Why People Stay

The Best and the Rest:

Revisiting the Norm of Normality of Individual Performance
This Research Briefing is a service from BC HRMAs research group. Our aim is to make it easier and
quicker for HR professionals to find and apply the latest and best people management insight to their
challenges and projects. This paper contains a concise and practical summary of a recent academic
finding that should shape your HR practices.
Summary

Key Definitions

In the field of human resources management,

organizational behaviour and industrial and
organizational psychology, the assumption that
individual performance follows a Gaussian
(normal) distribution the form of a bell curve
with the majority of performers clustered
around the mean has long influenced
organizational practice. The study, The Best and
the Rest: Revisiting the Norm of Normality of
Individual Performance by Ernest OBoyle Jr.
and Herman Aguinis, presents a new paradigm
for understanding why applying a normal
distribution, that assumes the majority of
individuals will perform in an average manner,
does not present an accurate picture of the way
that individual performance unfolds in an
organization. In fact, as some of us may witness
in our own organizations, the findings of this
research provide evidence that the majority of
work is carried out by a small number of people
that out-perform the rest. Thus, the picture of
performance is better represented by a
Paretian distribution, which resembles the
shape of a ski slope, where under-performers
are at the upper most starting point of the slope
and high performers are at the tip of the end.

BC HRMA 2011

Guasian Distribution - also known as a normal

distribution, the Guasian distribtion follows
the shape of a bell curve. That is to say that
results cluster around a mean and then fan
out into symetrical tails on either side. In the
case of individual performance, a Guasian
distribution would assume that there would
be a small number of non-performers and a
small number of high performers, while the
majority of indivuals would be average
performers, clustered around the mean.
Performers not conforming to the normal
distribution would be considered errors or
outliers and excluded from analysis or forced
to fit the distribution (e.g., "grading on a
curve").
Paretian Distribution - also known as a power
law distribution, Paretian distributions have
unstable means and a greater portion of
extreme events. They resemble a ski slope in
shape. In the case of individual performance,
the non-performers would be at the top of
the slope and the high performers at the
bottom. The Paretian distribution allows for
more extreme events and has fatter tails than
a normal distribution. In the case of individual
performance, the majority of performers will
fall below the mean.

- Research Briefing -

Page 1

Why People Stay

The typical performer would fall below the mean or average result. Thats right, in this model, the
typical performer is below average (speaking from a statistical perspective).
Background
Normal distributions have long been used as a model
for simplifying complex theories related to scientific
disciplines (e.g., laws of thermo dynamics,) economics
(e.g., stock pricing), human traits (e.g., height or
intelligence), among others. Perhaps most evident in
the study of performance and appraisal, we can see
how both the design of studies and the statistical
analysis used to measure performance yield to the
assumptions of a normal distribution.

Forced Response Formats - in this design situation,

a study may, for example, force a performance
rater to rate performance on a scale of 1 to 7, and
to "normalize" judgments in their distribution.

Deletion or Correction of Outliers - in statistical

analysis, a researcher may either delete or correct
outliers, based on the assumption that
performance should fit a normal distribution and
outliers are either the result of error or bias (e.g.,
the performance rater may be either too tough or
too soft) or an anomaly (e.g., a highly successful
sales person who simply has a natural talent in the
sales business and is not a fair comparator for the
average individual).

Assumptions of the Normal Distribution of

Performance
How does research based on the assumption
of the normal distrubiton of individual
performance translate into organizational
practice? Here are a few examples:
Distribution of Work / Work Objectives Work objectives (e.g., production targets) set
according to the "average" production time,
where the calculation of average excludes
outliers (e.g., people perceived to vastly over
or under perform) in order to maintain the
normal distribution assumption.
Performance Appraisal - Appraisal systems
which use rating systems and set limits on the
distribution of ratings; thus enforcing a
normal distribution.
Training & Development - Training initiatives
that focuses on improving the productivity of
the "average worker" or how to improve the
performance of the majority (e.g., the team).
Compensation - Salary grades based on the
assumption of a normal distribution of
performance for individual performers in a job
group or category which place the average
performer at the mid point and distribute pay
within the limits of a normal distribution.

