sources of social support and burnout, job satisfaction, and productivity

10
Sources of Social Support and Burnout, Job Satisfaction, and Productivity Caren Baruch-Feldman, Elizabeth Brondolo, and Dena Ben-Dayan St. John’s University Joseph Schwartz State University of New York at Stony Brook Social support has been identified as an important correlate of a variety of work outcomes. Support from different sources, including family, coworkers, and supervisors, was examined in 211 traffic enforcement agents (92 men, 119 women). Outcomes included subjective variables (burnout and job satisfaction) and an objective measure of productivity (number of summonses). Support was negatively associated with burnout and positively associated with satisfaction and productivity. A cluster of support variables accounted for 7% of the variance in burnout and productivity and 12% of the variance in job satisfaction. Family support was more closely associated with burnout than with satisfaction or productivity, whereas immediate supervisor support was related to satisfaction and productivity but not burnout. Results suggest that support may be associated with work-related outcomes through multiple pathways. Studies of the relationship of social support to work-related outcomes have emerged from research on perceived organizational support (Eisenberger, Cummings, Armeli, & Lynch, 1997; Eisenberger, Fasolo, & Davis-LaMastro, 1990; Eisenberger, Hun- tington, Hutchinson, & Sowa, 1986), leader–member exchange (Gerstner & Day, 1997; Schriesheim, Cas- tro, & Cogliser, 1999), and supervisor consideration (Downey, Sheridan, & Slocum, 1975), among other constructs. Overall, the data suggest that the per- ceived quality of workplace social support is associ- ated with a number of important work-related vari- ables, including burnout (Brown & O’Brien, 1998; Cherniss, 1980), job satisfaction (Eisenberger et al., 1997), and performance ratings (Gerstner & Day, 1997), and may moderate the effects of stress on burnout (Greenglass, Fiksenbaum, & Burke, 1994). However, recent reviews have illuminated the com- plexity of the association of support to workplace outcomes and suggested that the size and nature of the effects depend on a number of important vari- ables, including the source of support and the sub- jective versus objective nature of the outcome mea- sure, among other variables (Gerstner & Day, 1997; Schriesheim et al., 1999). The present study examined the relationship of social support to burnout, job satisfaction, and pro- ductivity in New York City Traffic Agents. Agents serve as a useful sample in which to investigate social support and its correlates because their work involves issuing summonses for vehicular and parking viola- tions, a structured and repetitive task. Early work on the association of supervisor support to productivity suggested that these effects could be seen most clearly in employees performing structured tasks (Downey et al., 1975). Their work also presents a significant emotional demand because agents fre- quently encounter members of the public who are angry about receiving summonses (Brondolo, Jellife, Quinn, Tunick, & Melhado, 1996). Most previous research on workplace social sup- port, particularly that emerging from the leader– member exchange model, has focused on the super- visor– employee relationship (see Schriesheim et al., 1999, for a review). However, recent studies suggest that other relationships, including those with friends, Caren Baruch-Feldman, Elizabeth Brondolo, and Dena Ben-Dayan, Department of Psychology, St. John’s Univer- sity; Joseph Schwartz, Department of Psychology, State University of New York at Stony Brook Portions of these data were presented at the Work, Stress, and Health Conference of the American Psychological As- sociation and the National Institute of Occupational Safety and Health, March 1999. This work was supported by National Institute of Mental Health Grant MH51930 and American Heart Association Grant 95006910 to Elizabeth Brondolo. This article is based, in part, on the doctoral dissertation of Caren Baruch-Feldman and submitted in partial fulfillment of the degree of Doctor of Philosophy at St. John’s University. We gratefully acknowledge the contributions of Thomas Jelliffe and Christopher Quinn, formerly of the New York City Department of Transportation, and Cary Cherniss of Rutgers University. Correspondence concerning this article should be ad- dressed to Elizabeth Brondolo, Department of Psychology, St. John’s University, Jamaica, New York 11439. E-mail: [email protected] Journal of Occupational Health Psychology 2002, Vol. 7, No. 1, 84 –93 Copyright 2002 by the Educational Publishing Foundation 1076-8998/02/$5.00 DOI: 10.1037//1076-8998.7.1.84 84 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Upload: joseph

Post on 27-Jan-2017

307 views

Category:

