their job (Clark and Oswald 1994). Qualitative studies have shown that unemployment leads to social withdrawal and social isolation. This can be attributed to, in part a lack of the financial resources needed to take part in social activities, and in part it is seen as a result of a loss of selfconfidence and a desire to avoid social contact that may well be damaging to ones self esteem. Some studies have shown that the unemployed tend to be segregated in networks in which a far higher proportion of their friends are unemployed than is the case for employed people (Gallie, Marsh and Yogler 1994). It is argued that these unemployed friends are less likely to offer strong psychological support or effective assistance in meeting financial problems or the difficulties of finding a job.
Therefore, such relationships amongst the unemployed are seen as offering few opportunities for alleviating the stress of unemployment.
This literature highlights the fact that social networks and social support do more than provide an individual with practical or instrumental support. At a basic level, social relations (the combination of the above) seem to help the individual develop their sense of self and their expectation about the world. Antonucci (1987) has suggested that supportive others, through the provision of support, enable individuals to feel efficacious, to have higher levels of self esteem, mastery and control which, in turn, influences the individuals health and well-being. So, in effect, the notion of social cohesion captures the idea of a world full of individuals who feel they are capable and competent and that the world is full of others who love/like them, believe in them, and can be counted on when needed.
The data come from the first six waves of the British Household Panel Study (BHPS), a nationally representative, random sample of over five thousand households across England, Wales and Scotland (South of the Caledonian Canal). Each year respondents are asked to detail their labour market movements over the preceding twelve months, allowing a continuous labour market history of each individual to be recorded since September 1990. The BHPS also includes a complete employment status history for each respondent. This additional information is susceptible to substantial recall error and is therefore only used as a covariant in the series of logistic regressions estimated in the next section.
4.2 The dependent variable:
The case for this analysis is each month starting from entry into the survey up to the end of wave 6 (1997) for all the men in the BHPS dataset (work was carried out for women but the approach is sufficiently different to warrant separate treatment). The dependent variable is binary, taking the value of unity if the individual has moved from unemployment (not working) into employment (working) in the current month, and zero if the respondent remains in the unemployed state. All other cases were excluded. The definition of unemployment used here does not restrain the sample to those actively seeking a job. This would have been an unnecessary constraint and excluded many unemployment exits.
4.3 Definitional issues:
In researching the impact of differing levels of social involvement amongst the unemployed it is important to distinguish between the concepts of social participation, social support and social networks. Researchers have been unable to reach definitional agreement about the meaning of these terms. In this paper, social network refers to the restricted measures of friendship carried in the BHPS, in specific the number of close employed friends the respondents has. It is impossible to collect data on the respondents entire network of associations, so data was collected for a maximum of three close friends about whom numerous questions were asked (see appendix for exact questions). The restriction to a sample of three close friends seems adequate giving the House and Kahn (1985) finding that the first five relationships are the most important and contribute most to the understanding of an individuals’ social relations.
It is, of course, possible to describe the structure of the network in a number of ways, for example we find that male respondents are more likely to have female close friends. Both men and women are also more likely to have friends of their own age. The unemployed do have employed friends, with 67% of the unemployed having some (one or more) employed friends by wave 6. This represents an eleven percent increase in 4 years and is probably reflecting the declining unemployment rate in the UK over the period.
Social networks are usually viewed as a vehicle through which social support can be provided but, unfortunately in the BHPS, the social support questions were asked in general and not in relation to the friendship questions. Definitions of social support echo those of social cohesion, for example it may refer to a process which leads individuals to feel loved, esteemed and valued (Cobb 1976) or, more specifically, to the receipt of certain types of aid. In the BHPS it refers to the perceived availability of five types of support ranging from instrumental aid to emotional support (see appendix for exact questions). It was found that the unemployed perceive themselves to face much weaker support networks than the employed, with 10.6 percent of the unemployed believing they have no-one to help in a crisis in contrast to only 3.8 percent of the employed (see table 1).
