the business case for equal opportunities

The business case for equal opportunitiesRebecca Riley, Hilary Metcalf and John Forth

ABSTRACT

It has long been argued that equality of opportunity brings business benefits and thatit is in employers’ interest to implement policy to promote equality of opportunity.Our analysis of the Workplace Employment Relations Survey 2004 found neitherlarge and widespread business benefits, nor large and widespread costs associatedwith Equal Opportunities policies amongst the establishments that implement these.Given the net benefits to society of equal opportunities policies, this suggests thatpublic and private benefits are likely to differ substantially and points to the need forpolicy intervention.

1 INTRODUCTION

The early 1990s saw the beginning of a shift from moral and social justice argumentsfor equal opportunities to an emphasis on business self-interest (Dickens, 1994).Successive governments have promulgated the business case for equal opportunitiesas a major tool to address inequality, arguing that individual businesses can reducelabour shortages, improve employee commitment and morale, reduce staff turnover,and increase sales through greater equality of opportunity (e.g. Age Positive, 2008;Women and Equality Unit, 2003).

Qualitative research shows that a range of benefits do occur (e.g. Bevan et al., 1999;Metcalf and Forth, 2000; Task Force on Race Equality and Diversity in the PrivateSector, 2004), but the evidence suggests that benefits to a specific organisation arecontingent on that organisation’s characteristics and circumstances (Dickens, 1994).Moreover, providing equality of opportunity incurs administrative, management andtraining costs; it may also incur other costs such as reduced morale and commitmentin the previously advantaged group (Holtermann, 1995). It is therefore unclear, apriori, whether an individual organisation will benefit from providing equality ofopportunity. In turn, it is unclear whether, on average, organisations benefit fromtheir own steps to improve equal opportunities.

This article seeks to provide more evidence on the net effect of equal opportunitiespolicies on own-business performance and, specifically, on productivity and profits. Itcontributes in three ways to the small body of quantitative research on this topic forGreat Britain. First, using the Workplace Employment Relations Survey 2004(WERS 2004) (Department of Trade and Industry, 2005), we provide more recentevidence than previously available on the relationship between equal opportunities

❒ The authors are Principal Research Fellows at the NIESR. Correspondence should be addressed toRebecca Riley, NIESR, London SW1P 3HE, UK; email: [email protected]

Industrial Relations Journal 44:3, 216–239ISSN 0019-8692

© 2013 The Author(s)Industrial Relations Journal © 2013 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and350 Main Street, Malden, MA 02148, USA.

policies and practices and business performance. Previous studies using WERS werebased on earlier surveys. Second, previous research has had to rely on subjectivemeasures of productivity and profitability. We extend the analysis to objective(accounts-based) data on labour productivity and profitability, which is possible dueto the introduction of a financial performance questionnaire in WERS 2004 and thelinking with the Office for National Statistics’ Annual Business Inquiry. Third, weexplicitly address the issue of causality, aiming to distinguish statistical associationsthat robustly identify the impact of equal opportunities on business performance fromstatistical associations that may not be causal.

The remainder of this article is structured as follows. The next section discusses howequal opportunities might affect business performance and the following sectionreviews available quantitative evidence. The data and methodology we use to measurethe impacts of equal opportunities policies on business performance are discussed insections 4 and 5. Results are reported in section 6. A final section discusses theimplications of our empirical analysis and draws some conclusions.

2 PROCESSES BY WHICH EQUAL OPPORTUNITIES POLICIES MAYAFFECT BUSINESS PERFORMANCE

There are a number of ways in which successful equal opportunities policies (i.e. thosewhich do, indeed, increase equality of opportunity) may improve business productiv-ity and/or profits. These include: improved recruitment; improved staff utilisation;improved morale and employee commitment; greater employee diversity; and cus-tomer approval; and increased share price (see Anderson and Metcalf, 2003). Each ofthese, and the conditions under which benefits may arise, is discussed below. We thendiscuss the costs to businesses of implementing equal opportunities policies. Tosummarise the discussion, while there are many ways in which businesses might derivepositive performance effects from their equal opportunities policies and practices, theeffect for each organisation is likely to be conditional on its characteristics and itsenvironment, such that a net benefit may not necessarily accrue.

2.1 Recruitment

Discrimination in recruitment will reduce the pool of workers from which an organi-sation draws and may mean that suitable candidates are rejected or do not apply. Itis argued that this results in a poorer match between recruits’ competence andjob requirements, leading to recruitment difficulties, skill shortages and lowerproductivity. Non-discriminating organisations will be able to recruit higher-qualityworkers from a larger pool (which includes those discriminated against), therebyreducing hiring costs. They may also benefit from lower labour costs: the wages ofworkers who are discriminated against are likely to be less than the value of theirmarginal product (Becker, 1971).

The benefits depend, to a large extent, on the tightness of the labour market and theextent to which the labour pool contains workers from the discriminated-againstgroup. Indeed, in slack labour markets reducing the pool of recruits may be beneficial,reducing recruitment costs. The benefits are also predicated on the assumption thatemployee performance is closely aligned to selection criteria and that recruiters aregood at recruiting to these criteria.

217The business case for equal opportunities

© 2013 The Author(s)Industrial Relations Journal © 2013 Blackwell Publishing Ltd

2.2 Staff utilisation

Lack of discrimination in the provision of training, development opportunities andpromotion should result in better utilisation of staff resources (through better match-ing of skills and jobs). The benefits depend upon the extent to which there is discretionover work allocation and the extent (and importance) of development and promotion.As with recruitment, this relies on the assumption of appropriate selection criteria inthe absence of discrimination.

2.3 Morale and employee commitment

It is argued that equal opportunities policies improve morale and employee commit-ment and thereby improve business performance. Certainly, equal opportunities poli-cies and practices have been found to be associated with reduced stress, staff turnover,absenteeism and grievances; improved psychological well-being, job performance andwork quality; and greater ‘organisational citizenship’; and so are assumed to improveemployee morale and commitment (see Judge et al., 2001; Meyer et al., 2002; Rhoadesand Eisenberger, 2002; Riketta, 2002; Thorsteinson, 2003; Wright and Bonett, 2002).

However, these effects may depend on the composition of the workforce and therequirements of the business. While it seems likely that equality of opportunity wouldenhance the morale and commitment of members of discriminated against groups (seeForth and Rincon-Aznar, 2008, for some supporting evidence), the effects on thosewho tend to benefit from discrimination is less clear and, indeed, the effects could benegative. The net effect on morale and commitment within a particular organisationmay depend on workforce composition. The consequent impact on business perform-ance will depend on other characteristics of the business. For example, reductions instaff turnover are beneficial when turnover is too high, but may have net costs ifturnover is low.

2.4 Employee diversity

Equality of opportunity may increase the diversity of an organisation’s employees,although this depends on the ethnic composition of the labour market. Increaseddiversity is professed to bring three types of benefits: customer approval, better serviceto diverse customer groups and greater innovation.

Customer approval is assumed to enhance sales and to be affected by diversity intwo ways. First, it is assumed that: customers support equality and disapprove ofdiscrimination; they therefore tend to approve more of organisations with a diverseworkforce, and this approval increases their custom. Barrington and Troske (2001)refer to a case in which a business lost the majority of its largest vendors following acampaign about the founder’s racist statements. However, some people do notsupport equality and so the net effect on customer approval is uncertain. The extentof any effect on custom is also unknown. Second, it is assumed that customers wish tosee or be served by people like themselves (Metcalf and Forth, 2000). There is noevidence to support this assumption. In any case, equality of opportunity might resultin staff who are dissimilar to their customer base. Moreover, this benefit can only bederived where diversity is visible (i.e. for certain groups and certain jobs).

Better understanding of a diverse customer/client base is assumed to stem fromgreater diversity and to enhance sales and service (Hon and Brunner, 2000). This may

218 Rebecca Riley et al.


be manifest through more effective personal contact with customers/clients, productdevelopment and marketing appropriate to diverse groups (see, e.g. Osborne, 2000),depending on the nature of the business and the composition of customers.

Diversity is also purported to increase innovation, although the evidence is mixed(Anderson and Metcalf, 2003). Different cultural backgrounds may produce differentexperiences, attitudes and approaches. It is therefore assumed that the range of ideasincreases with diversity.

At the same time, greater employee diversity may have costs in respect of employeerelations. It may reduce effective team working because of differences in how indi-viduals interact, hostility from prejudiced employees and increased communicationdifficulties (Jehn et al., 1999; Lang, 1986). At worst, it may result in harassment,antagonism and resentment, with consequent management costs, cultural diversitytraining costs, costs associated with reductions in morale and, potentially, legal costs.Certainly, management demands may increase (Robinson and Dechant, 1997;Shapiro, 2000; Thomas, 1991).

2.5 Shareholder approval

Just as diversity is assumed to increase customer approval, knowledge of the existenceof an equal opportunities policy or equality itself may result in share buyers’ approval.Even without effective implementation, equal opportunities policies may have signal-ling effects. In the United States, an announcement that a company is recognised bythe Department of Labor as having an exemplary affirmative action programme tendsto be followed by an immediate increase in stock price (Wright et al., 1995). This maybe a result of expected higher future sales or because of a publicity effect.

2.6 Costs of equal opportunities policies and practices

The implementation of equal opportunities policies and practices entails costs, someof which are identified in the discussion above as potential dis-benefits. These includedevelopment costs and continuing costs of training and dissemination. Some incurother costs, for example: increased job advertising costs and time to conduct selectionfairly; collection and analysis of data to monitor equal opportunities; specialist pro-vision to cater for a diverse workforce (e.g. workplace adjustments to accommodateemployees with mobility impairments); and reduced morale and increased grievancesif employees are not confident that discrimination is being dealt with effectively.

Costs will vary with the characteristics of the organisation and its circumstances.For example, costs associated with hiring and selection will be greater for organisa-tions with high turnover, while workplace adjustment costs are likely to be greater forthose occupying older buildings.

