Can you ‘know’ big data?: Jeffery C. Johnson’s Sunbelt keynote address

This week sees the 34th Sunbelt Conference, an annual week-long conference on social network analysis. Running since 1981, the conference brings together researchers from all career stages and encompassing all aspects of network analysis, incorporating sessions on both methodological and empirical applications.

Each year, Sunbelt provides a keynote presentation from one of the very leading experts in the field. This year, that honour was given to Jeffrey C. Johnson from East Carolina University.  Johnson’s presentation was entitled ‘Knowing Your Social Network Data and Measures’ and argued that it’s important when Analyzing Social Networks to have a strong understanding of both the data and techniques used.

Johnson has a background in anthropology and argued that he always brings an ethnographic perspective to his network analysis. He provided a detailed account of how his research findings have been shaped by the diligent attention paid to gathering strong and valid data. He famed his discussion around his research of South Pole stations and the relationship between team composition and morale.

Johnson argued that research design is highly important, showing diligent consideration of deciding both what data to collection and how to analyse it. He discussed the thoughtful consideration of what precise questions to ask, showing how giving slightly different options (ranking or rating of potential friends) can influence not only what sort of analysis can be operationalised, but also hold consequences for building rapport with research subjects. Similarly, he outlined the careful process of determining which precise measures to use rather than blindly selecting an existing technique. He demonstrated how a combination of multiple correspondence analysis and multi-dimensional scaling provided a valuable understanding of identifying if a core-periphery or clique structure was observed. The talk highlighted that a sound network analyst requires exploring the composition of the data and methods to the same degree that a detailed ethnographic study attempts to understand a community.

This brought Johnson to query whether the growing trend for analysing big data can provide as detailed an analysis as is possible with small data. He argued that, hitherto, big data analysis has been exploratory and atheoretical, providing only a descriptive account and being largely data-driven. They have focussed on, for instance, providing visualisations, identifying patterns providing rankings rather than theorising about what those patterns actually mean. Johnson argued that this is the next step for big data, but questioned whether it is possible to ‘know’ big data in the same way we can with smaller datasets. Johnson was not arguing from the position of a big data sceptic, but rather as someone who has actively explored big data in studies on textual information, news reports on Sudanese violence and the Reuters tracker.

During the introduction to the presentation, Russ Bernard stated that for as long has he’s known Jeff Johnson he’s always had the ability to turn the research he’s imagined into grants and publications. Let’s hope that he’s successful in his charge to add increased theoretical output from big data analysis.

Dave Griffiths, 20 February 2014

Advertisements
Posted in Uncategorized | Leave a comment

UKHLS and social connections data

The latest wave of Understanding Society, or the UK Household Longitudinal Study (UKHLS), includes the first module on social networks. The UKHLS was first run in 2009, emerging from the previous 1991-2008 British Household Panel Survey (BHPS), providing detailed information on all individuals in 30,000 UK homes. Those individuals in Wave A, or who entered from the BHPS, are tracked every year and interviewed along with all others within their household. Therefore, when households are stable across time (i.e., a married childless couple who always live together)  the same individuals are interviewed each year. But, when  households change (i.e., a married couple whose grown-up children move out of the parental home and have flatmates, partners and children of their own) different people become eligible for inclusion.

Individuals are interviewed annually, covering a wide range of subjects including, amongst others, employment, politics, health, income and education. In addition, there are various modules which are included once every three years, including cognitive ability. The most recently released wave, C, contains the first data from the social networks module. This is something I’ve been looking forward to immensely since hearing Raj Patel discussing what was included last May.

Wave C provides a range of variables which explore different elements of social networks. There is a particularly interesting set of questions relating to homophily, or the concept that ‘birds of a feather flock together’. Participants were asked what proportion of their friends are:

  •  of a similar age
  • of the same ethnicity
  • have a similar level of education
  •  have a job
  • have a similar income
  • live in their local area
  •  are also family members.

Respondents have five options: all similar; more than half; about half; less than half; or none. These categories perhaps introduce uncertainty surrounding the boundaries between categories, for instance, what proportion is required to move from about half to move than half? 60% could be viewed as about half if you have 10 friends (6/4 split) but less so if you have 30 (18/12 split). If 19 of our 20 friends have a similar level of education does that count as ‘all’ and, indeed, does the strength of relationship with that one other influence people’s choices.

Also, it is somewhat arbitrary what counts as ‘similar’. Should people who are 40 view those who are 49 as a similar age based on a similar position in the lifecourse, or limit the criteria to within 5 years? Do people with a PhD and a university degree have a similar level of education? If the graduate sees similar as those with a degree and the postgraduate sees it as someone with a higher degree, two people could easily differ in whether their friendship is homophilous. Therefore, such data should be seen as how people perceive their friends rather than necessarily subjectively measuring levels of homophily in society.

homophily

The data show trends we might have predicted. It appears that the youngest and oldest respondents were most likely to have all their friends of a similar age, perhaps indicating that those in maximal years of employment (20 to 59) have connections across the age range. Similarly, the social isolation of those in the least advantaged occupational groups can be seen with a greater tendency to have mostly friends with a similar level of income. Similarly, those with the lowest educational qualifications are most likely to have friends who with the same levels. Many of the social trends we might expect to see are observable.