(Table 1 about here)
An advantage of using panel data is that it provides a life span orientation to the analysis, so that we can see an individual’s level of support in the past given the future we know occurred. More generally, the idea is that people move through time influenced by specific events, circumstances and people which increment over time and influence their needs and expectations of supportive interactions or networks (convoy model of social relations1). By focusing on the unemployed men who entered the survey in year one, and seeing their job status 5 years later revealed that
most of them (68 percent) were employed. Of those who were unemployed in year 1 and were also unemployed in year 5, 20.3 percent perceived they had no-one (for example) to help in a crisis in year one. This contrasts to only 9.5% of those who became employed (see table 2). Therefore, those who exited unemployment were more likely to have understood themselves to have had someone to help in a crisis than those who did not.
(Table 2 about here).
In addition, it is interesting to compare the levels of perceived support amongst those who had no unemployment spell as compared to those who had. It can be seen that that the figures for those who had become employed in table II are very similar to the averages reported for the unemployed in table I. This seem to indicate that a spell of unemployment (even if it was far in the past) has a long-term effect on ones perception of support. The continuously employed sample reported much higher perceived support levels across all 5 categories.
The social participation index refers to an individuals activity and membership in societal groups through which friendships and social support may be activated e.g. political parties, religious organisations, fitness clubs etc. The intention is to single out those individuals who may be more sociable. As regards Putnam’s hypothesis (1995) on the decline of social capital, the data provides no evidence of overall
declines in sports and cultural associations. The only significant decline in organisation activity occurred in attendance at a religious service (20 percent drop over 5 years).
4.4 Measurement issues:
The first important issue to clarify is the reasoning behind the social support index. Its justification lies in the belief that the actual help provided is less important than the individual’s perception of the amount and quality of the support available. However, the perception of support as a psychological variable may be only partially related to the objective characteristics of the support exchanged. On the other hand, questions which identify actual particular examples of support run the risk of missing the actual perceived non-acceptance of this as support. Support may be given but it could be assessed as misguided, malintentional, or misinformed. There, thus, can be a mismatch between what is desired or expected and what is actually experienced. One argument (Gouldner 1960) is that people disassociate themselves from those who fail to provide support (relying on the assumption that the norms of reciprocity are widely accepted). This implies that some people develop a support reserve or a savings account of support “owed” to them (Antonucci and Akiyama 1987). Other psychological studies have shown that illusion is an important element in the maintenance of well-being (Taylor and Brown 1988). So, in this case, the non
availability of a social support structure (perceived and/or actual) could lead the unemployed to believe that they are incapable of work and prone to failure. It therefore seems that it is the perception of support which is of essential interest here.
There are many different techniques for measuring the social network. My concern is with measures that attempt to infer the strength of a relationship. Granovetter’s work was based around the fundamentally flawed measure of frequency of contact. The amount of time two people spent together was taken as a crude measure for the strength of an interpersonal tie. However, he defined the strength of a tie in a more inclusive manner as a "combination of the amount of time, the emotional intensity, the intimacy (mutual confiding), and the reciprocal services that characterise the tie" (Granovetter 1973: 1361). The problem with this measurement in practise is that it seems likely that some people will have close friends they do not see that often. This is verified in the BHPS where 1 in 5 people contact their ‘best friend’3 only once a month or less often.
This points to the importance of realising that friendship is judged on criteria internal to the character of the on-going relationship. It is possible with the BHPS to select out those friends who live close to the respondent or those the respondent contacts on a daily basis. But it proved more adequate to allow respondents define their close relationships themselves. In that sense, the measurement of the network employed here is far the most satisfactory available in a large scale quantitative survey. However, it must be noted that we do not know anything about how or why
the respondents choose these people as their closest friends. Yet, it is hoped that these quantitative assessments of an individuals social relations do take some account of the quality of the relationships amongst individuals. They are not based on counting the number of people one knows, or how often one sees them, or on how much support is actually received. Yet, as argued earlier, without engaging in fieldwork and participant observation, there is little one can say about how people define their close friends or the importance trust and obligation in the development of friendship and support.