3 EVIDENCE

The majority of evidence relating to the effects on business performance of equalopportunities policies is qualitative in nature, and, given the specificity of the likelyeffects, it is difficult to draw general conclusions from these studies about the averageimpacts of equal opportunities policies on the average business. A few studies providequantitative evidence on the relationship between equal opportunities policies andbusiness performance in Britain. In these studies, business performance is measured as



managers’ subjective view of their establishment’s productivity or profits comparedwith other establishments in the same industry. Pérotin and Robinson (2000) isperhaps the most oft-cited example. They found that managers’ ratings of labourproductivity at their workplace were higher in workplaces with a formal, written equalopportunities policy than in similar workplaces without a policy, after controlling forother factors. Elsewhere, for a range of specific equal opportunities practices, includ-ing composite indices, the relationship with productivity and performance usingsimilar data has been variously identified as positive, negative or zero (Dex et al.,2001; Forth and Rincon-Aznar, 2008; Gray, 2002; Pérotin and Robinson, 2000).There is also some quantitative evidence on the relationship between equal opportu-nities policies and factors that may affect business performance, such as employeecommitment (Dex and Smith, 2001; Forth and Rincon-Aznar, 2008) and employees’perceptions of fairness of treatment (Bryson, 2000; Forth and Rincon-Aznar, 2008),but here the evidence is again mixed.

It is not only the variation in findings for different measures of equal opportunitiesand performance that prohibits firm conclusions from being made from this body ofevidence. The reliance on subjective measures of performance which, as Forth andMcNabb (2007) discussed, may be prone to error or bias, also reduces one’s confi-dence in the conclusions. Moreover, these studies identify association and not cau-sality: they are consistent with equal opportunities practices being a consequence ofgood business performance rather than vice versa (e.g. if good performance providesthe resources to implement equal opportunities practices), and with other unobservedfactors resulting in both implementation of equal opportunities practices and changesin business performance.

4 DATA

We conduct our analysis using WERS 2004, a survey of employers and employeesyielding detailed information on the nature of work in 2,295 British workplaces.Besides earlier surveys in this series, this is the only data set, of which we are aware,that identifies both businesses’ use of equal opportunities policies and measures ofbusiness performance for a representative sample of British workplaces that is suit-able for quantitative analysis. Further, it contains detailed information on othermanagement practices and business characteristics. Information on workplaces’ locallabour market can be linked to the survey using workplaces’ postcodes.

4.1 Measuring equal opportunities policies

A range of equal opportunities indicators are available from WERS 2004. The exist-ence of a formal written policy on equal opportunities or managing diversity is identified,providing a general indicator of policy presence, as used, for example, in Pérotin andRobinson (2000). However, there is evidence of the ineffectiveness of formal writtenequal opportunities policies per se (Noon and Hoque, 2004). Therefore, we developadditional measures intended to be indicative of stronger policy commitment. Thereare numerous ways this can be done and previous research into the effectiveness andbusiness benefits of equal opportunities policies provides little guidance as to a set ofbest measures. We focus on whether an establishment reviews promotions or relativepay to identify indirect discrimination and whether an establishment tries to measurethe effects of its equal opportunities policies. We prefer an indicator that the workplace



reviews pay or promotion procedures to an indicator that the workplace reviewsrecruitment procedures, because the former practice is less common and, arguably, ismore likely to signify commitment to achieving equality of opportunity. We prefer anindicator of attempts to measure the effects of equal opportunities policies within theworkplace to an indicator of simple monitoring as the former implies more than meredata collection and is a strong indicator of policy commitment, and hence of policyeffectiveness and quality; indeed, this practice is both difficult and rare.

4.2 Measuring business performance

Business performance is measured (i) in terms of subjective assessments of the work-place’s comparative productivity and financial performance (assessed by the WERSrespondent, usually the HR manager or the owner, and recorded on a five-point scale)and (ii) by accounts-based measures of gross value added and profits. There are anumber of issues relating to accuracy and consistency of the subjective performancemeasures (see Forth and McNabb, 2007). We are able to separate out those respond-ents who interpret the subjective measure of financial performance in terms of prof-itability (rather than turnover, costs, or something else), which reduces the sample ofprivate sector workplaces by approximately a third. While the accounts-based meas-ures of gross value added and profits are to be preferred to the subjective performancemeasures all other things equal, they are only available for approximately 500workplaces. Therefore, we evaluate the impacts of equal opportunities on businessperformance using both the subjective and accounts-based measures of productivityand profits. We focus on workplaces in the private sector that trade externally, on thegrounds that public sector workplaces and workplaces that provide goods or servicessolely to other establishments in the same organisation are less likely to measureperformance accurately.1

Table 1 illustrates the incidence of equal opportunities policies, as measured by thethree indicators discussed above, for workplaces in the three different samples that aredistinguished by availability of the particular performance measure. A little less thantwo thirds of workplaces have a formal written policy on equal opportunities. Farfewer implement general practices to promote equality of opportunity. Consistentlyacross policy indicators and samples, the incidence of equal opportunities is higher inlarger workplaces and organisations, workplaces with union presence, and work-places with a relatively high representation of women or ethnic minority employees.

5 METHODOLOGY

We begin our exploration of the relationship between equal opportunities and businessperformance by augmenting empirical models of workplace productivity and profitswith indicators of equal opportunities policies and practices. This is in line with theapproach adopted in previous studies in which equal opportunities policies andpractices are assumed exogenous. The subjective indicators of above average perform-ance are modelled using a probit specification (models explaining variation in thisdichotomous performance indicator performed better than models explaining

1 There are a small number of public sector workplaces that trade externally (e.g. those belonging tonationalised industries or trading public corporations). Where we have accounts-based measures of busi-ness performance for these workplaces, we include them in our analysis.



Tab

le1:

The

inci

denc

eof

equa

lopp

ortu

niti

esin

UK

wor

kpla

ces

(%)

For

mal

wri

tten

polic

yon

equa

lopp

ortu

niti

esor

man

agin

gdi

vers

ity

Mea

sure

men

tof

the

impa

cts

ofeq

ualo

ppor

tuni

ties

polic

ies

Rev

iew

ing

ofpr

omot

ions

orre

lati

vepa

yto

iden

tify

indi

rect

disc

rim

inat

ion

(a)

(b)

(c)

(a)

(b)

(c)

(a)

(b)

(c)

All

wor

kpla

ces

6259

656

46

119

10In

wor

kpla

ces

wit

h:W

orkp

lace

size

:10–

4962

6062

76

810

97

Wor

kpla

cesi

ze:5

0–49

989

8886

119

819

1518

Wor

kpla

cesi

ze:�

500

9696

8830

2724

3633

22P

art

ofa

larg

eror

gani

sati

on78

7881

76

714

1315

Org

anis

atio

nsi

ze:�

100

8887

949

78

1717

20U

nion

pres

ence

8280

8512

8~

1311

14F

emal

eem

ploy

ees:

�50

%68

6568

76

914

1315

Eth

nic

min

orit

yem

ploy

ees:

�10

%81

7967

124

1616

919

Sam

ple

(num

ber

ofw

orkp

lace

s)15

9295

148

815

9295

148

815

9295

148

8

Sour

ce:

Wor

kpla

ceE

mpl

oym

ent

Rel

atio

nsSu

rvey

2004

and

Ann

ualB

usin

ess

Inqu

iry.

Not

es:

Fig

ures

are

wei

ghte

d;sa

mpl

e(a

)inc

lude

spr

ivat

ese

ctor

esta

blis

hmen

tstr

adin

gex

tern

ally

;sam

ple

(b)i

nclu

des

priv

ate

sect

ores

tabl

ishm

ents

trad

ing

exte

rnal

lyth

atin

terp

ret

finan

cial

perf

orm

ance

aspr

ofits

;sa

mpl

e(c

)in

clud

eses

tabl

ishm

ents

trad

ing

exte

rnal

lyfo

rw

hich

we

have

acco

unts

-bas

edin

form

atio

non

finan

cial

perf

orm

ance

,sam

ple

wei

ghts

corr

ecte

dfo

rsa

mpl

ese

lect

ion

bias

(see

For

than

dM

cNab

b,20

07);

~ex

clud

edfo

rdi

sclo

sure

reas

ons

(Mic

ro-d

ata

Ana

lysi

sU

ser

Supp

ort

ON

Sre

gula

tion

s:pu

blis

hed

data

item

sm

ust

refe

rto

am

inim

umof

10es

tabl

ishm

ents

).



variation in the 5-category indicator) and the accounts-based measures of performanceare modelled using linear regression. We include controls for workplace characteristics,employees’ skills, market conditions and competitiveness, and industrial relations andHR management; important here insofar as they correlate with both equal opportu-nities and business performance. Small sample sizes limit the number of significantcovariates in the models of accounts-based measures of performance. We exclude fromall models measures of employees’ commitment and morale, and workforce composi-tion (e.g. by gender or ethnicity), which, as discussed above, may be influenced by equalopportunities policies. As such, their inclusion might mask any potential policy effect.Instead, we control for factors that are likely to influence employees’ commitment andmorale (industrial relations and HR practices) and factors that are likely to influencethe composition of employees in the workplace (measures of workforce composition inthe industry and local labour market), but which are unlikely to be affected by theindividual establishment’s policy on equal opportunities.

Models of subjective and accounts-based productivity and profits (excluding equalopportunities indicators) are reported in Table 2; the observed relationships largelyaccord with expectations. We are better able to explain variation in the accounts-based, than the subjective, measures of business performance. This is not entirely dueto differences in outcome measures, but is also explained by differences in the samplesof workplaces. Thus, we are better able to explain variation in the subjective measuresof performance in the sample for which we have accounts-based data than in thelarger sample for which we have subjective measures of performance.

In a second step we compare business performance among those establishmentsthat operate equal opportunities policies to business performance among a matchedsample of establishments that do not. The matched sample for each performancecomparison is selected on the basis of a propensity score (the propensity to operateequal opportunities policies), predicted using the variables included in the model inTable 2. The advantage of this approach over the augmented business performancemodel is that it focuses only on those establishments which differ in their equalopportunities policies, but which can be regarded as similar in terms of the factors thatdetermine business performance. Estimates obtained using this approach may there-fore better approximate the causal effects of equal opportunities on businessperformance.

