ENFORCING AFFIRMATIVE ACTION IN NORTHERN IRELAND EMPLOYMENT: COMPARING THE EFFECTS OF AGENCY NEGOTIATION, NGO ACTIVISM, AND LITIGATION
Raya Muttaraka, Anthony F. Heathb, Heather Hamillb and Christopher McCruddenc
a
Department of Political and Social Sciences, European University Institute, Italy
Tel:+39-55-4685635, Fax:+39-055-4685201, Email:
[email protected] b
Department of Sociology, University of Oxford, UK
Email:
[email protected],
[email protected] c
Faculty of Law, University of Oxford, UK, Email:
[email protected]
Corresponding author: Raya Muttarak
1
Abstract
We investigate the effects of different methods of securing affirmative action in employment in Northern Ireland. We compare the effects of agreements negotiated between the legal enforcement agency (initially the Fair Employment Commission) and employers, with the effects on employers of litigation, and NGO pressure exerted through the MacBride campaign in the United States. We also examine whether there were differences between the voluntary and legally-enforceable agreements negotiated by the Commission, and whether the effects of agreements declined after 2000 with the establishment of the Equality Commission, in the place of the earlier FEC. We find that Commission agreements were effective in securing progress towards fair employment, both among employees as a whole and among professional and managerial employees in particular. We find a large difference between voluntary and legally-enforceable Commission agreements, with the former being more effective. We were unable to find any effects of litigation, or NGO pressure, on individual firms.
2
Introduction
Affirmative action in the labor market has been operated in a number of countries using significantly different approaches, ranging from India, where quotas have been established for the employment of members of the Scheduled Castes, Scheduled Tribes and Other Backward Classes in public sector employment, and in the United States where federal contractors are obliged to show evidence of affirmative action with respect to gender and race in order to qualify for government contracts. There are also programs in South Africa, Malaysia, Canada, Namibia, and Fiji, all with significant differences in how affirmative action is sought to be implemented. In Northern Ireland, on which this paper focuses, affirmative action has been an important feature of employment from the 1980s. Of these, the American affirmative action programs have been by far the most extensively researched. Several studies have focused on the effects of contractor status and have compared the employment growth of females and minorities in establishments that are federal contractors (and so subject to affirmative action) with that of non-contractors. (Under Executive Order 11246 employers with federal contracts are required to file reports indicating “underutilization” of women or minorities and are then obliged to address this by making corrective efforts including the use of written goals and timetables. Contractors may be sued and barred from federal contracts if they are judged not to be pursuing affirmative action.) Notable earlier studies were those by Ashenfelter and Heckman (1976); Goldstein and Smith (1976); Heckman and Wolpin (1976); Leonard (1984, 1990), Rodgers and Spriggs (1996) and Smith and Welch (1984).1
1
See Harper and Reskin 2005; Holzer 2007 and Holzer and Neumark 2000 for recent reviews.
3
While there is some disagreement about the size of the effects, there is general agreement that the United States federal contractor program as a whole has improved the proportions of African Americans and of women both in employment generally and in managerial posts specifically, although the gains have been greater in periods when the program received greater political emphasis from the federal government than in periods (such as the Reagan years) when government relied more on market forces to remedy disparities (Leonard 1990). The main tools available to enforcement officials for the United States federal contractor program were compliance reviews conducted by the Office of Federal Contract Compliance Programs (OFCCP).
Officials could survey employment practices in the
contractors selected for reviews (although the evidence suggests that reviews were not especially targeted) and request changes. They had the power to debar contractors and to require employers to provide back pay to workers or groups that had not been fairly treated. There is evidence that compliance reviews of federal contractors have been more effective than anti-discrimination lawsuits in encouraging firms to improve their share of under-represented groups (Hirsh 2009; Kalev and Dobbin 2006; see also Skaggs 2009 who shows that lawsuits have only temporary effects).
However, apart from the American
studies, there is relatively little hard evidence in other countries of the efficacy of affirmative action programs in the labor market. (See Agocs and Burr 1996, Agocs 2002 on Canada; Zwart 2000 on India; Bowmaker-Falconer 1998 on South Africa.) These studies largely review the nature of the programs rather than evaluate their effectiveness.
4
In this paper, we consider whether similar results to those found in the United States have occurred as a result of the somewhat different Northern Ireland affirmative action program. While there are some important differences in the programs (described below) there are also several similarities, especially with regard to compliance reviews and to lawsuits.
It is valuable to see how far the effects identified in American research on
affirmative action are generalizable to Northern Ireland, as this may well have implications for the wider extension of these approaches to other jurisdictions such as the European Union that is considering how to tackle issues of employment equality (for example, in the context of ethnic inequalities in the labor market). It would tend to show that these policy interventions have generalizable results and that their effectiveness is not restricted to the American context. In particular, we consider whether the two central hypotheses suggested by Kalev and Dobbin (2006) are supported by Northern Ireland experience. Their first hypothesis was that “[c]ompliance reviews, designed to alter organizational personnel routines, will have more lasting effects than lawsuits, which create incentives to avoid discrimination that may erode over time.”
