The role of leader-member exchange in high turnover work environments

The role of leader-memberexchange in high turnover work

environmentsPaula C. Morrow, Yoshinori Suzuki, Michael R. Crum

and Robert RubenCollege of Business, Iowa State University, Ames, Iowa, USA, and

Gregory PautschCollege of Business and Public Administration, Drake University, Des Moines,

Iowa, USA

Abstract

Purpose – To assess the role of leader-member exchange (LMX) in affecting voluntary turnover in ahigh turnover work context.

Design/methodology/approach – Following consideration of traditional predictors of employeeturnover, how LMX is related to voluntary turnover is examined among 207 over-the-road truckdrivers using a telephone survey.

Findings – Leader member exchange is found to be nonlinearly related to turnover such thatturnover is lowest when LMX is moderate (i.e. both “bad” and “good” LMX are associated with higherlevels of turnover).

Research limitations/implications – Findings indicate that LMX and other antecedents should beexamined for nonlinear relationships to turnover. This research may help to bridge the gap betweenturnover research and that associated with supervision and leadership.

Practical implications – These study results suggest that unrealistic expectations should not beformed regarding the power of any single factor (e.g. LMX) to reduce turnover.

Originality/value – This paper suggests that nonlinear relationships between antecedents ofturnover and turnover receive fuller consideration.

Keywords Leadership, Employee turnover, Commercial road vehicles, United States of America

Paper type Research paper

The role of leader-member exchange in high turnover work environmentsTurnover experts, both academic and practitioner, have long asserted that supervisionplays a meaningful role in voluntary employee turnover decisions. However, empiricalinvestigation documenting these relations has lagged (Griffeth and Hom, 2001).Relatively few supervisory-linked antecedents of turnover beyond satisfaction withsupervision have been explicitly studied (Griffeth et al., 2000). A morebehaviorally-oriented specification of supervisory antecedents of turnover would be

The Emerald Research Register for this journal is available at The current issue and full text archive of this journal is available at

www.emeraldinsight.com/researchregister www.emeraldinsight.com/0268-3946.htm

The authors wish to acknowledge the assistance of the anonymous firm that provided the datafor this study and James McElroy for helpful advice and criticism in manuscript preparation.Correspondence concerning this paper should be addressed to Paula C. Morrow, Department ofManagement, 2350 Gerdin Business Building, Iowa State University, Ames, Iowa 50011-1350.Electronic mail may be sent via the internet to [email protected].

Role of LMXin work

environments

681

Received June 2005

Journal of Managerial PsychologyVol. 20 No. 8, 2005

pp. 681-694q Emerald Group Publishing Limited

0268-3946DOI 10.1108/02683940510631444

beneficial. Leader-member exchange (LMX) theory (Liden et al., 1997; Maslyn andUhl-Bien, 2001), which explicates specific factors that play a role in determining thequality of supervisory-subordinate relations, has been identified as such a frameworkthat might shed more light on supervisory behaviors that affect turnover (Gaertner,1999). Moreover, meta-analytic findings, albeit based on a very small number ofsamples, indicate that the quality of leader member exchange is negatively related tointended turnover (Gerstner and Day, 1997) and actual turnover (Griffeth et al., 2000).More recently, however, Harris et al. (2004) have proposed that LMX may actually becurvilinearly related to turnover intentions and turnover, and that this possibility hasnot adequately been considered. Accordingly, the overarching purpose of the presentstudy is to increase understanding of how perceptions of supervisors affect turnoverbehavior by exploring how LMX is related to employee turnover in the truckingindustry. This research setting is particularly useful in that turnover rates are high incomparison to many other industries (thereby reducing customary problems withrestriction-in-range), and because turnover is less sensitive to general economicconditions (i.e. there is a long standing shortage of qualified truck drivers; Min andEmam, 2002), whose role in turnover decisions remains controversial (Williams andLivingstone, 1994).

The role of leader member exchange in turnoverLMX theory posits that the relationship between a supervisor and an employeedevelops as a result of work-related exchanges between these two individuals. Theserelationships can be characterized as high in quality or “good” (i.e. reflecting trust,respect and loyalty) or low in quality or “bad” (i.e. reflecting mistrust, low respect and alack of loyalty).

Achieving good relations between supervisors and subordinates helps to embedemployees within organizations, and thereby provides a disincentive for employees toquit. Accordingly, firms have sought to maximize LMX perceptions amongsubordinates as a strategy for reducing turnover. Bad LMX has similarly beenviewed as an undesirable attribute in an employment relationship and has beenobserved to explain employees’ quit decisions (Griffeth and Hom, 2001). Subordinatesmay come to judge their LMX relations as bad for a variety of reasons, ranging frompoor management skills of the supervisor to lack of person-supervisor fit(Kristof-Brown et al., 2005). Poor LMX may even be a function of situationalartifacts such as when opportunities to develop good supervisory/subordinaterelationships are curtailed (e.g. employees quit before sufficient time has elapsed for agood relationship to evolve, the nature of the work limits opportunity forsupervisory-subordinate interaction). Regardless of the underlying cause of poorLMX, it is generally recognized as a determinant of voluntary turnover (Griffeth andHom, 2001).

