The impacts of racial and religious diversity on union participation in the United Kingdom. Adam Azraei bin Nadzri Abstract: The purpose of this paper is to investigate the correlation between diversity and union participation in the United Kingdom by using the Quarterly Labour Force survey from the Office of National Statistics. The measure of diversity used in this paper were race, religion and religiosity. Race, religion and religiosity had positive effect on union participation at an individual level but only religiosity had a negative effect in the regional level. There is also a U-shape graph when observing union participation and atheism, proposing that diversity should not be measured in binary or categorical terms but in ordinal terms with relation to others. In conclusion, there is a statistically significant relationship between diversity and union participation. Introduction The purpose of this paper is to understand the impacts of diversity on union participation in the United Kingdom. According to the Calmford-Driffils hypothesis, a fully fragmented labour market or a fully centralized wage setting environment similar to the Scandinavian Corporatist model would result in stable macroeconomic labour structures that ensure low unemployment (Layard et. Al., 2005). Ever since Thatcher’s union busting legislation took its toll on union membership, union density as a percentage of the workforce in the UK has been progressively decreasing. Visser brought in arguments such as macroeconomic conditions as the reason for the union membership decrease (Visser, 2002), while Toubøl and Jensen brought in the idea that workplace union density acts as pressure for people to join unions (Toubøl and Jensen, 2014). The purpose of this paper is to investigate whether unions are dependent on issues pertaining to race. Schnabel and Wagner had analysed that union membership is strongly related to personal characteristics (Schnabel, 2002; Schnabel and Wagner, 2007) According to the Office for National Statistics, union participation as a percentage of the labour force has been declining since 1995, as can be seen in Figure 1. In the recent 2015 report by the ONS, one of the key findings was that non-UK nationals were less likely to participate in trade unions (ONS, 2016). This is an interesting development as it brings up the question of whether an increase in diversity brought through immigration results in a negative impact on union participation, reducing the bargaining power of unions, opening up implications towards inequality and the wage setting in workplaces that are influenced by collective action. Figure 1 Olson described union participation and retention using social custom theory in which members encourage union membership through negative reinforcement and punishment (Olson, 1965). Workers are compelled to join unions as to comply with norms in the fear of being ostracised in their community. However, Booth and Chatterji (1993) has shown that there is a minimum level of unionisation in which the reputation effect will not work. There are two hypotheses that could explain the impacts of diversity in reducing the reputational effect. Firstly, a more diverse environment would create newer norms in which the community could not punish collectively. Thus, reducing the cost of non-membership and reducing union participation. Secondly, an increase in diversity would mean that the share of unionised members from the total workforce has been diluted by an increase in immigrants who are not unionised. This is a likely hypothesis as, as can be seen from Figure 1, union membership has remained at a constant level as the workforce continues to grow in the UK. Description of variables and data sets I have used the data sets from the UK Quarterly Labour Force Survey from 2002 to 2014. To measure union participation, I have used the response to the question ‘Are you a member of a trade union or staff association?’. As I am not interested in labour participation in general and am only interested in the proportion of unionisation in the labour force, I have removed cases in which the question does not apply to the respondents and missing cases. To measure diversity, I have used three different measures which are race, religion and religiosity. The proxies that I chose are proportion of the majority race, proportion of majority religion and proportion of atheism. Further calculation method is described in Table 1. The reason that I am observing just the majority race and not segmenting the nationalities is to test the effects of physical diversity on union participation. The reason for testing the majority religion was to see if there is a correlation of identifying with a common religion with identifying with a union. Lastly, the reason for using atheism is to measure religiosity. The assumption is that being an atheist acts as a dummy proxy for not participating in a religious community, a more probable and consistent measure of low religiosity. I am interested in two levels which are the individual level and the regional level. At an individual level, I used a Probit model to determine the likelihood of being unionised given the characteristics of the respondent. Table 1 describes all the variables that I have taken account in the Probit model. The Probit model that I am estimating is given in the equation below: Pr (UnionParticipation = 1|Race, Religion, Atheist, Public, Manual, Qualifications) = 𝐺(𝛽0 + 𝛽1 Race + 𝛽2 Religion + 𝛽3 Atheist + 𝛽4 