HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003 21
The impact of HR practices on the performance of business units
Patrick M. Wright, Cornell University
Timothy M. Gardner, Brigham Young University
Lisa M. Moynihan, London Business School
Human Resource Management Journal, Vol 13 No 3, 2003, pages 21-36
This article examines the impact of HR practices and organisational commitmenton the operating performance and pro® tability of business units. Using a predictivedesign with a sample of 50 autonomous business units within the same corporation,the article reveals that both organisational commitment and HR practices aresignificantly related to operational measures of performance, as well as 代写留学生论文operatingexpenses and pre-tax pro® ts.Contact: Patrick M. Wright, Department of Human Resource Studies, Schoolof Industrial and Labor Relations, Cornell University, Ithaca, NY 14853-3901.
Email: pmw6@cornell.edu
Firms have increasingly recognised the potential for their people to be a source ofcompetitive advantage (Pfeffer, 1994). Creating competitive advantage throughpeople requires careful attention to the practices that best leverage these assets.This change in the mindset of executive decision-makers has spurred an increasingbody of academic research attempting to reveal a relationship between a ® rm’s HRpractices and its performance.Much of this research has demonstrated statistically significant relationshipsbetween measures of HR practices and firm profitability (Delery and Doty, 1996;Guthrie, 2001; Huselid, 1995). While these studies have been useful for demonstratingthe potential value created through HR practices, they have revealed very regarding the processes through which this value is created (Wright and Gardner,2002). Some authors have referred to this as the `black box’ problem, noting that theonceptual development of the mediating mechanisms through which HRM has animpact on pro® tability has thus far eluded empirical testing (eg Purcell et al, 2003).In addition, the vast majority of studies examining the relationship betweenHR practices and firm performance have been entirely cross-sectional in theirdesign. Again, while providing useful information, such designs are somewhatproblematic. In essence, cross-sectional designs preclude making any causal inferences
regarding the direction of the relationship. So, whi le we may bel ieve theHR practices are driving firm performance, we cannot rule out that the reverse isactually the case.
Thus, the purpose of this study is to examine the relationship between HR practicesand ® rm performance in a way that improves the causal inferences that can be drawn.This article goes beyond previous work in three ways. First, it examines thephenomenon at the business unit level, thus minimising the amount of potential`noise’ introduced when studying more heterogeneous HR systems across variousbusinesses within corporations. Secondly, it uses more proximal measures of businessunit performance rather than only the distal profitability or stock price measures.
Finally, it uses a predictive research design enabling more con® dent causal inferences.
The impact of HR practices on the performance of business units
PREVIOUS RESEARCH
The body of research examining the relationship between HR practices and firm
performance has grown exponentially over the past few years. The seminal work in this
area was produced by Huselid (1995), who examined the relationship between HR
practices and corporate turnover, profitability and market value. Huselid (1995)
surveyed senior HR executives in a sample of 968 publicly traded corporations in the
US regarding the percentage of employees who were covered by a set of HR practices
generally considered representative of a high-performance work system (HPWS). After
controlling for a number of variables, he found that his HR index was signi® cantly
related to the gross rate of return on assets (a measure of pro® tability) and Tobin’s Q
(the ratio of the market value of a ® rm to its book value). This study provided the
foundation for much of the research that followed.
Delery and Doty (1996) examined the relationship between HR practices and
pro® tability in a sample of banks in the US. In testing universalistic, contingency and
con® gurational approaches to HRM, they found that, in general, HR practices were
positively related to pro® tability. Guthrie (2001) examined the impact of HR practices
on turnover and ® rm productivity among a sample of ® rms in New Zealand. He noted
that HR practices had an impact on turnover, and that the relationship between
retention and productivity was positive when ® rms implemented high-involvement
HR practices, but negative when they did not.
Two major studies at the plant level have been conducted examining the relationship
between HR practices and firm performance. MacDuffie (1995) found that the HR
practice `bundles’ he measured were related to quality and productivity on auto
assembly lines. Meanwhile, Youndt et al (1996) discovered that human capitalenhancing
HR practices were related to operational performance among a sample of
manufacturing plants.
While much of the research on the relationship between HR practices and
performance has somewhat consistently revealed a signi® cant relationship, some recent
debates have emerged regarding the value of different approaches to studying this
phenomenon. Debates have arisen regarding the proper sources for gaining the most
valid reports of HR practice measures, the proper level of analysis and proximity of
performance measures, and the timing of measurement.
Sources of HR practice measures
Regarding the use of single respondent designs, Gerhart et al (2000b) provided evidence
calling into question the reliability of measures of HR practices stemming from single
respondents. They found single-rater reliabilities to be frighteningly low. These results
were largely replicated by Wright et al (2001). Together, these two articles (consisting of
four studies) suggested that the reliability of single raters may be close to zero.
Huselid and Becker (2000), in response to Gerhart et al`s (2000b) article, suggested that
in many cases single respondents (ie senior HR executives) were the best placed, and
perhaps the only ones quali® ed, to provide HR practice information across a number of
jobs. This led to the debate regarding the most valid source of HR practice information.
As noted above, Huselid and Becker (2000) defended their use of senior HR
executives as the most valid source of HR practice data. However, they also argued that
the construct to be measured should be the HR practices actually implemented in the
® rm rather than HR policies that were not necessarily carried out. This led Gerhart et al
(2000a) to suggest that, if one seeks to assess the actual practices, then using employees
as the source of HR practice data would be a more logical approach.