We contend, however, that strong or favorable LMX relations may also beproblematic in terms of subsequent turnover. Employees who experience strong LMXmay benefit from more supervisory attention (i.e. receive more training, guidance),especially during early employment periods, thereby improving their skill sets andenhancing their marketability. Such improvements in job skills may, in turn, make theemployee more attractive to other employers and ultimately lead to greater turnover.Viewing leader-member relations in this way implies that it may well be non-linearly

JMP20,8

682

related to turnover, where both unfavorable and favorable levels of LMX result inhigher levels of turnover.

Harris et al. (2004) tested this thesis with respect to intent to quit among twosamples (i.e. water management and distribution services employees). Their resultssupported the curvilinear relationship described above. Their study was limited,however, in that it incorporated only turnover intentions and not actual turnover.Moreover, it failed to control for one of the most well-established predictors ofvoluntary turnover, organizational commitment.

The role of leader member exchange in the trucking industryLMX may also be an overlooked explanation for the very high turnover rates commonto trucking and other high turnover work environments. As detailed below, truckdrivers and their supervisors are constrained in their opportunities to developfavorable supervisory-subordinate exchanges. Hence, the potential for drivers toexperience “bad” LMX is high and truck drivers may therefore give greaterconsideration to quitting than is typical in other work settings. Interestingly, eventrucking organizations where “good” LMX has been achieved, cannot be assured of lowturnover. There is such a shortage of drivers (Min and Emam, 2002), especially goodones with strong skill sets (e.g. experienced, excellent driving records), that drivers caneasily be recruited by competing trucking firms.

In summary, this study provides the opportunity to assess leader-memberrelationships and turnover in the commercial transportation industry, specificallyamong over-the-road truck drivers and their immediate supervisors, dispatchers.Commercial truck drivers exhibit much higher rates of turnover than many otheroccupations. For example, the total US voluntary turnover rate for September2002-August 2003 period was 19.2 percent. The turnover rate for the same period in theoccupational category housing drivers (i.e. trade, transportation and utilities)according to the US Department of Labor was 23.2 percent (Nobscot, 2004). Industrystatistics for trucking firms comparable to the ones used in this study are typicallyeven higher, with industry reports indicating that the annual turnover rate for largetruck loads firms during this period averaging 105.5 percent (Costello, 2004).

Predictors of turnoverIn order to explore the role of leader-member relations in turnover decisionsrealistically, traditional predictors of voluntary turnover were included in the study.Demographic factors (company tenure, number of dependents), organizationalcommitment, work management practices relevant to truck driving work (pay,consignment waiting time, time at home) are assessed as traditional predictors ofturnover. The role of LMX is then evaluated.

Demographic factorsPrevious studies have indicated that individual differences such as age, organizationaltenure, education and number of dependents (Cotton and Tuttle, 1986) can account forsignificant variance in turnover. Thus, the decision was made to treat demographics ascontrol variables in this study. With respect to the two demographic factors included inthis study, meta-analytic assessments by Griffeth et al. (2000) indicate that bothcompany tenure (r ¼ 20.20) and number of dependents (r ¼ 20.14) are negatively

Role of LMXin work

environments

683

related to turnover. Within the trucking industry, organizational tenure has beenshown to be a potent predictor of turnover (Min and Emam, 2002).

Organizational commitmentOrganizational commitment is a well-established predictor of turnover and turnoverintentions. The meta-analytic estimate for relations between turnover andorganizational commitment, derived from 67 samples and over 27,000 observations,is r ¼ 20.23, according to Griffeth et al. (2000).

Work management practicesTruck driving is a large occupational category, with over 721,000 people operatingheavy and tractor-trailer trucks in 2002 (Bureau of Labor Statistics, 2004). These truckdrivers, normally called over-the road drivers to distinguish them from light truck anddelivery service drivers, face a wide variety of working conditions including variableweather and traffic conditions, fatigue, and boredom. The industry is also highlyregulated with federal regulations limiting the amount of work that can be done inspecified time periods. Because of the unique requirements of truck driving work, workmanagement practices specific to this industry have evolved and are used in this study.

Total driver pay. In contrast to local drivers who are paid hourly or on a salary basis,over-the-road truck drivers are typically paid by the mile. Many companies will offerbonuses for meeting certain performance objectives (e.g. on-time delivery of freight,safe driving) and additional compensation for activities that decrease potential drivingtime such as waiting for trucks to be loaded or unloaded. A modest negativerelationship (r ¼ 20.09) between pay and turnover is supported by meta-analysis(Griffeth et al., 2000) and explicitly supported in a truck driving samples (Min andEmam, 2002; Shaw et al., 1998).

Average consignment wait time. Having to wait for trucks to be loaded or unloadedis one of the most frustrating aspects of truck driving work long periods of waitingcontributes to fatigue, “close calls” (near accidents), stress, and, because no miles arebeing accrued, a potential loss of income (Morrow and Crum, 2004).