Public + 𝛽5 Manual + 𝛽6 Qualifications + 𝑢) (1) At the aggregate level, I am interested in the effects of the environment on union participation. By using different regions of the United Kingdom outlined by the Labour Force Survey, I have created a Fixed Effects model to understand the relationship between union participation against the various measures of diversity. Table 1 describes the variables that I have taken into account. Below shows the equation that I am modelling: 𝛥UnionRateit = 𝛽1 𝛥Diversity + 𝛽4 𝛥PublicPercent it + 𝛽5 𝛥ManualPercent it + 𝛽5 𝛥QualificationsPercent it + 𝑢it (2) ‘Diversity’ covers three different measures of diversity which are: proportion of majority race, proportion of majority religion and proportion of atheism. These measures could not be regressed simultaneously due to endogeneity between the other regressors. Although including them would hold those characteristics constant, we would need to understand how one measure of diversity affects the others in order to get the full effect of the measure of diversity. However, one would argue to just regress it all together but it would be inaccurate since the variations in race and religion are not enough to do such regression and would render the effect insignificant when in fact there is a correlation between an increase in diversity with the union participation rate. It would be more sensible to regress individually and just acknowledge the biasness that arises when including other forms of measures of diversity until we would obtain a larger sample with more variations in the data. A sensible decision would be to compare the R2 of each regression and choose the measure with the highest R2 as the instrument to measure diversity. TABLE 1 Variables and Measurement Variable Label Measurment (values) Union Race Religion Atheist Public Manual Qualifications yes (1); no (0) white (1); non-white (0) christian (1); non-christian (0) yes (1); no (0) yes (1); no (0) yes (1); no (0) yes (1); no (0) Union participation rate Union Rate Proportion of White Workers Race Percent Proportion of Christian Workers Christian Percent Proportion of Atheist Workers Atheist Percent Proportion of Public Sector Workers Public Percent Proportion of Manual Workers Manual Percent Proportion of Workers with Qualifications Higher than NQF Level 4 Qualifications Percent Employed members as a share of employed wage and salary earners Employed white employee as a share of employed wage and salary earners Employed Chrisitan employee as a share of employed wage and salary earners Employed Atheist employee as a share of employed wage and salary earners Employed public sector employee as a share of employed wage and salary earners Employed manual-based employee as a share of employed wage and salary earners Employed employee with Level 4 Qualifications or higher as a share of employed wage and salary earners A: Individual-level data Union participation Race Religion Atheist Public Sector Worker Manual-based Occupation Qualifications Higher than NQF Level 4 B: Regional aggregates There are three things that I have included in the regression to be held constant with union participation. One of them is the proportion of manual workers, as manual workers tend to be highly unionised. I have used the NS-SEC measure to obtain the proportion of manual workers. Another variable is proportion of public workers. Lastly, I have included education level as I presumed that higher educated individuals are less likely to be unionised. For education, I have obtained the proportion of individuals obtaining an NQF Level 4 and above qualification as being a dummy variable for having higher education. The 2015 Data Set In understanding the current demographic of the UK, I have used the 2015 Quarterly Labour Force Survey to see if there is any pattern within the UK area. In 2015, by observing different areas of the UK using the same data set, it is interesting to note that there is a Northern Bias when it comes to union participation. However, what is interesting inside the data is that atheism does not have a strong effect in decreasing union participation. Despite having higher levels of atheism in the West Midlands, union participation was still just as high in West Yorkshire and the Rest of the Northeast. There are two hypotheses that I would like to propose. Firstly, union participation is not religion specific. We might expect that being part of a union would mean the person was brought up in a community of high religiosity that promotes community engagement. But as we can see here, this is not the case. In places where measures of religiosity were low (higher atheism), the same level of union membership can be maintained. Another hypothesis that I would like to propose is that union participation may be U-shaped when measuring diversity. From the graph, high union participation is observed at the ends, which represents two type of homogeneity. As religiosity decreases, union participation may begin to decrease, but as the proportion of those areligious communities increases, a new community develops. If it is rational to be in a trade union, we would see that despite not being active in a community, the self-encouragement of people similar to you would encourage union participation within that particular group, in line with Olson’s theory of social customs. Figure 2 In the cross-sectional view of Christians and white ethnicity, not much can be said other than a reduction in the proportion of being white