Many of the variables that explain business performance and used to compute thepropensity score correlate with the presence of equal opportunities. The mean pre-dicted propensity score among establishments that operate these policies is signifi-cantly higher than among establishments that do not (see Table A1). In estimating thepropensity score, we exclude variables that predict equal opportunities, but that donot predict business performance (Bhattacharya and Vogt, 2007; Heckman andNavarro-Lozano, 2004). We use nearest neighbour matching with replacement,excluding from the matched sample those establishments operating equal opportuni-ties whose propensity score is greater (less) than the maximum (minimum) estimatedpropensity score observed for the controls and those for whom we cannot find acontrol with an estimated propensity score within a range of 0.002. These commonsupport criteria result in the loss of between a third and half of establishments withformal written policies on equal opportunities, depending upon the performancemeasure, far less for other equal opportunities measures (see Table A1). Surveyweights are used in estimating the propensity score; in comparing means in thematched sample, we use survey weights for the treated, ignoring the weights on the



Tab

le2:

Mod

els

ofbu

sine

sspe

rfor

man

ce

Inde

pend

ent

vari

able

:

Dep

ende

ntva

riab

le

Subj

ecti

vein

dica

tor

ofab

ove

aver

age

prod

ucti

vity

perf

orm

ance

Subj

ecti

vein

dica

tor

ofab

ove

aver

age

finan

cial

perf

orm

ance

(pro

fits)

Log

gros

sva

lue

adde

dpe

rem

ploy

eeL

ogpr

ofits

per

empl

oyee

Coe

ffici

ent

pV

alue

Coe

ffici

ent

pV

alue

Coe

ffici

ent

pV

alue

Coe

ffici

ent

pV

alue

Indu

stri

alre

lati

ons

and

HR

MU

nion

pres

ence

-0.0

61(0

.296

)-0

.059

(0.4

40)

-0.1

10(0

.029

)-0

.073

(0.0

55)

Par

tici

pati

onin

retu

rns

0.07

9(0

.050

)0.

047

(0.3

57)

0.09

2(0

.028

)0.

079

(0.0

08)

Par

tici

pati

onin

cont

rol

0.04

3(0

.275

)-0

.048

(0.3

59)

-0.0

31(0

.079

)In

depe

nden

cein

wor

k0.

041

(0.1

73)

0.06

2(0

.056

)C

ultu

reof

head

offic

e:U

nite

dSt

ates

0.30

5(0

.051

)0.

367

(0.0

21)

0.26

2(0

.017

)C

ultu

reof

head

offic

e:ot

her

non-

UK

0.27

0(0

.067

)0.

180

(0.0

43)

0.12

0(0

.080

)In

fluen

ces

onem

ploy

eech

arac

teri

stic

sT

TW

Apo

pula

tion

:%et

hnic

min

orit

y-0

.003

(0.0

96)

-0.0

03(0

.012

)In

dust

ryem

ploy

men

t:%

ethn

icm

inor

ity

-0.0

55(0

.011

)-0

.018

(0.4

87)

Indu

stry

empl

oym

ent:

%fe

mal

e0.

006

(0.0

20)

0.00

3(0

.259

)M

arke

tco

ndit

ions

and

com

peti

tive

ness

Ext

ent

dem

and

depe

nds

onqu

alit

y:m

ediu

m0.

090

(0.1

50)

Ext

ent

dem

and

depe

nds

onqu

alit

y:hi

gh0.

147

(0.0

13)

Stat

eof

the

mar

ket:

mat

ure

-0.0

73(0

.216

)-0

.054

(0.4

63)

Stat

eof

the

mar

ket:

decl

inin

gor

turb

ulen

t-0

.121

(0.0

31)

-0.1

36(0

.049

)-0

.137

(0.0

07)

Stat

eof

the

mar

ket:

mat

ure,

decl

inin

gor

turb

ulen

t-0

.105

(0.0

09)

Tra

ding

inth

ein

tern

atio

nalm

arke

t0.

123

(0.0

75)

-0.1

00(0

.233

)Sk

ills

Man

ager

ial/p

rofe

ssio

nalo

ccup

atio

ns(%

ofem

ploy

ees)

0.00

4(0

.002

)R

outi

neun

skill

edoc

cupa

tion

s(%

ofem

ploy

ees)

-0.0

00(0

.892

)A

ge16

–21

(%of

empl

oyee

s)-0

.002

(0.0

70)

Est

ablis

hmen

tch

arac

teri

stic

sO

wne

rshi

p:pa

rtly

fore

ign

-0.1

01(0

.371

)O

wne

rshi

p:pr

edom

inan

tly

fore

ign

-0.2

30(0

.080

)E

stab

lishm

ent

size

:50

empl

oyee

sor

mor

e0.

043

(0.3

99)

0.07

5(0

.254

)O

rgan

isat

ion

size

:100

empl

oyee

sor

mor

e-0

.108

(0.0

60)

0.20

2(0

.007

)0.

017

(0.6

70)

0.01

6(0

.585

)P

art-

tim

ew

orki

ng(%

ofem

ploy

ees)

-0.0

02(0

.062

)-0

.003

(0.0

00)

-0.0

01(0

.237

)Y

oung

esta

blis

hmen

t-0

.055

(0.3

23)

-0.0

53(0

.210

)R

-squ

ared

(pse

udo

for

prob

itm

odel

s)0.

085

0.08

80.

368

0.36

3Sa

mpl

e(u

nwei

ghte

d)13

2782

744

444

8

Sour

ce:

Wor

kpla

ceE

mpl

oym

ent

Rel

atio

nsSu

rvey

2004

and

Ann

ualB

usin

ess

Inqu

iry.

Not

es:

Est

imat

ion

take

sin

toac

coun

tsu

rvey

wei

ghts

;su

bjec

tive

(acc

ount

s-ba

sed)

perf

orm

ance

mod

elle

das

apr

obit

(lin

ear

regr

essi

on);

prob

itco

effic

ient

ssh

own

asm

argi

nal

effe

cts;

inde

pend

ent

vari

able

sin

clud

em

ajor

SIC

indi

cato

rsan

da

cons

tant

term

;par

tici

pati

onin

retu

rns

vari

able

cons

truc

ted

from

fact

oran

alys

isof

indi

cato

rsof

perf

orm

ance

-rel

ated

pay,

profi

t-re

late

dpa

y,em

ploy

eesh

are

sche

mes

;par

tici

pati

onin

cont

rolv

aria

ble

cons

truc

ted

from

fact

oran

alys

isof

indi

cato

rsof

brie

fing

betw

een

man

ager

san

dw

orke

rs,j

oint

cons

ulta

tive

com

mit

tees

,and

qual

ity

circ

les;

inde

pend

ence

inw

ork

vari

able

cons

truc

ted

from

fact

oran

alys

isof

indi

cato

rsof

the

exte

ntof

vari

ety

inw

ork,

disc

reti

onin

wor

k,co

ntro

love

rpa

ceof

wor

kan

dde

sign

ofw

ork;

sam

ple

incl

udes

priv

ate

sect

ores

tabl

ishm

ents

trad

ing

exte

rnal

ly;s

ampl

efo

rsu

bjec

tive

finan

cial

perf

orm

ance

incl

udes

only

thos

ees

tabl

ishm

ents

who

rega

rdfin

anci

alpe

rfor

man

ceas

profi

ts;s

ampl

efo

rac

coun

ts-b

ased

busi

ness

perf

orm

ance

mea

sure

sin

clud

espu

blic

and

priv

ate

sect

ores

tabl

ishm

ents

trad

ing

exte

rnal

ly.



matched controls. Comparing mean differences in the determinants of business per-formance between establishments with and without equal opportunities in thematched sample, the matching exercise appears more successful on the sample forwhich we have objective measures of business performance and for the two measuresof equal opportunities indicative of stronger policy commitment (see Table A1). Wediscuss the sensitivity of our results to alternative match methods in the next section.

The two approaches discussed so far are unlikely to yield estimates of the causalimpacts of equal opportunities on business performance if either the relationshipbetween these is truly circular, in the sense that business performance determinesuptake of equal opportunities and equal opportunities affect business performance, orif we are unable to identify from theory and measure in the data all the factors thatmay coincide with equal opportunities and business performance. To deal with thesepossibilities, we jointly model business performance and uptake of equal opportuni-ties as a function of the covariates in Table 2. For this approach to be successful (interms of yielding causal impact estimates), we need to identify factors which appear toinfluence whether or not establishments operate equal opportunities policies, butwhich are unrelated to business performance. WERS 2004 records the gender and thetraining of the HR manager in the workplace. We consider these as potential instru-mental variables. One might speculate that women HR managers (being from atraditionally discriminated against group) and highly trained HR managers (graspingthe specifics of equal opportunities policies and practices) are more likely to imple-ment effective equal opportunities policies. At the same time, it seems unlikely thatthese factors themselves should have any bearing on business performance. To test thevalidity of using the gender and occupational training of the HR manager as instru-mental variables, we first test whether these are correctly excluded from the models ofbusiness performance in Table 2. We find neither attributes of the HR manager to bestatistically significant in explaining business performance (individually or jointly; seetest for exogeneity in Table A2).2 Next, we assess the relevance of the gender andoccupational training of the HR manager in a probit model of equal opportunitiesuptake (including the covariates of business performance). Establishments with HRmanagers or owners that are qualified in personnel management are more likely tohave implemented equal opportunities policies and practices on all three equal oppor-tunities measures considered. These correlations are statistically highly significant (seetest for weak instruments in Table A2). The gender of the HR manager or owner istypically a statistically significant predictor (on its own and jointly with the qualifi-cations of the HR manager) of whether establishments have a formal written policy onequal opportunities or whether establishments measure the impacts of their equal oppor-tunities policies; where it is not (i.e. where it appears to be a weak instrument) we donot use it as an instrument for the policy (see Table A2). The gender of the HRmanager is not a statistically significant predictor of reviewing practices to identifyindirect discrimination. Thus, we do not use the gender of the HR manager as aninstrument for reviewing practices.

