The complexities of using survey data to understand the influence of unions

Author: Rhys Davies, Alex Bryson, Samuel Jones
Institution: University of Cardiff, UCL
Type of case study: Research

About the research

The downward trend in trade union membership in the UK and the USA is well recognised and has been the subject of continued debate amongst trade unionists and labour researchers for some time. 

Based upon membership returns submitted annually by individual trade unions to the Certification Office, trade union membership within the UK peaked in 1979 at approximately 13.2 million.  Estimates published by the UK Government based upon the Labour Force Survey places the current number of union members within Great Britain at approximately 6.7 million.  Current estimates suggest that approximately 1 in 4 employees are a member of trade unions (BEIS 2018).  However, persistent geographical variations remain.  Union density in England ranges from 18% in London and the South East to approximately 28% across the regions of Northern England.  Among the devolved nations, density is estimated to be 28% in Scotland, 31% in Wales and 35% in Northern Ireland.

In the context of a body of literature that demonstrates the wide variety of benefits associated with union membership among workers, it is important to understand the reasons that underpin such differences in union density between those living in different parts of the UK.   Earlier research conducted by the team had already examined union membership in Wales using a range of data sources.  The research team wanted to extend this to examine spatial variations in trade union membership in the UK for more detailed geographical areas and for different aspects of trade union membership:

  • Union density: The percentage of those in employment who are a trade union member.
  • Union presence: Whether or not a trade union or staff association is present within a workplace.
  • Union coverage: Whether the pay and conditions of employees are agreed in negotiations between the employer and a trade union.

In undertaking this work, analysis demonstrated the inaccuracies associated with official statistics produced by Department for Business, Energy and Industrial Strategy, which led to the department producing revised official statistics.


Research funding and partners

This research was  supported by the Wales Institute of Social & Economic Research, Data & Methods (WISERD) under its Civil Society: Changing perspectives on Civic Stratification and Civil Repair strand funded by the Economic and Social Research Council (ESRC).  The research formed part of a wider project entitled Trade Union Membership, Associational Life and Wellbeing; a mixed methods and interdisciplinary study of the determinants of trade union membership.  WISERD’s work on union membership will continue to be supported via its new ESRC funded research programme: Civil Society: Changing perspectives on Civic Stratification and Civil Repair.

Methodology

The researchers’ current work has the explicit aim of unpicking reasons behind geographical variations in trade union membership in the UK. 

Their analysis draws upon data contained within

The research team gave particular attention to the measurement of trade union membership within these datasets for the purposes of comparative analyses. 

Membership has been examined not just in relation to union density, but also union presence (% people employed in workplaces where unions are present) and union coverage (% of people whose pay and conditions are agreed in negotiations between the employer and a trade union). 

Questions used to measure Trade Union Membership, Presence and Coverage

The researchers found that using data from the four different surveys highlighted immediate differences between the questions asked and thus what the data showed.

Within the UK, the official source of data used in the measurement of trade union membership is the Labour Force Survey (LFS).  A question on trade union membership was introduced into the LFS in 1989 and has been asked in the fourth quarter (Q4) every year since 1992.Questions on trade union presence and recognition were added in 1993, and a question on collective agreements was introduced in 1996.The union questions were revised substantially from 1999 affecting the consistency of time‐series data for trade union presence and collective agreements.

While all surveys used the same question regarding trade union membership (‘Are you a member of a trade union or staff association?’), the surveys varied in whether and how they asked about presence and coverage. The order of questioning and cases where certain questions were only asked depending on the answers to previous questions also muddied the waters.

Within the LFS, in order to establish levels of union presence those who are not union members are asked if others in their workplace are members of trade unions.  There are various issues with how trade union membership is addressed in the LFS that can cause particular problems for the estimation of union presence.