Driver time at home. The amount of time an over-the-road truck driver is able to gethome is one of the most important aspects of work to drivers. The more frequently adriver gets home, the greater the probability of securing adequate recovery time betweenwork stints (e.g. sleeping at home, sleeping at night) and the more drivers perceive thatthey can be home when they want to be (i.e. achieve control over their schedules).Moreover, starting a driving assignment with inadequate recovery time is common andassociated with the frequency of “close calls” (near accidents) and perceptions of fatigue(Crum and Morrow, 2002). In addition, time at home is frequently cited by drivers as oneof the most important “quality of life” dimensions (Rodriguez and Griffin, 1990;Stephenson and Fox, 1996) and work-family conflict is a well-documented driver ofdecisions to quit (Griffeth and Hom, 2001). Thus the more frequently a company can getits drivers home, thereby improving quality of work life and post work recovery, thelower the turnover rate (De Croon et al., 2004; Shaw et al., 1998).

Leader-member relationsLMX. Supervisory support and satisfaction with supervision have long been proposedas possible means of reducing turnover directly or indirectly through heightened job

JMP20,8

684

satisfaction (Griffeth and Hom, 2001; Mowday et al., 1982) and/or organizationalcommitment (Gaertner, 1999; Griffeth and Hom, 2001). However, the precisemechanisms through which subordinate perceptions of supervisors might affect jobsatisfaction or organizational commitment are not well understood. In addition,relationships with leaders have recently been identified as a neglected antecedent ofturnover (Maertz and Griffeth, 2004). Lastly, the specific behaviors supervisors mightengage in to promote retention have not been identified (Griffeth et al., 2000).

The LMX construct has the potential to ameliorate this situation since itacknowledges that LMXs can entail more than just contractual arrangements andidentifies some specific behaviors supervisors might engage in to enhance subordinateperceptions of the relationship (Scandura and Graen, 1984). More specifically, the LMXmodel asserts that supervisors determine the work roles assigned to subordinates andthose individuals whom supervisors like or trust will be afforded better roles andopportunities. Over time, subordinates assigned these key roles will develop closer,higher quality LMX relationships with their supervisors and receive more support andrewards from those supervisors (Dienesch and Liden, 1986; Liden et al., 1997).

The impact of low quality LMX relationships on subordinates has been shown tohave undesirable consequences (e.g. lower objective performance, satisfaction,organizational commitment, role clarity, and stronger turnover intentions) throughmeta-analysis (Gerstner and Day, 1997). However, actual turnover was not shown to besignificantly related to LMX in the Gerstner and Day meta-analysis and has only beenshown to be weakly related to LMX in a more recent examination (r ¼ 20.23, threesamples) by Griffeth et al. (2000). The studies from which these meta-analyticconclusions are derived have seldom considered non-linear relationships between LMXand turnover, and this fact may explain the weak linkage observed between LMX andturnover.

Highly favorable subordinate perceptions of LMX may extract a similar price interms of higher turnover. Maertz and Griffeth (2004) identified eight motivationalforces associated with turnover. Two are relevant here: “alternative” and “calculative”.An alternative force reflects the extent to which individuals believe that they can besuccessful in obtaining an alternate job and that this belief plays a major role in thedecision to quit. Subordinates who establish solid relations with their supervisors arelikely to believe, and even expect, that they can replicate this relationship in anotherfirm. Thus a subordinate, in this case a driver, who believes an alternate employercould provide an enhanced work situation along some other lines (e.g. better pay,hours) would not be reluctant to assume the risk of quitting (Steel, 2002). The drivershortage (Min and Emam, 2002) too would add to drivers’ confidence in this regard.Calculative motivations to quit are, as the name suggests, a cost-benefit assessmentconducted by drivers concerning the ability of a driver to “get ahead” within his or herpresent firm. There may come a time as suggested by Harris et al. (2004) when thebenefits associated with good LMX have reached a plateau. The dispatcher/supervisormay no longer have any remaining inducements to offer to keep the employee with thefirm. Accordingly, we hypothesize that both low and high LMX assessments bydrivers will also be associated with high levels of turnover.

LMX within the trucking industry. Within the trucking industry, the relationshipbetween drivers and supervisors, who are commonly referred to as “dispatchers”, takeson special importance because truck drivers typically work with little direct

Role of LMXin work

environments

685

supervision. A driver’s dispatcher is the primary contact a driver has with people in theemploying organization. Dispatchers help coordinate the movement of trucks andfreight and thus play a major role in determining drivers’ assignments, schedules androutes. Communication between drivers and dispatchers is often limited to telephoneand electronic means when drivers are working. Hence developing a strong, favorablerapport with drivers is not easy to do simply because dispatchers have littleopportunity to develop personal and informal relations with drivers.

Few studies have examined relations between drivers and dispatchers. One studydid find a relation between dispatchers’ listening skills and responsiveness to driverconcerns and driver turnover rates (Keller and Ozment, 1999). Another study foundthat dispatcher responsiveness to driver concerns was associated with driver operatingperformance in the field and driver relationships with external customers of the carrierfirm (Keller, 2002). These findings imply that good LMX might be desirable, but it isdifficult to draw conclusions regarding turnover on the strength of one study.Moreover, even if and when favorable LMX relations do evolve, the exchange elementsa dispatcher has at his or her disposal to reward drivers, such as better workassignments (e.g. more lucrative back hauls, less urban routes) are likely to becomelimited as dispatchers seek to maintain multiple LMX linkages. It is also possible thatthe demands of business may curtail dispatchers’ latitude with respect to the nature ofthe work assignments.