Social isolation, as well as social connections are explored in the wider survey. Nearly 14% of respondents stated they never go out socially with friends. Nearly a third of those who don’t socialise claimed they were too busy and around a quarter stated health, illness or disability reasons. Caring responsibilities and financial reasons were also common causes of low sociability. Similar data is in the survey for why people don’t visit their families.

  • Too busy  32.7%
  • Health  24.5%
  • Caring responsibilities  18.6%
  • Financial reasons  18.3%
  • No-one to see  8.4%
  • No access to public transport  7.0%
  • No access to a car  6.3%
  • Anxiety/lack of confidence  6.0%
  • Nowhere to go  5.0%
  • Fear of crowds  2.7%
  • Fear of crime  1.7%

Other social capital related questions include sense of belonging to the local neighbourhood, attitudes and opinions towards neighbours, membership and activity within in a range of social organisations and also whether people are a member of a social media website, and if so how long they engage with such sites. Perhaps unsurprisingly there is a marked effect of younger people being more likely to use social media, with much lower engagement with such websites for older people.

social_media_age

The BHPS regularly asked for basic demographic information on up to three friends, with more detailed depictions of the best friend asking about occupations and other information. This appears to have been dropped from the UKHLS. However, both surveys offer a range of within-household social connections which can be operationalised to ascertain patterns of homophily within homes and, as households change, potentially homophily between members of connected households. Therefore, a survey like Understanding Society offers far more opportunities for analysing social networks than what is incorporated within the networks module.

Dave Griffiths, 19 Februrary 2004

Posted in Uncategorized | 1 Comment

Social Stratification Research Seminar – 2013

September saw the 2013 edition of the long-running Social Stratification Research Seminar, at the University of Cambridge. As discussed in an earlier blog, this is an event which has been running annually since 1968 and brings together researchers with diverse interests in the area of stratification. The event always has an international scope and, this year, included presenters from Italy, Canada, the Netherlands and Sweden alongside many UK-based researchers.

A nice feature of this year’s seminar was the wealth of data sources discussed, with presentations utilising nine UK social surveys, two European multi-country surveys, Canadian data and utilising the range of sources available through the Minnesota Population Centre. A theme running through many of the papers was that empirical findings from survey datasets did not support the rhetoric commonly claimed by politicians and social commentators, with a number of speakers making the point that popular view was not supported by the evidence.

In keeping with the spirit of open access, below follows a description of all 15 papers presented. These are my comments and reflections of the papers and any misrepresentations are mine. As always, please correspond with the authors of the presentations if you’re interested in their research, rather than allowing my appraisal of their papers to reflect their work.

Roxanne Connelly opened the seminar with her paper ‘The Micro-Meritocracy: The distribution of ‘merit’ throughout ‘big’ class, ‘micro’ class and gradational representations of the social structure’. Connelly examined the relationship between occupation, parental occupation and cognitive ability using the 1970 British Cohort Study, utilising newly released fine-grained occupational data. These unit groups were converted to microclass and compared with big class and CAMSIS effects.

Cinzia Meraviglia, Deborah de Luca and Harry Ganzeboom discussed the construction of the ICAM scale in their paper, ‘Social distance, status and prestige: Towards a unique measure?’. Meraviglia used data on household income and cultural consumption from the European Social Survey to provide sensitivity analyses of the ICAM (international CAMSIS) scale, ISEI and national CAMSIS schemes. She argued that despite the theoretical rationales for stratification schemes differing they essentially pick up the same underlying effects and therefore the selection of the occupational measure is more trivial than the literature suggests. The principal message of the presentation was that if different scales are achieving broadly similar results we should rethink the theoretical contribution each measure attempts to convey.

Yaojun Li and Anthony Heath discussed The social mobility of ethnic minorities in Britain (1992-2011): changes over time and across generations. Using data from the General Household Survey, British Household Panel Survey (BHPS) and Understanding Society they explored the social mobility outcomes for members of minority ethnic group, broken down by generation of migration. They argued the patterns of migration differed to that suggested by the tabloid press with an over-representation of salariat workers. Overall, they found that social mobility is declining by slow degrees for all ethnic groups, although at differing rates by ethnicity, gender and migratory status.

Marina Shapira presented ‘Do you speak your grandma’s language? Intergenerational communication and attainment of immigrant children in Western countries of immigration’. Shapira used PISA data to explore the educational attainment of immigrant children across Europe, examining the languages spoken in the home and migration status. She argued there was much variation between countries, and types of country studied, finding much difference in attainment levels between native and immigrant schoolchildren, largely with a small but important difference between the 1st and 2nd generation immigrants. The relationship between living with grandparents and attainment appeared to interact with migration status and languages spoken.

Included with the first day was a strand from the ESRC-funded project ‘Is Britain Pulling Apart?’. The first paper in this strand was Paul Lambert Dave Griffiths, Richard Zijdeman and Erik Bihagen’sSocial distance of family and friends: Socio-economic and socio-demographic patterns‘. Lambert provided correspondence analysis and associational models on a range of variables from both the BHPS and from national censuses over time from IPUMS-I data. He argued that it is possible to differentiate between the effects of stratification and other socio-economic and socio-demographical effects through using such methods. He ended by suggesting the data shows that Britain isn’t pulling apart and that internationally the data doesn’t suggest society is as complex and ever-changing as sometimes suggested.