22 HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003
Outcomes and level of analysis issues
Dyer and Reeves (1995) reviewed much of the existing research on the relationship
between HR practices and performance and proposed that measures of performance
could be broken down into four categories. First, employee outcomes deal with the
consequences of the practices on employees such as their attitudes and behaviour,
particularly behaviour such as absenteeism and turnover. Organisational outcomes
focus on more operational measures of performance such as productivity, quality and
shrinkage, many or all of which would be precursors to profitability. Financial/
accounting outcomes refer to the actual ® nancial performance measures and include
expenses, revenues and pro® tability. Finally, market-based outcomes re¯ ect how the
® nancial markets value a ® rm, particularly stock price or variations of it.
Rogers and Wright (1998) reviewed the empirical research on the HR performance
relationship, surveying 29 studies reporting 80 `effect sizes’ (ie reported statistical
relationships between HR practice and performance measures), and noted two
particularly relevant trends. First, although strategic HRM focuses largely on the link
between HR and business strategy, the largest bulk of research had been conducted at
the corporate level of analysis. A lesser amount of research has used the establishment
level. Almost entirely ignored was research on the link between HR and performance
at the business unit level of analysis.
Secondly, with regard to the types of performance outcomes, they found that very
few studies had examined HR outcomes (three effect sizes examined turnover), many
had used accounting and ® nancial market measures, and the largest number of effect
sizes was observed for organisational outcomes (productivity, quality, service etc.)
However, interestingly, while 34 effect sizes used these organisational outcomes, 68 per
cent of them (25) were gathered from surveys, with only a small number coming from
company records (seven) or public databases (two). Recognising the limitations of
performance measures derived from informant surveys, we chose to utilise the
company’s archival records.
One is hard-pressed to separate the choice of outcomes from the choice of level of
analysis. For instance, Becker and Huselid (1998) argue that the corporate level of
analysis is valid because this enables the examination of shareholder wealth (a ® nancial
market outcome), which is the corporation’s raison d’tre. However, Huselid and Becker
(2000) recognised potential methodological issues at this level as they suggested that
one reason for the low reliabilities in the Gerhart et al (2000a) study was the inclusion of
large diversi® ed corporations. They noted that the original Huselid (1995) study had an
average company size of approximately 4,000 employees.
On the other hand, Wright et al (2001) questioned the usefulness and validity of
research at the corporate level of analysis. They noted that, given the potential for huge
variations in HR practices across business units and sites, the potential for gaining
accurate and valid measures of HR practices was quite low. In addition, Rogers and
Wright (1998) suggested that conceptually, studying the link between HR and business
strategy suggests focusing at the business unit level of analysis.
Regardless of the level of analysis, numerous authors have suggested the need to
better understand the processes through which HR practices might have an impact on
performance (Becker and Huselid, 1998; Dyer and Reeves, 1995; Hutchison et al, 2002;
Wright and Gardner, 2002). While a number of models have been proposed (eg Becker
and Huselid, 1998; Dyer and Reeves, 1995; Truss and Gratton, 1994), very little
empirical research has examined multiple potential linkages (Wright and Gardner,
2002). Dyer and Reeves’ (1995) categorisation of outcomes suggests that (a) some
Patrick M. Wright, Timothy M. Gardner and Lisa M. Moynihan
HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003 23
outcomes, such as HR outcomes, are more proximal to HR practices than others, and
(b) the impact that HR practices have on more distal outcomes are through the impact
on more proximal outcomes. Together, these two points suggest that to understand
how HR practices affect pro® tability, one would need to see what impact they have on
proximal outcomes (eg HR outcomes) that have an impact on more distal outcomes (eg
organisational outcomes) and consequently have an impact on the most distal
outcomes (eg pro® ts). Given the paucity of research on HR outcomes alone, and the
lack of research examining multiple outcomes in a causal chain, the existing research
base presents little empirical data to shed light on the causal process through which HR
practices affect performance.
Timing of measurement
While not obvious to most, the timing of measurement in much of the research on the
impact of HR practices on performance has precluded drawing firm, causal
conclusions of this relationship. Very few studies have used simple, cross-sectional
designs that present problems in drawing causal inferences. However, many of the
studies accepted as being somewhat predictive are not true predictive designs. For
instance, Ichniowski et al (1997) used monthly performance data from steel ® nishing
lines over a three-year period. However, they measured HR practices by asking
respondents after the three-year production period to recall which HR systems were in
place at different points during the timeframe. Similarly, Guthrie used performance
data from 1996-97 but asked respondents during that time to report the practices that
existed during 1995-96. Given the potential problems noted above with regard to the
unreliability of single-rater responses, compounded with the memory requirements to
report practices that existed from one to three years in the past, such retrospective
designs are problematic for drawing causal conclusions.
Others, while not using purely cross-sectional designs, gathered contemporaneous
data. For instance, Delery and Doty (1996) gathered HR practice data during 1992 and
used the year-end performance data. Because the year-end data includes performance
from months prior to and concurrent with the HR practice measure, it is dif® cult to
draw firm causal conclusions. Huselid (1995) gathered both contemporaneous and
subsequent year performance data and reported only the subsequent year data in his
study in order to provide more conservative effect size estimates.
