Documente Academic
Documente Profesional
Documente Cultură
MOTIVATING LANGUAGE IN
EMPLOYEE ABSENTEEISM
Jacqueline Mayfield
Milton Mayfield
Texas A&M International University
This study investigates the relationship between strategic leader language (as embodied in Motivating
Language Theory) and employee absenteeism. With a structural equation model, two perspectives
were measured for the impact of leader spoken language: employee attitudes toward absenteeism and
actual attendance. Results suggest that leader language does in fact have a positive, significant rela-
tionship with work attendance through the mediation effect of worker attendance attitude.
Yet within this context, all absenteeism is not created equal. Some
employee absences cannot be avoided, such as in the cases of serious per-
sonal or family member illness. These times of missed work can often be
recognized as long in duration (defined as more than 1 or 2 days) on review
of organizational records. However, research also indicates that an esti-
mated 52% of total employee absences are discretionary in nature. These
absences are the result of factors such as stress, personal needs, and entitle-
ment mentality (Cascio, 2000; VanDerWall, 1998). These avoidable cases
often appear as 1- or 2-day occurrences in organizational documents.
Such discretionary absences have been categorized as “utility maximi-
zation and choice” or the body of research that approaches absenteeism
as an outcome of individual decision making and inherently influenced
by motivational states (Harrison & Martocchio, 1998). Employee moti-
vation has been closely associated with managerial behaviors, including
explicit and implicit communication practices. Previous investigations
have uncovered links between avoidable absences and negative employee
perceptions of the workplace, including procedural and distributive jus-
tice, and similarly, with low organizational commitment in workers (De
Boer, Bakker, Syroit, & Schaufeli, 2002; Farrell & Stamm, 1988;
Harrison & Martocchio, 1998). All of these undesirable employee atti-
tudes have also been moderated by leader behavior (Cole & Kleiner,
1992; Levin & Kleiner, 1992; Robbins, 2005; Yukl, 2006). However, the
458 JOURNAL OF BUSINESS COMMUNICATION
METHODOLOGY
to the manifest variables because the assumption is that the latent variable
causes changes in the manifest variables. Intuitively, it might seem that
the arrows would run from the manifest variables to the latent variables
due to calculation source. However, the causal diagram represents
expected relationships between the variables—not mathematical opera-
tions. Therefore, since the manifest variable scores are theorized to origi-
nate from the latent variable state, it is appropriate to place the arrows
running from the latent variables to the manifest variables.
Testing a SEM is a two-stage procedure. The model’s overall quality is
evaluated during the first stage. In this stage, the proposed model is com-
pared to the actual data, and the model’s quality is assessed through a set
of measures. Model acceptability occurs if most of these evaluations are
at or above accepted guidelines.
The most longstanding measure is a chi-square test. This test probes for
significant differences between the proposed model and the data. However,
the chi-square test can be misleading because it is sensitive to large sam-
ple sizes, and risks showing a misleading significant difference even when
only trivial discrepancies exist. To remedy this problem, a rule-of-thumb
applies that a model can be considered acceptable if the chi-square to
degrees-of-freedom ratio is equal to or less than 2.5.
Multiple fit indices are also available for model testing. These fit indices
all range from 0 to 1, with results closer to 1 pointing to better models. The
466 JOURNAL OF BUSINESS COMMUNICATION
goodness-of-fit index (GFI) is the most commonly applied fit index. A GFI
of .90 is required to indicate a good fit between model and data, and a fit
of .95 or better is considered to be a very good fit. The other fit indices
impose more stringent calculation methods, and so a somewhat lower fit is
allowable for these measures. A score of .85 is required as a minimum for
model acceptability, and a score of .90 is considered to be good. Preferably,
multiple indices are scrutinized for any major discrepancies between the
measures because each fit index is calculated somewhat differently.
Two other model fit measures examine the variance not accounted for
by the model. These measures are the root mean square average error and
the standardized root mean residual. Lower scores on the measures imply
a better fit between the model and the data, with a score of 0 indicating a
perfect fit. Scores are generally considered to be acceptable when they are
below 0.10 and good when they are below 0.05.