  1. Most significantly, approximately one third of responses to the LFS are made by a proxy respondent. Questions regarding the quality of data when provided by proxy respondents, who are typically partners of the target respondent, have existed for some time. It may be particularly difficult for proxy respondents to be aware of the union membership status of others at the target respondent’s workplace.  
  2. Under GDPR, trade union membership is considered to be one of several Special Categories of data (alongside race, political opinions, religious beliefs, health, sex life and sexual orientation). Trade union membership may therefore be regarded as a sensitive and personal issue which could have an impact on how the question is answered, especially if by a proxy respondent and consequently under-reported.
  3. Home workers: the LFS does not ask any questions about union presence among people who usually work from home or at a site on the same grounds as their home (it is likely that at least some of these workers also have an office they work in from time to time). The Workplace Employment Relations Survey (WERS) and Skills and Employment Survey (SES) do not exclude homeworkers from questions regarding union presence.
  4. Approximately 15% of the respondents to the LFS either
  1. did not give a response for the union membership question and are thus not asked about union presence question
  2. or they did not give a response to the union presence question.

Generally, in statistical analysis, missing data such as homeworkers/non-responders would not be included as a negative response. However, official estimates of union presence treated these groups as having answered as if there was no union present in their workplace.

This consequently skewed the results in the official statistics.

Once the researchers excluded these two groups from calculations, estimates of union presence rose by approximately 7-8 percentage points, bringing the LFS figures more in line with those derived from other sources.

Official statistics produced by the Department for Business, Energy and Industrial Strategy (BEIS) also provide an estimate of the coverage of collective agreements derived from the LFS. Collective agreement coverage was historically defined as the proportion of employees in the labour force whose pay and conditions were ‘agreed in negotiations between the employer and a trade union’. 

Assessing ‘collective agreement’ in the available data is also problematic as the LFS uses different wording for the relevant question compared to the other surveys, asking whether pay and conditions are ‘directly affected‘ by employer / trade union agreements. The focus of the similar question in other surveys is on whether unions are recognised by the management for pay negotiations, although they also ask respondents to asses the effectiveness of unions in these negotiations.

The researchers noted that official measures based up on the LFS gave a figure of approximately 30% of workplaces where trade unions affect pay and conditions.  However, when data from the SES and BHPS/USoc were used they found that around 50% of workplaces recognised unions in pay and conditions negotiations.

Furthermore, the researchers analysed WERS data (the survey also asks about recognition of trade unions in pay and conditions negotiations), they discovered a differential of between 16 and 20 percentage points compared the LFS-derived estimates.

BEIS itself has also explored problems of under-reporting of trade union coverage within official statistics. The department has investigated why estimates are lower in statistics derived from the LFS (31% in 2011) than when estimates are derived from the Annual Survey of Hours and Earnings (ASHE – 47% in 2011).

ASHE is different to the LFS in being derived from questions asked directly of employers. The relevant question in ASHE is ‘Was the employee’s pay set with reference to an agreement affecting more than one employee (for example, pay may be agreed collectively by a trade union or workers’ committee)?’ which allows for a wider definition of pay agreements (neither ‘collective agreement’ or ‘trade unions’ are specifically mentioned in the question).

Data used from the UK Data Service collection

Labour Force Survey

Workplace Employment Relations Survey

Skills and Employment Survey

British Household Panel Survey

Understanding Society

Messages

The research raised concerns regarding the quality of official UK statistics created by the Department for Business, Energy and Industrial Strategy (BEIS) in relation to estimates of union presence and union coverage. These statistics are used by the UK Government, the CBI, the Trade Union Movement and international organisations such as the OECD and ILO

Having assessed the evidence, BEIS agreed that estimates of union presence were being calculated incorrectly and that revisions would be made to future publications of trade union membership.  The meaning of union coverage was also clarified within these publications.    

The research also produced an interactive UnionMaps website which maps trade union density, presence and coverage for small geographical areas across the UK.

Findings

Trade union presence

Analysis of Labour Force Survey (LFS) data conducted during his research raised concern regarding the quality of official statistics related to trade union presence.  

Respondents to the LFS who do not provide a valid response to the question on trade union presence were still included in the population of employees upon which official estimates were based. The estimates therefore treat them as if they had reported in the LFS that no other employees at their workplace were members of trade unions or staff associations.