Documents


13 download

TRANSCRIPT

Page 1: Sources of social support and burnout, job satisfaction, and productivity

Sources of Social Support and Burnout,Job Satisfaction, and Productivity

Caren Baruch-Feldman,Elizabeth Brondolo, and Dena Ben-Dayan

St. John’s University

Joseph SchwartzState University of New York at Stony Brook

Social support has been identified as an important correlate of a variety of work outcomes.Support from different sources, including family, coworkers, and supervisors, was examined in211 traffic enforcement agents (92 men, 119 women). Outcomes included subjective variables(burnout and job satisfaction) and an objective measure of productivity (number of summonses).Support was negatively associated with burnout and positively associated with satisfaction andproductivity. A cluster of support variables accounted for 7% of the variance in burnout andproductivity and 12% of the variance in job satisfaction. Family support was more closelyassociated with burnout than with satisfaction or productivity, whereas immediate supervisorsupport was related to satisfaction and productivity but not burnout. Results suggest that supportmay be associated with work-related outcomes through multiple pathways.

Studies of the relationship of social support towork-related outcomes have emerged from researchon perceived organizational support (Eisenberger,Cummings, Armeli, & Lynch, 1997; Eisenberger,Fasolo, & Davis-LaMastro, 1990; Eisenberger, Hun-tington, Hutchinson, & Sowa, 1986), leader–memberexchange (Gerstner & Day, 1997; Schriesheim, Cas-tro, & Cogliser, 1999), and supervisor consideration(Downey, Sheridan, & Slocum, 1975), among otherconstructs. Overall, the data suggest that the per-ceived quality of workplace social support is associ-ated with a number of important work-related vari-ables, including burnout (Brown & O’Brien, 1998;

Cherniss, 1980), job satisfaction (Eisenberger et al.,1997), and performance ratings (Gerstner & Day,1997), and may moderate the effects of stress onburnout (Greenglass, Fiksenbaum, & Burke, 1994).However, recent reviews have illuminated the com-plexity of the association of support to workplaceoutcomes and suggested that the size and nature ofthe effects depend on a number of important vari-ables, including the source of support and the sub-jective versus objective nature of the outcome mea-sure, among other variables (Gerstner & Day, 1997;Schriesheim et al., 1999).

The present study examined the relationship ofsocial support to burnout, job satisfaction, and pro-ductivity in New York City Traffic Agents. Agentsserve as a useful sample in which to investigate socialsupport and its correlates because their work involvesissuing summonses for vehicular and parking viola-tions, a structured and repetitive task. Early work onthe association of supervisor support to productivitysuggested that these effects could be seen mostclearly in employees performing structured tasks(Downey et al., 1975). Their work also presents asignificant emotional demand because agents fre-quently encounter members of the public who areangry about receiving summonses (Brondolo, Jellife,Quinn, Tunick, & Melhado, 1996).

Most previous research on workplace social sup-port, particularly that emerging from the leader–member exchange model, has focused on the super-visor–employee relationship (see Schriesheim et al.,1999, for a review). However, recent studies suggestthat other relationships, including those with friends,

Caren Baruch-Feldman, Elizabeth Brondolo, and DenaBen-Dayan, Department of Psychology, St. John’s Univer-sity; Joseph Schwartz, Department of Psychology, StateUniversity of New York at Stony Brook

Portions of these data were presented at the Work, Stress,and Health Conference of the American Psychological As-sociation and the National Institute of Occupational Safetyand Health, March 1999. This work was supported byNational Institute of Mental Health Grant MH51930 andAmerican Heart Association Grant 95006910 to ElizabethBrondolo. This article is based, in part, on the doctoraldissertation of Caren Baruch-Feldman and submitted inpartial fulfillment of the degree of Doctor of Philosophy atSt. John’s University.

We gratefully acknowledge the contributions of ThomasJelliffe and Christopher Quinn, formerly of the New YorkCity Department of Transportation, and Cary Cherniss ofRutgers University.