6 RESULTS

Tables 3 and 4 report our estimates of the average effect on business performance ofhaving in place a particular equal opportunities policy or practice among those who

2 Throughout this article we use the 5 per cent threshold to denote statistical significance, unless specifiedotherwise.



Tab

le3:

Pro

duct

ivit

yan

deq

ualo

ppor

tuni

ties

Out

com

eva

riab

le:

Equ

alop

port

unit

ies

mea

sure

For

mal

wri

tten

polic

yon

equa

lop

port

unit

ies

orm

anag

ing

dive

rsit

y

Mea

sure

men

tof

the

impa

cts

ofeq

ual

oppo

rtun

itie

spo

licie

s

Rev

iew

ing

ofpr

omot

ions

orre

lati

vepa

yto

iden

tify

indi

rect

disc

rim

inat

ion

Coe

ffici

ent

pV

alue

Coe

ffici

ent

pV

alue

Coe

ffici

ent

pV

alue

Subj

ecti

vein

dica

tor

ofab

ove

aver

age

prod

ucti

vity

perf

orm

ance

Dif

fere

nce

inm

eans

-0.1

21(0

.009

)-0

.018

(0.8

13)

0.00

3(0

.965

)P

robi

tm

odel

,exo

geno

usE

O-0

.165

(0.0

05)

-0.0

34(0

.652

)-0

.007

(0.9

15)

Dif

fere

nce

inm

eans

,pro

pens

ity

scor

ees

tim

ates

-0.0

49(0

.290

)0.

005

(0.9

67)

-0.0

16(0

.851

)P

robi

tm

odel

,end

ogen

ous

EO

-0.2

16(0

.295

)0.

104

(0.8

36)

0.25

7(0

.329

)L

oggr

oss

valu

ead

ded

per

empl

oyee

Dif

fere

nce

inm

eans

0.07

3(0

.104

)0.

028

(0.8

34)

-0.0

83(0

.114

)L

inea

rre

gres

sion

,exo

geno

usE

O-0

.029

(0.5

13)

0.03

6(0

.693

)-0

.124

(0.0

69)

Dif

fere

nce

inm

eans

,pro

pens

ity

scor

ees

tim

ates

-0.0

17(0

.627

)0.

076

(0.6

26)

-0.0

95(0

.162

)L

inea

rre

gres

sion

,end

ogen

ous

EO

-0.2

18(0

.130

)0.

304

(0.3

02)

-0.0

88(0

.597

)

Sour

ce:

Wor

kpla

ceE

mpl

oym

ent

Rel

atio

nsSu

rvey

2004

and

Ann

ualB

usin

ess

Inqu

iry

Not

es:

Est

imat

ion

take

sin

toac

coun

tsu

rvey

wei

ghts

;pro

bit

coef

ficie

nts

show

nas

mar

gina

leff

ects

;dif

fere

nce

inm

eans

mod

elgi

ves

the

sim

ple

diff

eren

cein

busi

ness

perf

orm

ance

betw

een

esta

blis

hmen

tsw

ith

and

wit

hout

equa

lopp

ortu

niti

es(E

O);

prob

itan

dlin

ear

regr

essi

onm

odel

sof

busi

ness

perf

orm

ance

incl

ude

the

cont

rols

show

nin

the

mod

els

inT

able

2;pr

open

sity

scor

ees

tim

ates

take

into

acco

unt

surv

eyw

eigh

tsin

esti

mat

ing

the

prop

ensi

tysc

ore

and

surv

eyw

eigh

tsfo

rth

etr

eate

din

esti

mat

ing

the

diff

eren

cein

mea

ns;p

redi

ctio

nof

the

prop

ensi

tysc

ore

isba

sed

onth

eco

ntro

lssh

own

inT

able

2;pr

open

sity

scor

ees

tim

ates

gene

rate

dus

ing

near

estn

eigh

bour

mat

chin

gw

ith

repl

acem

ent,

calip

er0.

002;

EO

sele

ctio

neq

uati

ons

inth

een

doge

nous

EO

mod

els

incl

ude

the

cont

rols

used

toex

plai

nbu

sine

sspe

rfor

man

cein

Tab

le2

and

addi

tion

alin

stru

men

ts:

the

gend

eran

doc

cupa

tion

alqu

alifi

cati

onof

the

HR

man

ager

/ow

ner

(equ

atio

nfo

r‘R

evie

win

g’ex

clud

esth

ege

nder

ofth

eH

Rm

anag

er/o

wne

r,as

does

equa

tion

for

‘Mea

sure

men

t’in

the

subj

ecti

vesa

mpl

e).



Tab

le4:

Pro

fits

and

equa

lopp

ortu

niti

es

Out

com

eva

riab

le:

Equ

alop

port

unit

ies

mea

sure

For

mal

wri

tten

polic

yon

equa

lop

port

unit

ies

orm

anag

ing

dive

rsit

y

Mea

sure

men

tof

the

impa

cts

ofeq

ual

oppo

rtun

itie

spo

licie

s

Rev

iew

ing

ofpr

omot

ions

orre

lati

vepa

yto

iden

tify

indi

rect

disc

rim

inat

ion

Coe

ffici

ent

pV

alue

Coe

ffici

ent

pV

alue

Coe

ffici

ent

pV

alue

Subj

ecti

vein

dica

tor

ofab

ove

aver

age

finan

cial

perf

orm

ance

(pro

fits)

Dif

fere

nce

inm

eans

0.04

2(0

.452

)0.

064

(0.5

47)

0.05

5(0

.525

)P

robi

tm

odel

,exo

geno

usE

O-0

.053

(0.4

58)

0.05

4(0

.639

)-0

.026

(0.7

92)

Dif

fere

nce

inm

eans

,pro

pens

ity

scor

ees

tim

ates

0.00

8(0

.900

)0.

178

(0.2

53)

-0.1

24(0

.269

)P

robi

tm

odel

,end

ogen

ous

EO

0.18

9(0

.266

)0.

474

(0.0

00)

0.12

6(0

.706

)L

ogpr

ofits

per

empl

oyee

Dif

fere

nce

inm

eans

0.04

1(0

.249

)0.

005

(0.9

59)

-0.0

58(0

.140

)L

inea

rre

gres

sion

,exo

geno

usE

O-0

.032

(0.2

51)

0.05

0(0

.483

)-0

.093

(0.1

35)

Dif

fere

nce

inm

eans

,pro

pens

ity

scor

ees

tim

ates

0.02

8(0

.331

)-0

.069

(0.7

34)

-0.0

58(0

.568

)L

inea

rre

gres

sion

,end

ogen

ous

EO

-0.1

67(0

.027

)0.

176

(0.7

81)

-0.1

36(0

.251

)

Sour

ce:

Wor

kpla

ceE

mpl

oym

ent

Rel

atio

nsSu

rvey

2004

and

Ann

ualB

usin

ess

Inqu

iry.

Not

es:

See

note

sto

Tab

le3;

Equ

atio

nfo

r‘R

evie

win

g’ex

clud

esth

ege

nder

ofth

eH

Rm

anag

er/o

wne

r,as

does

equa

tion

for

‘For

mal

wri

tten

polic

y’in

the

acco

unts

-bas

edsa

mpl

e.



have these in place [the ‘average treatment effect on the treated’ (ATT)]. Table 3concerns workplace productivity and Table 4 concerns workplace profitability. Whenthe outcome measure refers to subjective performance, the ATT measures the per-centage point difference in the probability of reporting above average performanceassociated with operating equal opportunities. With the accounts-based performancemeasures the ATT measures the per cent difference in outcomes (gross value added orprofits per head) associated with operating equal opportunities. We report in bracketsthe probability that the ATT is zero, based on the estimated standard error andcentral estimate of the ATT.

For each business performance measure and each equal opportunities measure wereport estimates of the ATT from four different models, distinguished by the identi-fying assumptions that these involve. The first of these is the simple difference in meanbusiness performance between establishments with and without equal opportunitiespolicies, essentially a cross-tabulation of the data. The second of these is the estimatedmarginal effect of having an equal opportunities policy within a probit or linearregression model of business performance, equivalent to the models reported in Table2, augmented with an indicator of equal opportunities. In these models the presenceof equal opportunities policies is assumed exogenous, given the other influences onbusiness performance included in the model. The third estimate of the ATT is thedifference in mean business performance between establishments with and withoutequal opportunities policies, within a matched sample of establishments (discussed inthe previous section). The fourth estimate of the ATT is the estimated marginal effectof having an equal opportunities policy within a probit or linear regression model ofbusiness performance, where the presence of equal opportunities is assumed to beendogenous. In this case the model of business performance in Table 2, including anindicator of equal opportunities, is estimated jointly with a probit model of equalopportunities uptake including the variables used to explain business performanceand additional instruments (discussed in the previous section).

6.1 Equal opportunities and workplace productivity

The first column in Table 3 reports estimates of the workplace productivity effects ofhaving a written policy on equal opportunities or managing diversity. For those work-places with formal policies, the percentage reporting above average productivity is12.1 percentage points less than for those workplaces without formal written policies.This difference is statistically significant and is not obviously attributable to differ-ences in observable influences on workplace productivity. Controlling for observableinfluences on workplace productivity in the simple probit model, the share of work-places with formal policies reporting above average productivity is 16.5 percentagepoints less than for workplaces without. But, in the matched sample, this differencefalls to 4.9 percentage points and is no longer statistically significant. The estimatedATT in the propensity score model is sensitive to the choice of caliper used in thematching. Matching within a wider caliper (0.01), the difference is larger at 9.4percentage points and is statistically significant (p value 0.013); only 32 observationsfrom the treatment group are lost in this case, but the sample is less balanced.Matching within a smaller caliper (0.001), the estimated ATT is qualitatively similarto the central case reported in Table 3; the ATT in this scenario is -0.006 (p value0.913), 522 observations are lost from the treatment group (compared with 319 in thecentral case) and only five covariates remain statistically different between the treat-



ment and comparison groups (compared to eight covariates in the central case).Treating the presence of a written policy on equal opportunities or managing diversityas endogenous, we find further support to suggest that there is little, if any, differencein perceived productivity performance between workplaces that have formal policiesand workplaces that do not. The correlation of the error terms in the two equationsof the endogenous model is not statistically different from zero (see Table A2).However, we note that the Wald test on which this conclusion is based is not aparticularly strong test of treatment exogeneity (Monfardini and Radice, 2008) andwe do not interpret this to mean that the probit model with the exogenous policyassumption provides the more robust estimate of the ATT.