Their second hypothesis was that “Compliance reviews conducted after
deregulation will have more modest effects than compliance reviews conducted earlier. This will not be the case for lawsuits, which are less sensitive to regime change” (Kalev and Dobbin 2006:865). Their evidence on changes in managerial diversity broadly supported these hypotheses. Both these hypotheses assume sociological processes of a general kind that ought in principle to apply in different legal jurisdictions and contexts, including Northern Ireland. It is undoubtedly the case that the legal and political context of Northern Ireland is very
5
different from that in the United States: in Northern Ireland the primary focus was the disadvantaged socio-economic position of Catholics whereas in the United States it was that of African Americans. But in both countries the disadvantage could be at least in part explained by the legacy of past (and present) discrimination and exclusion, and the aims of the legislation in both countries was to move towards a situation that represented fair employment. In addition, agreements made by the Fair Employment Commission (FEC) and later by its successor, the Equality Commission for Northern Ireland (ECNI), have many similarities with United States federal compliance reviews and were similarly designed to alter organizations’ personnel practices in order to facilitate the recruitment of the disadvantaged community. Similarly, in Northern Ireland individual litigation (in what are termed “industrial tribunals”) is broadly equivalent to lawsuits in the United States, and was similarly designed to enable individuals to seek redress for discrimination. However, the financial penalties tend to be much smaller in Northern Ireland. As with the U.S. federal contractor system, there were also changes in the political and legal regulatory system in Northern Ireland during the period studied. In particular, the change in the regulatory framework in 2000, although very different from the U.S. deregulation that occurred after 1980, might have been expected to result in less emphasis being given to inequalities between Protestants and Catholics, thus reducing the effectiveness of agreements negotiated by the Commission, but not of litigation. Specifically, in 2000, the FEC became the ECNI, taking over the FEC’s responsibilities regarding inequalities between Catholics and Protestants, but with a much expanded mandate than that of the FEC in other areas of inequality. The ECNI has responsibilities in the areas of gender, age, sexual orientation, race, and disability as well as that of the fair employment between the two
6
(religiously-defined) communities. One reasonable hypothesis would be that the ECNI would have to spread its resources more widely and thus more thinly, with a consequentially “lighter touch” approach to the operation of the affirmative action program.
Northern Ireland Fair Employment Legislation
There are a number of specific features of the Northern Ireland context that are unique and that we need to take into account in analyzing the data. Northern Ireland has, since 1989, had a remarkably innovative, legally based program of affirmative action. This aims to use legal enforcement measures to ensure that both communities in Northern Ireland (Catholics and Protestants) are ensured “fair participation” in employment. The affirmative action program was established by the Fair Employment (Northern Ireland) Act 1989 (amending and substantially replacing the previous 1976 fair employment legislation which essentially prohibited discrimination but which did not require any significant positive action to promote fair employment). This legislation established the Fair Employment Commission (FEC) and initially required all concerns with 25 or more employees, both public and private, to prepare and submit to the Commission annual monitoring returns reporting inter alia the proportion of Protestants, Catholics and people of “non-determined” community affiliation in the employer’s full-time workforce. (There were various procedures for minimizing the number of non-determined employees, and in most concerns the proportion was rarely above 6% for any given year.) The FEC devoted considerable resources to checking the accuracy of these monitoring returns. After considerable debate, it was decided that these returns should be published, identifying all concerns by name and showing the overall proportions of
7
their workforces that were Catholic, Protestant and non-determined (but not breaking these proportions down by gender or occupational level).2 From 1992 the requirement to prepare annual monitoring returns was extended to all concerns with 11 or more employees. This legislation was subsequently modified by the Fair Employment and Treatment (Northern Ireland) Order 1998 (FETO) while the Northern Ireland Act 1998 (Section 73) established the Equality Commission for Northern Ireland with its broader mandate. FETO also extended the monitoring requirement to part-time employees from 2001. In addition to undertaking regular monitoring of their workforces, the Northern Ireland legislation imposes on all employers, both public and private, a duty to carry out every three years regular reviews of the composition of their workforces in order to determine whether there is “fair participation”, and to undertake remedial action where fair participation has not been achieved. The program thus applies to all concerns (above a certain size) and not simply to federal contractors as under the U.S. federal contractor requirements. (Throughout this paper we use the term “concerns” to describe the relevant employers, rather than “firms” since we are dealing with public sector bodies as well as private sector firms.) Fair participation is not defined in the legislation but the FEC interpreted the concept to mean that fair participation will be present in a concern when “under-representation” of a particular community has been eliminated. Under-representation should not be taken to mean a simple numerical disparity between the composition of the workforce in the particular concern and that of the Northern Ireland labor force as a whole but depends on the availability of suitably-qualified individuals in the area from which the concern might reasonably be expected to recruit for particular jobs in the concern (McCrudden 1992). 2
See for example ECNI (2009).
8
The legislation applies to concerns where Protestants are under-represented just as it does to concerns where Catholics are under-represented. The legislation is thus symmetrical and applies equally to both communities. It is in this respect radically different from the more familiar U.S. federal contractor program, which is directed at reducing ethnic minority (particularly African American) and female under-representation. A major tool made available under the legislation to the FEC (now Article 13 of FETO), and from 2000 to the ECNI, was, where the Commission considered it was necessary, to establish agreements to improve the representation of the under-represented group. In considering whether to work towards an agreement, the Commission could use its investigatory powers under the 1989 Act and the results of the triennial reviews that employers were required to submit (“Section 31 reviews” under the 1989 Act, “Article 55 reviews” under FETO 1998). While the majority of agreements have been established to remedy Catholic underrepresentation, there have also been a substantial number designed to remedy Protestant under-representation in specific concerns. (See Table 1 below.) These affirmative action agreements typically included process requirements and substantive requirements.