Correspondence concerning this article should be ad-dressed to Elizabeth Brondolo, Department of Psychology,St. John’s University, Jamaica, New York 11439. E-mail:[email protected]

Journal of Occupational Health Psychology2002, Vol. 7, No. 1, 84–93

Copyright 2002 by the Educational Publishing Foundation1076-8998/02/$5.00 DOI: 10.1037//1076-8998.7.1.84

84

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

Page 2: Sources of social support and burnout, job satisfaction, and productivity

family, and coworkers, may have important implica-tions for burnout (Dignam, Barrera, & West, 1986;Greenglass et al., 1994). For example, some datasuggest that family support may be more effectivethan workplace support in buffering the effects ofwork stress on burnout in teachers (Greenglass et al.,1994).

There has been much less research on the associ-ation of family support to job satisfaction and pro-ductivity, and only limited research determining ifthe correlates of different sources of support varydepending on the nature of the outcome measure(e.g., burnout vs. productivity; Gerstner & Day,1997). The structure of the agent workforce makespossible comparisons of the correlates of differentsources of social support. Specifically, comparisonscan be made of the correlates of support from bothdifferent worksite sources (i.e., coworkers, immedi-ate supervisors, and unit supervisors) and personalsources (i.e., family and friends).

Previous research on the work-related correlates ofsocial support has emphasized subjective outcomes,including supervisor ratings of employee perfor-mance (Gerstner & Day, 1997). In studies of socialsupport and productivity, the use of supervisor rat-ings presents some difficulties, because these ratingsare not uniformly linked to objective indices, and theassociations are at least partly dependent on the qual-ity and duration of the relationship between the su-pervisor and the employee (Duarte, Goodson, &Klich, 1993). An objective measure permits an eval-uation of productivity independent of the supervi-sor’s relationship with the person whose performancehe or she is evaluating (Landy & Farr, 1983; Vec-chio, 1998). For Traffic Agents, the number of sum-monses issued over the course of a month is a salientand objective measure of productivity. The use oftwo subjective measures (i.e., burnout and job satis-faction) permits examination of the relative magni-tude of the association of support to different types ofwork-related outcomes.

Relatively few studies of the relationship of sup-port to job satisfaction or productivity have examinedor controlled for the effects of personality character-istics, despite their potential influence on both per-ceptions of support and reports of distress or satis-faction (Lepore, 1998; Lepore, Allen, & Evans,1993). For example, anger-related traits or negativeaffectivity may influence the respondent’s attitudestoward relationships in general and perceptions ofsupport from supervisors or coworkers in particular(Houston & Kelly, 1989) and may be associated withthe exacerbation of distress responses, including the

experience of burnout (Brondolo et al., 1998; Smith& Tziner, 1998).

We were able to assess and control the effects of traitanger on perceived support and work-related outcomes.Trait anger was included as a measure of the moregeneral construct of the tendency to experience negativeaffect, because our previous research indicated that an-ger was the emotion most closely related to the types ofstressors agents experience, and trait anger was associ-ated both cross-sectionally and prospectively with theexperience of burnout (Brondolo et al., 1998). Be-cause the direction of the relationship betweenstress and personality characteristics is unknown,we examined models both with and without theinclusion of this personal characteristic.

The bulk of recent literature on leader–memberexchange has focused on individual perceptions ofsupport, because researchers have recognized thatsupervisors do not offer the same level of support toall employees (Dansereau, Graen, & Haga, 1975).However, even if there are variations in the relation-ships between a supervisor and his or her employees,some supervisors may provide more support thanothers. Examining work-unit average ratings of sup-port permits us to determine the degree to whichdifferences among supervisors account for variationsin outcome above and beyond the effects associatedwith the individual relationship between supervisorand agent. To our knowledge, no prior study hascompared individual versus work-unit perceptions ofsupport on a range of work-related outcomes.

In sum, this study investigated multiple sources ofsupport (i.e., family and friends, coworkers, immedi-ate supervisors, and unit supervisors) and examinedtheir association to subjective variables, includingburnout and job satisfaction, as well as an objectiveand salient measure of productivity (i.e., the rate ofsummons issuance). Both individual and work-unitperceptions of support were evaluated. The influenceof individual differences in trait anger on the percep-tions and correlates of social support was evaluated.Gender differences in the correlates of support wereevaluated in a sample in which men and women wereequally represented and performed the same tasks.