Excluding these groups from calculations of union presence (as is normal practice with the treatment of missing data), would increase estimates of trade union presence derived from the LFS by approximately 7-8 percentage points and would be consistent with estimates derived from other sources, roughly equivalent to 2 million employees.

 

Trade union coverage

The researchers discovered a significant difference between BEIS estimates of trade union coverage derived from the Labour Force Survey and other data sources considered in their work (SES, BHPS/USoc and WERS).

They observed that the relevant questions in SES, BHPS/USoc and WERS asked whether workplaces recognised unions in pay and conditions negotiations, whereas in the LFS the question refers to pay being directly affected by trade unions.

Their analysis suggests that it is the wording of the question in the LFS which largely contributes to under-estimates of trade union coverage when figures are derived from the LFS. 

 

Supporting a better understanding of trade union membership

The researchers also produced a statistical compendia which provides the most comprehensive statistical portrait of regional variations in trade union membership, presence and coverage in the UK undertaken to date.  The analysis explores both trends in membership and how membership varies among different groups within the population and by selected job and workplace characteristics. 

The researchers also produced the UnionMaps website which presents estimates of trade union membership for across Unitary Authority and Local Authority Districts of Great Britain. The estimates provide on the maps cover the three main measures of union membership, i.e.:

  • Trade union density
  • Trade union presence
  • Trade union coverage


Example UnionMaps search for Camarthen, Wales

Findings for policy

The Department for Business, Energy and Industrial Strategy (BEIS) is responsible for the production of statistics regarding the membership of trade unions within the UK.

Trade Union Membership (TUM) was first published in its current form in 2004 and has been published each year since.

The majority of the statistics are compiled from data provided from the Labour Force Survey (LFS), run by the Office for National Statistics (ONS). Trade union membership statistics are presented by:

  • occupation
  • industry
  • descriptors such as age, sex, ethnicity and geography.

Statistics are given on trends in membership numbers, trade union presence within a workplace, and collective agreement coverage.

The statistics are used by policy colleagues for drafting advice about trade union membership to Ministers, and in communication with the International Labour Organization (ILO).

The Confederation of British Industry (CBI) uses the statistics for research, policy development and briefing members about the state of union membership in GB.

The Trades Union Congress (TUC) uses the statistics to identify areas of low membership, for example, in particular regions and industries.

The data are also re-presented by OECD on its labour statistics pages, which provide comparable statistics across the member countries. These open data are also available (alongside other international macrodata) from the UK Data Service.

As a result of this research, the researchers submitted their evidence to the UK Statistics Authority with a recommendation that BEIS reconsider how they calculate the TUM statistics.

Trade Union Presence:

Those respondents to the LFS who do not provide a valid response to the question on trade union presence are still included in the population of employees upon which estimates are based.  They are therefore treated as if they would have reported in the LFS that no other employees at their workplace were members of trade unions or staff associations.  If these groups were excluded from calculations of union presence (as is normal practice with the treatment of missing data), estimates of trade union presence derived from the LFS would increase by approximately 7-8 percentage points and would be consistent with estimates derived from other sources.

The researchers also recommended that figures on trade union coverage derived from LFS be more clearly explained.

Trade Union Coverage: 

Official measures of trade union coverage based upon the LFS are estimated to be approximately 40% lower than those derived from other sources, including the ONS Annual Survey of Hours and Earnings.  This problem appears to stem from the wording of the LFS union coverage question. Whilst other labour market surveys generally ask respondents whether or not trade unions are recognised by management in the negotiation of pay, the LFS asks respondents for their assessment of whether trade unions or staff associations directly affect pay. Comparisons suggest that emphasis in the LFS upon pay being ‘directly affected’ contributes to the estimation of relatively low levels of trade union coverage among employees from this source.  There are inconsistencies in the way that BEIS portrays what is actually meant by union coverage as derived from the LFS in the publication of its official statistics.  