As can be seen by this detailed analysis of the designs, some of the seminal studies
in the HR performance literature fail to provide predictive designs that allow the
drawing of more con® dent causal inferences. Concurrent and retrospective designs are
particularly weak for drawing causal conclusions because they may be subject to
implicit performance theories, suggesting that knowledge of ® rm performance can
in¯ uence reports of HR practices. For instance, a study by Gardner and Wright (2003)
presented executives and graduate students with ® ctitious descriptions of high and
low-performing companies and found evidence that their reports of HR practices can
be in¯ uenced by knowledge of the company’s past performance.
This article seeks to provide more de® nitive causal inferences by (a) using business
units as the level of analysis, (b) using multiple employees as the sources of HR practice
measures, (c) assessing HR, organisation and financial outcomes, and (d) using a
predictive design where the operational and financial performance measures
temporally follow the gathering of the HR and employee attitude data.
The impact of HR practices on the performance of business units
24 HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003
HYPOTHESES
To date, Becker and Huselid (1998) offer the most logical and de® nitive model of the
processes through which HR practices affect ® rm performance. They suggest that HR
practices have a direct impact on employee skills, motivation, job design and work
structures. These variables elicit certain levels of creativity, productivity and
discret ionary effort , which subsequent ly translate into improved operating
performance. This has an impact on pro® tability and growth, which in turn have a
direct impact on the ® rm’s market valuation.
The model we suggest in this article diverges slightly from the basic Becker and
Huselid (1998) model ± not so much in logic as in the actual variables measured. We
base our hypotheses on job performance theory (Campbell, 1990). According to this
theory, per formance is behaviour; people’ s actions have an impact on the
organisation’s goals. This can be positive or negative, and the behaviour can be either
prescribed as part of the job or outside the prescribed duties.
Researchers examining various task elements and role behaviour in both micro and
macro organisational behaviour literature seem to agree on three categories of job
behaviour relevant to organisational performance. First, in-role behaviour refers to
behaviour expected of employees, largely based on job requirements and commonly
accepted norms. This has also been referred to as `core task pro® ciency’ (Campbell,
1990). In essence, such behaviour entails doing what one was hired to do.
Extra-role behaviour consists of behaviour going outside the requirement for the
job and which has a positive effect on organisational performance. For instance,
helping others, redesigning processes to be more ef® cient or deviating from standard
operating procedures when necessary to serve a good customer might exemplify
extra-role behaviour. This has sometimes been referred to as citizenship behaviour
(Organ, 1988), prosocial behaviour (Brief and Motowidlo, 1986), organisational
spontaneity (George and Brief, 1992) and discretionary behaviour (MacDuf® e, 1995).
In essence, extra-role behaviour consists of going beyond the call of duty for the good
of the organisation.
Finally, counter-productive (or dysfunctional) behaviour usually consists of
activities, in-role or extra-role, that are speci® cally or implicitly aimed at harming the
organisation (Sackett and DeVore, 2000). For example, theft of materials, sabotage or
strikes are speci® cally aimed at harming the organisation’s performance, while `time
theft’ (eg spending time on personal errands or phone calls) is implicit.
The attitudes of core workers can have considerable influence on these three
categories of work behaviour in organisations. Because attitudes include behavioural
as well as affective and cognitive components (Fishbein and Ajzen, 1972), they are
important antecedents of employee participation and role behaviour in work
environments. In fact, a recent meta-analysis found that a number of business unitlevel
outcomes were positively associated with employee attitudes (Harter et al, 2002).
The present research examines the effects of a key work attitude ± organisational
commitment ± on a variety of performance outcomes of central importance to
organisational effectiveness. It assumes that these outcomes are influenced by the
different categories of job behaviour discussed above.
In addition to examining the outcomes of commitment, we posit that HR
practices are an important lever driving this type of attitude. Prior research at the
individual level of analysis supports the notion that the management practices of an
organisation influence individual employee feelings of commitment (eg Konovsky
and Cropanzano, 1991; Meyer and Allen, 1997). There are a number of ways an
Patrick M. Wright, Timothy M. Gardner and Lisa M. Moynihan
HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003 25
organisation’ s HR practices can foster a collective level of commitment in its
workforce. First, we suggest that the initial impact of HR practices on employees’
commitment to the organisation begins with selection and staffing. When firms
invest in selecting the most highly skilled people, and providing them with
increased skills through continuous training and development opportunities,
employees ® nd a workplace ® lled with well quali® ed co-workers. This makes for a
posit ive work environment by enabling them to focus on serving their own
customers successfully, doing their own job well and not having to constantly clean
up the mess of other co-workers.
Additionally, by using valid performance management systems and monetary
incentives to elicit high performance, employees can see a more direct line of sight
between their behaviour and their personal outcomes. This creates a positive work
environment where individuals feel fairly and equitably rewarded for their efforts.
Finally, having open communications and participatory systems enables employees
to both understand the organisation’s competitive position and be able to participate in
processes to help improve it. This creates a positive work environment where people
feel they are listened to and respected. An environment created by the systems
discussed above is one where people are unlikely to want to leave; they identify with
the organisation personally and want to see it succeed. This describes the construct of
organisational commitment (Porter et al, 1974).
Virtanen (2000) argues that the social nature of commitment includes such issues
as consistency of observable behaviour and loyalty, together with ideology,
conviction and value systems. Thus, commitment in¯ uences an employee’ s view of
obligations, utilities and emotions in any work situation, and thus has an impact on
the behaviour of employees. Consequently, employees who are committed to an
organisation should be motivated to exhibit higher quality in-role behaviour, exhibit
a greater volume of positive extra-role behaviour and engage in less counterproductive
behaviour relative to those who are not committed. This role behaviour
is likely to have an impact on a number of operational performance measures. For
instance, committed employees following safety rules (in-role) are less likely to be
injured and are unlikely to either exploit minor injuries or make spurious or
f icti tious injury claims (counter-productive), resul ting in fewer workers’
compensation claims.