Vernon GaylePaul Lambert, Dave Griffiths and Mark Tranmer presented ‘Measuring the influence of others: Exploring social connections in contemporary Britain‘. Gayle argued that survey analyses frequently miss out intra-household connections which can appear in panel studies, exploring the differing types of ties which can be found within homes. It was argued that incorporating household ties can provide strong improvements to models and it cannot be known a priori which types of connections, or means of measuring them, is most appropriate without performing a sensitivity analysis of the potential methods.

Mark Tranmer discussed ‘Methods for the analysis of social distance patterns and trends’. Tranmer argued that ignoring a level in statistical analyses can be consequential and applied this logic to occupational analysis. He modelled the difference in CAMSIS scores between husbands and wives, using the BHPS, between 1992 and 2008. His initial findings suggest the difference in stratification positions between couples were increasing, but warned further analysis is required into these trends.

Dave Griffiths and Paul Lambert presented ‘Domains of social distance’, which looked at various methods for measuring levels of social interactions between groups. The preliminary and exploratory analysis utilised Understanding Society to construct networks of social groups more likely to interact and used methods such as QAP regression and E-I Index analysis to test levels of cohesion in contemporary Britain.

The second day saw presentations from researchers representing institutions in Canada, the Netherlands, Switzerland and the UK, providing both analyses of those countries and international comparisons.

The first Friday presentation was ‘Gender and the pay of university graduates: Evidence from Canada’s National Graduate Survey’ by Mike Smith. Smith used data from the National Graduates Survey to build a longitudinal dataset of destinations of Canadian graduates from 1986 to 2000 to explore gender differences in their careers two and five years after graduation. He found for Bachelors students that the field of study for first degrees made a large difference to gender pay gaps, although as these pointed in different directions this is often missed as it can be evened out across the whole workforce. He argued that part of these differences can be accounted to differences in the Canadian economy over the 20 years studied, with differing employment rates and GDP at different time-points and a dot-com boom in the 1990’s which increased salaries in the male-dominated computing sector. He also argued that during the 1990’s women were more likely than men to be outside of employment and education, thereby acquiring less human capital to complement their university qualifications. This was presented as evidence that the gender pay gap can affect those in advantaged occupations and early in their careers, contradicting arguments that pay discrimination is generally found amongst low and un-skilled workers.

Nichola Shakleton critiqued ministerial claims about child obesity in her paper ‘Is there really a link between low parental income and childhood obesity?’. Using the Millennium Cohort Study she found poverty was only associated with childhood obesity when there was an absence of contextualising variables studied. Shakleton showed that incorporating parental education, social class and other demographical factors eliminated the association between low pay and obesity. Comparison with data from Growing Up in Scotland supported the findings, suggesting that the income effect claimed by politicians was created by an underlying effect of parental education.

Harry Ganzeboom and David Nikoloski considered stratification measures in their paper ‘Occupational Homogamy and occupational mobility as measures of occupational stratification’. They argued that intergeneration and homogamy measures of stratification, such as ISEI and CAMSIS respectively, are created through measuring entirely different processes yet produce broadly similar results. Using the European Social Survey, they compared intergenerational mobility models (which should be linked to socio-economic status and measures such as ISEI) and homogamy models (which should be linked to social distance and measures such as CAMSIS). It was argued that analysis of high residual off-diagonal combinations could aid understanding of whether processes of homogamy or reproductions were dominant forces of stratification.

Franz Buscha and Patrick Sturgis addressed the differing views on mobility in their paper ‘Inter-cohort trends in intergenerational mobility in England and Wales: Income, status and class (InTIME)‘. They looked at the debate between economists and sociologists regarding relative levels of social mobility, using the ONS Longitudinal Survey to build a large longitudinal dataset of the occupations of individuals and their parents, which was matched to the New Earnings Survey Panel Dataset to imputed likely income levels. They argued that the ‘by slow degrees’ argument of increasing social mobility was found when looking at the CAMSIS scores of individuals, but the income-derived measure was more consistent with a ‘constant flux’ of changing levels of mobility rather than an increase in social reproduction.

Robin Samuel explored ‘How cognitive and different non-cognitive characteristics affect labour market outcomes in Switzerland’. Using PISA data he looked at how cognitive and non-cognitive characteristics related to job satisfaction and earnings. Samuel argued that cognitive ability was often a stronger explanatory than non-cognitive controlling variables. He argued that both cognitive and non-cognitive processes were important, although the non-cognitive effect was smaller than other studies have suggested.

Lindsay Richards presented ‘Patterns of Connectedness, Economic Hardship and Psychological Effects, exploring the relationship between life satisfaction, financial situation and connectedness’. Latent class analysis of social capital from the BHPS identified six classes of social supported. Household income and satisfaction scores were compared for these groups, showing an effect for social capital with regard to other wellbeing positions. The effect of higher income upon satisfaction was shown to differ by latent class.

Dave Griffiths, Richard Zijdeman and Paul Lambert provided the final paper of the seminar,’Bakers’ sons aren’t butchers: Analysis of microclass mobility in the 19th century‘. Making use of the linked 19th century census in Norway and the USA from NAPP, they explored whether contemporary microclass theories could be observed in historical datasets. Expanding on the HISCO, HISCAM and HISCLASS developments in historical occupational analysis, a microclass scheme sensitive to 19th century occupations was discussed.