4.5 The explanatory variables:
The social support index runs from 0 to 5 depending on the number of situations in which the individual feels there is support available to them. Most people (80 percent) perceive the availability of full support. The social participation scale was simplified into three dummies. 14 percent of the unemployed respondents are members only in an organisation and 39 percent are active members. Three dummies were also created based on the social network measure with an average of only 37 percent of the unemployed sample naming all three of their close friends as employed. The importance of these social resources patterns on ones job chances was then tested before and after controlling for other factors.
Included in the models is the respondents age at the start of the current unemployment spell. We would expect from previous literature a U-shaped relationship between age and unemployment proneness. We take the socio-economic background of the individual’s father (Goldthorpe class) as the indicator of the cultural capital of the household of origin, and the highest educational qualification in 1991 as the indicator of individuals educational capital. Socio-demographic controls:
Included at this point is whether or not the respondents attended a private school as an indicator of one form of potential social capital (6 percent of the unemployed sample fell into this category). The fathers Goldthorpe class measure was dropped for dummies indicating whether the respondents came from a two earner household, a one breadwinner situation or an unemployed background. A housing tenure dummy was included to control for any spatial concentration of the unemployed. Research has shown that there is a tendency for this section of the population to cluster together in large council housing estates. This, of course, has major implications on the social network. A health dummy was included to control for any possible physical impairment preventing the individual from working. Finally, dummy variables were included to control for domestic cycle effects. Previous studies (for example, Morris 1990) have shown that one’s marital status and the present of children have an effect on the probability of gaining employment. In particular, the present of young children is strongly associated with weak male labour force status.
Using the work history data collected at wave 2 in the BHPS it is possible to control for previous unemployment experiences. A variable was included measuring, in months, the length of an individual’s previous unemployment spell. In addition, regional monthly unemployment rates were included to take account of the economic
environment at that time. These were based on travel to work areas for the first 4 years of the BHPS but, for the final two waves, they were only available at a regional level.
The tables in this paper report the effects (log odds) of the independent variables on the general probability of entering employment from unemployment in a given month (estimated from logistic regressions). A model performance statistic is included, -2 the log of the likelihood, which measures how well the estimated model fits the data. A model that fits the data well is one that results in a high likelihood of the observed results. This translates to a small value for -2LL (if a model fits perfectly, the likelihood is 1, and -2 times the log likelihood is 0).
Model 1 in table 3 enters the social variables one at a time and tests their effect separately and then together on the probability of exiting unemployment. The social support scale has a significant positive influence on this probability. This positive influence increases as one moves from perceiving no available support to perceiving some social support in all 5 situations. Interestingly, the effect of being a member in an organisation has a stronger effect on unemployment exits rates than being active. This may be due to the fact the those who are members in an organisation are more likely to have all employed close friends (45 percent of members had all employed
close friends compared to 34 percent of active members). The excluded dummy was neither being active nor being a member of any organisation.
(Table 3 about here).
In the third column, the effect of having employed close friends was examined. The excluded group are those who have no employed close friends. In relation to this group, those with some or all close employed friends are significantly more likely to exit unemployment. Those having all three close friends employed have the highest probability of exiting. However, -2LL has risen in these two models compared to the social support model. Finally, when all these variables are entered simultaneously, the models fit improves but the effect of the social support index is now insignificant.
As well as these variables being related to each other they may also be capturing the effects of other more conventional covariates. In model two, controls are added for some possible demographic influences on unemployment exit rates i.e. age, education and family background. In all cases the goodness of fit of the models improved yet with the friendship variable remaining highly significant and displaying the same association as before on the probability of exit from unemployment.
membership/activity in organisations dummies are significant. The other variables display the expected results e.g. as age increases the likelihood of exit from
unemployment decreases at an increasing rate. Those in the sample who have a higher educational qualification are much more likely to exit than those with some other qualification (excluded group those with no formal qualifications). The effects of fathers occupations status (as measured by Goldthorpe class when the respondent was aged 16) shows a significant fall in the probability of exit as the fathers occupational status increases. One possible explanation is that those from ‘better’ social backgrounds have a lower probability of exit due to their higher expectations which may lead them to refuse the first job on offer.