Businesses with committed employees should also experience higher productivity
as their employees seek to better execute required behaviour, go beyond the job to
devise more ef® cient ways of working (extra-role) and are not likely to shirk or `freeride’
(counter-productive). Such businesses should also experience higher quality
performance as employees are more likely to execute job behaviour well (in-role) and
less likely to sabotage order deliveries (counter-productive). Finally, committed
employees are far less likely to steal or damage goods (counter-productive) resulting in
inventory shrinkage for the business.
Because workers’ compensation claims, productivity, quality and shrinkage all have
a direct impact on the costs of an organisation, by in¯ uencing these variables employee
commitment should have an impact on operating expenses. Profitability is largely
determined as the difference between revenue and expenses, leading to the conclusion
that, if commitment has an impact on operating performance which, in turn, has an
impact on expenses, then it should subsequently be related to pro® tability.
The impact of HR practices on the performance of business units
26 HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003
METHOD
Overview
This article reports on research that examines the relationships of both HR practices
and organisational commitment with various operational measures of performance
using a predictive research design. Employee attitude surveys were conducted and
related to subsequent performance measures collected in the three to nine-months
timeframe after the survey data was collected.
Sample
The sample consisted of 50 business units of a large food service corporation with
operations in the United States and Canada. The corporation provided products to
meet all the needs of restaurants and other food service companies (eg food, napkins,
silverware etc). In each of the units (internally they were referred to as `companies’), we
used the survey responses from employees in three core jobs ± MAs or Merchandising
Associates (ie salespeople), delivery drivers and warehouse employees. These three
jobs had the most direct impact on the customer from sale to delivery. In fact, the
company prided itself on having its army of MAs, who were extremely knowledgeable
about its products and could work effectively with their customers to ® nd the right
products, as one source of competitive advantage. However, no matter how well MAs
sold, customers were extremely dissatis® ed if the delivery of their orders was either not
on time or incorrect; these responsibilities fell to the warehousers who prepared the
deliveries and the drivers who actually delivered the orders.
The sample consisted of 5,635 respondents. Each company had an average of 38.30
MAs, 34.96 drivers, 39.44 warehousers and 112.70 employees. In order to eliminate the
possibility of common method variance (or percept-percept correlations which are
biased by collecting two measures from the same source using the same method at the
same time), we used the reports of HR practices from half of the respondents from each
organisation and the reports of organisational commitment from the other half.
This company’s management philosophy, emphasising structural ways to create an
entrepreneurial environment, presented a unique opportunity to study the relationship
between HRM and performance in a controlled ® eld setting. First, the company tried to
keep every business between $350 million and $700 million in revenues, with a
corresponding employee count of between 250 and 600 employees. If a company grew
beyond the $700 million mark, it was then divided into two. This created a sample
where size (both in terms of revenue and employees) was strongly controlled. Also, the
basic products and IT were largely uniform across all the units. While regional
differences may have resulted in different volumes and mixes of products, the products
available for sale were uniform. Similarly, while localised changes might be made to IT,
the basic systems were largely uniform. Thus, again the sample controls for
performance differences were due to products and technology.
However, while size, technology and products provided little opportunity for
variance, considerable differences existed with regard to HR strategies. Guided by the
corporate principle of earned autonomy, business units were largely free to manage
their employees however they saw ® t. Minimal uniformity in HR practices existed with
regard to legally mandated benefits, but the majority of HR practices (eg specific
selection processes and practices, pay systems, performance management systems,
training and development strategies and practices) were left to the business units to
design, develop and implement.
Patrick M. Wright, Timothy M. Gardner and Lisa M. Moynihan
HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003 27
Corporate control over the business units came from monitoring the operational and
® nancial results we report here. This is analogous to the `® nancial’ controls as opposed
to `strategic’ controls discussed in the corporate strategy literature (Rowe and Wright,
1997). Thus, this sample provided a unique opportunity to study the HR performance
relationship, where many sources of extraneous variance are controlled through design
(thus negating the need for statistical controls), while the major focus of variance
/concerned the phenomenon of interest ± HR practices.
We should note that this lack of corporate HR authority to dictate HR practices in
the business units both drove the decision to implement `climate’ surveys and
exempli® ed the problem in implementing them. The senior vice president of HRM
sought to use the survey as a means of providing empirical data, demonstrating the
positive business benefits that could be gained by managing people in a more
progressive way. He hoped (and hopes) this data will encourage business units to
develop more progressive HR practices. However, because he could not mandate
anything, not all companies participated in the survey (although the numbers were
progressively growing each year).
MEASURES
HR practices
Employee respondents in each work unit were asked whether or not nine speci® c HR
practices existed for their job category (1 = `Yes’, 2 = `No’, 3 = `I don’t know’). `I don’t
know’ responses were classi® ed as `No’ (see Table 1, opposite, for the complete listing).
The choice of the nine items was based on a compromise between what the
researchers wanted to examine and what the company was willing to allow to be
asked. We were allowed to add some actual HR practice items to the survey in return
for a reduced fee (the company was originally only interested in assessing the climate).