To summarize, this research seeks to assess the effects of LMX after controlling fortraditional antecedents of turnover. While virtually all employees seek “good”relationships with their supervisors, we assert that “bad” and “good” relationshipsmay be predictive of employee turnover behavior.

MethodSample and procedureSurvey responses were secured from a medium-sized trucking firm based in themidwest employing approximately 724 over-the-road drivers and experiencing anannual driver turnover rate around 80 percent. This carrier has more than 500 vehicles(class-8 tractors) and 1,500 trailers in its inventory.

A number of steps were undertaken to solicit driver involvement in the project. All724 drivers were made aware of an upcoming telephone survey via a notice included intheir 6 December 2002 payroll statement which was delivered to drivers’ homeaddresses. A copy of the instrument was provided and drivers were told that a toll freenumber would be sent to them in a Qualcomm system message within the next fewdays. A fleet-wide message to this effect was sent on 9 December, with remindermessages transmitted again on 16th and 30th December. Drivers were informed thatthe deadline for calling the toll free number to participate was 5 January 2003.

Initially 294 of the 724 drivers (40.6 percent) responded. Drivers who worked inteams (N ¼ 38), drivers who experienced dispatcher (supervisor) changes during thedata collection period (N ¼ 48), and surveys with excessive missing data (N ¼ 1) wereeliminated. Thus usable responses were acquired from 207 of 724 drivers (28.6 percent).Company records identified the dispatcher (referred to as team leader by the company)assigned to each driver. Fourteen of the 18 dispatchers employed at the time wereincluded in the study and the average number of drivers assigned to the dispatcherwas 45.79. Drivers ranged in age from 26 to 66 years, with a mean age of 46.87 years

JMP20,8

686

and over 95 percent were male. These driver characteristics are representative of otherover-the-road commercial truck drivers samples (Crum et al., 2001).

MeasuresVoluntary turnover. Turnover status was derived from company records by trackingresignations for approximately one year (i.e. 9 December 2002 to 31 December 2003).For the 207 drivers participating in the study, 75 (36.2 percent) quit and 132 (63.8percent) remained.

Company tenure and number of dependents. This information was taken fromcompany records. Company tenure at termination or 31 December 2003 rangedfrom 3 to 203 months, with a mean company tenure of 40.14 months.

Organizational commitment. Affective organizational commitment was measuredusing Balfour and Wechsler’s (1996) 3-item measure. Drivers were asked to report theirlevel of agreement using a 5-point strongly disagree to strongly agree response formatto three statements:

(1) “I feel a strong sense of belonging to this organization”;

(2) “The people I work for do not care about what happens to me” (reverse-coded);and

(3) “I feel like ‘part of the family’ at this organization”.

Balfour and Weschler reported an internal consistency estimate of 0.81 in a sample ofpublic sector employees while Kacmar et al. (1999) obtained further evidence of scale’sreliability and validity. The measure demonstrated a Cronbach a of 0.74 (Table I) in thepresent study.

Total driver pay. The total pay represents the average weekly pay of a driver in 2002and was obtained from company records. The average pay in this sample was $834 perweek, which compares favorably to 2002 national truck driver averages of $627 perweek (Bureau of Labor Statistics, 2004).

Average consignment wait time. This variable was also obtained from companyrecords. It reflects the average number of minutes a driver had to wait for his/her truckto be unloaded at destination (consignment) sites. Such waiting time is stressful todrivers because of federal regulations limiting the consecutive hours they may work,

M SD 1 2 3 4 5 6 7

Company tenure 40.14 38.97Number of dependents 1.42 1.41 0.10Organizational commitment 3.60 0.98 20.11 20.09 (0.74)Total driver pay 834.26 200.29 0.35 0.04 20.19Avg. consignment wait time 139.98 68.79 20.05 0.04 20.00 0.07Driver time at home 27.16 18.37 0.22 0.09 20.15 20.02 0.07LMX 3.73 0.87 20.09 20.07 0.46 .04 0.12 20.11 (0.94)Turnover 1.36 0.48 20.36 0.01 0.05 20.23 0.16 20.05 0.13

Notes: N=207; means and standard deviations of scaled variables were based on a 1 to 5 scale metric;turnover was coded 1=stayers and 2=leavers; Cronbach as were reported along diagonal; correlations$ ^ 0.13 (two-tailed) are significant to p , 0.05; correlations $ ^ 0.18 (two-tailed) are significant atp , 0.01; LMX=leader-member exchange

Table I.Descriptive statistics

Role of LMXin work

environments

687

which in turn affects their compensation since drivers are paid by the mile driven. Inthis sample, drivers reported waiting an average of 139.98 minutes per stop.