None of the models of gross value added per employee suggest that there is astatistically significant relationship between having a formal written equal opportu-nities policy and workplace productivity. Looking at gross value added per employee,this is on average 7.3 per cent higher among workplaces with formal equal opportu-nities policies in comparison with workplaces without formal written policies.Although this difference is nearly statistically significant at the 10 per cent level,it—critically—turns negative and moves further from statistical significance in themodels where we control for other influences on gross value added. The simpledifference in mean log gross value added between workplaces with and without equalopportunities policies stands in complete contrast to the correlations in the dataregarding firms’ subjective evaluation of productivity performance. This contrastdoes not reflect differences in samples. The tendency for workplaces with formalwritten policies on equal opportunities to report relatively poor productivity perform-ance, as measured by the subjective indicator, is also evident in the accounts-basedsample [the difference is 18.6 percentage points (p value 0.060)].

Estimates of the workplace productivity effects associated with measuring theimpacts of equal opportunities policies in the workplace and with reviewing promotionprocedures or relative pay rates to identify indirect discrimination are reported in thesecond and third columns of Table 3, respectively. Both are intended to be generalindicators of equal opportunities policy and practice. We consistently find no evidenceof a statistically significant relationship between either of these measures of equalopportunities in the workplace and workplace productivity (columns two and three).All models, using either measure of workplace productivity, produce ATT estimatesthat are statistically no different from zero. We emphasise that, in these samples, thenumbers of workplaces with these practices are relatively small (see Table 1), whichmay reduce the likelihood of finding statistically significant policy impacts, even ifthese are genuinely different from zero.

In summary, we find little robust evidence that equal opportunities have a netimpact (either positive or negative) on workplace productivity, once one hasaccounted for differences between establishments that do and do not operate equalopportunities policies and once one considers accounts-based information onperformance.

6.2 Equal opportunities and workplace profitability

The first column in Table 4 reports estimates of the effects on workplace profits ofhaving a written policy on equal opportunities or managing diversity. None of themodels show a statistically significant relationship between equal opportunities poli-cies and the subjective indicator of financial performance. The results are much the



same for the relationship between equal opportunities policies and profits peremployee, except in the model that treats having a formal written policy asendogenous. In this model, we find that having such a policy appears to be associatedwith a statistically significant reduction in profits per employee of 16.7 per cent.However, we suggest that this is interpreted with some caution, as this result stands instark contrast to all the other estimates in column one of Table 4. Further, includingadditional instruments selected from the factors that do not correlate with profits peremployee in this sample, but which do predict equal opportunities, we find no statis-tically significant association between formal written policies and profits peremployee.3 We note that the propensity score estimates are qualitatively similar whenwe use different caliper widths (0.01 and 0.001); that is, there is, on average, nodifference in profits per employee between workplaces with and without formalwritten policies in the matched sample.

We generally find no evidence of a statistically significant relationship betweenworkplace profits and either measuring the impacts of equal opportunities policies in theworkplace or reviewing procedures to identify indirect discrimination. This mimicsthe findings regarding the relationship between these practices and workplaceproductivity. The exception is the estimated effect of measuring the impacts of equalopportunities policies on the subjective indicator of financial performance in themodel that treats equal opportunities as endogenous. In this model, we find thatestablishments that measure the impacts of their equal opportunities policies are 47percentage points more likely to report above average financial performance thanestablishments that do not make these measurements. The magnitude of this effectwould seem difficult to attribute to policy alone and is inconsistent with all otherevidence from our analysis.

7 DISCUSSION AND CONCLUSIONS

The analysis presented in this article has sought to provide quantitative evidence onthe average relationship between equal opportunities policies and practices andbusiness performance in the workplace. The assessment of two equal opportunitiespractices which might be expected to indicate a commitment to effective equal oppor-tunities (measuring effectiveness and reviewing pay or promotion), as well as a broadpolicy (a written policy), and the use of several outcome measures using severalidentification approaches, are intended to help us to draw robust conclusions aboutthis relationship. The evidence we have presented suggests that it is difficult to arguethat the net benefits to businesses associated with implementing equal opportunitiespolicies and practices are large and widespread among the establishments whichimplement these. Similarly, the evidence does not support the notion that equalopportunities policies and practices place disproportionate net burdens on businesses.We find some strong and statistically significant relationships between subjectiveindicators of business performance and equal opportunities policies. But, as we haveargued above, these are unlikely to reflect the causal impacts of policy.

3 Including ‘Trading in the international market’ and ‘Independence in work’ as instrumental variables (inaddition to those in the central case), we find an ATT of -0.119 (0.121). Although on statistical groundsthese factors may be regarded as instrumental variables, on theoretical grounds it is more difficult to justifythese as exogenous to profits (and they do influence subjective measures of financial performance, see Table2). Hence, we do not include these factors as instrumental variables in the central case reported in Table 4.Nevertheless, we believe this exercise illustrates the sensitivity of the results.



Although we suggest that there is little evidence that equal opportunities policiesand practices result in a net cost or benefit to employers on average, this is not to saythat no employers will derive net benefits from implementing equal opportunitiespolicies and practices or that none will see a net cost. Indeed, as we have discussed, therelationship between equal opportunities and business performance is complex andthe net benefits to an organisation of these practices may be positive or negative,depending on the organisation’s characteristics and its circumstances. Also, wecannot rule out that specific equal opportunities practices other than those measuredhere may be associated, on average, with enhanced workplace performance or netcosts. We have assessed three indicators of general practice, rather than assessing theimpact of more specific equal opportunities practices, such as those targeted atparticular groups.

While the WERS 2004 enables us to examine the performance effects of equalopportunities policies and practices in considerable depth, the analysis we have under-taken also has significant limitations and it is important to bear these in mind. Thesestem primarily from data limitations, in respect of variables and sample sizes, exac-erbated by the pattern of implementation of policies and practices. Productivity andprofit measures based on accounts data are only available for a small subset of theWERS 2004 sample. This reduces the likelihood of identifying performance effectsand limits the possibilities for looking at subsets of the data, which are preferredwhere equal opportunities policies and practices are highly coincident with otherfactors that correlate with business performance, such as workplace size. Subjectivemeasures of performance are available for a larger sample of workplaces, but thesemay be prone to measurement error, thus reducing the reliability of the findings.Separately, it appears important to evaluate the relationship between equal opportu-nities and business performance using a range of models, in order to reach robustconclusions.

Our findings are somewhat in contrast to those of Pérotin and Robinson (2000), thestudy which is probably closest in method to that adopted here. Using WERS 1998,they find a positive and statistically significant relationship between having a formalwritten policy on equal opportunities and labour productivity. They use an orderedprobit model of the subjective productivity ranking and treat the policy as exogenous.In a similar model we find a negative and statistically significant relationship betweenequal opportunities and labour productivity using WERS 2004, which we do notinterpret as a causal impact. These differences in results appear to arise because of adifferent bivariate association between the equal opportunities policy and the subjec-tive labour productivity ranking in the 1998 and 2004 WERS. The association inWERS 2004 is negative, but the association is positive in WERS 1998. This shows upirrespective of whether one uses the full five-point scale as per Pérotin and Robinson(2000) or a binary variable as in this article, or whether one controls for otherinfluences on labour productivity, and is unrelated to the weighting scheme. This isunlikely to suggest that anything has specifically changed in respect of equal oppor-tunities though, as the same pattern of results occurs using, for example, ‘unionrecognition’ in place of equal opportunities. We are, of course, unable to evaluatewhether the bivariate association with accounting measures of performance has alsochanged, but the change in the bivariate association with the subjective measure mightlead to further suspicion over the validity of the perceptual data.

Equality of opportunity in the labour market may bring economic and socialbenefits. Notably, it may increase the supply of labour and improve the efficiency with



which human resources are used, reducing labour costs and raising aggregate income.It may also help to reduce social inequalities. Individuals, society at large and indi-vidual businesses may all share in these benefits. At the same time, the evidencepresented in this article suggests that, on average, individual employers do not nec-essarily gain (nor lose) from implementing policies and practices to promote equalityof opportunity. The implication is that there is likely to be a difference between theprivate and public costs and benefits of equal opportunities policies and practices,suggestive of market failure and pointing to the need for policy intervention. Analternative interpretation of our results is that equal opportunities policies and prac-tices are ineffectual, that is, that they do not actually succeed in influencing interme-diate outcomes such as morale, commitment and equality, and therefore that they donot influence business performance. In this case, policy intervention is perhaps lessjustified.

An overriding concern in this article has been the complexity of the linkagesbetween equal opportunities policies and practices and business outcomes.Although not strictly essential to the derivation of business benefits, there is theexpectation that policies and practices affect equality. However, this assumption hasnot yet been proven. In section 2, we identified a large number of routes by whichequal opportunities policies and practices might affect profits and productivity.However, which of these routes is important is not known. Moreover, the likelyrange of linkages, combined with data limitations, will have reduced the potentialfor detecting effects. These difficulties suggest two important areas for furtherresearch. First, we need more evidence on the impact of equal opportunities policiesand practices on equality in the workplace. This is likely to be difficult to find,because equality is difficult to measure. Second, further evidence on the extent andnature of business benefits could be gained through examining in more detail theeffects of equal opportunities policies and practices on outcomes that are interme-diate to business performance. It is important that such analysis considers thepotential endogeneity of policy.

Acknowledgements

Funding from the Department for Work and Pensions is gratefully acknowledged.We would also like to thank Helen Bewley, Anthony Johnson, James Mitchell, JohnPurcell, Lucy Stokes and Liz Such variously for comments, discussion and assistancewith WERS, and the MAUS team at Office for National Statistics (ONS) for facili-tating the research undertaken there.

We acknowledge the Department of Trade and Industry; the Economic and SocialResearch Council; the Advisory, Conciliation and Arbitration Service; and the PolicyStudies Institute as the originators of the 2004 Workplace Employment RelationsSurvey data, and the Data Archive at the University of Essex as the distributor of thedata. The National Centre for Social Research was commissioned to conduct thesurvey fieldwork on behalf of the sponsors. None of these organisations bears anyresponsibility for the analysis and interpretations of the data here.