The
former usually required concerns to introduce changes in the way in which they conducted their personnel functions, particularly by formalizing advertising, hiring, promotion, dismissal, etc. The latter involved requirements to take specified measures deemed most appropriate for that concern, such as the use of advertising targeted to the under-represented group, the use of statements in advertisements particularly welcoming applications from the under-represented group, and (frequently) the adoption of specified numerical goals that concerns committed themselves to achieving in order to reduce under-representation,
9
together with timetables by which these numerical targets would be achieved. No reverse discrimination or quotas was permitted or could be required by the FEC/ECNI. (A major exception to this approach involves affirmative action in the police force where a form of quota system has been in operation since the implementation of the Patten Report of 1999.) There were three sorts of agreement, namely “legally-enforceable” (Article 13) agreements, “formal” agreements approved by the FEC executive, and “voluntary” agreements negotiated with Commission staff.
In practice, the great majority of the
agreements (around two-thirds) have been “voluntary” agreements with most of the remainder being “legally-enforceable” Article 13 agreements. (See Table 1 below.) Our fieldwork at the ECNI has suggested that Article 13 agreements generally followed investigations by the Commission or situations in which the employer was considered more likely to be resistant, while voluntary agreements generally followed Article 55 reviews, and involved situations in which the employer was considered less likely to be resistant. The substantive content of these two main types of agreement were essentially the same. Ultimately, the difference between the “legally-enforceable” (Article 13) agreements and “voluntary” or “formal” agreements is that legally-enforceable agreements are backed up by sanctions if the agreement is not complied with, although in practice the Commission has primarily employed persuasion rather than enforcement even with regard to the enforcement of “legally-enforceable” agreements (again similarly to the U.S. enforcement activity under the contractor program). In summary, the most notable features of the Northern Ireland legislation which distinguish it from the U.S. contractor program approach are:
10
Its symmetrical character – the legislation applies both to Catholic and to Protestant under-representation The application of the duties and powers of the regulator to all concerns, above a certain size, irrespective of whether they are government contractors or not The use of different types of legally-binding and voluntary agreements to achieve compliance and redress under-representation;
American NGO Activism and Fair Employment in Northern Ireland
Another very important strand of affirmative action in Northern Ireland is provided by the campaign to establish and secure compliance with the MacBride Principles. This was a campaign primarily by U.S.-based activists from the Irish-American community, and some human rights groups, both to put pressure on the British Government to act more decisively on fair employment in Northern Ireland, and to put pressure on American corporations with subsidiaries in Northern Ireland to adopt a set of anti-discrimination and weak affirmative action principles. These were called the MacBride Principles, and were named after and sponsored by Sean MacBride, a controversial Irish statesman who had been chief of staff of the (illegal) Irish Republican Army during the 1930s, then Minister of Foreign Affairs in the Irish Republic, founder of Amnesty International, and recipient of the Nobel Peace Prize in 1974. The MacBride Principles were launched in 1984 and supported by leading U.S. pension funds, most notably the New York City and State retirement funds which held stock in U.S. firms operating in Northern Ireland. The Principles presented American companies
11
with subsidiaries in Northern Ireland with the opportunity to commit themselves to nondiscrimination and affirmative action in their operations in Northern Ireland; the campaign was initially directed at U.S. firms and later expanded to non-U.S. multinational companies with Northern Ireland operations. These Principles had much in common with the content of the agreements that were reached with the FEC/ECNI. They included, inter alia:
Increasing the representation of individuals from under-represented religious groups in the work force, including managerial, supervisory, administrative, clerical and technical jobs All job openings should be publicly advertised and special recruitment efforts should be made to attract applicants from under-represented religious groups The abolition of job reservations, apprenticeship restrictions and differential employment criteria, which discriminate on the basis of religion or ethnic origin The appointment of a senior management staff member to oversee the company’s affirmative action efforts and the setting up of time tables to carry out affirmative action principles
Companies were invited to indicate their acceptance of the Principles by “signing” them. The model chosen was that previously adopted in the Sullivan Principles relating to South Africa (McCrudden 1999) although, unlike in South Africa, there was no substantial move to force divestment from companies that were unwilling to sign the Principles. The MacBride Principles had no legal force in Northern Ireland, although there was the risk of economic sanctions from U.S. state and local government. For example, some U.S.’ jurisdictions (such
12
as New York City) provided that prospective City contractors with operations in Northern Ireland could lose contract bids if their Northern Ireland subsidiaries were not implementing the MacBride Principles. In addition, supporters of the MacBride Principles used the tactics of shareholder activists to increase pressure on American parent companies, in particular by securing the support of institutional investors for shareholder resolutions requiring corporations to sign the Principles. Each signatory to the MacBride Principles was required to report annually to an independent (private) monitoring body on its progress in implementing the Principles. (This was specified in the 1986 amplified version of the Principles.) The body monitoring the operation of the MacBride Principles was the (Washington, DC based) Investor Research Responsibility Center (IRRC). The IRRC has generously supplied us with their database. This details which companies signed the Principles, and in which year. We can thus carry out an analysis of the effect of the Principles analogous to that of the FEC/ECNI agreements. A comparison of their effect with that of the FEC/ECNI agreements should be instructive. On the one hand, the risk of economic pressure in the case of the MacBride Principles might be expected to strengthen their effects, although that effect might be expected to decline over time as the “Troubles” began to recede (particularly following the Good Friday/Belfast Agreement of 1998) and pressure from activists accordingly being reduced. On the other hand, the greater institutional authority and legal power of the FEC/ECNI to check on implementation of their agreements might be expected to increase their long-term effectiveness.
13
Litigation Activity and Pressures to Adopt Affirmative Action
A third set of activities that may have increased the pressure to adopt affirmative action, analogous to lawsuits in the United States, is provided by cases alleging discrimination brought against concerns by individual complainants. From 1990 these were heard by the Fair Employment Tribunal (FET), a specialized section of the Industrial Tribunals, which hear complaints of discrimination on the basis of religion or political opinion. The FET has the power, if it finds in favor of the complainant, to make a financial award (with no upper limit specified). Access to the FET register of cases and decisions was negotiated and secured.