First, we examined some of the seminal studies described above (eg Becker and
Gerhart, 1996; Delery and Doty, 1996; Huselid, 1995; Ichniowski et al, 1997) to identify
the items that had been used in past research. We identi® ed approximately 25 items
that we wanted to use originally. However, space limitations required us to cut back the
items to 15. The company then refused to allow us to ask some questions they deemed
too sensitive in terms of potentially priming employees to wonder why they did not
have these practices (eg gainsharing and pro® t-sharing). In the end, we used items that
represented the major areas of HR practices: selection, pay for performance, training
and participation.
One training item, `On average, how many hours of formal training do employees in
this job receive each year?’, was originally written in a different response format from
the rest of the HR practice items and was re-coded to comply with the `Yes’/’No’
dichotomous response format of the other practice items. If the number of training
hours entered was equal to or greater than 15, that response was coded as 1 = `Yes’.
Hours below 15 were coded as 0 = `No’, as such low levels arguably did not represent
signi® cant investment in employee training.
Consistent with previous research, we used an additive index of these HR practices
(eg MacDuf® e, 1995; Youndt et al, 1996). Because there was no reason to believe these
practices should be conceptualised as a unidimensional construct (Delery, 1998), interrater
reliability was deemed the most appropriate reliability assessment. Intra-class
correlations were computed for this scale at the job group level because differences in
The impact of HR practices on the performance of business units
28 HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003
HR practices existed across the groups (MAs had the highest scores, followed by
drivers and warehousers, respectively).
For each individual, a ratio was calculated of the number of practices they stated
were present, divided by nine. The business unit index of HR practices was created by
taking the mean of this ratio for the half of the employees providing information about
this measure. These measures illustrated that using a single respondent (ICC (1), which
assessed the reliability of a single respondent measure), would result in extremely low
reliability, but that by using multiple respondents (ICC (2), which assessed the
reliability of using aggregated multiple respondents), the reliability of the measures is
more than adequate (average item ICC (1) = .07, average item ICC (2) = .77; scale ICC
(1) = .13, scale ICC (2) = .89).
Organisational commitment
Five items were used from two different organisational commitment scales (Meyer and
Allen, 1997; Porter et al, 1974). Sample items included: `I feel a strong sense of
belonging to this organisation’, `I am willing to work harder than I have to to help this
company succeed’ and `I am proud to be working for this company’. Items were
Patrick M. Wright, Timothy M. Gardner and Lisa M. Moynihan
HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003 29
SELECTION AND STAFFINGa
1Applicants undergo structured interviews (job-related questions, same questions
asked of all applicants and rating scales) before being hired.
2 Quali® ed employees have the opportunity to be promoted to positions of greater pay
and/or responsibility within the company.
3Applicants for this job take formal tests (paper and pencil or work sample) before
being hired.
TRAINING
4 On average, how many hours of formal training do employees in this job receive
each year?b
PAY FOR PERFORMANCE
5 Employees in this job regularly (at least once a year) receive a formal evaluation of
their performance.
6 Pay raises for employees in this job are based on job performance.
7 Employees in this job have the opportunity to earn individual bonuses (or
commission) for productivity, performance or other individual performance
outcomes.
PARTICIPATION
8 Employees in this job are involved in formal participation processes, such as
quality improvement groups, problem-solving groups, roundtable discussions, or
suggestion systems.
9 Employees in this job have a reasonable and fair complaint process.
a With the exception of those marked, the response option for these questions was, `Yes’, `No’, or `I don’t know’
b Response option was `Hours ___________’
TABLE 1 HR practice questions
averaged to create one index per person, then aggregated to the business level using
half the sample of employees providing information about commitment. Again,
aggregating over respondents resulted in good reliability ( = .86, ICC (1) = .07, ICC (2)
= .78).
Performance
Six measures of performance were provided from archival company records. Each
measure was for a six-month period, beginning three months after the administration
of the attitude survey (see Figure 1). These measures represent ed the major
performance measures tracked by the corporate headquarters as indicators of a
businesses success. `Workers’ compensation’ was the workers’ compensation expenses
incurred during the six months divided by sales; the lower the number the better.
`Quality’ was measured as 100,000 pieces per error, where each piece represented a
carton. `Shrinkage’ was measured as the percentage of inventory loss, including loss
due to spoilage, warehouse outs, inventory adjustments, cycle count adjustments,
warehouse damage, delivery shorts, delivery damage, samples shrinkage and sales
return damage. `Productivity’ was assessed as payroll expenses for all employees,
divided by the number of pieces; the lower the number the better. `Operating expenses’
consisted of all relevant business operating expenses, including warehouse, occupancy,
delivery, selling, data processing and general and administrative expenses. Finally,
`Pro® tability’ was assessed as the operating pre-tax pro® t of the business unit as a
percentage of sales where operating pre-tax pro® t was calculated as sales ± (cost of
goods sold + operating expenses + cash discounts).
RESULTS
Because of the small sample size, we examined the relationships among the relevant
variables using only bivariate correlations. We were less interested in interpreting any
speci® c results than in understanding the overall pattern of results in how HR practices
and employee commitment relate to more proximal performance measures. The
descriptive statistics and intercorrelations are provided in Table 2 (opposite).
In order to disguise the data, the overall mean was subtracted from each company’s
actual value on each variable. Consequently, the standard deviations represent the
actual variations in the real units for each variable, but the mean is always set to zero.