Driver time at home. The home time measure reflects the average amount of time adriver was home during a given week as reported by drivers and recorded in companyrecords. Drivers in this sample reported being home an average of 27.16 hours perweek, consistent with company policy of getting drivers home every other weekend.

Leader-member relations. Recent treatments of LMX emphasize both work-relatedand social bases of exchange relationships (Liden et al., 1997). Four conceptualcomponents of LMX have been recognized: affect (friendship and liking), loyalty(feeling of mutual support and obligation), contribution (performing at a level beyondminimal standards), and professional respect (appreciation of job knowledge andcompetence). A 12-item measure formulated by Liden and Maslyn (1998), whichcompares favorably to an earlier LMX-7 measure (Graen and Uhl-Bien, 1995), andrecently used again by Maslyn and Uhl-Bien (2001), was employed in the presentstudy. Drivers were asked to record their level of agreement using a (1) stronglydisagree to (5) strongly agree set of response options to statements such as “I like myteam leader very much as a person” and “I am impressed with my team leader’s jobknowledge”. The Cronbach a was 0.94 (Table I).

Data analytic strategyIn the current study we examined how LMX was related to turnover only after theeffects of traditional predictors of turnover (i.e. demographic factors and organizationalcommitment) and work management practices were considered using logistichierarchical regression. Logistical regression was employed due to the dichotomousdependent variable, turnover (Morrow et al., 1999). Specifically, the traditionalpredictors or turnover are entered in step 1, followed by work management practices instep 2. Then, LMX is examined by entering the linear LMX term, the squared LMXterm (which assesses the possibility of a U shape or an inverted U shape), and a cubedLMX term (which allows for the possibility of a second bend or “S” shapedrelationship). Means, standard deviations, and intercorrelations of all unstandardizedstudy variables are reported in Table I.

ResultsResults of the hierarchical regression analysis (binary logit) are reported in Table II.The results indicate that company tenure and average consignment wait time variablesattained the “correct” or expected signs and were statistically significant. Thus, asexpected, a driver’s turnover decision is likely to be affected by the duration of time thedriver has been with the company, and by the amount of consignment wait time thedriver experiences. More specifically, a driver is more likely to leave a company ifhe/she:

. has been with the company for a relatively short period of time; and

. experiences excessive consignment wait times on a regular basis.

The variables representing the number of dependents, organizational commitment,total driver pay, and the amount of driver time at home failed to demonstrate statisticalsignificance, thus indicating that these variables have little or no impact on driverturnover. One possible reason why time at home was not significant in this study is

JMP20,8

688

that the variable measures the amount of time a driver was home during a given week,and does not necessarily reflect when the home time occurred. Drivers may prefer to behome as much as possible during weekends when they can be with their family, butmay not feel so strongly about being home during weekdays, when family membersmay be busy. A driver’s satisfaction (or utility) with home time may be measured moreadequately by the amount of home time during weekends.

The LMX and turnover findings are evident in step 3. The findings associated withthe linear LMX term (b ¼ 210.97, p # 0.05) suggest that part of the relationshipbetween LMX and turnover is negative. The LMX squared term (b ¼ 3.55, p # 0.05)indicates a positive and significant component to the relationship (i.e. a U-shape).Finally, the inclusion of the cubed LMX term, which tests for a second bend in thecurve, was marginally supported (b ¼ 20.34, p ¼ 0.058). In addition, these resultsindicate that the model with the LMX terms provides the best overall fit to the data (i.e.the likelihood-ratio test between step 2 and step 3 is statistically significant). Figure 1displays the relationship between LMX and turnover.

The results indicate that after controlling for traditional turnover predictors in thetrucking industry, both “bad” LMX (the initial part of the graphed relationship whereLMX is in 1.25-2.50 range) and “good” LMX (where LMX is around 4.50) are associatedwith higher levels of turnover. There is a slight downward trend between 4.75 and 5.0,which is only marginally significant and awaits further replication for confirmation.

In summary, the results reported in this paper imply the following. First, drivers aremore likely to quit the company if they:

Step 1 Step 2 Step 3Coeff. Wald Coeff. Wald Coeff. Wald

Demographic controlsCompany tenure 20.037 21.742 * * * 20.037 17.613 * * * 20.040 17.978 * * *

Number ofdependents 0.074 0.430 0.059 0.250 0.067 0.301Organizationalcommitment 0.016 0.009 0.001 0.000 20.277 1.640

Work managementpracticesTotal driver pay – – 20.001 1.982 20.001 2.227Average cons.wait time – – 0.006 5.764 * * 0.006 5.469 * *

Driver time at home – – 0.007 0.492 0.010 0.716Team leader

relationshipsLMX – – – – 210.974 4.165 * *

LMX squared – – – – 3.551 4.041 * *

LMX cubed – – – – 20.344 3.596 *

Sample size 207 207 207Log-likelihood 2116.007 2111.897 2107.55Likelihood-ratio testa – 0.042 * * 0.0337 * *

Notes: *p , 0.10; * *p , 0.05; * * *p , 0.01; acompares the fit of the model in the column with that ofthe model estimated in the previous step ( p-values)

Table II.Estimation results for

predicting turnover

Role of LMXin work

environments

689

. are relatively new employees who have been with the company for only a shortperiod of time; and

. experience excessive amount of consignment wait times.