Method

Participants

The sample included 211 Traffic Enforcement Agents whoprovided informed consent to participate in this study. These211 agents represented 51% of the agents on roster in the workgroups tested during the survey administration. The sample

85SOCIAL SUPPORT AND WORK PERFORMANCE

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

Page 3: Sources of social support and burnout, job satisfaction, and productivity

included 92 men and 119 women, with a mean age of 35.4years (range � 21–62 years). The majority of participantswere African American (n � 130), with smaller groups ofLatinos (n � 36), Whites (n � 25), Asians (n � 11), NativeAmericans (n � 1), and those of mixed ethnicity (n � 8). Themajority of participants (85%) had a high school diploma.

Agents were housed in district offices throughout the fiveboroughs of New York City. Each district office couldaccommodate one or more work units, each of which wascomposed of several squads. An immediate supervisor ledeach squad, and the immediate supervisors in a work unitreported to a unit supervisor.

Agents were drawn from three job titles. Foot patrolagents (n � 71, 33.6%) patrolled on foot, whereas motor-ized patrol agents (n � 75, 35.5%) traveled in cars. Theprimary responsibility of both motorized and foot patrolagents was to issue summonses for vehicular and parkingviolations, with motorized patrol agents expected to issuemore summonses than foot patrol agents because they patrola larger area. Business district enforcement agents (n � 65,30.8%) were responsible for improving traffic flow throughhigh volume areas in midtown Manhattan. They were re-quired to issue summonses but at a lower rate than otherenforcement groups. Agents were assigned to a particulardistrict office and work unit but might change their enforce-ment group assignment (i.e., business district, foot patrol, ormotorized patrol) daily.

Measures

Perceived social support. Four scales were used to as-sess perceived support from coworkers, immediate supervi-

sors, unit supervisors, and family. Scales measuring co-worker and supervisory support were drawn from Karasek’sJob Content Survey (Karasek et al., 1985) and were modi-fied to be applicable for use with the traffic agents. Itemswere rated on a 4-point Likert-type scale ranging from 1 �not at all to 4 � very much. The items measure perceptionsof caring and support and the absence of hostility (see Table1). Scale items were subjected to factor analysis. Each scalehad one factor, with the exception of the Immediate Super-visor scale, which had two factors. One factor containedpositive items, reflecting the supervisor’s display of positiveand caring behaviors; the other contained two items assess-ing the supervisor’s use of criticism. Given that these factorsappeared to represent two dimensions of the same ability tocommunicate care, all items for the Immediate Supervisorscale were retained, and this scale had good internal con-sistency (� � .86). The Family Support scale (� � .91), theCoworker Support scale (� � .87), and the Unit Supervisorscale (� � .91) also had good internal consistency. Test–retest reliability assessed in a different, smaller sampleindicated moderate reliability for each scale over a 4-monthperiod (r � .48–.64).

Control Variables

Demographics. A brief measure inquired about age,gender, ethnicity, and education.

Work unit size. The structure of a work unit (i.e., thenumber of employees and the number of supervisors) mayaffect the quality of social support potentially available toemployees (Dubinsky, Yammarino, & Jolson, 1994; Green,

Table 1Items Included in the Social Support Scale

Scale Item

Family Support When something goes wrong at work, I can talk it over with my friends or family.My friends/family care about how I feel about my job.My friends/family help me feel better when I’ve had a hard day at work.My friends/family are interested and proud when something good happens at work.

Coworker Support My coworkers care about me.People I work with are competent in doing their jobs.People I work with take a personal interest in me.I am exposed to hostility and conflict from the people I work with. (reverse coded)People I work with are friendly.The people I work with encourage each other to work together.People I work with are helpful in getting the job done.

Immediate Supervisorand Unit Supervisora

My supervisor is concerned about the welfare of those under him or her.My supervisor pays attention to what I’m saying.My supervisor exposes me to hostility and conflict.My supervisor is helpful in getting the job done.My supervisor is successful in getting people to work together.My supervisor gives me credit for things I do well.My supervisor criticizes me for small things.My supervisor backs me up if there is a problemMy supervisor cares about me.My supervisor appreciates me.

a These items were administered twice to each participant, once referring to the immediate supervisor and once to the unitsupervisor.

86 BARUCH-FELDMAN, BRONDOLO, BEN-DAYAN, AND SCHWARTZ

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

Page 4: Sources of social support and burnout, job satisfaction, and productivity

Anderson, & Shivers, 1996; Steers & Rhodes, 1978). Largerunits may attenuate the amount of support a supervisor cangive to any single agent. Therefore, data taken from agencyrecords on work unit size (i.e., the number of agents as-signed to the unit) and the total number of available imme-diate and unit supervisors were included to control theeffects of access to support.