This work contains statistical data from ONS which is Crown copyright andreproduced with the permission of the controller of HMSO and Queen’s Printer forScotland. The use of the ONS statistical data in this work does not imply the endorse-ment of the ONS in relation to the interpretation or analysis of the statistical data.



This work uses research data sets which may not exactly reproduce National Statisticsaggregates.

References

Age Positive (2008), Age Isn’t An Issue: Employers’ Guide to Your 21st Century Workforce(London, Department for Work and Pensions).

Anderson, T. and H. Metcalf (2003), The Wider Benefits of Diversity (London, CIPD).Barrington, L. and K. Troske (2001), ‘Workforce Diversity and Productivity: An Analysis of

Employer-employee Matched Data’. The Conference Board, Economics Program WorkingPaper Series, EPWP #01-02.

Becker, G. (1971), The Economics of Discrimination, 2nd ed (Chicago, IL, University ofChicago Press).

Bevan, S., S. Dench, P. Tamkin and J. Cummings (1999), Family-Friendly Employment: TheBusiness Case Department for Education and Employment Research Report RR136 (London,Department for Education and Employment).

Bhattacharya, J. and W. Vogt (2007), ‘Do Instrumental Variables Belong in PropensityScores?’, NBER Technical Working Paper 343.

Bryson, A. (2000), ‘Have British Workers Lost their Voice, or Have they Gained a New One’.London, Policy Studies Institute Research Discussion Paper No. 2.

Department of Trade and Industry (2005), Workplace Employment Relations Survey: Cross-Section, 2004 [computer file]. 2nd ed. Colchester: The Data Archive [distributor], April 2007.SN: 5294.

Dex, S. and C. Smith (2001), Which British Employers Have Family-friendly Policies? Analysisof the 1998 Workplace Employee Relations Survey. Research papers in management studies,WP 17/2001. Cambridge: The Judge Institute of Management Studies.

Dex, S., C. Smith and S. Winter (2001), Effects of Family-friendly Policies on Business Per-formance, Research papers in management studies, WP 22/2001. Cambridge: The JudgeInstitute of Management Studies.

Dickens, L. (1994), ‘The Business Case for Women’s Equality: Is the Carrot Better Than theStick?’, Employee Relations, 16, 8, 5–18.

Forth, J. and R. McNabb (2007), Innovations in WERS 2004: The Collection of ObjectiveData on Workplace Performance. WERS 2004 Information and Advice Service TechnicalPaper No. 1.

Forth, J. and A. Rincon-Aznar (2008), Equal Opportunities, Employee Attitudes and WorkplacePerformance Employment Relations Research Report No.38 (London, BERR).

Gray, H. (2002), ‘Family Friendly Working: What a Performance! An Analysis of theRelationship Between the Availability of Family Friendly Policies and EstablishmentPerformance’. London, Centre for Economic Performance Discussion Paper No 529.

Heckman, J. and S. Navarro-Lozano (2004), ‘Using Matching, Instrumental Variables, andControl Functions to Estimate Economic Choice Models’, Review of Economics and Statis-tics, 86, 1, 30–57.

Holtermann, S. (1995), ‘The Costs and Benefits to British Employers of Measures to PromoteEquality of Opportunity’, Gender, Work and Organization, 2, 3, 102–112.

Hon, L. and B. Brunner (2000), ‘Diversity Issues and Public Relations’, Journal of PublicRelations Research, 12, 4, 309–340.

Jehn, K., G. Northcraft and M. Neale (1999), ‘Why Differences Make A Difference: A FieldStudy of Diversity, Conflict and Performance in Work Groups’, Administrative ScienceQuarterly, 44, 4, 741–763.

Judge, T., C. Thoresen, J. Bono and G. Patton (2001), ‘The Job Satisfaction-Job PerformanceRelationship: A Qualitative and Quantitative Review’, Psychological Bulletin, 127, 376–407.



Lang, K. (1986), ‘A Language Theory of Discrimination’, Quarterly Journal of Economics, 101,2, 363–382.

Metcalf, H. and J. Forth (2000), The Business Benefits of Race Equality at Work ResearchReport 177 (London, DfEE).

Meyer, J., D. Stanley, L. Herscovich and L. Topolnytsky (2002), ‘Affective, Continuance, andNormative Commitment to the Organization: A Meta-analysis of Antecedents, Correlates,and Consequences’, Journal of Vocational Behavior, 61, 20–52.

Monfardini, C. and R. Radice (2008), ‘Testing Exogeneity in the Bivariate Probit Model: AMonte Carlo Study’, Oxford Bulletin of Economics and Statistics, 70, 2, 271–282.

Noon, M. and K. Hoque (2004), ‘Equal Opportunities Policy and Practice in the UK: Evalu-ating the “Empty Shell” Hypothesis’, Work, Employment and Society, 18, 3, 481–506.

Osborne, E. (2000), ‘The Deceptively Simple Economics of Workplace Diversity’, Journal ofLabor Research, 21, 3, 463–475.

Pérotin, V. and A. Robinson (2000), ‘Employee Participation and Equal Opportunities Prac-tices: Productivity Effect and Potential Complementarities’, British Journal of IndustrialRelations, 38, 4, 557–583.

Rhoades, L. and R. Eisenberger (2002), ‘Perceived Organizational Support: A Review of theLiterature’, Journal of Applied Psychology, 87, 698–714.

Riketta, M. (2002), ‘Attitudinal Organizational Commitment and Job Performance: A Meta-analysis’, Journal of Organizational Behavior, 23, 257–266.

Robinson, G. and K. Dechant (1997), ‘Building A Business Case for Diversity’, Academy ofManagement Executive, 11, 3, 21–31.

Shapiro, G. (2000), ‘Employee Involvement: Opening the Diversity Pandora’s Box?’, PersonnelReview, 29, 3, 304–323.

Task Force on Race Equality and Diversity in the Private Sector (2004), Race Equality: TheBenefits for Responsible Business (London, IPPR).

Thomas, R. (1991), Beyond Race and Gender: Unleashing the Power of Your Total Workforce byManaging Diversity (New York, AMACOM).

Thorsteinson, T. (2003), ‘Job Attitudes of Part-time vs. Full-time Workers: A Meta-analyticReview’, Journal of Occupational and Organizational Psychology, 76, 151–177.

Women and Equality Unit (2003), Business Case for Equality and Diversity (London, Depart-ment of Trade and Industry).

Wright, P., S. Ferris, J. Hiller and M. Kroll (1995), ‘Competitiveness Through Management ofDiversity: Effects on Stock Market Valuation’, Academy of Management Journal, 38, 1,272–287.

Wright, T. and D. Bonett (2002), ‘The Moderating Effects of Employee Tenure on the Relationbetween Organizational Commitment and Job Performance: A Meta-analysis’, Journal ofApplied Psychology, 87, 6, 1183–1190.



AP

PE

ND

IXT

able

A1:

Det

ails

ofpr

open

sity

scor

em

atch

ing

Out

com

eva

riab

le:

Subj

ecti

vein

dica

tor

ofab

ove

aver

age

prod

ucti

vity

perf

orm

ance

Subj

ecti

vein

dica

tor

ofab

ove

aver

age

finan

cial

perf

orm

ance

(pro

fits)

Equ

alop

port

unit

ies

mea

sure

:F

orm

alpo

licy

Impa

ctm

easu

rem

ent

Rev

iew

ing

For

mal

polic

yIm

pact

mea

sure

men

tR

evie

win

g

Mea

npr

open

sity

scor

ein

unm

atch

edsa

mpl

e(s

tand

ard

erro

r):

Tre

ated

0.76

3(0

.012

)0.

135

(0.0

14)

0.19

4(0

.013

)0.

734

(0.0

17)

0.10

5(0

.014

)0.

210

(0.0

22)

Con

trol

s0.

357

(0.0

17)

0.05

1(0

.002

)0.

094

(0.0

03)

0.34

2(0

.018

)0.

036

(0.0

02)

0.07

6(0

.004

)O

bser

vati

ons

inun

mat

ched

sam

ple:

Tre

ated

1013

178

275

615

9414

4C

ontr

ols

278

1149

1052

190

733

683

Off

supp

ort:

Tre

ated

319

1220

213

1614

Mea

ndi

ffer

ence

inm

atch

edsa

mpl

e(p

-val

ue):

Indu

stri

alre

lati

ons

and

HR

MU

nion

pres

ence

0.00

6(0

.874

)-0

.100

(0.3

63)

-0.2

73(0

.000

)-0

.026

(0.5

51)

-0.1

18(0

.237

)-0

.030

(0.6

59)

Par

tici

pati

onin

retu

rns

0.18

1(0

.001

)-0

.082

(0.6

16)

-0.1

45(0

.235

)0.

221

(0.0

07)

0.10

9(0

.600

)0.

001

(0.9

91)

Par

tici

pati

onin

cont

rol

-0.1

26(0

.013

)-0

.337

(0.0

12)

-0.3

39(0

.000

)-0

.178

(0.0

11)

-0.2

93(0

.102

)-0

.145

(0.4

01)

Inde

pend

ence

inw

ork

0.18

5(0

.027

)0.

322

(0.0

69)

0.30

9(0

.001

)0.

040

(0.6

74)

-0.1

59(0

.344

)0.

309

(0.1

40)

Cul

ture

ofhe

adof

fice:

Uni

ted

Stat

es-0

.102

(0.0

16)

0.01

1(0

.617

)0.

012

(0.5

66)

Cul

ture

ofhe

adof

fice:

othe

rno

n-U

K-0

.013

(0.5

80)

-0.0

93(0

.396

)-0

.014

(0.6

25)

Influ

ence

son

empl

oyee

char

acte

rist

ics

TT

WA

popu

lati

on:%

ethn

icm

inor

ity

Indu

stry

empl

oym

ent:

%et

hnic

min

orit

y0.

169

(0.4

04)

0.39

3(0

.410

)1.

09(0

.008

)0.

250

(0.3

36)

-0.5

30(0

.408

)-0

.250

(0.6

70)

Indu

stry

empl

oym

ent:

%fe

mal

e4.

99(0

.004

)5.