The seminar has been running since around 1968, usually in Cambridge although which visits to Cardiff, Stirling and Utrecht. Next year’s seminar will, again, be organised by Bob Blackburn and Paul Lambert  in early September.

Posted in Uncategorized | Leave a comment

Exploring the Influence of Others at the Understanding Society Conference

This week sees the 2013 Understanding Research Conference, at the University of Essex. Understanding Society, which started in 2009, is the largest longitudinal household survey in the UK, interviewing people living in 40,000 households. This biennial conference will see 50 papers, from 35 research institutions and attract 250 delegates, focusing not only on the use of Understanding Society data but longitudinal household studies from across the globe.

Our project will contribute a paper to Friday’s session on ‘Social Connections and Cultural Groups’. Prof. Vernon Gayle will be talking about ‘Measuring the influence of others: Exploring social connections in contemporary Britain’. A non-technical summary is included below and can be downloaded, alongside the PowerPoint slides, from our website. The methods we are exploring in this paper will be utilised throughout our project when asking the question ‘Is Britain Pulling Apart?’

Dave Griffiths, 25 July 2013

Exploring the influence of others: Modelling social connections in contemporary Britain

 Prof. Vernon Gayle, Prof. Paul S. Lambert, Dr Dave Griffiths and Dr Mark Tranmer

Do birds of a feather flock together? Or do opposites attract? Between spouses and friends there are notable structural similarities. In countries such as Britain, two hundred years ago we would have expected that a farm worker would have married another famer worker. A hundred years later we could reasonably have expected that a coal miner’s best friend would have worked in the same pit (and maybe even on the same shift).

In recent decades the forms that social interactions take have changed in many ways, but the patterns of similarity amongst spouses and friends persist. Nationally representative data show that a sizable proportion of bus drivers are married to cleaners, and we know that many academics are married to other academics. Multiple observations lead us to conclude that young Goths show a preference for friendships with other Goths, and that Christians often restrict their dating to others with the same beliefs.

It is reasonable to assume that individuals with the same social values, views and political beliefs are more likely to gravitate towards spouses and friends that are similar. But are these connections selected of similarity, or do connected people become more similar as time goes by? Consider the following simple illustration.  If a man moves in with his vegetarian girlfriend, and she does all of the cooking and the shopping then we expect that his diet may soon become largely vegetarian. In this situation it is not initial similarity that has led to the outcome, but a process of assimilation (or dependency). If the same man avidly watches the Test Match, then we see no obvious reason to suspect that his girlfriend will become a cricket fan.

As a result of their design, datasets such as Understanding Society (the UK Household Longitudinal Study) routinely collect information on other individuals who have connections with the survey respondent. These additional sources of data are often not exploited in analyses. In the presentation we illustrate how a number of groupings within the contemporary household can be identified. We then showcase how information on spouses and on other household members can be included in analyses of a series of outcomes (e.g. smoking, obesity, health, sport and voting). We conclude that the extra information that comes from other connected people is beneficial and improves our understanding of a wide range of things that are consequential in contemporary social life.

Posted in Uncategorized | Leave a comment

Social Stratification Research Seminar 2013

To my mind, one of the nicer features of work in higher education is the opportunity to participate in collaborative seminars and conferences which, at their best, provide convivial social meetings, generate feedback that help improve our own analyses, and allow us to better understand the endeavours of others and our subject area more generally. Not all conferences have all of these qualities, of course (I found some rather pessimistic views at this blogsite!). However in recent years I’ve been happy to participate regularly in a couple of meetings which I think are consistently of a very good standard. Thus as the call for papers for the latest version of the annual Social Stratification Research Seminar  is circulated (deadline for paper proposals: 26 June 2013; meeting dates: 11-13 September 2013), it seems worth making a few comments on this long-standing – and I believe very successful – seminar.

 

First organised in 1968 (we think so – though nobody that I’ve spoken to is 100% sure that’s the right year) in Edinburgh, the Social Stratification Research Seminar has taken place every year since, with just a couple of exceptions. It is normally held in early September in Cambridge, nowadays with a link to the Department of Sociology, what was previously the Department of Applied Economics. The seminar is also sometimes known as the ‘Cambridge Stratification Seminar’ and alternatively the ‘Stratification and Inequality Seminar’. For most of its history the seminar has been organised by Bob Blackburn, nowadays an Emeritus Reader at Cambridge and a Fellow of Clare College (the college within Cambridge where the seminar has often been held). In the last decade or so, the meeting has occasionally been held outside of Cambridge. Bob Blackburn continues to be on the organising committee, whilst I’ve taken the lead on meeting arrangements when we’ve had recent meetings in Cambridge and, on two occasions, in Stirling. We’ve also in the recent past had three meetings in Cardiff that were organised by Ken Prandy, and one at Utrecht that was organised by Richard Zijdeman.  Nowadays, I also maintain a webpage that describes the seminar and hosts the call for papers, and on that site we also have a limited archive of selected papers or slides downloadable from some of the previous meetings.