In model three of table three, controls are added for another set of possible influences on unemployment exit rates. The first control is the dummy which highlights a private school education (the old schools ties argument). In addition, dummies were included to signal the level of employment in the respondents household of origin and the characteristics of their household of destination. Even after controlling for all these, the friendship measure remained highly significant. The models fit did deteriorated but only slightly.
The main argument is, of course, that it is the individual employment history (issues of state dependency or the influence of past states on the present condition) and the local employment possibilities that ultimately influence ones chances of finding a job. Even after controlling for these historical details, the friendship dummies remain significant (this time at the 0.01 level). The differential effect of having some
versus all employed close friends has declined but, as compared to having none, the effect is still strong. It still seems that employed close friends exert a significant influence all of their own on unemployment exit rates. The best fitting model includes all these control and the social network measure only. Table 4 presents the full final model.
In table 4 we find that the local authority tenant dummy correlates strongly with employment status. 62 percent of those who exited unemployment at some point in the 6 years of the panel were owner occupiers compared to 45 percent of those who did not. The influence of such informal factors on unemployment exit rates is strong. In other words, the world of social contacts which grows up around unemployment does much to reproduce the pattern. It seems that the reduced contact with the world of work explains much of the position of the long term unemployed.
(Table 4 about here).
Finally, figure 1 presents a visual illustration of these results. For illustrative purposes the probability of entering employment for a man, aged 25, as estimated from the final model is plotted. So, for example, an educated young man, living in a low risk unemployment area and having been unemployed for a year has a probability of finding a job of 9 percent. If this man were married the probability of exit increases to 14 percent, if he was active in some social organisation, his
likelihood of becoming employed increases to 18 percent. If some of his close friends worked, his chance of finding a job increases to 26 percent. The downward sloping line indicates that the longer the individual has been unemployed the smaller the effect of these variables.
(Figure 1 about here).
The above graph clearly illustrates the strong effect informal social factors have on employment chances. However, these social resource measures may be acting as a proxy for some other characteristic. Maybe all the above states (marriage, group activity, friendship) are indicative of ones psychological status, sociability level, or individual motivation to find a job. Obviously the one clear link is between perceived social support and psychological well-being. But what is the link between, for example, having employed close friends and psychological well-being and can it help us explain the relationship between these friends and ones chances of finding a job?
We can not infer from the above finding that these employed friends actually helped the unemployed find jobs since the employed sample were not asked how they found their jobs. In addition, we can’t relate the support question to these three friends because the support questions were not asked in relation to these close friends. It
does seem most likely that the influence of these informal social factors is indicative of the importance of psychological status. The next section links these differing sociability levels to psychological health scores.
The BHPS contains the 12-item version of the General Health Questionnaire (GHQ12). It is used as a general indicator of psychological well-being. The questions ask respondents how they have been feeling over the last few weeks. The items concern concentration, lost sleep, usefulness, decisiveness, strain, overcoming difficulties, enjoyment, problems, depression, confidence, worthlessness and happiness. What follows is a brief and simply analysis of the effect of differing social resource patterns on psychological well-being. For this purpose, the GHQ is scored by rating each response according to whether each of the symptoms is simply present or absent, yielding an additive score of the number of symptoms, giving a possible maximum of 12. This score was then multiplied by -1 so that, the lower the score the worse an individual’s psycho-social well-being. Table 5 reports the results of multivariate analysis using this score (ordered probits with the GHQ-12 as the dependent variable). A dummy variable (EMPLOYED) was entered for those who found a job to test the effect of exiting unemployment on psychological well-being. As expected, it proved insignificant. Previous research has highlighted that psychological well-being does not instantly recover after finding a job. We also see the usually inverted U shape effect with age, so that the middle aged suffer more psychological distress than either the young or the old. The unemployed who attended a private school are particularly prone to poor
psychological well-being as are those ill of health, the widowed, separated and divorced. The effect of the job history variables are small but significant with the unemployed who live in a high risk unemployment area showing less distress than others (probably because the norm is to be unemployed). In addition, the longer term unemployed seem to find a balance and display, as expected, a higher score.
(Table 5 about here).