and Christian is correlated with lower union participation. We can use the ideas brought forward through social customs theory to explain the decrease. Another possibility is that the number of trade union members has remained constant or rather has not kept up with the increase in the labour force via market liberation and immigration. Figure 3 Figure 4 Before running the regression, an important factor to include that differs from one region to another is the effect of education and manual workers. As can be seen, there is a negative relationship between the proportion of highly-educated individuals and trade union membership. The ONS reports that the unionised workforce are less likely to be manual workers and have lower forms of qualifications. Despite what the ONS reports, I have found that regions with higher proportions of manual workers are more likely to be unionised. This disparity between the report of the ONS and my study may be a result of the historical context of those regions. Regions with a higher proportion of manual workers would historically have higher levels of community-centric behaviour despite the workers not being unionised. As I am just observing the correlations between averages and the ONS is observing individual data in joining unions, this disparity may not be contradictory at all, as there may be effects of historical norms just as proposed by Olson. Figure 5 There is no clear correlation despite being a negative trend with education. Removing outliers of Central and inner London, we can see that the effects of education is not really that clear. Figure 6 A probit regression analysis for individual decision of union participation TABLE 2 Determinants of probability of joining unions in 2015 (probit) No of observations: 35,343 Constant Diversity Variables Race Religion Atheist Background Variables Public Sector Worker Manual-based Occupation Qualifications Higher than NQF Level 4 - 1.4044 * Beta t-values Average Marginal Effect 0.07676 * 0.2064 * 0.09446 * 2.46 5.36 2.39 0.01929 0.05188 0.02374 69.78 - 0.95 9.18 0.3055 - 0.004721 0.03974 1.216 * - 0.1879 0.1581 * * significant at 0.05 level Observing at an individual level, all the measures of diversity are statistically significant. However as described in the previous section in observing atheism on the impacts of union participation, there is a positive correlation between being an atheist and being part of a union. This would provide some evidence to reject that union participation is based on religiosity. However, religion is still relevant in determining the likelihood of union participation with respondents professing Christianity being highly likely to join unions. It is interesting to note that the effects of being white on union participation are not as strong as being a Christian, which would mean physical appearance is less likely to be an impediment to union participation. This hypothesis is supported in the ONS 2015 Trade Unions Report, showing that those who identify as Black British account for the second largest proportion of union members after those who identify as White British. Despite the previous regional diagram showing a correlation between areas with high manual workers and union participation, the probit analysis of individual decision to join unions shows that there is not a strong correlation between being in a manual-based occupation and the likelihood of joining unions. We would argue for the relevance of controlling for manual-based occupation due to the fact that manualbased occupations have different effects at regional and individual levels, therefore it should still be a relevant variable to be put into the UK context. A pooled regression analysis of the British Unionization trends TABLE 3 Impacts of diversity on union density (fixed effects) No of observations: 20 x years = 280 Dependent Variable Change in density level Ethnicity Coefficients t-values Intercept Proportion of White Workers Proportion of Christian Workers Proportion of Atheist Workers Proportion of Public Sector Workers Proportion of Manual Workers Proportion of Workers with Qualifications Higher than NQF Level 4 2 - 0.0447 0.2307 * 0.5900 * 0.1449 * - 0.1874 * 0.6641 - 0.62 3.24 10.80 3.81 - 7.10 Measures of diversity Religion Religiosity Coefficients t-values Coefficients t-values 0.05926 * 2.14 0.1815 * 0.1231 * 5.58 0.4640 * 0.2045 * - 0.05466 7.97 5.64 - 1.40 0.6882 - 0.1066 * 0.4847 * 0.2081 * - 0.09839 * 10.68 - - 4.81 8.24 5.61 2.73 0.6793 * significant at 0.05 level From the results of regressing the different measures of diversity respectively, we can see that all the measures of diversity are statistically significant. At an aggregate level, we can observe that atheism has the reverse effect when regressing over time. This further complicates things as it may show that union participation is a function of not only individual rationality but also the characteristics of those around the individual. At a regional level, religion predicts better the model of diversity with union participation with an R2 of 0.6882. However, an interesting fact is that there is a change in the sign of the coefficient for the proportion of atheism. From the probit model, there is a higher probability for the respondent to join a trade union compared to if the respondent was not an atheist. Two factors can help explain this inconsistency. Firstly, it is important to understand what each regression is measuring. The probit model explains the correlation of the