Second, although the variables such as number of dependents, organizationalcommitment, total pay, and amount of time at home should affect driver turnovertheoretically, this study did not find evidence to support these theoretical expectations.Third, the relationship between LMX and turnover is complicated. Clear evidence of aU-shaped relation was observed, with the possibility that the relationship is S-shaped(i.e. beyond some high level of LMX, turnover begins to decline again).

DiscussionThe results of this study underscore the complex role supervision plays in affectingemployee turnover decisions.

Implications for understanding employee turnoverThe findings associated with the first analysis were both surprising and disappointing.The failure of a major attitudinal antecedent of turnover, organizational commitment,to demonstrate a significant relationship with turnover suggests that current thoughtregarding the antecedents of turnover may not generalize to all employment contexts.Our study results are consistent with those of Lee et al. (1996, p. 18) where some nurses“explicitly stated that they were satisfied with their jobs and organizationallycommitted when they quit”. It would seem that further refinements regarding thecontextual applicability of so-called “established” predictors of turnover is still merited.Organizational commitment may be a less potent predictor of turnover when alternatejobs are readily available, as is the case for both truck drivers and nurses. Stateddifferently, the findings presented here support calls for greater attention tonon-attitudinal causes of turnover such as employment conditions (Hom and Kinicki,2001; Lee et al., 1996).

Alternatively, the absence of a relationship between organizational commitmentand turnover reported in this study may herald a new understanding of how

Figure 1.The relationship betweenLMX and turnoverprobability

JMP20,8

690

organizational commitment functions in high turnover work environments (i.e. thisfinding may be indicative of a boundary condition to the often observed negativeassociation between organizational commitment and turnover). Most turnover researchhas focused on this behavior within a single organization, as opposed to comparativestudies of turnover across organizations or industries. Hence, relatively little is knownabout whether turnover is differentially related to outcomes, depending on its base ratelevel. In this regard, the results of this study may be restricted to other high turnoverwork environments (i.e. environments where reducing turnover is most critical) and notbe generalizable to firms with low or average levels of turnover. We look to futurestudies to consider this issue in more depth.

Another implication of these findings entails the potential value of investigatingnon-linear relationships between antecedents of turnover and turnover (i.e. otherpredictors of turnover may function in a manner similar to that observed here inconjunction with LMX). In this study, when employees (i.e. drivers) found that theirrelationships with their supervisors (i.e. dispatchers) were lower or higher than theirexpectations, they were energized to quit. These findings are also consistent with thoseof Harris et al. (2004) who explain their findings in terms of push-pull dynamicswherein poor LMX employees are “pushed” out of the organization and good LMXemployees are “pulled” out by attractive competitors. This explanation seems quitereasonable in the trucking industry where not only alternate employment is plentiful,but there is little in the way of downside transaction costs. Indeed hiring bonuses aremore the norm (Min and Emam, 2002).

An intriguing, unresolved implication of these results is how trucking firms shouldapproach the training of dispatchers in light of these findings. Following therelationship between LMX and turnover shown in Figure 1, it would seem thatencouraging favorable LMX should be emphasized. Turnover falls from near 50 toaround 15 percent as LMX perceptions improve. This might be construed as a “push”factor helpful in retaining employees. Beyond an intermediary level, LMX appears tolose its capacity to affect turnover and the “pull” factors associated with perceivedbenefit of alternate employment may become operational. This would suggest thatfirms focus on the internal work practices related to turnover they can control (e.g.consignment wait time). The downward turn at very high levels of LMX (i.e. .4.5),while still tentative, suggest that LMX may become influential in retarding turnover,thereby inferring that concern for LMX should not be abandoned.

Implications for understanding leader-member exchangeLMX research is extensive and has shown considerable utility in predicting a widevariety of employee behaviors (Liden et al., 1997, for a review). It has also demonstratedutility when paired with other variables in interaction effects (Kacmar et al., 2003). Still,effect sizes for LMX constructs have been only small to moderate (Griffeth et al., 2000).Failure to consider LMX as a nonlinear variable vis-a-vis other variables may helpexplain why amounts of explained variation have been relatively small.

LimitationsThere are several limitations of this study, not the least of which was the lack ofsupervisor (dispatcher) perceptions. In effect, we were able to assess only one memberof the LMX dyad. Our inability to secure supervisor data precluded exploration of

Role of LMXin work

environments

691

other factors that affect LMX quality such as the role of reciprocity (Uhl-Bien andMaslyn, 2003). Offsetting this limitation is the fact that the focus of the study was ondriver perceptions and corresponding behavior measured at a later point in time.Consequently, perceptions held by dispatchers, while indirectly important in terms ofthe reciprocity issue, would not be expected to be directly related to employee behavior.Moreover, it would have been impractical to control for the effects of dispatcher/driverdyad combinations because of the large numbers of drivers assigned to eachdispatcher. In the present sample, the average number of drivers per dispatcherexceeded 45, which is common in the trucking industry (Keller and Ozment, 1999).Indeed, span of control may be an independent and another overlooked predictor ofturnover, worthy of further study.