An Initial Investigation of Affirmative Action in Northern Ireland
Before turning to sophisticated analysis, we begin with examining trends in the share of Catholics and Protestants in concerns comparing between those with agreement and those without as presented in Figure 1. ______________________ Figure 1 about here ______________________
Looking at concerns without agreement (which represent the majority of concerns in Northern Ireland), it can be seen that there was a general decline in the Protestant share of the workforce and a corresponding increase in the Catholic share (reflecting wider demographic changes). We also find that there was a greater increase in the Catholic share
14
in concerns that had Catholic under-representation agreements, and an increase rather than a decrease in the Protestant share in concerns that had Protestant under-representation agreements. Given a decline in Protestant population and changing political climate in Northern Ireland over the past decade, one could expect likewise a change in a workforce composition. Figure 2 shows a distribution of concerns in different years by their proportion of Catholic employees. A particularly interesting result is that there appears to be increased integration of non-agreement concerns as well as of the concerns which had signed agreements. In other words, the extent of under-representation declined in concerns that did not have agreements as well as in concerns that did. Non-agreement concerns, then, also showed a trend towards greater integration. As Figure 2 shows, the distribution of concerns shifted from a bimodal one in 1990 to a unimodal one in 2005 with fewer firms located in the two tails of the distribution. ______________________ Figure 2 about here ______________________
There are various possible explanations for this change in the distribution. It could be that concerns were anxious to avoid the risk of being targeted for investigation and a legallyenforceable agreement and therefore took steps to improve the representation of the underrepresented group. We could think of this as a “spillover” effect of the affirmative action program. However, alternative mechanisms might also have been at work. For example, there might have been a general reduction in what has been termed the “chill factor” and workers might have become more willing to apply for jobs in concerns where they would be
15
in the minority. Or it could also be that, as the socio-political climate changed and the hiring of the previously under-represented group became more acceptable, employers were more able or willing to hire on other than religious criteria. Industrial restructuring might also have led to the decline or exit from the labor market of more “extreme” firms (especially in manufacturing) and their replacement more integrated ones in services.
Issues
In this paper, we look not only at overall levels of community representation in the concern but also at representation in management, on which the American research has tended to focus. One possibility is that concerns targeted for agreements by the Commission did increase employment of the under-represented group but did so only at lower-level occupations and continued to exclude them from managerial positions. Furthermore, we would like to check whether effects are the same for voluntary and legally-enforceable agreements, and we plan to explore and control for the effects of litigation or of MacBride activity. In this paper, we seek to address, therefore:
Whether Commission agreements help to promote fair participation with respect to access to higher-level occupations as well as to the generality of occupations Whether there are differences between voluntary and legally-enforceable agreements in helping to promote fair participation, and whether there are differences between the different types of agreement in this respect; or between Catholic and Protestant agreements
16
Whether the effects of agreements declined after 2000 with the transfer of responsibility for the program from the FEC to the ECNI with its broader mandate Whether litigation initiated against concerns at the Fair Employment Tribunal have had any effect on community representation within the relevant concern, and how any such effects compare in magnitude with those of agreements Whether non-governmental activity, in the shape of the campaign surrounding the MacBride Principles, has had any impact on fair employment in those concerns that signed the Principles Whether the effects of the MacBride Principles declined over time as the political climate in Northern Ireland improved and activist pressure reduced.
In the next section we describe the dataset and statistical methods we employed. Following this, we then describe the over-time patterns of our key independent variables, namely voluntary and legal agreements, FET cases and “signing” the MacBride Principles. We conclude with a discussion of our findings and their wider significance.
Data
We rely, in part, on an examination of the monitoring returns supplied since 1990 by concerns in Northern Ireland. The monitoring information is, to some extent, publicly available in the annual publication by the Commission of monitoring information, which gives the community composition of each of the relevant concerns in Northern Ireland in overall terms. We were fortunate, however, in being given access to the data in an electronic
17
form by the ECNI, together with information about the size of the concern, its industry classification (for example, manufacturing or services), its sector classification (public or private), its location, the year in which an agreement was concluded, the nature of the agreement (whether it was to address Catholic or Protestant under-representation), and its legal status (Article 13, formal or voluntary).
In addition to the overall numbers of
Protestants, Catholics and non-determined employees for each year within each concern, the ECNI provided us with information on the community composition at three broad occupational levels, namely professional and managerial (SOC 1, 2 and 3), intermediate (SOC 4 and 5), and routine (SOC 6,7,8 and 9)3. (Disclosure issues prevented us from using a more detailed classification.) The data do not constitute a sample but are the complete population of eligible concerns within Northern Ireland. There is no unit non-response although there are some exclusions.
There is also some item non-response due to the need to maintain
confidentiality. No data could be provided to us in a form that might permit in principle the identification of any individuals. Data based on small cell sizes could not therefore be released. Before release to us the data were anonymized by the ECNI. Rigorous procedures were followed in order to ensure that individuals could not be identified from the data supplied to us. (In practice this meant, for example, combining occupational categories into larger aggregations.) Over the period as a whole (1990 to 2005) some new concerns entered the market while others merged, were taken over, or went out of business. 3
These concerns can be
SOC is based on a Standard Occupational Classification 2000 (ONS 2000). SOC can be divided into nine major groups: SOC1 Managers and officials; SOC2 Professional occupations; SOC3 Associate professional and technical occupations; SOC4 Administrative and secretarial occupations; SOC5 Skilled trades occupations; SOC6 Personal service occupations; SOC7 Sales and customer service occupations; SOC8 Process, plant and machine operatives; and SOC9 Elementary occupation.