The Bayesian information criterion (BIC) is a final assessment mea-
sure. Lower measure values are associated with better accuracy of a
model’s fit with the data. While the measure is most useful for comparing
different models, it can also be used to determine model adequacy. A
positive score suggests a poor fitting model, and negative scores imply a
better fitting model.
After determining a model’s overall quality, path significance and
strength must be examined. This analysis is performed using a t test. If a
path tests as significant, the relationship strength can be examined through
standardized path estimates. These path estimates are similar to standard-
ized beta weights in regression analysis.
RESULTS
none of these tests were significant at the .05 level—indicating that worker
demographic characteristics do not influence variable scores.
While sample results appear to be generalizable across a range of work-
ers, an important note needs to be made about the cross-sectional nature
of the study. Because this information was collected at one point in time,
the study’s analysis methods do not guarantee that the relationships per-
form causally. The SEM analysis method employed can only indicate
significant relationships, not fully determine the direction of these rela-
tionships. The expected direction of the relationship is based on theory,
and awaits later testing through different means (such as a time-series
study, an experimental design, or other more sophisticated analytical pro-
cedures) to fully establish causality.
As for reliabilities, all measures fall within acceptable levels. For the
motivating language scale, item reliabilities were all above .92. The
absenteeism measure’s reliability was .81. A reliability for the measure of
days missed could not be calculated because it was a single-item mea-
sure. However, the sample results were compared with a prior work (Kim
et al., 1995), and this study’s absenteeism results were very similar to
prior findings. The absenteeism results also lend evidence of the mea-
sure’s validity since the study by Kim et al. found a strong similarity
between their sample’s self-reported absenteeism and company absentee-
ism records (Price, 1997). Item interrelationships and means are pre-
sented in Table 1. Reliabilities are presented in Table 2. A comparison
between this study’s absenteeism results and those in Kim et al.’s study
is presented in Table 3.
Because the attitude toward attendance scale is new, it is appropriate to
more carefully scrutinize the scale. Thus, the new measure was tested
using factor analysis, item-to-total (scale score) correlations, and indi-
vidual item reliabilities. It was predicted that the items should all load on
one factor, have high item-to-total correlations, and high individual reli-
abilities. All of these expectations were met. The factor analysis strongly
indicated a single underlying latent factor, with all items appreciably load-
ing there. In addition, all item-to-total correlations were high (.53 or
greater), as were individual item reliabilities (all above .71). Further infor-
mation on analysis results is presented in Table 4.
The SEM analysis results indicated a good fit between the proposed
model and the data. While the chi-square test was significant at the .05
level, the chi-square to degrees of freedom ratio was 1.7 (31/18)—well
below the recommended rule-of-thumb of 2.5. Additionally, all fit indices
were at or above .95, with several approaching the 1.00 maximum score.
Mayfield, Mayfield / MOTIVATING LANGUAGE AND ABSENTEEISM 469
Note: Numbers in bold are covariances. Numbers in italics are correlation coefficients. Item means are presented
in the last row of the table.
Confidence Interval
The error tests showed similarly good scores, with the root mean square
error of approximation at 0.05, and the standardized root mean residual
equal to 0.04. In all, these statistics strongly support the hypothesized
model as well suited to represent the actual variable relationships. Model
adequacy measures are presented in Table 5.
470 JOURNAL OF BUSINESS COMMUNICATION
Note: Kim et al.’s (1995) study focused on short term absenteeism. Therefore, the authors
only reported on absenteeism of 3 days or less. The current study included all absenteeism
lengths in data collection and structural equation modeling analysis. For validation pur-
poses, the comparison between the studies was limited to 3 days to match construct con-
ceptualizations. The percentages of days absent in the current study sample are based on
respondents with 3 or fewer days absent.
Note: Eigenvalue = 2.19. Percentage of variance accounted for by factor = 55%. Chi-
square test: χ2 = 0.13, df = 2, p = .94.