Another caveat concerning these finding is that they were observed in an industrycontext where alternate jobs are readily available. This study’s results need to bereplicated under less favorable employment conditions. In addition, the turnover rate inthe sample was lower than in the organization as a whole and low by industrystandards.

Future directionsAn important next step would be to specify why, how and when a “bad”, “moderate”,and “good” LMX occurs. This study also points to the importance of continuingresearch in work environments like trucking where leader-member interactions arelimited by the nature of the job and other comparable jobs are plentiful (e.g.telecommuting work, evening employment such as in retailing and in the hospitalityindustry). There are many reasons to anticipate high turnover in such environments.Common advice to managers seeking to manage turnover in such environments is toinsure that people are treated well (Lawler, 2003). These study results suggest thatwhen it comes to curbing turnover such advice still holds but that unrealisticexpectations should not be formed regarding the power of any single factor (e.g. LMX)to reduce turnover.

Lastly, and from a broader perspective, this research can perhaps serve as anintegrating bridge between two large research traditions, that associated with turnoverand that associated with supervision and leadership. Interpersonal relationships atwork have long been thought to influence people’s work-related decisions, but theamount of published empirical research explicitly examining how perceptions ofsupervisory behavior affects turnover has been limited, perhaps because of a failure todetect and report statistically significant study findings (i.e. the so-called file drawerproblem) and/or a failure to derive theoretical explanations of human behavior that arenon-linear. In other words, it is not that supervision has not been examined in relationto turnover, but that the nature of the relation may not been modeled correctly. We lookforward to seeing more research on this important topic.

References

Balfour, D. and Weschler, B. (1996), “Organizational commitment: antecedents and outcomes inpublic organizations”, Public Productivity and Management Review, Vol. 29, pp. 256-77.

Bureau of Labor Statistics (2004), Truck Transportation and Warehousing, Bureau of LaborStatistics, Washington, DC, available at: www.bls.gov/oco/cg/cgs021.htm (accessed 18May 2004).

JMP20,8

692

Cotton, J. and Tuttle, J. (1986), “Employee turnover: a meta-analysis and review with implicationsfor research”, Academy of Management Review, Vol. 11, pp. 55-70.

Costello, R. (2004), “Presentation to food shippers of America, Lake Buena Vista, Florida(February 2004)”, American Trucking Associations’ Trucking Activity Report.

Crum, M.R. and Morrow, P.C. (2002), “The influence of carrier scheduling practices on truckdriver fatigue”, Transportation Journal, Vol. 42, pp. 423-39.

Crum, M.R., Morrow, P.C., Olsgard, P. and Roke, P.J. (2001), “Truck driving environments andtheir influence on driver fatigue and crash rates”, Transportation Research Record,Vol. 1779, pp. 125-33.

De Croon, E.M., Sluiter, J.K., Blonk, R.W.B., Broersen, J.P.J. and Frings-Dresen, M.H.W. (2004),“Stressful work, psychological job strain, and turnover: a 2-year prospective cohort studyof truck drivers”, Journal of Applied Psychology, Vol. 89, pp. 442-54.

Dienesch, R.M. and Liden, R.C. (1986), “Leader-member exchange model of leadership: a critiqueand further development”, Academy of Management Review, Vol. 11, pp. 618-34.

Gaertner, S. (1999), “Structural determinants of job satisfaction and organizational commitmentin turnover models”, Human Resource Management Review, Vol. 4, pp. 479-93.

Gerstner, C.R. and Day, D.V. (1997), “Meta-analytic review of leader-member exchange theory:correlates and construct issues”, Journal of Applied Psychology, Vol. 82, pp. 827-44.

Graen, G.B. and Uhl-Bien, M. (1995), “Relationship-based approach to leadership: development ofleader-member exchange (LMX) theory over 25 years: applying a multi-level,multi-domain perspective”, Leadership Quarterly, Vol. 6, pp. 219-47.

Griffeth, R.W. and Hom, P.W. (2001), Retaining Valued Employees, Sage, Thousand Oaks, CA.

Griffeth, R.W., Hom, P.W. and Gaertner, S. (2000), “A meta-analysis of antecedents and correlatesof employee turnover: update, moderator tests, and research implications for the newmillennium”, Journal of Management, Vol. 26, pp. 464-88.

Harris, K.J., Kacmar, K.M. and Witt, L.A. (2004), “An examination of the curvilinear relationshipbetween leader-member exchange and intent to turnover”, Journal of OrganizationalBehavior, Vol. 26, pp. 363-78.

Hom, P.W. and Kinicki, A.J. (2001), “Toward a greater understanding of how dissatisfactiondrives employee turnover”, Academy of Management Journal, Vol. 44, pp. 975-87.

Kacmar, K.M., Carlson, D.S. and Brymer, R.A. (1999), “Antecedents and consequences oforganizational commitment: a comparison of two scales”, Educational and PsychologicalMeasurement, Vol. 59, pp. 976-94.