18
identified from the ECNI database and we retain in our dataset all concern-years for which there are valid data. However, we exclude the concerns with fewer than 25 employees, since we do not have data on these smaller firms for the whole of our study period. (Smaller concerns were also much less likely to be targeted for agreements by the Commission.) We also consider only the composition of the full-time workforce (again because data on parttime workers is not available for the whole period). Concerns that signed “formal” agreements are excluded since there are too few of them to perform useful statistical analysis. The total number of concerns included in the dataset is 4,306 and the maximum number of concern-years is 16 years. In addition, we have linked information pertaining to tribunal cases and MacBride activity to this dataset based on information from the Commission. We were granted access to the Fair Employment Tribunal (FET) register of cases and decisions. This gives details of the cases filed against each employer and the Tribunal’s decisions on each case. The Investor Research Responsibility Center (IRRC) has given us access to their files showing which Northern Ireland concerns signed the Principles, and when they did so. Definitions and data sources for all variables used in the analysis as well as means and standard deviations (based on all concern-years) are shown in Table 1. ______________________ Table 1 about here ______________________
19
Method
A standard statistical method to investigate the effects of compliance reviews in the American literature is the fixed effects model. In a fixed effect model one essentially controls for unmeasured time-constant characteristics of firms by looking at deviations over time from the firm’s average level of, say, African American representation. One thus investigates in effect whether African American representation in the particular firm was higher after the compliance review than it had been before the review (conditional on the other time-varying factors included in the model). This method has been used by Kalev and Dobbin (2006) and Kalev et al. (2006) in two important papers that look at the effects of compliance reviews (and of other time-varying measures) on black representation in management. We follow essentially the same strategy. In the case of overall Catholic employment, the dependent variable is the log odds of the ratio of Catholic to non-Catholic employees in the concern measured annually one year after the independent variables (yeart+1).
The log-odds of Catholic to non-Catholic
employees in yeart is included on the right-hand side as a control.
ln
Catholic Non - Catholic
j
Z j ,t β1
M j ,t β2
Fj ,t β3 ln
j ,t 1
Catholic Non - Catholic
4
P
5
j ,t
(1)
j ,t
where j
= 1,…, N representing each concern
Zj,t
= a vector of two Catholic agreement variables, specifically a dummy
for a voluntary agreement coded one from the year the concern signed such an
20
agreement, otherwise zero;
and a dummy for a legally-enforceable agreement
(Article 13) coded one from the year the concern signed the agreement, otherwise zero. Mj,t
= a dummy variable for acceptance of the MacBride Principles by
concern j in year t coded one from the year that the principles were signed, otherwise zero. Fj,t
= a vector of Fair Employment Tribunal cases, specifically a dummy
variable for the first tribunal case filed against concern j in year t coded one from the year that the case was filed, otherwise zero; and a continuous variable (time-varying) of the accumulating number of tribunal cases filed against concern j in year t ln
Catholic Non - Catholic
P
= log odds of Catholics in concern j, in year t j,t
= dummy variables for each year of observation, omitting 1990
We then repeat the analysis taking as our dependent variable the log odds of the ratio of Protestant to non-Protestant employees, and including the two types of Protestant agreements as independent variables. We
also
undertake
two
parallel
analyses
of
the
log
proportion
of
professional/managerial employees in each concern. Here we focus on the proportion of professional and managerial employees who are Catholic, and the proportion who are Protestant. The dependent variable in these analyses is ln
mij
mij rj
where
= number of professional/managerial employees from community i in concern j
21
rj
= total number of professional/managerial employees in concern j
An important issue that arises in the Northern Ireland context when applying a fixed effects model of this kind is how to deal with the symmetrical character of the Northern Ireland legislation. As described earlier, in Northern Ireland the legislation requires concerns where either Catholics or Protestants are under-represented to move towards fair participation. In effect, this means that concerns at both extremes of the distribution are being pressed to move towards the center of the distribution. As noted above, in Figure 2, we find evidence for increasing integration with the distribution of both agreement and nonagreement firms shifting from the two extremes towards the center. It is thus possible that we achieve different estimates for the effects of the time-varying controls depending upon whether we include all concerns in the analysis or only those concerns that are in the same tail of the distribution. One way to deal with this issue, therefore, which arises from the symmetrical character of the legislation, is to restrict the analysis to specified selections from the dataset, for example those in a particular tail of the distribution. In this study we find that a reasonable strategy is to limit the analyses to the subsamples of concerns with less than 36% Catholics and less than 41% Protestants respectively. (Different cut-off percentages for the two communities are employed because of the differing shares of the two communities in the workforce as a whole.) The models are estimated using fixed effects estimates.4
4
We also compared the results from fixed effects model with exact matching and propensity score matching (available from the authors) and found that this strategy gave very similar results. Fixed effects models are also more flexible than propensity score matching since they allow the simultaneous inclusion of a number of timevarying predictors (which is what we need to do in this analysis where the focus is on comparing the effects of different types of agreement with those of tribunal cases).
22
The Pattern of FEC/ECNI Agreements
Our first step is to analyze the pattern of FEC/ECNI enforcement activity, the numbers of tribunal cases and MacBride signatures. First, we consider the number of Catholic under-representation agreements, and Protestant under-representation agreements, in both cases distinguishing “voluntary” from “formal”, and “legally enforceable” agreements, and the total number of concerns making monitoring returns in each year from 1990 to 2005. ______________________ Table 2 about here ______________________
As we can see, while overall the ratio of Catholic to Protestant agreements is just below 2.5:1, in the earlier part of the period, up until the mid 1990s, Catholic underrepresentation agreements predominated, with the number of Protestant under-representation agreements rising in the later years. “Voluntary” agreements outnumber legally-enforceable (Article 13) agreements (more so in the case of Protestant under-representation agreements than with Catholic agreements) and there were in total only four formal agreements, all of which were reached before 2000 (i.e. under the FEC). There have also been some other important changes over time. In the first part of our period, up to and including 1995, there were almost equal numbers of voluntary and legallyenforceable agreements. From 1996 to 2000 there was an increased volume of agreements overall with voluntary agreements outnumbering legally-enforceable agreements.