As can be seen in Table 2, the ® rst link in the hypothesised causal chain shows a
relationship between HR practices and organisational commitment. The observed
The impact of HR practices on the performance of business units
30 HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003
HR
practices
Expenses
Time 1 Time 2
(3-9 months)
Organisational Pro® ts
commitment
Operational performance
Workers’ compensation
Quality
Shrinkage
Productivity
FIGURE 1 Hypothesised model
Patrick M. Wright, Timothy M. Gardner and Lisa M. Moynihan
HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003 31
Mean
.0
.0
.0
.0
.0
.0
.0
.0
HR practices
Commitment
Workers’
compensation/sales
Payroll per piece
Payroll per error
Shrink
Operating expense
Pro® t
SD
.11
.27
.001
.26
223.37
.001
.01
1.03
HR Practices
.55**
-.27²
-.20
.42**
.27²
-.40**
35*
Commitment
-.44**
-.44**
.27²
-.27²
-.50**
.32*
* p < .05
** p < .01
² p < .10
To protect the con® dentiality of the company studied, all variables have been centred by subtracting their mean.
Operating
expense
-.66**
Shrink
.46**
-.43**
Piece per
error
-.52**
-.40**
.58**
Payroll per
piece
-.21
.33*
.77**
-.40**
Workers’
compensation
.50**
-.32*
.25²
.62**
-.37**
TABLE 2 Correlations of HR practice, commitment, operational and ® nancial performance measures
correlation of r = .55 (p < .01) demonstrates support for this relationship. We should
note that this correlation is based on concurrent measures, but because each measure
came from a different subset of employees, it is not subject to percept-percept bias.
The next hypothesised set of relationships was between these two variables and the
four operational performance measures. The closer proximal relationships revealed
that organisational commitment was strongly and significantly related to workers’
compensation expenses (-.44; p < .01) and productivity (-.44; p < .01), and was
marginally related to quality (.27; p = .06) and shrinkage (-.27; p = .06). More distally,
HR practices were strongly and signi® cantly related to quality (.42; p < .01), marginally
related to workers’ compensation expenses (-.27; p = .06) and shrinkage (-.27; p = .06),
and unrelated to productivity (.20; n.s.).
The HR practices and organisational commitment were also strongly and
signi® cantly related to operating expenses (-.40 and -.50, respectively; both p < .01) and
pro® tability (.35 and .32, respectively; both p < .05). Completing the causal chain, the
operating performance measures of workers’ compensation, productivity, quality and
shrinkage were all strongly and signi® cantly related to expenses (.62, .77, -.40 and .46
respectively; all p < .01) and pro® tability (-.37, -.40, .58 and -.43, respectively; all p <
.01), and expenses were strongly and signi® cantly related to pro® tability (-.66; p < .01).
While not conducting a thorough path analysis, the results seem to indicate that HR
practices have an impact on operational performance at least in part through their
impact on employee commitment (due to HR’s weaker relationships relative to
commitment), and the impact commitment has on profitability is largely through
operational performance (because of its weaker relationships relative to the operational
performance measures).
DISCUSSION
The results of this study reveal a detailed, predictive model of HR’ s impact on
pro® tability, consistent with the model hypothesised by Becker and Huselid (1998). The
measures of HR practices were demonstrated to be highly reliable, unlike past research
in this vein (Wright et al, 2001). In addition, these measures were assessed prior to the
later performance measures, thus providing a bit more weight to the (albeit not perfect)
drawing of causal inferences. The study also demonstrated the relationships between
both HR practices and organisational commitment, with a number of operational and
® nancial performance measures. Finally, the unique nature of the company studied
enabled us to control for a number of sources of extraneous variance that would exist in
cross-company, and particularly cross-industry, studies to provide a much cleaner test
of the impact HR has on ® nancial performance.
The results revealed that HR practices were strongly related to organisational
commitment. While the observed relationship’s reliance on cross-sectional data precludes
making any causal attributions, it is important to note that using separate samples for each
of the two measures eliminates common method variance as an explanation. It should also
be noted that using employees as the source of the HR practice measures ensures that the
measure represents the actual practices rather than the espoused policies of the business
(Huselid and Becker, 2000; Wright et al, 2001). Finally, using multiple employees provided
a psychometrically sound measure of these practices ± something that has rarely been
observed in the past (Gerhart et al, 2000a, 2000b; Wright et al, 2001).
These measures of practices and employee attitudes were strong predictors of
operational performance measures used within the company to track business unit
performance. The correlations ranged from .20 to .44 (in absolute values) so that even
The impact of HR practices on the performance of business units
32 HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003
the non-signi® cant relationships were strong and in the expected direction, and their
non-significance was likely due to the small sample size. Thus, the study tends to
support the hypothes ised relationships of both HR practices and employee
commitment with business unit operational performance.
Finally, both HR practices and employee commitment were strongly and
significantly related to operating expenses and profitability. While the relationship
between HR practices and pro® tability has been demonstrated before at the corporate
level (Delery and Doty, 1996; Guthrie, 2001; Huselid, 1995) and at the establishment
level (MacDuffie, 1995), this is the first study to demonstrate it at the level of the
business unit.
Thus, it seems that, when employees are managed with progressive HR practices,
they become more committed to their organisation. At least in part, this commitment
leads them to exhibit proper role behaviour (and thus lower workers’ compensation
costs, higher quality and higher productivity) and to not engage in dysfunctional
behaviour (that would result in shrinkage). Again, in part, these operational performance
outcomes result in lower overall operating expenses and higher pro® tability.