00(0

.272

)11

.4(0

.000

)6.

67(0

.003

)1.

14(0

.846

)0.

310

(0.9

35)

Mar

ket

cond

itio

nsan

dco

mpe

titi

vene

ssE

xten

tde

man

dde

pend

son

qual

ity:

med

ium

-0.0

39(0

.374

)-0

.139

(0.1

98)

-0.0

80(0

.429

)E

xten

tde

man

dde

pend

son

qual

ity:

high

0.07

1(0

.144

)0.

067

(0.5

80)

0.03

3(0

.742

)St

ate

ofth

em

arke

t:m

atur

e0.

029

(0.4

34)

-0.1

69(0

.025

)-0

.036

(0.6

58)

-0.0

16(0

.784

)-0

.061

(0.6

08)

0.09

1(0

.446

)St

ate

ofth

em

arke

t:de

clin

ing

ortu

rbul

ent

-0.0

27(0

.488

)0.

132

(0.1

06)

0.02

5(0

.786

)-0

.076

(0.1

38)

-0.0

36(0

.784

)0.

025

(0.7

53)

Stat

eof

the

mar

ket:

mat

ure,

decl

inin

gor

turb

ulen

tT

radi

ngin

the

inte

rnat

iona

lmar

ket

-0.1

02(0

.004

)-0

.123

(0.0

54)

-0.1

68(0

.025

)-0

.068

(0.1

62)

-0.2

23(0

.043

)-0

.054

(0.2

29)



Tab

leA

1:C

onti

nued

Out

com

eva

riab

le:

Subj

ecti

vein

dica

tor

ofab

ove

aver

age

prod

ucti

vity

perf

orm

ance

Subj

ecti

vein

dica

tor

ofab

ove

aver

age

finan

cial

perf

orm

ance

(pro

fits)

Equ

alop

port

unit

ies

mea

sure

:F

orm

alpo

licy

Impa

ctm

easu

rem

ent

Rev

iew

ing

For

mal

polic

yIm

pact

mea

sure

men

tR

evie

win

g

Skill

sM

anag

eria

l/pro

fess

iona

locc

upat

ions

(%of

empl

oyee

s)2.

38(0

.194

)10

.0(0

.015

)6.

08(0

.023

)R

outi

neun

skill

edoc

cupa

tion

s(%

ofem

ploy

ees)

Age

16–2

1(%

ofem

ploy

ees)

0.58

7(0

.736

)-1

.11

(0.6

90)

-0.8

90(0

.794

)E

stab

lishm

ent

char

acte

rist

ics

Ow

ners

hip:

part

lyfo

reig

n0.

070

(0.0

11)

-0.0

48(0

.617

)0.

074

(0.1

86)

Ow

ners

hip:

pred

omin

antl

yfo

reig

n-0

.129

(0.0

25)

-0.1

64(0

.201

)0.

031

(0.6

55)

Est

ablis

hmen

tsi

ze:5

0em

ploy

ees

orm

ore

-0.2

89(0

.000

)-0

.454

(0.0

00)

-0.4

28(0

.000

)-0

.185

(0.0

00)

-0.2

63(0

.036

)-0

.346

(0.0

00)

Org

anis

atio

nsi

ze:1

00em

ploy

ees

orm

ore

-0.0

02(0

.939

)-0

.110

(0.3

00)

-0.0

93(0

.051

)0.

058

(0.1

68)

0.02

2(0

.869

)-0

.054

(0.4

81)

Par

t-ti

me

wor

king

(%of

empl

oyee

s)8.

40(0

.010

)9.

20(0

.106

)7.

55(0

.358

)Y

oung

esta

blis

hmen

tIn

dust

rySI

C2

-0.0

34(0

.154

)-0

.012

(0.2

37)

-0.0

05(0

.352

)-0

.011

(0.3

93)

SIC

30.

010

(0.5

52)

0.03

2(0

.438

)0.

002

(0.9

10)

0.00

7(0

.753

)0.

127

(0.0

98)

0.01

2(0

.255

)SI

C4

-0.0

42(0

.372

)-0

.086

(0.2

21)

0.10

9(0

.198

)0.

026

(0.6

46)

0.11

8(0

.379

)0.

110

(0.3

15)

SIC

50.

043

(0.0

29)

-0.0

12(0

.565

)0.

037

(0.3

75)

0.05

7(0

.009

)0.

017

(0.6

12)

-0.0

24(0

.752

)SI

C6

-0.0

11(0

.705

)0.

033

(0.5

03)

-0.0

63(0

.049

)-0

.027

(0.4

36)

-0.0

50(0

.665

)-0

.031

(0.3

58)

SIC

70.

062

(0.0

10)

0.07

0(0

.281

)0.

047

(0.3

38)

0.05

2(0

.010

)0.

011

(0.6

00)

-0.0

25(0

.541

)SI

C8

-0.0

17(0

.625

)0.

035

(0.6

64)

0.00

0(0

.996

)-0

.041

(0.4

85)

-0.0

33(0

.678

)0.

078

(0.5

30)

SIC

90.

001

(0.1

78)

SIC

100.

012

(0.3

27)

0.01

3(0

.498

)0.

011

(0.4

20)

0.00

6(0

.316

)0.

016

(0.3

23)

SIC

110.

029

(0.2

35)

0.06

4(0

.478

)0.

133

(0.0

18)

-0.0

09(0

.808

)-0

.052

(0.4

61)

-0.0

76(0

.476

)SI

C12

-0.0

17(0

.396

)0.

015

(0.7

17)

-0.0

32(0

.498

)0.

056

(0.0

03)

-0.0

53(0

.506

)-0

.040

(0.6

21)



Tab

leA

1:C

onti

nued

Out

com

eva

riab

le:

Log

gros

sva

lue

adde

dpe

rem

ploy

eeL

ogpr

ofits

per

empl

oyee

Equ

alop

port

unit

ies

mea

sure

:F

orm

alpo

licy

Impa

ctm

easu

rem

ent

Rev

iew

ing

For

mal

polic

yIm

pact

mea

sure

men

tR

evie

win

g

Mea

npr

open

sity

scor

ein

unm

atch

edsa

mpl

e(s

tand

ard

erro

r):

Tre

ated

0.78

9(0

.026

)0.

216

(0.0

77)

0.38

6(0

.076

)0.

827

(0.0

30)

0.23

3(0

.085

)0.

540

(0.0

86)

Con

trol

s0.

341

(0.0

29)

0.06

0(0

.006

)0.

076

(0.0

12)

0.34

1(0

.038

)0.

046

(0.0

06)

0.07

0(0

.010

)O

bser

vati

ons

inun

mat

ched

sam

ple:

Tre

ated

294

4570

296

4673

Con

trol

s12

437

037

312

437

337

4O

ffsu

ppor

t:T

reat

ed15

53

2314

43

27M

ean

diff

eren

cein

mat

ched

sam

ple

(p-v

alue

):In

dust

rial

rela

tion

san

dH

RM

Uni

onpr

esen

ce-0

.201

(0.0

60)

-0.1

12(0

.540

)-0

.089

(0.5

21)

0.28

8(0

.026

)-0

.391

(0.0

84)

-0.1

00(0

.367

)P

arti

cipa

tion

inre

turn

s-0

.059

(0.5

48)

0.10

4(0

.721

)0.

408

(0.4

73)

-0.3

16(0

.280

)-0

.628

(0.2

20)

-0.0

53(0

.858

)P

arti

cipa

tion

inco

ntro

l-0

.192

(0.1

62)

-0.6

03(0

.108

)-0

.223

(0.2

64)

Inde

pend

ence

inw

ork

Cul

ture

ofhe

adof

fice:

Uni

ted

Stat

es0.

018

(0.0

28)

-0.1

12(0

.398

)-0

.152

(0.2

41)

0.00

2(0

.091

)-0

.246

(0.2

79)

0.00

7(0

.610

)C

ultu

reof

head

offic

e:ot

her

non-

UK

0.06

7(0

.608

)0.

012

(0.9

59)

0.00

2(0

.213

)0.

168

(0.1

90)

0.13

0(0

.366

)-0

.014

(0.3

36)

Influ

ence

son

empl

oyee

char

acte

rist

ics

TT

WA

popu

lati

on:%

ethn

icm

inor

ity

-2.4

7(0

.256

)-1

.01

(0.7

20)

0.02

1(0

.995

)-6

.92

(0.0

02)

-0.7

86(0

.907

)-0

.053

(0.9

90)

Indu

stry

empl

oym

ent:

%et

hnic

min

orit

yIn

dust

ryem

ploy

men

t:%

fem

ale

Mar

ket

cond

itio

nsan

dco

mpe

titi

vene

ssE

xten

tde

man

dde

pend

son

qual

ity:

med

ium

Ext

ent

dem

and

depe

nds

onqu

alit

y:hi

ghSt

ate

ofth

em

arke

t:m

atur

eSt

ate

ofth

em

arke

t:de

clin

ing

ortu

rbul

ent

-0.0

04(0

.968

)-0

.094

(0.6

94)

0.02

9(0

.935

)St

ate

ofth

em

arke

t:m

atur

e,de

clin

ing

ortu

rbul

ent

-0.1

06(0

.582

)-0

.118

(0.5

59)

0.11

6(0

.434

)T

radi

ngin

the

inte

rnat

iona

lmar

ket



Tab

leA

1:C

onti

nued

Out

com

eva

riab

le:

Log

gros

sva

lue

adde

dpe

rem

ploy

eeL

ogpr

ofits

per

empl

oyee

Equ

alop

port

unit

ies

mea

sure

:F

orm

alpo

licy

Impa

ctm

easu

rem

ent

Rev

iew

ing

For

mal

polic

yIm

pact

mea

sure

men

tR

evie

win

g

Skill

sM

anag

eria

l/pro

fess

iona

locc

upat

ions

(%of

empl

oyee

s)R

outi

neun

skill

edoc

cupa

tion

s(%

ofem

ploy

ees)

1.28

(0.7

05)

3.81

(0.5

82)

-17.

