Frequency versus time lost measures of absenteeism: Is the voluntariness distinction an urban legend?

• Author: Gary Johns,Raghid Al Hajj
• Date: 01 April 2016
• DOI: http://doi.org/10.1002/job.2055
• Published date: 01 April 2016

Frequency versus time lost measures of

absenteeism: Is the voluntariness distinction an

urban legend?

GARY JOHNS*AND RAGHID AL HAJJ

Department of Management, John Molson School of Business, Concordia University, Montreal, Quebec, Canada

Summary We investigate a long-standing methodological rule of thumb, the idea that the frequency of absenteeism from

work approximates an expression of voluntary behavior while total time lost better reflects involuntary behav-

ior and ill health. Conducting original meta-analyses and using results from existing meta-analyses, we deter-

mine that time lost and frequency are equally reliable, that the relationship between them approximates unity

when corrections for measurement artifacts are applied, and that there is very little evidence for differential

criterion-related validity predicated on the voluntariness distinction. We supply new meta-analytic estimates

of the reliability of absenteeism adjusted for aggregation period and determine that most extant meta-

analyses of the correlates of absenteeism have markedly under-corrected for unreliability. Our results question

the basic construct validity of the time lost–frequency distinction, and they contradict the practice of using

“trigger points”that factor absence frequency into attendance monitoring and associated discipline systems

so as to discourage short-term absenteeism, assumed to be volitional. We conclude that the idea that time lost

and frequency reflect different degrees of voluntariness is an unsupported urban research legend. Copyright ©

2015 John Wiley & Sons, Ltd.

Keywords: absence; absenteeism; time lost; frequency; reliability; meta-analysis; trigger points

In 2009, the edited book Statistical and Methodological Myths and Urban Legends was published (Lance & Van-

denberg, 2009). The basic premise of the book is that many commonly advocated research rules of thumb are in fact

socially constructed assumptions underpinning well meaning but scientifically incorrect research practices. In this

article, we use meta-analytic evidence to investigate a long-standing and currently held theoretical and methodolog-

ical rule of thumb, the idea that the frequency of absenteeism from work approximates an expression of voluntary

behavior while total time or total days absent better reflects involuntary behavior, especially ill health. Respectively,

these metrics are known as the frequency index and the time lost index (Chadwick-Jones, Nicholson, & Brown,

1982). Fundamentally, we examine the construct validity of the frequency–time lost distinction.

The Frequency–Time Lost Distinction

Time lost refers to the total number of days an individual is absent from work over some period of time, excluding

vacation days, jury duty, and the like. Frequency refers to the number of inceptions of absence over the same time

frame, irrespective of the duration of those inceptions. For example, a person who misses 10 days in a year would

score 10 on the time lost index. The frequency associated with this figure could range from 1 (10 days in a row

missed)to10(10 isolated occasions of absence). These indexes are usually expressed as counts, although they

can also be expressed as rates to account for differences in the number of days worked. The large majority of

*Correspondence to: Gary Johns, Department of Management, John Molson School of Business, Concordia University, 1455 de Maisonneuve

Blvd. West, Montreal, Quebec H3G 1M8, Canada. E-mail: gary.johns@concordia.ca

Copyright © 2015 John Wiley & Sons, Ltd.

Received 09 March 2015

Revised 04 September 2015, Accepted 12 September 2015

Journal of Organizational Behavior, J. Organiz. Behav. 37, 456–479 (2016)

Published online 7 October 2015 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/job.2055

Research Article

published absence research relies on frequency and/or time lost measures, and the voluntariness distinction is very

commonly invoked to justify measurement choice (e.g., Beemsterboer, Stewart, Groothoff, & Nijhuis, 2009; Berry,

Lelchook, & Clark, 2012; Davey, Cummings, Newburn-Cook, & Lo, 2009; Deery, Walsh, & Zatzick, 2014;

Demerouti, Bouwman, & Sanz-Vergel, 2011; Diestel & Schmidt, 2012; Gellatly & Allen, 2012; Magnavita &

Garbarino, 2013; Shapira-Lishchinsky, 2012; Shapira-Lishchinsky & Tsemach, 2014; ten Brummelhuis, ter

Hoeven, de Jong, & Peper, 2013).