The relatively large effect sizes observed here are due to the nature of the design,
which points to both the strength and weakness of this study. Kerlinger (1973) notes
that the purpose of research design is to maximise the experimental variance,
minimise error variance and control systematic variance. The `earned autonomy’
philosophy of the corporate headquarters provided a setting which allowed for
considerable true variance in HR practices. The constrained size, technology and
products controlled the systematic variance. In addition, the use of multiple raters for
the HR and commitment measures minimised error variance (due to measurement
error), resulting in a design that maximised the possibility of ® nding support for the
hypothesised relationships. With much of the systematic and error variance
eliminated through design and methodology, the variance explained by HR practices
could constitute a larger percentage of the total variance explained. In essence, this
mimics a laboratory study conducted in the field, enabling us to more specifically
tease out the nature of the relationships we sought to study. Future research with
additional waves of data from this organisation will help us understand these
mediating relationships.
However, because these factors created the equivalent of almost a laboratory
study in a field setting, they also lead to the same criticisms that are levelled at
laboratory studies. Most importantly, one could easily criticise the generalisability of
the results. Large, cross-industry studies such as Huselid’ s (1995) are subject to
considerably more systematic and error variance, but their results are also
signi® cantly more generalisable.
We should also note that, while the design minimises a number of sources of
extraneous variance (eg size), it does not eliminate all sources. For instance, each
company was located in geographically distinct markets, and local labour and
customer market conditions may vary widely. It is impossible to tell how this variance
might affect the results but we must recognise that it may have some in¯ uence.
Another weakness of this study is its failure to actually assess the behavioural
constructs we use to hypothesise the relationships between HR/commitment and
performance. We suggest that employees are less prone to engage in counterproductive
behaviour and more likely to exhibit both proper in-role and discretionary
behaviour. However, we were unable to actually measure this, and can only assume
they existed based on the performance outcomes of such behaviour.
In addition, we should note that, while this study is predictive in that the measures
Patrick M. Wright, Timothy M. Gardner and Lisa M. Moynihan
HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003 33
of HR practices and commitment temporally preceded the measures of performance, it
is not a true longitudinal design. Consequently, we cannot draw firm causal
conclusions. However, the design is similar to a predictive validation in a selection
context in that we were assessing the predictors temporally prior to the dependent
variables. Certainly, this does not prove causation, but provides a more rigorous design
than either a cross-sectional or a retrospective design. We continue to work with this
company and hope to tease out causal direction by examining how past performance
relates to reports of HR practices and commitment, and how those relationships
compare with the kind reported here.
An additional limitation is that, while the reports came from employees who should
be best placed to report the actual HR practices that exist (as opposed to the policies
that are supposed to be implemented but may not be), there is still room for error.
Employees may not be completely accurate and, in some cases, may not be completely
knowledgeable about practices that have an impact on others. By using multiple
employees, this problem is minimised, but not eliminated.
PRACTICE IMPLICATIONS
The results of this study support the notion that businesses which manage employees
with more progressive HR practices can expect to see higher operational performance
as a result. In fact, we noted previously that the senior vice president of HR sought
this data as a way to in¯ uence these autonomous business units to more effectively
manage their people. Consequently, he has used this data in presentations at meetings
with company presidents to demonstrate the kind of performance they might gain by
developing and implementing proven HR practices. In addition, the company has
now developed a portal to aid business unit presidents whose units are not meeting
their performance goals. The website helps them identify their key performance
de® ciencies (eg workers’ compensation costs) and then provides information on the
practices they can implement that should help increase their performance on this
performance measure.
In summary, this study used a highly controlled setting and sample, and a
predictive design to better analyse the processes through which HR practices might
have an impact on firm profitability. Our results seem to indicate support for the
hypothesised model. Future research should focus on providing more detailed and
more generalisable ® ndings to add to the knowledge base, exploring how ® rms can
leverage people as a source of competitive advantage.
REFERENCES
Becker, B. and Gerhart, B. (1996). `The impact of human resource management on
organizational performance: progress and prospects’ . Academy of Management
Journal, 39, 779-801.
Becker, B.E. and Huselid, M.A. (1998). `High-performance work systems and firm
performance: a synthesis of research and managerial implications’ in Research
in Personnel and Human Resource Management, 16. G.R. Ferris (ed). Greenwich, CT:
JAI Press.
Brief, A. and Motowidlo, S. (1986). `Prosocial organizational behaviors’ . Academy of
Management Review, 11, 710-725.
Campbell, J. (1990). `Modeling the performance prediction problem in industrial and
The impact of HR practices on the performance of business units
34 HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003
organizational psychology’ in Handbook of Industrial and Organizational Psychology
(2nd ed). M. Dunnet te and L. Hough (eds). Palo Al to, CA: Consult ing
Psychologists Press.
Delery, J.E. (1998). `Issues of ® t in strategic human resource management: implications
for research’. Human Resource Management Review, 8, 289-310.
Delery, J.E. and Doty, D.H. (1996). `Modes of theorizing in strategic human resource
management: tests of universalistic, contingency and con® gurational performance
predictions’. Academy of Management Journal, 39: 4, 802-835.
Dyer, L. and Reeves, T. (1995). `HR strategies and ® rm performance: what do we know
and where do we need to go?’ International Journal of Human Resource Management, 6:
3, 656-670.
Fishbein, M. and Ajzen, I. (1972). `Attitudes and opinions’. Annual Review of Psychology,
23, 487-544.