 

I personally have only been attending the seminar for the last 15 years – far fewer than several of the meeting regulars – but in that time the meetings have had a number of characteristics which distinguish them, and, I believe, contribute to their success. First, the seminar group is always a fairly small one – typically between 15 and 30 participants – and sessions comprise a single continuous plenary, to which everyone attends. Thus, throughout the meeting, everyone is present for all of the papers, and this works very well I think in gradually building up a corpus of mutual understanding and constructive engagement between participants. Of course, we are never quite able to completely enforce the expectation that everybody should attend every session – the large international conferences have a lot to answer for, in my opinion, as many of us learn from those meetings the rather rude habit of dropping in just for the sessions in which we ourselves are presenting or are most involved. At the stratification seminar, we try to deploy reproachful frowns and murmurs of disappointment when colleagues leave early, arrive late, and so forth – that seems to show them.

 

Another important characteristic of the seminar is that papers are allocated a reasonably long time for their presentation and discussion. Our usual pattern is to schedule two papers per 90 minute session, with four 90 minute sessions per day. Thus each paper is usually given 45 minutes presentation time, a format that usually allows both for the presenter to develop their material at a reasonable length, and for the audience to ask relatively probing questions in an extended discussion period. Often papers presented to the meeting are in a fairly well-developed state, but it’s also the case that the seminar format is well suited to presentations of work that is still very much in progress, and open for discussion. 

 

The stratification seminar has also historically been characterised by its eclecticism of coverage. Participants themselves are drawn from a wide range of institutions, from a mix of more junior and more senior positions, and from various nationalities (in the 2009 meeting for example, every paper on the first day was given by a presenter of a different national background – if you’re willing, that is, to regard Scotland and England as different). The content of papers is also typically fairly wide-ranging on the general theme of stratification and inequality. With an average of 14-16 papers per meeting, the volume of activity in stratification research is such that it would actually be quite easy to organise the seminar in such a way that every paper was on a very similar topic. On the contrary, however, the stratification seminar has a reputation for incorporating some fairly diverse materials from across the gamut of sociology and social policy as related to stratification and inequality. Positive consequences of this approach include that the meeting proves interesting and engaging on a range of topics, and that the membership of the meeting evolves gradually through time, with (hopefully) no great barriers to prevent new faces coming along and presenting something potentially a bit different to what has come before.

 

Traditionally, many but not all of the papers given at the seminar involve quantitative data analysis, often using medium- and large-scale sample surveys or other large scale data resources. In many of the academic social sciences, such methodologies are a minority specialism, and it is not uncommon to see our ‘non-quants’ colleagues flapping anxiously when presented with quantitative research analysis, disengaging, and running away. One important feature of the stratification seminar is an unspoken premise that papers with statistical content are worthy of consideration – something that may seem an unnecessary courtesy, but in fact is not shared in all social science environments. Not everybody at the meetings has particularly extended knowledge of statistical approaches, and accordingly we don’t necessarily all fully understand every methodology that is used, but there is an acceptance nevertheless that relatively complex statistical methods might be appropriate ways to investigate social systems, whilst there is usually sufficient methodological expertise within the audience to submit most results to often quite robust informed critique.

 

Finally, another characteristic of the meeting is that the atmosphere of the group strikes – to my mind – a nice balance between supportive and friendly, but also inquisitive and, when relevant, critical. For most of us, the meeting is an enjoyable, collegial activity, but most years it is typically not without a few more demanding criticisms being raised of papers, with the odd blunt question or robust response (I’m told this happened a lot more in the past; to some, the modern modes of academic debate have become a bit too soft and friendly by comparison!).

 

Are their downsides to our convivial meeting? As a small group running a similar meeting over a long period of time, arguably the seminar is at risk of being a little conservative, both in its membership and its values. Sociologists at the meeting have I think a natural impulse to try to avoid this, but there is probably nevertheless a drift in this direction. We also consistently welcome new participants to the meeting, though from the outside it’s easy to imagine the seminar group might seem a little hard to approach. Sometimes, also, the eclectic coverage of topics means that some papers aren’t really being given to the most informed or critical audience for their topic, and some papers must seem to some participants or little relevance. Additionally, we could note that are some shared views held by a number of the most regular participants to the meeting  that are not all that widely held in other areas (I am not necessarily commenting obliquely here on the widespread endorsement of CAMSIS scales by members of the seminar, but to outsiders this is one example that must seem obvious!).

 

None of these worries keep me awake at night, though organisational issues on the seminar sometimes do! In some regards, the seminar is caught between a rock and a hard place insofar as, in trying to be a small, engaging meeting, it is never likely to generate a great deal of income or attract large scale funding (the seminar arguably has a very successful cousin in the same area, the meetings arranged by the ISA RC28 group, another conference which I personally rate very highly, but one that operates on a much larger scale and is able to generate significant income to support hosting and running costs).  One possible threat to the stratification seminar, indeed, is the fairly widespread outlook – often linked to obtaining funding to allow personal participation – that it is only appropriate to attend a meeting if one is presenting one’s own paper to it. This poses a numerical threat to the meeting – if people only attend if they will be presenting a paper, then we will have at most 16 participants, when a slightly higher number would be desirable for better engagement. Usually a few additional people attend papers who might be co-authors on presented papers or who might be local to the venue and thus easily able to attend, but an increasingly common refrain is that a potential participant could only get funding to attend the meeting if their paper is presented. Still, from this perspective, if you’re reading this note and thinking about the seminar, please do consider coming along as a delegate one year even if you don’t particularly have a paper presentation in mind.

