Is it worth it to win the Talent War? Evaluating the Utility of
Performance-Based Pay
Michael C. Sturman, Cornell University
Charlie O. Trevor, University of Wisconsin-Madison
John W. Boudreau, Cornell University
Barry Gerhart, University of Wisconsin-Madison
Although the business press suggests that "winning the talent war," the attraction and
retention of key talent, is increasingly pivotal to organization success, executives often report
that their organizations do not fare well on this dimension. We demonstrate how, through
integrating turnover and compensation research, the Boudreau and Berger (1985) staffing
utility framework can be used by I-O psychologists and other HR professionals to address this
issue. Employing a step-by-step process that combines organization-specific information about
pay and performance with research on the pay-turnover linkage, we estimate the effects of
incentive pay on employee separation patterns at various performance levels. We then use the
utility framework to evaluate the financial consequences of incentive pay as an employee
retention vehicle. The demonstration illustrates the limitations of standard accounting and
behavioral cost-based approaches and the importance of considering both the costs and
benefits associated with pay-for-performance plans. Our results suggest that traditional
accounting or behavioral cost-based approaches, used alone, would have supported rejecting a
potentially lucrative pay-for-performance investment. In addition, our approach should enable
HR professionals to use research findings and their own data to estimate the retention patterns
and subsequent financial consequences of their existing, and potential, company-specific
performance-based pay policies.
The ability to achieve competitive advantage through people depends in large part on the
composition of the work force. This, in turn, is a function of who is hired, how they are developed, and
who is retained- the latter of which is the focus of this study. Voluntary employee turnover can be either
dysfunctional or functional for the organization, depending on who leaves (Boudreau, 1991; Boudreau &
Berger, 1985; Hollenbeck & Williams, 1986; Trevor, 2001). Both low and high performers are generally
more likely to leave an organization than are average performers (Jackofsky, 1984; Trevor, Gerhart, &
Boudreau, 1997; Williams & Livingstone, 1994). Thus, organizations often will shed poor employees
(functional turnover), but will also fail to retain star employees (dysfunctional turnover). It appears,
however, that organizational practices can influence the performance distribution of leavers.
Specifically, though high performers typically may leave the organization more often than do average
performers, they do not necessarily do so. Although research consistently reports that an organization’s
pay system affects the probability of voluntary turnover (Dreher, 1982; Gerhart & Milkovich, 1992;
Griffeth, Hom, & Gaertner, 2000; Harrison, Virick, & William, 1996; Porter & Lawler, 1968; Schwab,
1991; Steers & Mowday, 1981; Trevor et al., 1997), the probability of high-performer turnover is
particularly sensitive to the strength of the pay-for-performance link (Trevor et al., 1997). Consequently,
organizations may be able to design compensation systems to enhance organizational value by targeting
retention efforts at the dysfunctional high performer turnover.
This may, in fact, be increasingly happening as organizations in the United States and abroad are
progressing toward linking pay more strongly to performance (Milkovich & Newman, 2002). Although
many organizations have expanded their use of plans that reward team, business unit, and corporate
performance (Milkovich & Newman, 2002), the predominant basis for pay-for-performance continues to
be individual performance (Hewitt Associates, 2002; IOMA, 2002), and survey data indicate that
companies believe individual pay-for-performance programs are effective (IOMA, 2002). Although there
are concerns about the wisdom of pay-for-performance (e.g., Kohn, 1993; Pfeffer, 1998), particularly for
individual performance, research reviews find ample evidence that pay-for-performance is associated
with higher performance at both the individual (Jenkins, Mitra, Gupta, & Shaw, 1998) and organizational
levels of analysis (Gerhart, 2000). Such research, however, has not explicitly examined the mechanisms
through which pay-for-performance plans affect individual behaviors to influence the organizational
bottom line. One such mechanism involves pay-for-performance’s effects on performance-specific
turnover, and the associated costs and benefits that contribute to organizational financial performance.
The professional HR literature suggests that influencing the retention of high performers in
particular is a crucial matter. Many articles cite the increasing difficulty in obtaining and keeping top
talent (e.g., Bartlett & Ghoshal, 2002; Branch, 1998; Chambers, Handfield-Jones, Hanking, & Michaels,
1998; Rich, 1999). A report based on interviews of over 5,000 executives and managers (McKinsey &
Company, 1998), for example, found that 65% of executives believed that they had insufficient talent in
the ranks of their top 300 leaders while only 10% strongly believed that their companies retained most
of their high performers. Even with the recent economic slowdown, organizations face increased
pressures to attract and retain top talent in their most pivotal talent areas. The Bureau of Labor
Statistics projects that, by 2010, the labor supply will grow by 17 million (Fullerton & Toosi, 2001) and
labor demand will increase by 22.2 million (Berman, 2001), indicating that labor shortages will play
increasing roles in the future. Moreover, even if a company is reducing employee headcount, voluntary
attrition is often the first and most attractive option (Sherwyn & Sturman, 2002). Each of these
circumstances highlights the potential benefits of managerial investments that particularly facilitate top-
performer retention.
Few would debate the merits of a performance-based pay practice that, all else equal, resulted
in greater retention of high performers. Unfortunately, all else is far from equal when changing an
organization’s pay systems. Because such changes will affect total labor costs, individual employee pay
levels, and subsequent employee behaviors, the critical question becomes one of whether the benefits
of such a practice outweigh the costs. We propose that although the potential retention benefits of
incentive pay have been recognized, they have yet to be quantified in dollar terms. Moreover,
researchers have failed to adequately address actual costs of performance-based bay. Our goal here is
to provide the first empirical cost-benefit assessment of the viability of performance-based pay. Our
approach should contribute to the pay-for-performance literature by specifying the circumstances that
affect the success of pay-for- performance plans.