Trait anger. Personality traits may influence both per-ceptions of support and reports of burnout or satisfaction.Previous research on traffic agents indicated that anger-related traits were significant predictors of burnout, risk forconflict, and other variables (Brondolo et al., 1998). Traitanger was measured using the Spielberger Trait AngerExpression Inventory (Spielberger et al., 1985), which con-tains 14 items designed to tap the tendency to experienceangry feelings.

Outcome Variables

Burnout. Burnout was assessed with the Emotional Ex-haustion subscale of the Maslach Burnout Inventory(Maslach & Jackson, 1981). The Emotional Exhaustionsubscale, which assesses feelings of being emotionallyoverextended and exhausted by one’s work, has an alpha inthis sample of .93. To minimize participant burden, we didnot include the Personal Accomplishment and Depersonal-ization subscales. The items in these subscales are lessrelevant for this sample, whose job responsibilities do notemphasize the cultivation or maintenance of relationshipswith the public during the course of their patrol duties.

Job satisfaction. Job satisfaction was assessed withitems from the Job Content Survey (“Generally speaking, Iam very satisfied with this job,” “ I am generally satisfiedwith the kind of work I do in this job,” and “Most people onthis job are very satisfied with the job” ). The Cronbach’salpha for this scale was .79. Items on both the Burnout andJob Satisfaction scales were rated on a 1–7 scale (1 � agreeto 7 � disagree).

Productivity. Productivity was defined as the number ofsummonses issued each day adjusted for the number ofhours worked on that day. Because there can be substantialday-to-day variations in the rate of summons issuance (de-pending, in part, on the agent’s assigned patrol), the dailyrecords for 1 month (June 1994) were used. The dependentvariable is the number of summonses issued for every daythe agent worked in that month, adjusted for the number ofhours worked. To examine the internal consistency of themeasure of productivity (i.e., the consistency in summonsissuance across the days measured), we correlated the av-erage number of summonses issued on even-numbered dayswith the average number of summonses issued on odd-numbered days adjusted for number of hours worked. Thesplit-half correlation was .96 (p � .000). Using the Spear-man–Brown correction, we estimated that the reliability ofthe mean for the entire month is 0.98. These data suggestthat a 1-month period is adequate to obtain a reliable mea-sure of productivity.

Analytic Plan

To evaluate the unique and joint contributions of differentsources of social support to the intensity of burnout and job

satisfaction, we used a series of six multiple regressionanalyses for each dependent variable. First, four separatemultiple regression analyses were performed (one for eachmeasure of support) to determine the unique effects of eachtype of support on burnout or job satisfaction, controllingfor age, sex, work unit size, and number of supervisors.Four separate analyses were needed to compensate for thepotential effects of multicollinearity because the supportvariables were correlated with one another. In the fifthanalysis, a cluster containing the four measures of socialsupport (family support, coworker support, immediate su-pervisor support, and unit supervisor support) was enteredafter inclusion of the control variables. Entering the vari-ables in a cluster or group can limit the effects of multicol-linearity when estimating the overall association of supportto the outcome variable. The magnitude of the change in theadjusted R2 associated with the addition of the supportcluster allows an estimate of the overall strength of therelationship of social support to either burnout or satisfac-tion, above that associated with the control variables. Ex-amination of the standardized parameter estimates associ-ated with each measure of support provides an estimate ofthe relative importance of each individual measure of sup-port to the prediction of the outcome variables. The purposeof the sixth analysis was to determine whether the inclusionof trait anger diminishes the relationship of support to workoutcome. This analysis was a replication of the fifth, withthe inclusion of trait anger as a covariate.

We performed six analyses to evaluate the relationship ofsupport to productivity but used a different analytic method.Advances in statistical procedures, specifically mixed-model analyses of variance, have made it possible to morerigorously examine ecological momentary assessment dataincluding repeated measures collected from the same personunder field conditions (Schwartz & Stone, 1998). PROCMixed, a relatively new statistical procedure from SASInstitute, was used in these analyses (Littell, Milliken,Stroup, & Wolfinger, 1996).