 

Likewise, the small scale of the meeting means that it is usually run on something of a shoestring budget. Usually the seminar is a self-funded activity, with costs kept to a minimum, e.g. with a small conference fee and asking delegates simply to pay the cost-price of any accommodation arranged etc. We should highlight here that at several recent meetings we have been very grateful to part-sponsorship by the institutions hosting the meeting, who have kindly provided us with seminar or lecture rooms for free or at greatly reduced prices (this is true of the Department of Sociology, University of Cambridge; the Department of Sociology, University of Utrecht; Cardiff School of Social Sciences, Cardiff University; and the School of Applied social Science, University of Stirling). Occasionally the seminar organisers have also secured money for catering and/or travel bursaries from research grants, as happens this year when the 2013 meeting is part-sponsored by the ESRC’s funding of the ‘Is Britain Pulling Apart’ project. One outcome this year will be the rare innovation of free sandwiches for participants on, not just one, but both days. Perhaps that might swing the balance to attract you to the 2013 meeting!

 

Anyway, if you attend in 2013 in Cambridge, or indeed another year in the future of this seminar group, I’ll look forward to seeing you there!

 

Paul Lambert

 

Posted in Uncategorized | 1 Comment

Class and social distance as markers of stratification

We are currently in a boom time for public discussion of social class, largely thanks to the BBC Great British Class Survey, which received over 7 million views on the BBC website and was the most shared world news story on the New York Times website. Academically, it has also started numerous blogs (such as those by Colin Mills, Danny Dorling and ourselves), a feature on Radio 4’s Thinking Allowed and a meeting at the RSS.

The public understanding of social class is something which I always find interesting. I recently came across a section on the Guardian’s website asking ‘Are you Middle Class?’, which invited an open discussion of what constitutes ‘middle class’, which slightly predates the discussion from the BBC Class survey. Whilst the comments section of newspaper websites aren’t the best place for sensible, reasoned discussion, nor necessarily representative of wider public perception, I was interested to look at how the discussion flowed.

Somewhat inevitably, much of the discussion focussed on entangling socio-economic, socio-cultural and socio-political constructs, believing if they experienced social mobility, or voted differently to their peers, they were devoid of a social class. Occupations were rarely mentioned as a demarcation of class position. Indeed, most of the discussion focused on an ‘us and them’ stance, demarking the boundaries by characteristics they generally shared with people like themselves or distanced them from those they viewed as other.

For instance, the middle classes were often portrayed as eating cous cous, humus and vegetable boxes, sipping lattes and shopping at Marks and Spencers, IKEA, Waitrose or farmer’s markets. The working classes, meanwhile, are portrayed as eating beans on toast, burgers or other fried foods, often from fast food joints, and shopping at British Home Stores and Ebay. These differences were perhaps best summed up by the post asking if class can be measured if you eat both Waitrose muesli and Lidl cornflakes.

Other divisions between the middle and working classes included fashion (birkenstocks versus jogging trousers), newspaper readership (Guardian versus tabloids), car ownership (4×4 and Volvo versus white vans and Fords), accents and cultural identifiers such as frequenting bookmakers, having tattoos or littering. Many of the comments framed such dichotomies not just as ‘us /them’, but also as ‘acceptable/unpleasant’, defining their social class in terms of their difference from those they disliked. This is consistent to comments I’ve seen on similar websites and also the views Owen Jones quotes from Daily Mail readers.

It seems what was discussed wasn’t so much social class as social distance. Class and distance are two different, if related, concepts. Class refers to aggregating individuals into classifications (classes) based upon their socio-economic position. Social distance refers to the creation of social clusters based upon multidimensional forms of homophily and dissimilarity, cutting across socio-economic, socio-cultural and socio-political concerns.

We can think about the difference between class and social distance by thinking about music, particularly the way that musical tastes aren’t strictly bound by genre. Belle and Sebastian and the Fratellis are both Glaswegian Indie bands. Adapting a methodology recently suggested by David Beer and Mark Taylor, Spotify allows us to see the 20 artists listeners of those acts also most commonly listen to. What is interesting is that whilst the two bands share a genre, they do not share any common acts amongst the top 20 recommendations. The suggested artists themselves aren’t restricted to Indie (or Scottish) bands, but show links to common genres but different bands. Based around Scottish indie, US rock and Britpop, for instance, we see that Belle and Sebastian fans listen to Camera Obscura, Velvet Underground and Pulp, whereas Fratellis fans enjoy Franz Ferdinand, Wheatus and Toploader. It’s not the genre which is driving musical tastes but some multifaceted aspects of appreciation.
This is similar to how the Guardian readers view social class – it’s not so much the sharing of social positions and circumstances (social class) which is perceived but rather some multifaceted constructs of social connectivity (social distance). Whilst it is easy to be disappointed by the public understanding of social class, and the mixed ways it can be used in the popular press, it’s perhaps refreshing to view it as a mislabelling of a sociological construct and assume that when people present uninformed views on social class they are actually making valid comments about social distance.

Large-scale social surveys enable us to explore the various facets of social distance and piece together their interconnectivity. Such research could help us understand more fully the relationship between class and social distance, and ascertain how far the public misunderstanding of class is away from the sociological meaning.