Our results should also contribute to practice, as the likelihood that HR professionals would
apply the research findings to their own organizations should increase if these professionals are
provided with a viable technique for doing so. In this paper, we demonstrate such a technique. The
employee movement utility model of Boudreau and Berger (1985) provides the means to evaluate the
dollar value implications of various pay-for-performance strategies, which we illustrate with a step-by-
step application to a published turnover and pay-for-performance article. In doing so, we (a)
demonstrate how organizational representatives can use research findings, publicly available
compensation, and turnover data, or their own data to diagnose, inform, and evaluate their own
company-specific incentive pay decisions; and (b), demonstrate that this technique will often provide
different conclusions from typical decision models that use only traditional cost or accounting analysis.
Utility Analysis Applied to Pay Decisions
Utility analysis is a tool for cost-benefit analysis that helps quantify the impact of human
resource interventions (Cascio, 2000). Although utility analysis has been applied to numerous human
resource program areas, most applications have concentrated in the areas of employee selection and
training (Boudreau, 1991; Boudreau & Ramstad, 2003b; 1999). The Boudreau and Berger (1985)
framework represents one of the few applications to employee retention. Klass and McClendon (1996)
used that framework to examine the pay policy decision of whether to lead, lag, or match the market.
They gathered parameter information from published studies and simulated effects on employee
separation and offer acceptance patterns. Results for bank tellers suggested that a lag policy produced
higher payoffs, although “leading the market” (paying higher than the average) did enhance retention
and attraction of top candidates. The authors noted that these results did not necessarily suggest using
a particular pay policy, and showed how simulated reductions in citizenship behavior due to low pay
might change the results. This was an important initial application of employee movement utility
principles to decisions about pay.
In this paper, we focus on a different type of pay decision-how to allocate pay increases across
employees at different performance levels. Trevor et al. (1997) found that pay policies providing greater
pay growth for high performers (and less for low performers) substantially increased retention among
high performers, encouraged separation among low performers, and, thus, increased the value of the
work force. This is an appealing prospect, but it is unclear whether the enhanced workforce value would
offset the cost associated with such a reward system. Such costs are quite apparent using traditional
accounting or behavioral costing models, but such models have limited ability to reflect effects on
workforce value; furthermore, little data exists on the actual implications of these limitations (Boudreau
& Ramstad, 2003a; 2003b). It is also unclear to what extent the enhanced workforce value would
depend on such factors as the pay policy specifics, the retention pattern, and the variability in
performance. The Boudreau-Berger utility framework provides a method to address these questions.
Using the Boudreau and Berger (1985) separation/acquisition utility model, our paper presents a
model that captures the value associated with employee separations (turnover) and acquisitions (hires)
over time. The model estimates three components in each time period: (a) movement costs-the costs
associated with employee separations and acquisitions; (b) service costs-the pay, benefits, and
associated expenses required to support the work force; and (c) service value-the value of the goods
and services produced by the work force. The dollar-valued implications of a given pay plan, and of the
subsequent separation and acquisition patterns over time, are estimated by subtracting the movement
costs and service costs from the service value (i.e., subtracting the pay plan's costs from its benefits).
Figure 1 shows the steps necessary to compute this estimate and the tables we employ here to illustrate
these steps.
The Illustrative Case Study
We illustrate our approach using a scenario in which a hypothetical company is considering
implementing a pay-for-performance plan at the end of the year 2003. We assume that the company
does not currently relate pay to performance, so, under the current strategy, all employees would
receive the same pay increases over time. We compare the effects of this strategy with those of two
alternative strategies that place different emphases on pay-for-performance. We choose to evaluate the
implications of the three possible approaches over a 4-year period (2004 to 2007). Thus, because pay-
for-performance affects turnover differently at different levels of performance (Trevor et al., 1997), the
2007 workforce would reflect a different performance distribution under each of the three pay
strategies. By calculating the movement costs, service costs, and service values from 2004 to 2007, we
can estimate the cumulative effects of the pay strategies over the 4-year period. 1
We used a number of spreadsheets to make the necessary calculations, with each spreadsheet
corresponding to a table in this paper. The spreadsheets are available from the lead author upon
request, although the descriptions we provide here should be sufficient for many readers to create their
own. We also make a number of assumptions to perform the necessary calculations. These assumptions
are all based on published research (e.g., Trevor et al., 1997) or publicly available data (e.g., BLS, 2002).
First, we draw directly from the Trevor et al. (1997) study to estimate (a) the relationship between pay
growth, performance, and turnover that is captured in their survival analysis (see Appendix) and is used
to calculate the turnover probabilities at each performance level under each pay strategy; (b) the
baseline turnover probability necessary to compute those turnover probabilities that are specific to each
performance level-pay strategy combination; and (c) the performance distribution at the beginning of
our utility analysis timeframe.
It should be noted that the Trevor et al. (1997) data are from all 5,143 exempt employees hired
by a large petrochemical organization between 1983 and 1988. Furthermore, Trevor et al. (1997)
examined the effects of various strengths of pay-for-performance relationships based on archival data
on individuals' performance and pay levels; they did not specifically manipulate the pay-for-
1
The Boudreau and Berger (1985) model in its purest form would calculate the work force value in each
intervening year and apply a discount factor to equalize the time value of the dollar amounts. Although these
economic corrections can yield substantial changes to the estimated value (Sturman, 2000), such embellishments
do not have a significant effect in this case because the changes in dollar amounts are assumed to be linear, the
time frame is relatively short, and our focus is on the relative (vs. absolute) value of the different strategies. We
also did not have information about the organizational tax rate, so we report our results in pretax dollars. Aftertax
effects could be easily calculated by multiplying the final results by an appropriate aftertax proportion, but the
relative effects of the options would not be altered.