Impact

Following an assessment of the evidence, David Fry, Chief Statistician at BEIS acknowledged the researchers’ analysis, issuing the following responses:

On Trade Union Presence, we have investigated the calculation that has been used in the publication to derive this statistic, and have concluded that, as you say, the statistic does include in the population individuals who do not provide an answer to the question on whether there are people in the respondent’s place of work that are members of a trade union (TUPRES).

Having assessed these data, on balance we agree that this should be revised to exclude those who do provide an answer of ‘no’ to the question on trade union membership (UNION), but don’t provide an answer to TUPRES.

We will make this change and present a revised series in the next publication of TUMS, due towards the end of May.

[On Trade Union Coverage] we acknowledge that we could be clearer in the statistical bulletin about what is actually being measured by this statistic, and will revise our presentation of it to make clearer that we are talking specifically about agreements between the employer and the trade union/staff association directly affecting pay and conditions.

Letter from David Fry, Chief Statistician, Department of Business, Energy and Industrial Strategy dated 12th January 2018

 

Revised trade union membership statistics were published in May 2018.  The following statement was included within the Technical Annex (p51).  

Prompted by correspondence with Rhys Davies of the Wales Institute for Social and Economic Research at Cardiff University, BEIS reconsidered the calculations used to estimate union presence in the workplace from the Labour Force Survey data.

After examining the data, on balance BEIS decided that it would be more appropriate to exclude those who did not provide a valid response to the TUPRES question from the population used to estimate union presence.

The new method excludes this group from the estimates of union presence.

Trade Union Membership 2017; Department for Business, Energy and Industrial Strategy

 

The effect of the change has been to increase the rate of union presence by 8 percentage points, roughly equivalent to 2 million employees, a result which has stimulated debate within the trade union community.  In terms of the discussion and presentation of analysis for union coverage, explanatory text has been revised within statistical releases produced by BEIS to explicitly state that the figures for union coverage refer to those ‘whose pay was affected by a collective agreement’.

The statistical compendia was drawn upon heavily by the Welsh Centre for Public Policy in its recent report on The Value of Trade Unions in Wales.  

 

Online coverage

UK Government Forced To Revise Union Membership Figures After 20 Years Of Miscalculation. Campaign for Trade Union Freedom, May 2018

Trade Union Membership is Growing, but there is still work to do, blog post by Carl Roper (National Organiser, TUC) May 2018

 

In addition to a blog post and WISERD Research Note on this issue, results of the research also formed the basis of a conference paper delivered at the UK Data Service’s Labour Force Survey User Conference on 25 November 2016. The user conference brings data producers and data users (academics, postgraduate students, professional practitioners) to showcase research that is being carried out using LFS and Annual Population Survey data.

 

Publications

Beynon, H., Davies, R. and Davies, S. (2012), Sources of variation in trade union membership across the UK: the case of Wales, Industrial Relations Journal 43(3) pp 200-221 doi:10.1111/j.1468-2338.2012.00676.x

Davies R. (2016) Measuring Trade Union Membership: Harder than it might seem? WISERD Blog

Davies R. (2016) Making Sense of Official Estimates of Trade Union Membership, WISERD Research Note

Bryson A. and Davies R. (2018) Family, Place and the Intergenerational Transmission of Union Membership, British Journal of Industrial Relations doi: 10.1111/bjir.12435

Bryson A and Davies R. (2019) Accounting for geographical variance in the union satisfaction gap. Industrial Relations Journal. doi: 10.1111/irj.12243

Davies R., Bryson A and Jones S. (2019) Geographical Variations in Trade Union Membership, Cardiff: Wales Institute of Social & Economic Research, Data & Methods (WISERD), Cardiff University. 

This site uses cookies

Some of these cookies are essential, while others help us to improve your experience by providing insights into how the site is being used.

For more detailed information please check our Cookie notice


Necessary cookies

Necessary cookies enable core functionality. This website cannot function properly without these cookies.


Cookies that measure website use

If you provide permission, we will use Google Analytics to measure how you use the website so we can improve it based on our understanding of user needs. Google Analytics sets cookies that store anonymised information about how you got to the site, the pages you visit, how long you spend on each page and what you click on while you’re visiting the site.