χ2
31 (df = 18, p = .03)
BIC −72
Fit indices
GFI .98
Adjusted GFI .95
Bentler–Bonnett NFI .97
Tucker–Lewis NNFI .98
Bentler CFI .99
Error tests
Root mean square error average .05
RMSEA confidence interval .01 to .08
Standardized root mean residual .04
Note: GFI = goodness-of-fit index; NFI = normed fit index; NNFI = nonnormed fit index;
CFI = comparative fit index; RMSEA = root mean square error of approximation; BIC =
Bayesian information criterion.
by examining the path coefficients for all direct variable links. Indirect
effects are calculated by multiplying paths between connecting variables. In
the case of this model, the paths to be multiplied are from motivating
472 JOURNAL OF BUSINESS COMMUNICATION
motivating language process. Prior work has been mainly directed at mea-
suring the motivating language use–outcome link, and with a few excep-
tions, ignored the causal process by which these links were forged. While
such analysis is perhaps necessary in the early stages of theoretical devel-
opment, researchers must travel beyond simple input-output relationships
to a deeper understanding of how these processes operate if the benefits
of motivating language are to be fully realized. Fortunately, Sullivan’s
(1988) original conceptualization provides a rich foundation for such
examinations by assuming an attitudinal path between leader language
and employee behavior. And this study reveals an initial indication that
leader motivating language does in fact operate through the vehicle of
worker attitudinal change.
Similarly, this study extends the range of motivating language–related
outcome variables. In addition to worker performance, retention, innovation,
and job satisfaction, a new association has been shown to exist between
motivating language and worker absenteeism. Simply put, motivating lan-
guage now has more credibility to theoretically address some of the most
pressing outcome variable challenges with which management scholars con-
tend (Price, 1997; Staw, 1984). Equally encouraging to this prospect, study
respondents were drawn from a heterogeneous and wide ranging set of work
settings, thus expanding applicability. Moreover, the data were nearly evenly
balanced between male and female respondents, hence strengthening gener-
alizability inferences about motivating language and worker gender.
From a practice viewpoint and as discussed in this article’s background
section, absenteeism is a costly and pervasive organizational problem.
While not expected to be a panacea to a complex issue, motivating lan-
guage can now be adopted as a significant resource with which to promote
employee attendance. Furthermore, training investment and curriculum
design will benefit from these new insights about the communication-
attitude-behavior relationships.
Despite these contributions, there are study limitations that must be
acknowledged. The two major weaknesses are found with the study’s
cross-sectional nature and the sample’s demographic characteristics. As
discussed earlier, a cross-sectional study does not easily permit causal
statements. Instead, we can only infer based on variable linkages and
existing theory. And while motivating language has considerable theoreti-
cal foundation for making such inferences, non-causal studies are still
constrained in their predictive power.
Additionally, the study sample consisted of a relatively young set of
respondents in the early stages of their careers. While it is not expected that
Mayfield, Mayfield / MOTIVATING LANGUAGE AND ABSENTEEISM 475
APPENDIX
MEASURES USED
MOTIVATING LANGUAGE
The examples below show different ways that your boss might talk to you.
Please use the following selections to choose the answer that best matches your
perceptions, and then place an X in the bracket beside the appropriate response.
APPENDIX (continued)
Empathetic Language
Meaning-Making Language
(continued)
Mayfield, Mayfield / MOTIVATING LANGUAGE AND ABSENTEEISM 477
APPENDIX (continued)
ABSENTEEISM
Approximately how often were you absent from your job in the past month?
REFERENCES
Cascio, W. (2000). Costing human resources: The financial impact of behaviors in orga-
nizations (4th ed.). Cincinnati, OH: South-Western College.
Churchill, G. A. (1979). A paradigm for developing better measures of marketing con-
structs. Journal of Marketing Research, 16, 64-73.
Cole, T. C., & Kleiner, B. H. (1992). Absenteeism control. Management Decision, 30,
12-16.