Gardner, T and Wright, P. (2003). `The HR-® rm performance relationship: is it only
in the mind of the beholder?’ Center for Advanced HR Studies Working paper,
Ithaca, NY.
George, J. and Brief, A. (1992). `Feeling good ± doing good: a conceptual analysis of the
mood at work-organisational spontaneity relationship’. Psychological Bulletin, 112,
310-329.
Gerhart, B., Wright, P.M. and McMahan, G.C. (2000a). `Measurement error and
estimates of the HR ® rm performance relationship: further evidence and analysis’.
Personnel Psychology, 53, 855-872.
Gerhart, B., Wright, P.M., McMahan, G.C. and Snell, S.A. (2000b). `Measurement error
in research on human resources and ® rm performance: how much error is there and
how does it in¯ uence effect size estimates?’ Personnel Psychology, 53, 803-834.
Guthrie, J. (2001). `High-involvement work practices, turnover and productivity:
evidence from New Zealand’. Academy of Management Journal, 44, 180-192.
Harter, J.K., Schmidt, F.L. and Hayes, T.L. (2002). `Business unit level relationship
between employee satisfaction, employee engagement and business outcomes: a
meta-analysis’. Journal of Applied Psychology, 87, 268-279.
Hutchison, S., Kinnie, N. and Purcell, J. (2002). `Bringing policies to life: discretionary
behavior and the impact on business performance’. Paper presented at University of
Bath School of Management, 10-11 April.
Huselid, M.A. (1995). `The impact of human resource management practices on
turnover, productivity and corporate ® nancial performance’. Academy of Management
Journal, 38: 3, 635-672.
Huselid, M.A. and Becker, B.E. (2000). `Comment on measurement error in research on
human resources and ® rm performance: how much error is there and how does it
in¯ uence effect size estimates?’ Personnel Psychology, 53, 835-854.
Ichniowski C., Shaw, K. and Prennushi, G. (1997). `The effects of human resource
management practices on productivity’. American Economic Review, 87, 291-313.
Kerlinger, F.N. (1973). Foundations of Behavioral Research. New York: Holt, Rinehart and
Winston.
Konovsky, M.A. and Cropanzano, R. (1991). `Perceived fairness of employee drug
testing as a predictor of employee attitudes and job performance’. Journal of Applied
Psychology, 76: 698-707.
MacDuffie, J.P. (1995). `Human resource bundles and manufacturing performance:
organizational logic and ¯ exible production systems in the world auto industry’ .
Industrial and Labor Relations Review, 48, 197-221.
/Meyer, J.P. and Allen, N.J. (1997). Commitment in the Workplace: Theory, Research and
Patrick M. Wright, Timothy M. Gardner and Lisa M. Moynihan
HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003 35
Application, Thousand Oaks, CA: Sage Publications.
Organ, D. (1988). Organizational Citizenship Behavior: The Good Soldier Syndrome,
Lexington, MA: Lexington Books.
Pfeffer, J. (1994). Competitive Advantage Through People: Unleashing the Power of the
Workforce, Boston: Harvard Business School Press.
Porter, L.W., Steers, R.M., Mowday, R.T. and Boulin, P.V. (1974). `Organizational
commitment, job satisfaction and turnover among psychiatric technicians’. Journal of
Applied Psychology, 59, 603-609.
Purcell, J., Hutchinson, S., Kinnie, N., Rayton, B. and Swart, J. (2003). Understanding the
Pay and Performance Link: Unlocking the Black Box, London: CIPD.
Rogers, E.W. and Wright, P.M. (1998). `Measuring organizational performance in
strategic human resource management: problems, prospects and performance
information markets’. Human Resource Management Review, 8, 311-331.
Rowe, W.G. and Wright, P.M. (1997). `Related and unrelated diversi® cation and their
effect on human resource management controls’. Strategic Management Journal, 18,
329-338.
Sackett, P. and DeVore, C. (2000). `Counterproductive behaviors at work’ in Handbook of
Industrial, Work and Organizational Psychology. N. Anderson, D. Ones, H. Sinagil and
V. Chockalingam (eds). Thousand Oaks, CA: Sage.
Truss, C. and Gratton, L. (1994). `Strategic human resource management: a conceptual
approach’. The International Journal of Human Resource Management, 5, 662-686.
Virtanen, T. (2000). `Commitment and the study of organizational climate and culture’
in Handbook of Organizational Culture and Climate. N. Askenasy, C.P.M. Wilderom and
M.F. Peterson. (eds). Thousand Oaks, CA: Sage.
Wright, P.M. and Gardner, T.M. (2002). `Theoretical and empirical challenges in
studying the HR practice ± ® rm performance relationship’ in The New Workplace: A
Guide to the Human Impact of Modern Working Practices. D. Holman, T. Wall, C. Clegg,
P. Sparrow and A. Howard (eds). Chichester: John Wiley and Sons.
Wright, P., Gardner, T., Moynihan, L., Park, H., Gerhart, B. and Delery, J. (2001).
`Measurement error in research on human resources and firm performance:
additional data and suggestions for future research’. Personnel Psychology,54, 875-902.
Youndt, M.A., Snell, S. A., Dean, J.W. and Lepak, D.P. (1996). `Human resource
management , manufacturing strategy and firm performance’ . Academy of
Management Journal, 39: 4, 836-866.
The impact of HR practices on the performance of business units
36 HUMAN RESOURCE MANAGEMENT JOURNAL, VOL 13 NO 3, 2003