After
1999, however, the number of agreements declined quite sharply and all except one were
23
voluntary agreements. There has, then, been a marked shift over time in favor of voluntary agreements, with a tailing away of agreement activity in the years since the establishment of the ECNI. Commission staff explained to us that the FEC started a program of 47 formal investigations, using its section 12 investigatory powers, in 1991/92 and that these were generally followed by Article 13 agreements. These were then followed by consideration of the Article 55 periodic reviews of large employers as these started to become available from 1993 onwards. Subsequently the FEC’s business plan was to move to considering reviews of employers in the medium size range and then of smaller employers. In the period just before the ECNI came into existence, efforts were concentrated on finalizing the agreements that were currently being discussed with employers. After the establishment of the ECNI the priority was on implementing these agreements. As part of the ECNI compliance activities, the Commission continues to seek reviews from employers to ensure compliance with the Article 55 review requirement and, where appropriate, to discuss findings with employers in relation to fair participation and affirmative action. As might be expected, we also find a clear pattern for the FEC/ECNI to seek agreements with concerns that had more skewed community profiles of their workforces5: the lower the proportion of Catholics in a concern, the higher the risk of that concern being targeted by the Commission to negotiate a Catholic under-representation agreement, and there were no concerns with more than 35% Catholics which signed such an agreement. Conversely the lower the proportion of Protestants within a concern, the higher the risk of
5
This result is based on Cox proportional hazard models estimating the risks of signing an agreement based on time-varying and time-constant characteristics of concerns and their geographical location. The covariates included in the models are: concern size, proportion of Catholics (Protestants) in a concern, proportion of Catholics in the district where the concern is located, number of locations, change of Catholic (Protestant) composition in the concern and change of Catholic composition in the district, industry type and number of locations of concerns. Results are available from the authors.
24
that concern being targeted by the Commission to negotiate a Protestant under-representation agreement, and there were no concerns with more than 40% Protestant which signed such an agreement. Other significant predictors of signing an agreement with the Commission were the size of the concern (the Commission targeting larger concerns), the sector (the Commission targeting the public sector and industry rather than services) and the composition of the district within which the concern was located. Next, we look at the changing numbers of tribunal cases and MacBride “signatures” (which we term “agreements”) over time. Tribunal activity builds up in the 1990s and remains thereafter at a high level. MacBride activity also builds up slowly and well over half the MacBride agreements are signed after 2000, a very different pattern from that of the FEC/ECNI agreement activity. Moreover, there is not nearly such a clear relationship between tribunal or MacBride activity and the composition, size, sector or location of the concern as there was in the case of Commission agreement activity. In effect, then, neither tribunal nor MacBride activity is as precisely targeted as Commission activity was. This, of course, is not at all surprising given the processes involved – targeting of subsidiaries of American firms in the case of MacBride activity, and individual grievances in the case of tribunal activity. This might possibly have implications for the size of the effects of the activity: if the Commission were targeting concerns where it believed that progress towards fair employment could be made but was not in practice being made, then we might expect to see larger effects than in the case of un-targeted activity, where some of the concerns involved might already be making progress towards (or have already achieved) fair employment.
25
The “Effects” of Affirmative Action Agreements
We now turn to using fixed effects models in order to investigate the effects of the two main types of Commission agreement (“voluntary” and “legally enforceable”), FET cases, and MacBride Principles agreements. In Table 3 we look at the effects on overall employment within each concern as well as the effects on professional/managerial employment. ______________________ Table 3 about here ______________________
First we discuss the effects of different legal activities on overall employment. As explained above, when investigating the effects of Catholic Commission agreements we restrict the analysis to those concerns with less than 36% Catholic employees, and when investigating the effects of Protestant agreements we restrict the analysis to concerns with less than 41% Protestant employees. Table 3 shows a very clear pattern of results. First, we find significant and positive effects of year both on the share of Catholic employees in concerns where there were less than 36% Catholics and on the share of Protestant employees where there were less than 41% Protestant employees. This reflects the general move towards increasing integration, with a declining number of non-agreement firms being located in the two tails of the distribution, shown in Figure 26. 6
Thus if we include all concerns, we find that the dummy variables for each year show positive coefficients for Catholic employment, reflecting the growing size of the Catholic labor force in Northern Ireland. But if we
26
Turning next to the various measures of Commission enforcement activity, we find significant effects both for voluntary Catholic agreements (on the growth of Catholic employment) and for voluntary Protestant agreements (on the growth of Protestant employment). In contrast, although the sign is in the predicted direction, the effects of legally-enforceable (Article 13) agreements are not significant. Moreover, in the case of Catholic agreements, the magnitude of the coefficient for the legally-enforceable agreements is clearly smaller than that for the voluntary agreements. In the case of MacBride activity, the sign of the coefficient is in the “wrong” direction but the coefficients are in any case nowhere near significant.