PROC Mixed permits analysis of the data despite differ-ences among agents in the number of days worked over thecourse of the month (M � 19.8, SD � 3.74, range � 1–28).This procedure also permits us to accommodate day-to-dayvariations in the enforcement group to which the agentswere assigned. This accommodation was necessary, becausethere were substantial differences among enforcementgroups in average productivity and in the standard deviationof productivity. On average, motorized patrol agents had adaily issuance rate of 44.4 summonses (SD � 10.26), footpatrol agents had a rate of 33.5 (SD � 9.96), and businessdistrict enforcement agents had a rate of 7.6 (SD � 3.49)summonses. Of the 211 agents included in the analysis, 45switched enforcement groups at least once during themonth. Using enforcement group as a within-person controlvariable permitted adjustment of the day’s summons issu-ance rate to reflect that day’s assignment to a particularenforcement group. This is a more sensitive procedure thanthe use of weighted averages because both the number ofobservations (days) and the proportion of observations ineach enforcement group varied among agents, as is often thecase in the unbalanced designs common to field studies.PROC Mixed implicitly adjusts for the differential samplingvariabilities of the means, whereas weighted averages donot reflect this adjustment. Differences in the means andstandard deviations of productivity for the three enforce-

87SOCIAL SUPPORT AND WORK PERFORMANCE

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

Page 5: Sources of social support and burnout, job satisfaction, and productivity

ment groups meant that violations of homoscedasticity ofvariance for within-subject analyses using enforcementgroup as the within-subject variable were violated. There-fore, productivity scores were subject to a square-root trans-formation that substantially improved, although it did noteliminate, the heteroscedasticity. When evaluating thePROC Mixed results, comparisons were made among threevariance models as recommended by Wolfinger and others(Wolfinger, 1993; Wolfinger & Chang, 1995). Because theestimated autocorrelation was statistically significant ineach case, and the combination of compound symmetry andautoregressive error structures produced a better fit than thatobtained with compound symmetry or autocorrelationalone, the combined compound symmetry and autoregres-sive error structure (using the spatial power error structure)was used in all analyses. The F and p values reported herewere those obtained after specifying the combined errorstructure. All analyses were performed both with and with-out transformations, and all reported effects were significantboth with and without transformation. Except where noted,results obtained with transformed data are reported.

Results

Relationships Among Variables

Means, standard deviations, and zero-order corre-lations among variables are presented in Table 2. Forzero-order correlations, the measure of productivityreflects the square root of the average level of sum-mons issuance, uncorrected for variations in jobassignment.

Overall, agents felt well supported by family andfriends and moderately supported by coworkers, im-mediate supervisors, and unit supervisors. Agentshad moderate levels of burnout (M � 3.3) and jobsatisfaction (M � 4.2).

Indices of perceived support from family, cowork-ers, immediate supervisor, and unit supervisor weresignificantly and positively related, with the weakestrelationship emerging between family support andunit supervisor support (r � .18) and the strongestrelationship emerging between unit and immediatesupervisor support (r � .52).

Burnout was significantly and negatively related tojob satisfaction (r � –.40, p � .001). Zero-ordercorrelations suggested that burnout was not associ-ated with productivity. However, when the relation-ship of burnout to productivity was examined usingPROC Mixed, permitting control of variations inproductivity associated with enforcement group, asmall negative relationship emerges (B � –0.08,SE � 0.04), t(202) � –2.15, p � .04. The findingsindicate that higher levels of productivity were asso-ciated with lower levels of burnout. Job satisfactionwas not associated with productivity.

Trait anger was weakly negatively correlated with Tab

le2

Cor

rela

tion

sA

mon

gSo

cial

Supp

ort,

Wor

kO

utco

mes

,an

dC

ontr

olV

aria

bles

Var

iabl

e

Soci

alsu

ppor

tW

ork

outc

ome

Con

trol

vari

able

s

12

34

56

78

910

M3.

322.

972.

792.

953.

384.

1128

.34

42.5

08.

781.

67SD

0.57

0.50

0.53

0.62

1.53

1.44

18.1

717

.83

2.07

0.44

1.Fa

mily

supp

ort

—.2

2**

.20*

*.1

8**

�.2

1**

.17*

.06

�.0

5�

.02

�.1

32.