1(0

.011

)A

ge16

–21

(%of

empl

oyee

s)E

stab

lishm

ent

char

acte

rist

ics

Ow

ners

hip:

part

lyfo

reig

nO

wne

rshi

p:pr

edom

inan

tly

fore

ign

Est

ablis

hmen

tsi

ze:5

0em

ploy

ees

orm

ore

Org

anis

atio

nsi

ze:1

00em

ploy

ees

orm

ore

-0.0

27(0

.343

)-0

.041

(0.8

79)

-0.0

13(0

.886

)0.

180

(0.1

50)

-0.3

16(0

.072

)0.

030

(0.7

79)

Par

t-ti

me

wor

king

(%of

empl

oyee

s)4.

31(0

.329

)15

.1(0

.021

)21

.6(0

.087

)5.

42(0

.334

)0.

038

(0.9

97)

5.45

(0.3

89)

You

nges

tabl

ishm

ent

-0.0

38(0

.182

)0.

004

(0.3

50)

0.00

3(0

.358

)-0

.006

(0.3

28)

0.00

3(0

.359

)0.

001

(0.7

78)

Indu

stry

SIC

20.

002

(0.3

51)

0.00

2(0

.303

)-0

.006

(0.5

63)

0.00

1(0

.252

)SI

C3

-0.0

19(0

.369

)-0

.006

(0.3

49)

0.00

2(0

.236

)-0

.035

(0.2

73)

-0.4

61(0

.024

)-0

.000

(0.9

61)

SIC

40.

155

(0.0

80)

0.25

4(0

.133

)0.

135

(0.2

70)

0.15

5(0

.040

)0.

081

(0.4

37)

0.25

5(0

.067

)SI

C5

0.02

6(0

.146

)-0

.004

(0.3

25)

0.00

6(0

.788

)0.

051

(0.3

29)

SIC

60.

135

(0.2

47)

-0.1

18(0

.333

)-0

.001

(0.7

92)

0.10

8(0

.397

)0.

155

(0.2

75)

-0.0

03(0

.211

)SI

C7

-0.0

03(0

.102

)0.

006

(0.2

08)

0.24

9(0

.290

)-0

.021

(0.5

37)

0.00

4(0

.223

)-0

.160

(0.1

76)

SIC

8-0

.120

(0.1

59)

-0.1

67(0

.285

)0.

010

(0.9

50)

-0.3

64(0

.003

)0.

089

(0.1

47)

0.00

2(0

.928

)SI

C9

SIC

100.

002

(0.1

09)

-0.2

90(0

.190

)-0

.072

(0.2

07)

0.00

7(0

.358

)SI

C11

0.10

9(0

.089

)0.

050

(0.5

73)

-0.0

16(0

.853

)0.

106

(0.1

20)

0.29

0(0

.186

)-0

.202

(0.0

61)

SIC

120.

033

(0.4

67)

0.06

5(0

.188

)0.

085

(0.2

32)

0.08

6(0

.028

)-0

.051

(0.6

08)

0.07

0(0

.481

)

Sour

ce:

Wor

kpla

ceE

mpl

oym

ent

Rel

atio

nsSu

rvey

2004

and

Ann

ualB

usin

ess

Inqu

iry.

Not

es:

Pre

dict

ion

ofth

epr

open

sity

scor

eis

base

don

the

cont

rols

show

nin

Tab

le2

and

take

sin

toac

coun

tsur

vey

wei

ghts

;one

-to-

one

near

estn

eigh

bour

mat

chin

gw

ith

repl

acem

enti

mpo

sing

com

mon

supp

ort;

trea

ted

obse

rvat

ions

off-

supp

ort

incl

ude

thos

ew

hose

esti

mat

edpr

open

sity

scor

eis

grea

ter

(les

s)th

anth

em

axim

um(m

inim

um)

esti

mat

edpr

open

sity

scor

eob

serv

edfo

rth

eco

ntro

lsan

dth

ose

for

who

mw

eca

nnot

find

aco

ntro

lw

ith

anes

tim

ated

prop

ensi

tysc

ore

wit

hin

ara

nge

of0.

002;

mea

ndi

ffer

ence

sbe

twee

nth

etr

eate

dan

dco

ntro

lsin

the

mat

ched

sam

ple

are

calc

ulat

edus

ing

surv

eyw

eigh

tsfo

rth

etr

eate

dan

dar

esh

own

inbo

ldfa

cew

hen

stat

isti

cally

sign

ifica

ntat

the

5%le

vel.



Tab

leA

2:D

etai

lsof

inst

rum

enta

lvar

iabl

esan

alys

is

Out

com

eva

riab

le:

Subj

ecti

vein

dica

tor

ofab

ove

aver

age

prod

ucti

vity

perf

orm

ance

Subj

ecti

vein

dica

tor

ofab

ove

aver

age

finan

cial

perf

orm

ance

(pro

fits)

Equ

alop

port

unit

ies

mea

sure

:F

orm

alpo

licy

Impa

ctm

easu

rem

ent

Rev

iew

ing

For

mal

polic

yIm

pact

mea

sure

men

tR

evie

win

g

Tes

tfo

rex

ogen

eity

ofth

eIV

s:H

Rm

anag

er/o

wne

rfe

mal

ec2 (1

)=

0.42

(0.5

18)

c2 (1)

=0.

21(0

.646

)H

Rm

anag

er/o

wne

rqu

alifi

edc2 (1

)=

0.24

(0.6

24)

c2 (1)

=0.

65(0

.419

)H

Rm

anag

er/o

wne

rfe

mal

ean

dqu

alifi

edc2 (2

)=

0.53

(0.7

67)

c2 (2)

=0.

70(0

.703

)T

est

for

wea

kIV

s:H

Rm

anag

er/o

wne

rfe

mal

e,c2 (1

)9.

71(0

.002

)2.

91(0

.088

)0.

03(0

.865

)7.

88(0

.005

)5.

90(0

.015

)0.

50(0

.478

)H

Rm

anag

er/o

wne

rqu

alifi

ed,c

2 (1)

14.4

(0.0

00)

10.5

(0.0

01)

15.6

(0.0

00)

20.3

(0.0

00)

8.33

(0.0

04)

6.06

(0.0

14)

HR

man

ager

/ow

ner

fem

ale

and

qual

ified

,c2 (2

)20

.4(0

.000

)11

.5(0

.003

)16

.6(0

.000

)28

.0(0

.000

)13

.9(0

.001

)7.

59(0

.023

)T

est

for

EO

exog

enei

ty:

Cor

rela

tion

ofer

ror

term

s0.

084

(0.7

94)

-0.1

88(0

.784

)-0

.427

(0.4

07)

-0.3

96(0

.152

)-0

.977

(0.0

00)

-0.2

20(0

.619

)

Out

com

eva

riab

le:

Log

gros

sva

lue

adde

dpe

rem

ploy

eeL

ogpr

ofits

per

empl

oyee

Equ

alop

port

unit

ies

mea

sure

:F

orm

alpo

licy

Impa

ctm

easu

rem

ent

Rev

iew

ing

For

mal

polic

yIm

pact

mea

sure

men

tR

evie

win

g

Tes

tfo

rex

ogen

eity

ofth

eIV

s:H

Rm

anag

er/o

wne

rfe

mal

eF

(1,4

21)

=2.

09(0

.149

)F

(1,4

25)

=2.

61(0

.107

)H

Rm

anag

er/o

wne

rqu

alifi

edF

(1,4

21)

=1.

27(0

.261

)F

(1,4

25)

=2.

30(0

.130

)H

Rm

anag

er/o

wne

rfe

mal

ean

dqu

alifi

edF

(2,4

20)

=1.

69(0

.185

)F

(2,4

24)

=2.

37(0

.095

)T

est

for

wea

kIV

s:H

Rm

anag

er/o

wne

rfe

mal

e,c2 (1

)3.

95(0

.047

)15

.6(0

.000

)0.

84(0

.358

)2.

15(0

.142

)19

.1(0

.000

)0.

89(0

.345

)H

Rm

anag

er/o

wne

rqu

alifi

ed,c

2 (1)

16.8

(0.0

00)

13.0

(0.0

00)

7.63

(0.0

06)

15.8

(0.0

00)

10.9

(0.0

01)

7.98

(0.0

05)

HR

man

ager

/ow

ner

fem

ale

and

qual

ified

,c2 (2

)17

.7(0

.000

)26

.8(0

.000

)8.

41(0

.015

)16

.4(0

.000

)27

.6(0

.000

)8.

84(0

.012

)T

est

for

EO

exog

enei

ty:

Cor

rela

tion

ofer

ror

term

s0.

527

(0.1

67)

-0.7

24(0

.363

)-0

.104

(0.7

61)

0.66

5(0

.066

)-0

.563

(0.8

48)

0.21

4(0

.541

)

Sour

ce:

Wor

kpla

ceE

mpl

oym

ent

Rel

atio

nsSu

rvey

2004

and

Ann

ualB

usin

ess

Inqu

iry

Not

es:

pV

alue

sin

brac

kets

;tes

tfo

rex

ogen

eity

ofth

eIV

sis

aW

ald

test

ofth

ehy

poth

esis

that

the

mar

gina

leff

ect

ofth

eIV

isze

ro(a

sop

pose

dto

non-

zero

)in

the

mod

elsp

ecifi

edin

Tab

le2;

test

for

wea

kIV

sis

aW

ald

test

ofth

ehy

poth

esis

that

the

mar

gina

lef

fect

ofth

eIV

isze

roin

apr

obit

mod

elof

equa

lop

port

unit

ies

sele

ctio

nth

atin

clud

esth

eco

vari

ates

used

inth

eou

tcom

em

odel

spec

ified

inT

able

2;th

eco

rrel

atio

nof

the

erro

rte

rms

refe

rsto

the

corr

elat

ion

betw

een

the

erro

rte

rmin

the

equa

tion

for

equa

lopp

ortu

niti

esse

lect

ion

and

the

erro

rte

rmin

the

equa

tion

for

busi

ness

perf

orm

ance

.The

test

ofw

heth

erth

isco

rrel

atio

nis

zero

isa

test

ofw

heth

erth

eeq

ualo

ppor

tuni

ties

mea

sure

isex

ogen

ous

toth

ebu

sine

sspe

rfor

man

cem

easu

re.



the business case for equal opportunities

Documents