Finally, all the social variables proved highly significant and contributed most to the models fit. As expected, a high psychological health score indicates the perception of strong support networks. Interestingly, membership and activity in an organisation deteriorates one’s well-being. This may be due to the financial pressure exerted on the unemployed who still participate in these groups. It may also be explained in part by the suggestion that the unemployed join similar clubs. In addition, the unemployed who have close employed friends are ‘happier’ than those who don’t. Therefore, having all unemployed friends implies higher psychological distress levels in addition to decreasing probabilities of finding employment.
(Table 6 about here)
This paper investigated the effects of differing levels of social involvement amongst the unemployed (Table 6 provides a summary). It represents the first verification, using a nationally representative random sample, of the importance of informal social processes in influencing unemployment exit rates. Unemployed men who have close employed friends are significantly more likely than those who do not to find a job. Those who find jobs are also more likely to have perceived the existence of strong social support networks and be members of some social group. However, it was noted that there may be some unobserved characteristic which makes these unemployed individuals (who have employed friends) more predisposed to do well in the labour market and which also allows them form stronger and more numerous social relations.
To investigate this further an analysis of psychological distress scores was undertaken. Indeed, it was found that the unemployed who have all employed close friends are far most likely to display low psychological stress than either those who have some or no employed close friends. Therefore, the unemployed whose close friends are all unemployed are doubly worse off since these individuals are more likely to have high psychological distress scores and less likely to receive effective assistance in meeting the difficulties of finding a job. This work has emphasised the social and psychological aspects of unemployment and shown the crucial role social resource patterns play, in addition to tradition
economic accounts, in the determination of who finds a job. The segregation of the unemployed into networks in which a far higher proportion of people are unemployed directs these people down a road of depression and isolation. Social policies must address such problems. The harm caused by council housing schemes which have produced ghettos of excluded groups is one obvious area in need of change. This work has explicitly shown that social resource concerns have wider ranging implications and stronger estimated effects than previously thought.
End Notes 1
see Antonucci and Akiyama 1987.
The notion that economics behaviour is ‘embedded’ in a social context first dates to Polanyi 1957.
Respondents were asked in alternative years about their “best friend” and not their “three closest friends”.
Table 1: Respondent’s employment status by forms of support.
Employment Status Men only.
Proportion of respondents who claim there was nobody available to listen
help in a crisis
really appreciate them
Number of observationsc
Source: British Household Panel Survey, 1991, 1993, 1995. Average of the three year. a
The employed are defined as those in part-time, full-time and self-employment, an average of 65% of the respondents were in this state.
This category refers to the self-declared unemployed and those on government training schemes, for these three years a mean value of 8% were unemployed.
Includes all person-year observations for the three years of the survey when the support questions were asked.
Table 2: Employment status of respondents who were unemployed in 1991 by forms of support.
Employment Status 1995 Men only.
Proportion of respondents who claim there was nobody available to listen
help in a crisis
really appreciate them
Number of observationsa
Source: British Household Panel Survey, 1991-1995. a
Of the 572 men unemployed in the first year of the BHPS, 407 remained in the survey by year 5.
Table 3: Effect of measures of social cohesion on the probability of entering employment from unemployment in a given month. Logistic regressions
‘Social’ variables social support index
Models : I 1.14 ** (4698)
II control demographic + 1.12 (3809)
III socio-demographic + 1.11 (4163)
IV empl history 1.10 (3861)
organisation - member -active
1.46 1.30 (4971)
1.55* 1.35* (3956)
1.47** 1.30** (4323)
1.35* 1.27* (4024)
employed close friends -some -all
2.00*** 2.38*** (4984)
1.77*** 2.22*** (3936)
1.89*** 2.07*** (4302)
1.69*** 1.68*** (3494)
social support index member organisation active organisation some employed friends all close friends employed
1.06 1.42 1.24 1.97*** 2.29*** (4698)
1.04 1.51** 1.33** 1.80*** 2.25*** (3761)
1.07 1.46** 1.24* 1.78*** 2.01*** (4126)
1.08 1.36* 1.23* 1.58** 1.61** (3841)
Note: -2(Log Likelihood) in brackets. * p