Dave Griffiths, 7 June 2013

Posted in Uncategorized | Leave a comment

Review of ‘Is Britain Pulling Apart?’ workshop

Last week we staged the first dissemination event for our ‘Is Britain Pulling Apart?’ project. Held at the Royal Statistical Society, the event comprised presentations from the project team, a thought-provoking lecture from Danny Dorling and a series of PechaKucha presentations from researchers in the field.

Paul Lambert (Univ. Stirling) began the event, discussing ‘Social Relations, Social Connections and Social Distance’. Lambert defined the key concepts within the project as being: social relations – the character of a tie between two people or groups; social connections – the measurement of social ties; and social distance – the social structure that is revealed through analysing those ties. These areas have been extensively analysed in the UK in relation to occupational categories, but less so across other domains. A thorough exploration of their usage in occupational research was presented, providing much thought for other spheres which can be analysed from methods such as association modelling, social interaction distance analysis and social network analysis.

Richard Zijdeman (Utrecht Univ. and International Institute of Social History) talked about ‘Social Relations in the past: Using historical data to test theories’. He argued that researchers should be motivated to make stronger use of historical analysis, with the wealth of data sources and data linkage projects opening new avenues for exploring longer term trends. Zijdeman argued that data going back to the 1970’s isn’t showing historical change over time, which requires much longer analysis. He argued that much of the social sciences is, essentially, trying to answer questions which we solved in the past, such as female labour market participation. He concluded that emerging sources and methods means ‘historical data are currently up to a level where they can contribute to contemporary debates in sociology and perhaps even will make us rethink the research questions we are asking’.

Danny Dorling (Univ. Sheffield) was invited to give a guest lecture, on ‘Social Relations in Britain in 2025’. Dorling provided a compelling study of social mobility in the UK, providing a nice preview of his recent New Statesman article ‘How Social Mobility Got Stuck’. Dorling’s central argument was that increasing our understanding of the top end of society can help inform our knowledge of how society works. He argued that whilst the top 1% are often viewed as a coherent group there is actually greater differences between them than exist within the other 99%. Dorling talked of the increasing proportion of income the top 1% are earning, arguing those countries which can control how much the top 1% earn are usually more organised overall, as demonstrated by different social indicators. Whilst the rich are getting richer, the poorer were portrayed as getting poorer, with unemployment having a different meaning today than in the 1980s, the return of food banks and the rising numbers of people without sufficient savings to repair their boilers. Despite these social issues, the economy is rarely seen as more important than immigration as a political issue. Increasing integration of ethnic groups can be seen, but Dorling called for more assertive mating and social mixing between economic groups as there is increasing division between the rich and poor.

During the lively questions section Dorling suggested that all people believe they are underpaid, chiefly as they only think about those earning more than they are and ignoring how the poor are being left behind. He also argued that rising tuition fees, combined with educational expansion rising faster than the number of traditional graduate jobs, will generate a cohort of highly educated, under-employed and highly annoyed individuals, questioning whether this injustice will bring about change or whether younger people are more tolerant of inequalities and income differences as they’ve grown up in a society where it is the norm. He argued higher property taxes would be an effective method of redistribution as the majority of the global superrich own property in the UK and assortative mating means many dual high-income households are creating a divide in economic resources. He summed up his comments by suggesting that we currently rightly focus much attention on poverty, but we should also think more about the wealthy and what they show about our society.

After lunch, Mark Tranmer (Univ. Manchester) talked about ‘Modelling Variations in Social Connections’, focusing on the relationships between multilevel modelling and social network analysis. He argued that egonets can provide rich data for analysing, amongst other things, homophily in terms of which types of people connect to which others. However, Tranmer argued that social network dependencies are an important consideration as people named as a friend by multiple egos, for instance, could unduly influence results. He proposed using Multiple Membership Multiple Classification (MMMC) Models to overcome this problem, presenting simulation data demonstrating where variance in outcomes could be attributed. Tranmer identified further progression of this work to explore measuring the chances of connections occurring which were not present within the data (i.e., basing probability of a tie on all possible ties, not just on all observed ties).

Dave Griffiths (Univ. Stirling) talked of ‘Network analysis of social distance’, arguing that social surveys have not been fully utilised as sources of social network data. Two approaches were discussed. Firstly, using household panel surveys, such as Understanding Society, to create linkages across households and connect individuals to other respondents they have a shared social connection with outside their household. Secondly, connecting social groups through overly-common social interactions between members was shown to uncover elements of social structure. An example was given of sports participation, with an apparent distance between those playing team sports and those partaking in more individualistic pursuits. Griffiths showed how cross-comparisons with political attitudes demonstrated how dimensions of homophily can be measured and assessed through networks of aggregated connections.

There followed a number of PechaKucha style presentations from researchers working, in various ways, in the areas of social connections, social distance and stratification. PechaKucha presentations involve 20 slides, each displayed for 20 seconds, enforcing a fast-pace to the discussions. Despite being a relatively new, uncommon and potentially difficult presentation format, it worked very well for providing other viewpoints on methods, data and questions which can be undertaken in the area. Therefore, we thank the speakers who all excelled at the challenges of such presentations.

Liana Fox (SOFI, Stockholm Univ.) presented on ‘Re-evaluating Historical Poverty Trends in the United States’, arguing that poverty measures need to be updated to take account of changes in family units, breakdown of expenditure and income composition. She showed, for instance, that the number of traditional nuclear families has halved since the measures were developed with no account given to cohabiting partners. Analysis showed contemporary derived measurements produced broadly similar overall levels of household poverty, but with discrepancies appearing where not necessarily expected. The poverty of older people was particularly shown to be misrepresented by existing measures.