Perhaps this assumption of a normal distribution may

be partially related to the lack of other models for
understanding and / or predicting individual
performance on the job, or lack of research negating
the normal distribution. Regardless, whether individuals consciously believe that the workforce is
primarily made up of average performers (performers that cluster around the mean in a normal
distribution) or not, from recruitment to performance management to leadership development, the
assumption that the majority of the workforce will be made up of average performers has had a
tremendous impact on the decisions and practices that organizations make and apply on a daily basis
(see sidebar).

BC HRMA 2011

- Research Briefing -

Page 2

Why People Stay

This study provides an alternate model for considering the distribution of individual performance
that of a Paretian distribution, also known as a power law distribution. Paretian distributions have
unstable means and a greater portion of extreme events. They resemble a ski slope in shape. In the
case of individual performance, the non-performers would be at the top of the slope and the high
performers at the bottom. In this type of distribution, most performers will typically fall below the
mean.

This diagram shows

the difference
between a Gaussian
(normal) distribution
and a Paretian
(power) distribution.
The primary
difference is that in a
Paretian distribution
the majority of the
sample is below
average.

Hypotheses and Results

The researchers posited the following hypothesis:

The distribution of performance is better modeled with a Paretian curve than a Gaussian
(normal) curve.

The hypothesis was applied to five different studies, in which a Chi-square analysis was used to force
the data to fit a Gaussian (normal) curve and a Paretian curve. The studies included 198 samples
and 633,263 researchers, entertainers, politicians, and amateur and professional athletes.
The results, based on the five separate studies, indicated that in over 93% of cases a Paretian
distribution more closely represented the distribution of performance than a Gaussian distribution.
That is to say, that most performance was generated by a small group of superstars; in other
words, its no coincidence that a large number of your favourite entertainment awards or athletic
awards go to a small group of people.

BC HRMA 2011

- Research Briefing -

Page 3

Why People Stay

In accordance with a Paretian distribution, the picture of performance resembles a ski slope. The
typical performer would fall below the mean or average result. So there would be approximately 80%
below average, 10% around the middle, and 10% exceeding; meaning that the assumption that the
typical performer is average is a myth of the normal distribution assumption. The results lend
themselves to the 80 / 20 rule, which assumes at 80% of the work is done by 20% of the people.
The researchers identified three main areas for examination based on their findings.
1. Performance Measurement & Management
The assumption of a normal distribution means that the majority of performance management
systems are based on the mean or average worker (e.g., the performance of this individual should
represent performance at the middle of the performance spectrum). Thus, if most people were to be
rated below this level of performance, the result would be determined to be an anomaly and
excluded, or either the rater or the instrument (or both) would be determined to be at fault. Not so,
according to the results of the study. Following a Paretian distribution, the majority of performers
would reside in the lower end of the performance spectrum.
To shift to a Paretian perspective, future research and innovation in performance measurement
would benefit from the development of instruments that allow for the identification of top
performers that make the greatest contributions to organizational results.
2. Utility Analysis
Utility analysis is built on the assumption of normality and specifically the standard deviation of
performance. If we assume that the distribution of performance is normal, then we assume that
performance increases at a steady and predictable rate. For example, if performance was measured
in dollar values, and each standard deviation was equivalent to $20, the performance above one
standard deviation from the mean would increase by $20, at two standard deviations from the mean
it would increase to $40, etc. A Paretian distribution says that this is not the case, that the reality is
that the performance of those individuals above the mean increases at a rate more rapid and
extreme than this.
Many leadership and training and development practices that are grounded in the Gaussian
distribution have focused on shifting the performance of the majority of workers (those clustered
around the mean), rather than the workers that do the majority of the work as displayed in a
Paretian distribution. Practices that fail to influence those top performers that generate the majority
of performance are likely to result in minimal impacts in the organization. Despite small numbers, the
impact of increasing the performance of individuals at the extreme positive end of performance will
result in a much greater impact to the organization.