The result is
unchanged if we exclude Commission agreements and tribunal activity from the models and look at the effects of MacBride activity on its own or if we analyze the full set of concerns rather than the two subsets. We also tested the interactions between Commission and MacBride agreements as well as between Commission agreements and FET cases. (See Appendix A) We find no evidence that Commission agreement concerns that signed the MacBride Principles or experienced tribunal cases made greater progress towards fair employment than did other concerns. Finally, although two of the four coefficients for tribunal activity are in the predicted direction, none is significant. Again, excluding the other main predictors does not change the story.
include only concerns that are in the tail of the distribution where Protestants are under-represented, then we find positive coefficients for Protestant employment, reflecting the tendency of such concerns to move towards the center of the distribution, whether or not they have had agreements with the Commission. In turn we found that this affects the estimates for our variables of interest, Zj,t in equation 1 above (the estimates being substantially smaller when we restrict the analysis to concerns in a given tail of the distribution than when we include the whole distribution).
27
As noted earlier, one question is whether Commission agreement activity might have increased the share of the under-represented group, but only at lower levels of the occupational hierarchy. We now turn to examine the effects on professional and managerial employment. Here we select subsamples of concerns with less than 36% Catholics or less than 46% Protestants in professional/managerial positions rather than the entire workforce because our objective is to examine the effects of different legal activities on the representation of Catholic (Protestant) employees in professional/managerial positions. The results are very similar to that of overall employment, at least with respect to the various measures of compliance activity. Once again we find that the largest effects are for voluntary agreements negotiated by the Commission (although in the case of Protestant agreements the coefficient, though positive, is not significantly different from zero). As before, the coefficients for the legally-enforceable agreements and for tribunal cases are of modest magnitude and not significant. Surprisingly, however, we do find a significant negative effect of MacBride activity on the share of Protestants in management. One possibility is that these agreements were generally interpreted as requiring an increase in the Catholic share in management (as opposed to an increase in the share of the under-represented group). There is also one interesting difference with respect to the year dummies. Whereas in the case of employment we found significant and positive effects of year both on the share of Catholic employees in concerns where there were less than 36% Catholics and on the share of Protestant employees where there were less than 41% Protestant employees, in the case of professional and managerial employees we see negative and non-significant effects of year on the share of Catholics and Protestants. This shows that the general tendency towards
28
increased integration within non-agreement concerns that we described earlier does not apply in the same way to professional and managerial positions as it does to the overall workforce. This might be due to the rather different dynamics of the demand and supply of higher-level employees (Osborne and Shuttleworth 2004; McQuaid and Hollywood 2008), or it might reflect a greater resistance to change among managers when recruiting higher rather than lower level employees. Our next step is to see whether the magnitude of these effects changes over time. We hypothesized that the effects of Commission agreements would decline after the establishment of the ECNI in place of the FEC. And we also hypothesized that the effects of MacBride activity would decline with the diminution of the “Troubles”, with the conclusion of the Good Friday/Belfast peace agreement of 1998. Conversely, following Kalev and Dobbin (2006), we expect the effects of tribunal cases to be similar over time. Table 4 compares the effects of Commission agreements, MacBride agreements and Tribunal cases in the 1990s and from 2000 onwards. To construct these measures we simply decompose the variables used in Tables 3 into two; thus, for FEC voluntary agreements, agreements starting in the years 1990-99 are scored 1 from the year of agreement, otherwise zero, and similarly for FEC legally-enforceable Article 13 agreements.
For ECNI agreements, voluntary
agreements starting in the years 2000 onwards are scored 1 from the year of agreement, otherwise zero. Note that there were no formal or legally enforceable agreements established by ECNI and so we have no corresponding variable to include.
______________________ Table 4 about here ______________________
29
Dividing effects by period, does not change any of our earlier conclusions with respect to MacBride agreements or tribunal activity.
However, in the case of Catholic
voluntary agreements, there is a very clear refutation of our hypothesis that the effects of agreements negotiated by the ECNI would be smaller than those negotiated by the FEC: the coefficient actually increases from 0.057 to 0.114. We do not see a comparable increase in the case of Protestant agreements, although here we must remember that we are dealing with rather small numbers of agreements, especially since 2000. Turning finally to effects on the distribution of professional and managerial employees, we find a fairly similar picture. In the case of Catholic voluntary agreements, the coefficient for those negotiated by the ECNI becomes non-significant but the magnitude is effectively the same as that for the FEC agreements, so we cannot support the hypothesis that the effectiveness of voluntary Catholic agreements declined.
And in the case of the
voluntary Protestant agreements, we find no effects of neither the FEC nor the ECNI agreements similar to that reported in Table 3.