Two notions underpin the argued distinction between absence frequency and time lost. One is that frequent and

shorter duration absences reflect a lack of assiduity and represent active withdrawal from work. The other is that total

time lost is biased by genuine sickness (and perhaps other factors) beyond the absentee’s control. Shorter absences

are less likely to require medical certification, while longer absences, contributing disproportionately to time lost, are

more likely to require such proof. The origins of these ideas have been attributed to Fox and Scott (1943), who stud-

ied absenteeism in three US industrial companies during World War II. Given its persistence for over 70 years, it is

worth reading the legend’s evident genesis in the original (p. 3):

We decided accordingly that, in computing absences, any absence of a number of consecutive days should be

scored as one absence only…Since, in the particular study we were making, we were not directly concerned with

sickness, it seemed wise to take figures which would…minimize successive days of absence and maximize fre-

quency of absences, especially absences without permission.

No data were presented to support this choice. A quotation from Gibson (1966, p. 122) provides another slant on

the idea:

The high-identification worker could be expected to be absent relatively less frequently but, when absent, he can

be expected to be absent for a relatively longer period of time as compared to a worker with low or negative work

identifications, who can be expected to be prone to frequent absences. Since absence of longer duration becomes

subject to examination for legitimacy, questionable absences could be expected to be relatively short in duration.

Given the preceding discussion, time lost is thought to be a less-than-optimal criterion for predictors reflecting vol-

untary motives, and frequency is assumed to be less than optimal for involuntary predictors. However, scholars have

never argued that all of the true-score variance in frequency is due to voluntary sources and all of the corresponding

variance in time lost is attributable to involuntary sources; given the construction of the measures, empirical overlap

is clearly expected (Gellatly, 1995; Hammer & Landau, 1981). Rather, on the margin, frequency is thought to be a

more valid criterion for attitudinal and motivational predictors. As Steel (2003, p. 245) put it, “the core idea has been

that [the] different construction approaches tend to either emphasize or de-emphasize a measure’s exposure to vol-

untary sources of absence variance.”However, we have been unable to find any well articulated theory supporting

the idea that absences of short duration are less likely to be due to health problems or more likely to be behavioral

manifestations of work attitudes, although Froggatt (1970) offered an extended discussion of the latter position. The

distinction between voluntary and involuntary behavior is obviously not entirely clear cut. However, the degree of

control a person has over the behavior is an important determinant of voluntariness, as is the behavior’s susceptibility

to random events. Models that position absenteeism as active withdrawal from work stress the role of attitudes, mo-

tives, and intentions (Fichman, 1984; Harrison, Newman, & Roth, 2006). Medical models of absenteeism are more

predicated on random susceptibility and lack of control (Johns, 1997a).

As Steel (2003) describes, there have been occasional attempts to construct purified measures of absenteeism to

extract voluntariness, sometimes by combining organizational absence codes from personnel records into what are

thought to be more or less motivationally meaningful categories (e.g., Burton, Lee, & Holtom, 2002; Kohler &

Mathieu, 1993). These exercises have been diligent, but they ultimately fall prey to the idiosyncratic coding schemes

that characterize attendance measurement and are seldom generalizable to other organizations. The codes also con-

tain unknown attribution errors, both at source and in the assignment of codes to categories (Hammer & Landau,

1981; Nicholson & Martocchio, 1995; Steel, 2003). Hence, at-source coding errors are one reason why coded

FREQUENCY–TIME LOST MEASURES OF ABSENTEEISM 457

Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. 37, 456–479 (2016)

DOI: 10.1002/job