Dalton, D. R., & Mesch, D. J. (1991). On the extent and reduction of avoidable absenteeism:
An assessment of absence policy provisions. Journal of Applied Psychology, 76, 387-393.
De Boer, E. M., Bakker, A. B., Syroit, J. E., & Schaufeli, W. B. (2002). Unfairness at work
as a predictor of absenteeism. Journal of Organizational Behavior, 23, 181-197.
478 JOURNAL OF BUSINESS COMMUNICATION
DeVellis, R. F. (2003). Scale development: Theory and applications (2nd ed.). Thousand
Oaks, CA: Sage.
Farrell, D., & Stamm, C. L. (1988). Meta-analysis of the correlates of employee absence.
Human Relations, 41, 211-227.
Frayne, C., & Latham, G. (1987). Application of social learning theory to employee self-
management of attendance. Journal of Applied Psychology, 72, 387-393.
Friedman, T. L. (2007). The world is flat 3.0: A brief history of the twenty-first century.
New York: Picador.
Gaudine, A. P., & Saks, A. M. (2001). Effects of an absenteeism feedback intervention on
employee absence behavior. Journal of Organizational Behavior, 22, 15-29.
Gellatly, I. R. (1995). Individual and group determinants of employee absenteeism: Test of
a causal model. Journal of Organizational Behavior, 16, 469-485.
Harrison, D. A., & Martocchio, J. J. (1998). Time for absenteeism: A 20-year review of
origins, offshoots, and outcomes. Journal of Management, 24, 1-37.
Harrison, D. A., & Shaffer, M. A. (1994). Comparative examination of self-reports and
perceived absenteeism norms: Wading through lake Wobegon. Journal of Applied
Psychology, 79, 240-251.
Johns, G. (1978). A multivariate study of absence form work. In Academy of Management
Meeting Proceedings (pp. 69-73). Nriarcliff Manor, NY: Academy of Management.
Johns, G. (2003). How methodological diversity has improved our understanding of
absenteeism from work. Human Resource Management Review, 13, 157-184.
Kim, S. W., Cyphert, S. T., and Price, J. L. (1995). A suggested self-reported measure of
absenteeism (Working Paper). Department of Sociology, University of Iowa.
Levin, J. M., & Kleiner, B. H. (1992). How to reduce organizational turnover and absen-
teeism. Work Study, 41, 6-9.
Lu, V. (1999, August 15). Rising sick days cost billions. The Toronto Star, pp. A1, A10.
Martocchio, J. J., & Harrison, D. A. (1993). To be there or not to be there? Questions,
theories and methods in absenteeism research. Research in Personnel and Human
Resources Management, 11, 259-328.
Mayfield, J., & Mayfield, M., (1995). Learning the language of leadership: A proposed
agenda for leader training. Journal of Leadership Studies, 2, 132-136.
Mayfield, J., & Mayfield, M., (1998). Increasing worker outcomes by improving leader-
follower relations. Journal of Leadership Studies, 5, 72-81.
Mayfield, J., & Mayfield, M., (2002). Leader communication strategies: Critical paths to
improving employee commitment. American Business Review, 20, 89-94.
Mayfield, J., & Mayfield, M. (2007). The effects of leader communication on a worker’s
intent to stay: An investigation using structural equation modeling. Human Performance,
20, 85-102.
Mayfield, J., Mayfield, M., & Kopf, J. (1995). Motivating language: Exploring theory with
scale development. Journal of Business Communication, 32, 329-344.
Mayfield, J., Mayfield, M., & Kopf, J. (1998). The effects of leader motivating language on
subordinate performance and satisfaction. Human Resource Management, 37, 235-248.
Mayfield, M., & Mayfield, J., (2004). The effects of leader communication on worker
innovation. American Business Review, 22, 46-51.
Pettit, J. D., Jr, Goris, J. R., & Vaught, B. C. (1997). An examination of organizational
communication as a moderator of the relationship between job performance and job
satisfaction. Journal of Business Communication, 34, 81-97.
Mayfield, Mayfield / MOTIVATING LANGUAGE AND ABSENTEEISM 479