Marina Shapira (Univ. Stirling) discussed ‘Do you speak your grandma’s language: intergenerational communication and attainment of immigrant children in Western countries of immigration’. Shapira argued that European countries are increasingly likely to be defined as an ‘immigration country’, with increasing numbers of migrants in the working age population. Studies have often shown that first generation migrants perform better than second generation migrants. Shapira tested whether those second generation migrants who speak their host language at home perform worse than those speaking another language as this could create a barrier in communication with wider family members, especially grandparents. Language patterns were inconsistent overall, but there was evidence that the presence of grandparents in the home had a positive impact.

Raj Patel (Univ. Essex) presented on the potential of using Understanding Society for studying social connections. He provided an overview of Understanding Society and its design structure, as well as giving indicators of the vast size of the dataset. For instance, in Wave 2 there were over 74,000 cases including 1,147 Bangladeshis, 1,584 people who had seen ballot in the previous year and 1,077 who worked from home. Patel talked about the various modules which will appear in the survey, including questions about best friends, social networking sites, socialising and local neighbours. Homophily came through as a strong theme with questions about proportions of friends of the same age, ethnicity, educational level, geographical space and income. Given there was much discussion during the day regarding changing household composition, the question about what proportion of friends are family members appears very intriguing.

Ebru Soytemel (Univ. Oxford) talked about ‘The social space of gentrification: An exploratory analysis of neighbourhood belonging and social distance with Multiple Correspondence Analysis’. Her work focused on the Golden Horn neighbourhoods of Istanbul, an area containing traditional working class families, upper class gentrification and also Kurdish forced migrants. Soytemel argued this was an area with three social groups living separately together. Even within the same streets people didn’t mix with those from the other side of the road. The gentrified classes were shown to have fewer relatives living in the neighbourhoods and much less engagement with the area than those with stronger and deeper-rooted ties in the area. Soytemel argued that multiple correspondence analysis facilitated looking across occupational and income definitions to thinking about the geography of class relations. She argued that gentrification both created new boundaries of social polarisation and eroded social cohesion.

Gemma Catney (Univ. Liverpool) presented ‘Neighbourhood social relations: How has ethnic group segregation changed in England and Wales, 2001-2011?’. She argued that integration is an important policy issue and there is much debate about diversity, although the UK is actually segregated more by economic and occupational concerns than ethnicity. Mixed ethnicity nearly doubled between 2001 and 2011 and there was a growing likelihood of multi-ethnic couples. Segregation was shown to be declining for all ethnic groups. Catney critiqued the media argument of ‘white flight’ from urban areas showing the trend across all ethnic groups of a reduction in their share living in Inner London and a movement towards smaller urban and rural-urban areas.

Pierre Walthery (Univ. Manchester) talked about ‘Making the most of what we have: triangulating family composition in longitudinal studies’. This was based on his work with e-Stat, looking at the disparity of measures in longitudinal studies. He presented a pilot study looking at measuring changes in household composition from the British Household Panel Survey and the Millennium Cohort Study, presenting issues of variables having different coding and uneven sample sizes. Analysing the changes to household composition for child he presented a sensitivity analysis which demonstrated the harmonisation process across the surveys appeared stable. Walthery argued this showed what is possible in the area and that he was looking for further data and indicators to test. The Stata do and data files facilitating these mergers are openly available.

Those six PechaKucha presentations lasted less than 45 minutes, providing much to think about, in terms of data, methods, results and analysis, at very fast pace.

After a short break, Vernon Gayle (Univ. Edinburgh) concluded the event by summing up thoughts on ‘Social Connections’, presented through a variety of personalised tales and depictions of social trends. Gayle composed a sailing metaphor for social life, stating that tides are easy to predict long into the future, but wind patterns are almost impossible to accurately predict and suggesting social research needs to take into account what is likely and what is mere prediction. He gave examples of the 29 bus to Wood Green following the same route for generations, but an unlikelihood of babies to be born within the sounds of Bow Bells anymore. Gayle discussed the large increase in divorce rates, changes to marital status and its implications for employment status and household composition (factors in many empirical studies). He also questioned how we perceive social positions, asking if two nurses are seen in the same situation if one is married to a consultant and the other to a porter and, indeed, if our attitudes change whether it is male or female nurses being discussed. This talk was designed to generate discussion amongst participants and raised questions about the incorporation of migration into studies of social connections and also the importance of controlling for geography.

A number of themes seem to emerge from the various presentations and discussions. This included: the necessity to incorporate analysis of the changing migration and ethnic patterns in the UK; the implications of changes to marriage, cohabitation and household composition; and the requirement to understand the difference between short-term and long-term changes in social trends. Techniques used multiple correspondence analysis, association modelling, social network analysis and multilevel modelling, often making adjustments to suit the study of social connections. A vast array of datasets where discussed, including Understanding Society, the census and historical data such as marriage records. This provides much thought and breadth to considerations of analysing contemporary and longitudinal patterns of social connections and social distance.

Dave Griffiths, 24 May 2013

Posted in Uncategorized | Leave a comment