Discussion of Findings
The most striking positive finding of the research reported in this paper is that voluntary Commission agreements have been more effective than the legally-enforceable Commission agreements. This finding applies to Catholic and to Protestant agreements, to both time periods, and to their effects both on overall employment in the concern and to the share of professional and managerial employees. It thus seems to be a very robust result. Apart from this positive result, we have found little evidence for direct effects of individual
30
Fair Employment Tribunal activity, and not a great deal of evidence for the effect of legallyenforceable Commission agreements or for MacBride agreements. The lack of effect of tribunal activity on composition in contrast to the effectiveness of Commission agreements is in line with the American research discussed earlier and supporta the kinds of explanation offered by Kalev and Dobbin (2006) in the American context. They have emphasized the important role of institutional reforms brought about by compliance activity of the OFCCP, finding for example that having affirmative action plans, a full time diversity staff, and diversity committees are each highly effective in increasing the entrance of women and minorities into management. They also find mentoring programs effective, especially for non-white women. The same arguments might apply in the Northern Irish context. Thus the FEC and ECNI in their agreements with concerns have sought to change concerns’ procedures for advertising, recruitment, and redundancy, to establish clear responsibility for managing change within the concern, and for monitoring and auditing the changes. In contrast, financial incentives to avoid discrimination may not be so effective if they are not accompanied by institutional reform. Furthermore, it should be recognized that direct discrimination, which makes up the bulk of complaints in the Fair Employment Tribunal may only be one among many other factors that account for under-representation of a particular community. That is to say, while direct discrimination in hiring, promotion or redundancy decisions would, other things being equal, lead to under-representation, it may well not quantitatively be the most important factor. Lack of applications for posts at a particular concern, whether due to the “chill factor” (a term used in Northern Ireland to indicate that, for example, a new Catholic recruit in a predominantly Protestant concern might expect a chilly reception from co-workers) or
31
due to lack of knowledge about the vacancy (as a result of selective advertising or word of mouth recruitment) may well be more important. There is some evidence, although not from the Northern Ireland context, in support of this hypothesis. Heath and Li (2007) in attempting to explain the ethnic minority employment gap in Britain have found that self-reported discrimination and job refusals can explain only around 10 - 25% of the gap. If this analysis is correct, then reforms of the kind sought by the Commission in their agreements, requiring wider advertising and outreach, can potentially play a bigger role than reforms designed to eliminate direct discrimination at the point of application. So even if the financial incentives of tribunal cases (the costs of defending them as well as the financial penalties imposed) lead concerns to reform their selection procedures (in itself an untested assumption), this may not in itself make a great deal of difference to the degree of underrepresentation in the concern. This is not in any way intended to deny the importance of direct discrimination in hiring, promotion or firing (although unfortunately no field experiments of discrimination of the sort that have been carried out on racial discrimination have been attempted in Northern Ireland). Rather our point is that additional processes, some of which might be regarded as constituting indirect discrimination or deriving from prior beliefs about likely discrimination (such as the “chill factor”, may well be even more widespread but have so far not proven to be significant elements in tribunal activity. But if this argument for explaining differences in the effectiveness of tribunal and agreement activity is correct, how might we explain the differences in the effectiveness of voluntary and legally-enforceable agreements, or between Commission and MacBride agreements, all of which aim to tackle recruitment practices?
32
One explanation for the different results for voluntary and legally-enforceable agreements might be that those concerns where legally-enforceable agreements were negotiated were more resistant to change (which might well have been why they were selected for legally enforceable agreements in the first place). To be sure, a fixed effects model would take account of any time-constant characteristics of concerns, but it is possible that the degree of resistance would actually increase following the imposition of an agreement. Alternatively, it might be the case that effective implementation of reforms does require a degree of commitment that was lacking in the concerns selected for legallyenforceable agreements. Our qualitative research suggests that leadership from the top of an organization is crucial in the effective implementation of reforms. Actual implementation of reforms is bound to involve a degree of discretion on the part of lower-level employees: formal procedures, even if tightly specified, can never rule out discretion, and how that discretion is exercised may well depend upon the extent to which junior employees perceive that their seniors value the objectives. Voluntary agreements, where senior staff (with whom agreements are typically negotiated) have been persuaded of the legitimacy of the exercise, may thus be more wholeheartedly implemented than are legally enforceable agreements where the leadership of the concern had to be compelled to accept the intervention. This still leaves open the question of why the MacBride agreements appear to have been less successful than the voluntary agreements negotiated by the Commission. One possibility, as we suggested earlier, might be the different basis for targeting firms for MacBride agreements. As subsidiaries of large multinational companies, they may already have had in place more professionalized personnel functions. Initial MacBride signatories, American firms, may also have had more experience of anti-discrimination requirements,
33
and organizational sociology has also regularly shown that subsidiaries of large corporations tend to have more formalized procedures than do smaller stand-alone companies. Moreover it is important to recognize that our failure to find direct evidence of positive effects of legally-enforceable or of MacBride agreements on fair participation within individual concerns does not in any way imply that such activity was without value. Here, our earlier distinction between direct effects and spillover effects becomes highly relevant. The fact that the Commission had the power to impose legally-enforceable agreements, and that it was willing to exercise that power, might well have made its task of securing voluntary agreements considerably easier.
It might also have signaled to other, non-
agreement, concerns that the Commission “meant business”. The signaling effect of these agreements on other concerns might well have been important for the overall success of the program. Similar arguments might well apply in the case of the campaign for the MacBride Principles.
It has been plausibly argued by McCrudden (1999), and Osborne and
Shuttleworth (2004), that the campaign helped to overcome the British government’s resistance to implementing anything stronger than the 1976 Act – and it remains the case that the British government has been very reluctant to introduce anything approaching affirmative action programs in the rest of the United Kingdom, despite the fact that the ethnic minority employment gap in Britain is rather similar to the Catholic/Protestant employment gap in Northern Ireland at the height of the Troubles. Finally, it must be recognized that the relatively positive picture of the effect of voluntary agreements and of increased integration generally may well be conditional on the generally buoyant nature of the Northern Ireland economy over this period. Increased
34
integration may well be easier to achieve in the context of economic growth than retrenchment.
In addition, in the case of majority attitudes towards ethnic minorities,
recession and increased competition for jobs has regularly been found to increase a sense of threat and greater racial prejudice (Quillian 1995; Scheepers et al. 2002), and analogous processes may be at work in the case of Protestant/Catholic relations too. The 2008-9 recession may well therefore weaken progress towards integration and may dampen the effects of agreements on employment of the under-represented group.
35
Appendix A. Test of Interactions Between Commission Agreements, MacBride agreements and FET Cases Log odds of Catholic:nonCatholic
Log odds of Protestant;nonProtestant
Log proportion of Catholic in professional/managerial positions
Log proportion of Protestant in professional/managerial positions
Concerns with Concerns with Concerns with