likelihood of being unionised against being an atheist while the fixed effects model is modelling the regional choice rather than the individual choice. From Figure 2, we can observe that in regions with higher proportions of atheism, the proportion of those professing Christianity and the proportion of white ethnicity are lower. However, from the same figure, the correlation between being in an atheist community and union participation is very near to 0. Thus, just by observing these two results, we can see that the sign change in coefficients is to be expected. One thing to note is that being a Christian captures a large portion of the effects of having better qualifications in the workforce. Again, as shown in the fixed effects model above, there is a statistically significant positive relationship between manual workers and union participation at a regional level. Conclusions In conclusion, there are negative relationships at an individual level and at a regional level in being white and Christian towards the likelihood of being in a union. However, when observing other forms of diversity such as religiosity, this relationship slowly breaks down. It is important to note that the measures of diversity that I have been using are binary. Measuring diversity would be a complex issue to address as diversity would depend not only on how one views others on a relational basis but also on how others view one as well. A mathematical solution to this would be to use an index or a vector system to measure relational individual diversity while adding coefficients to signify certain aversion or preferences in diversity. A diversity index which incorporates all those measures of diversity would also avoid collinearity when regressing all the measures of diversity together. The first step in overcoming this problem is to statistically measure diversity in an ordinal manner instead of the nominal method of measuring diversity in data collecting. Secondly, research in diversity would need more cross-discipline integration in order to make decision-making more relevant in the human context. The need to understand human decision beyond the rational utility-maximising model is pivotal to develop new laws, regulations and mechanism that would benefit society. The results I have obtained are limited to the UK. It would be interesting to observe in the European context especially in Scandinavian countries in which wage centralisation is an integral part of the economy. One more external factor that I have yet to include is the impact of market liberation. Market liberation may affect the increase in diversity in the workplace as well as the weakening of trade union power. It is very difficult to measure openness to trade at a regional level. However, we can observe the effects of market liberalisation when conducting cross-country studies in the future. However, despite of these setbacks, the results are still significant and does show some form of correlation that should be continued and explored to understand the dynamics of union participation. For the purpose of policymaking, immigration affects more than just efficiency and output. It is important to note that there are evidence to show that trade unions do help in combatting inequality in the economy (Chauvel and Schröder, 2017). There is much that is needed to be understood on the types of diversity how it impacts certain labour market structures in order for policy-making to be beneficial towards society. Bibliography Booth, A.L. and Chatterji, M., 1993. Reputation, membership and wages in an open shop trade union. Oxford Economic Papers, pp.23-41. Department for Business, Innovation & Skills, 2016. Trade Union Membership 2015. London. Chauvel, L. and Schröder, M., 2017. A Prey‐ Predator Model of Trade Union Density and Inequality in 12 Advanced Capitalisms over Long Periods. Kyklos, 70(1), pp.3-26. Fitzenberger, B., Kohn, K. and Wang, Q., 2011. The erosion of union membership in Germany: determinants, densities, decompositions. Journal of Population Economics, 24(1), pp.141-165. Ibsen, F., Høgedahl, L. and Scheuer, S., 2013. Free riders: the rise of alternative unionism in Denmark. Industrial Relations Journal, 44(5-6), pp.444-461. Layard, R.M.N. et al., 2005. Unemployment : macroeconomic performance and the labour market 2nd ed., new ed.]., Oxford: Oxford University Press. Office for National Statistics. Social Survey Division, Northern Ireland Statistics and Research Agency. Central Survey Unit. (2015). Quarterly Labour Force Survey, 2002-2015. [data collection]. UK Data Service. Retrieved from https://discover.ukdataservice.ac.uk/series/?sn=2000026 Olson, M., 1965. Logic of collective action public goods and the theory of groups Rev. ed.. Schnabel, C., 2002. Determinants of trade union membership (No. 15). Diskussionspapiere/FriedrichAlexander-Universität Erlangen-Nürnberg, Lehrstuhl für Arbeitsmarkt-und Regionalpolitik. Schnabel, C. and Wagner, J., 2007. Union density and determinants of union membership in 18 EU countries: evidence from micro data, 2002/031. Industrial Relations Journal, 38(1), pp.5-32. Toubøl, J. and Jensen, C.S., 2014. Why do people join trade unions? The impact of workplace union density on union recruitment. Transfer: European Review of Labour and Research, 20(1), pp.135-154. Visser, J., 2002. Why fewer workers join unions in Europe: A social custom explanation of membership trends. British Journal of Industrial Relations, 40(3), pp.403-430.
© Copyright 2026 Paperzz