Kacmar, K.M., Witt, L.A., Zivnuska, S. and Gully, S.M. (2003), “The interactive effect ofleader-member exchange and communication frequency on performance ratings”, Journalof Applied Psychology, Vol. 88, pp. 764-72.

Keller, S.B. (2002), “Driver relationships with customers and driver turnover: key mediatingvariables affecting driver performance in the field”, Journal of Business Logistics, Vol. 23No. 1, pp. 39-64.

Keller, S.B. and Ozment, J. (1999), “Managing driver retention: effects of the dispatcher”, Journalof Business Logistics, Vol. 20 No. 2, pp. 97-119.

Kristof-Brown, A.L., Zimmerman, R.D. and Johnson, E.C. (2005), “Consequences of individuals’ fitat work: a meta-analysis of person-job, person-organization, person-group, andperson-supervisor fit”, Personnel Psychology, Vol. 58, pp. 242-81.

Lawler, E.E. (2003), Treat People Right, Jossey-Bass, San Francisco, CA.

Role of LMXin work

environments

693

Lee, T.W., Mitchell, T.R., Wise, L. and Fireman, S. (1996), “An unfolding model of voluntaryemployee turnover”, Academy of Management Journal, Vol. 39, pp. 5-36.

Liden, R.C. and Maslyn, J.M. (1998), “Multidimensionality of leader-member exchange: an empiricalassessment through scale development”, Journal of Management, Vol. 24, pp. 43-72.

Liden, R.C., Sparrowe, R.T. and Wayne, S.J. (1997), “Leader-member exchange theory: the pastand potential for the future”, in Ferris, G.R. and Rowland, K.M. (Eds), Research inPersonnel and Human Resources, JAI Press, Greenwich, CT, Vol. 15, pp. 47-119.

Maertz, C.P. and Griffeth, R.W. (2004), “Eight motivational forces and voluntary turnover: atheoretical synthesis with implications for research”, Journal of Management, Vol. 30,pp. 667-83.

Maslyn, J.M. and Uhl-Bien, M. (2001), “Leader-member exchange and its dimensions: effects ofself-effort and other’s effort on relationship quality”, Journal of Applied Psychology, Vol. 86,pp. 697-708.

Min, H. and Emam, A. (2002), “Developing the profiles of truck drivers for their successfulrecruitment and retention”, International Journal of Physical Distribution & LogisticsManagement, Vol. 33 No. 2, pp. 149-62.

Morrow, P.C. and Crum, M.R. (2004), “Antecedents of fatigue, close calls and crashes amongcommercial motor vehicle drivers”, Journal of Safety Research, Vol. 35, pp. 59-69.

Morrow, P.C., McElroy, J.C., Laczniak, K. and Fenton, J.B. (1999), “Using absenteeism andperformance to predict employee turnover: early detection through company records”,Journal of Vocational Behavior, Vol. 55, pp. 358-74.

Mowday, R.T., Porter, L.W. and Steers, R.M. (1982), Employee-Organization Linkages, AcademicPress, New York, NY.

Nobscot (2004), “Retention management and metrics”, available at: http://nobscot.com/survey/us_monthly_turnover_0803.cfm (accessed 27 May 2004).

Rodriguez, J.M. and Griffin, G.C. (1990), “The determinants of job satisfaction of professionaldrivers”, Journal of the Transportation Research Forum, Vol. 30, pp. 453-64.

Scandura, T.A. and Graen, G.B. (1984), “Moderating effects of initial leader-member exchange statuson the effects of leadership intervention”, Journal of Applied Psychology, Vol. 69, pp. 428-36.

Shaw, J.D., Delery, J.E., Jenkins, C.G. and Gupta, N. (1998), “An organizational-level analysis ofvoluntary and involuntary turnover”, Academy ofManagement Journal, Vol. 41, pp. 511-25.

Steel, R.P. (2002), “Turnover theory at the empirical interface: problems of fit and function”,Academy of Management Review, Vol. 27, pp. 346-60.

Stephenson, F.J. and Fox, R.J. (1996), “Driver retention solutions: strategies for for-hire truckload(TL) employee drivers”, Transportation Journal, Vol. 35 No. 4, pp. 12-25.

Uhl-Bien, M. and Maslyn, J.M. (2003), “Reciprocity in manager-subordinate relationships:components, configurations, and outcomes”, Journal of Management, Vol. 29, pp. 511-32.

Williams, C.R. and Livingstone, L.P. (1994), “Another look at the relationship between performanceand voluntary turnover”, Academy of Management Journal, Vol. 37, pp. 269-98.

Further reading

Hardie, B.G., Johnson, E.J. and Fader, P.S. (1993), “Modeling loss aversion and referencedependence effects on brand choice”, Marketing Science, Vol. 12 No. 4, pp. 378-94.

Mowday, R.T. (1996), “Equity theory predictions of behavior in organizations”, in Steers, R.M.,Porter, L.W. and Bigley, G.A. (Eds), Motivation and Leadership at Work, 6th ed.,McGraw-Hill, New York, NY, pp. 53-71.

JMP20,8

694