BC HRMA 2011

- Research Briefing -

Page 4

Why People Stay

Further, additional research needs to be conducted to understand how the performance of top
performers interact with and are impacted by the performance of the rest of the team in a variety of
team composition scenarios (e.g., primarily high performers on a team, a single high performer, etc.),
and how the rest of the team interacts with and impacts the performance of top performers.
3. Prediction of Performance
Basing prediction-of-performance research on the Gaussian (normal) distribution has resulted in
limited ability to gain insights into the antecedents of strong performance. Perhaps this is due to the
failure of a normal curve to represent trends in individual performance and, therefore, performance
improvements due to interventions. The researchers suggest that future investigations should first
determine if a normal distribution of a Paretian distribution best fits the data set (e.g., using a Chisquare analysis as in this research). After this determination, analytical techniques in line with the
best fit model should be selected.
Implications for HR
Given the findings of this research, do organizations now shift their focus away from the majority of
the current and potential workforce and to the superstars that have the greatest organizational
impact? The findings of the study further bring into conflict traditional perceptions of fairness and
equality in the workplace and the ongoing drive to increase individual, and therefore organizational,
performance. The pay-for-performance debate demonstrates the challenges in shifting away from
Gaussian assumptions and toward Paretian assumptions.
From the findings of this research we can examine a number of implications for HR practice as a
whole. From a high level perspective, consider the following:
Recruitment & Retention
The findings of this research do not mean that you should only employ top performers. From an
availability standpoint and a total rewards perspective this is simply not realistic. What the research
does imply is that you should have a strong understanding of the organizational structure and the
critical / high impact roles in your organization so that you can selectively source superstars for key
positions where they will make the most impact to the bottom line. An understanding of your
organizational structure will further help you to understand where a solid second string player will be
sufficient. Through this understanding you will facilitate your organization to manage the retention
demands (e.g., compensation, perquisites, flexibility) that will support you to keep that critical group
of top performers. If you need an analogy for selection and retention, consider building your sports
team under the salary cap.

BC HRMA 2011

- Research Briefing -

Page 5

Why People Stay

Performance Management
The findings of this research further speak to the need to
know who your top performers are so that you can
appropriately develop and take action to retain them.
Evaluate your performance management systems and
make adjustments to ensure that the systems dont
impose the traditional bell curve (e.g., ask raters to
distribute performers in this manner). Allow the
performance system to showcase top performers where
they exist so that efforts to develop them and retain
them can be targeted. Ensure that managers and
employees are trained to use the system thus limiting
bias, and support the success of the system with clear
definitions of ratings (in systems that are ratings based).
For HR professionals, growing evidence that the majority
of the workforce is, in fact, performing below the
statistical mean may be alarming. In a Paretian curve,
typical no longer means average, and this has
implications for the way that organizations make
decisions and the HR practices that they employ. It does
not mean that the overall workforce is not vital. All
members of your organization are valuable, but the
results of the research provide further evidence to

Performance Appraisals: Gaussian vs. Paretian

Consider a performance management system
that uses a five point rating scale. Assuming a
Guassian (normal) distribution of individual
performance, the mid-point of the scale, a
rating of three, would be considered an average
rating. While the ideal for the organization
would be for performance to be above the
average, the expectation would be that most
performers in an organization would be
clustered around the three rating, with fewer
and fewer fanning out to the more extreme
scores.
If one assumes a Paretian distribution of
individual performance then the ratings and
their descriptors need to shift. The expectation
would be that the typical performer exhibits
performance that is below the mean.
Acceptable performance would be rated as a
two (below the mid-point of the scale). Points
three, four and five would demonstrate the
performance that individuals can strive for.

suggest that the way that we design and focus our HR

efforts should be examined.

This article is based on the following research paper:

The Best & the Rest: Revisiting the Norm of Normality of Individual Performance
Ernest OBoyle Jr., Herman Aguinis,
Publisher: Personnel Psychology, 2012, 65, 79-119

BC HRMA 2011

- Research Briefing -

Page 6