Cow

orke

rsu

ppor

t—

.36*

**.2

0**

�.1

3.1

9**

.07

�.0

8�

.07

�.0

83.

Squa

dsu

perv

isor

supp

ort

—.5

2***

�.1

0.3

2***

.30*

**�

.26*

**�

.10

�.1

6*4.

Uni

tsu

perv

isor

supp

ort

—�

.25*

**.3

1***

.01

�.0

8�

.15*

�.1

35.

Bur

nout

—�

.40*

**.1

1**

�.0

3.1

1.3

5***

6.Jo

bsa

tisfa

ctio

n—

�.0

2�

.05

�.1

2�

.25*

**7.

Ave

rage

sum

mon

sis

suan

cera

te—

�.6

2***

�.0

3�

.08

8.N

o.of

agen

ts—

.59*

**.1

39.

No.

ofsu

perv

isor

s—

.09

10.

Tra

itan

ger

Not

e.A

vera

gesu

mm

ons

issu

ance

rate

sar

eno

tco

rrec

ted

for

vari

atio

nsas

soci

ated

with

job

assi

gnm

ent

chan

ges

over

the

cour

seof

the

mon

th.

*p

�.0

5.**

p�

.01.

***

p�

.000

1.

88 BARUCH-FELDMAN, BRONDOLO, BEN-DAYAN, AND SCHWARTZ

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

Page 6: Sources of social support and burnout, job satisfaction, and productivity

perceived support from family, immediate supervi-sors, and unit supervisors (all ps � .05) but wasunrelated to perceived coworker support. Trait angerwas positively associated with burnout and nega-tively associated with job satisfaction. In zero-ordercorrelations, trait anger was unrelated to productivity.However, PROC Mixed analyses, controlling for en-forcement group, indicated that trait anger wasweakly negatively associated with productivity (B �–0.37, SE � 0.12), t(202) � –3.04, p � .01. Exam-ination of untransformed summons issuance ratesrevealed that agents high in trait anger (i.e., withscores above the median) wrote fewer summons(M � 30.10) than those low in trait anger (M �27.99).

Age and Gender Differences

Demographic variables were independently asso-ciated with indices of support and work-related out-comes. Age was weakly positively correlated withunit supervisor support (r � .20, p � .01), negativelycorrelated with burnout (r � –.21, p � .005), andpositively correlated with job satisfaction (r � .22,p � .005). PROC Mixed analyses, with age as thepredictor variable, revealed a small but significantassociation of age and summons issuance rate (B �0.02, SE � .01), t(207) � 2.70, p � .01. Examinationof untransformed rates of summons issuance revealedthat younger agents (i.e., those below the median ageof 35 years) wrote fewer summonses (M � 28) thandid older agents (M � 31).

Men and women did not differ in level of per-ceived support from any source or in their levels ofburnout or job satisfaction. PROC Mixed analyses,conducted as described earlier, but with gender as thepredictor variable, revealed that gender was signifi-

cantly associated with productivity, F(1, 207) �7.71, p � .01. Examination of untransformed datarevealed that women (M � 30.1) issued more sum-monses than did men (M � 28.2). Therefore, age andgender served as control variables in the mainanalyses.

Perceived Support and Burnout, JobSatisfaction, and Productivity

Table 3 presents the independent relationship ofeach measure of support to each work-related out-come, controlling for age, gender, number of super-visors, and number of agents. The results indicatethat family and unit supervisor support were signifi-cantly negatively associated with burnout (p � .01).The relations for coworker support (p � .10) andimmediate supervisor support (p � .20) were in thesame direction but were not significant.

When a cluster containing all four support vari-ables was entered (following entry of the covariates),the overall R2 for the equation was 0.15, and theadjusted R2 was 0.12, p � .0001. The cluster ofsupport variables contributed 6.8% of the variance,F(8, 194) � 3.89, p � .005, above that associatedwith the control variables. Within the cluster, familysupport (� � –0.15), t(194) � –2.2, p � .04, and unitsupervisor support (� � –0.20), t(194) � –2.5, p �.02, remained significantly and negatively associatedwith the level of burnout.

With trait anger added as an additional covariate,the partial R2 associated with the support clusterdecreased slightly, but the cluster was still signifi-cantly associated with burnout (partial R2 � 0.05),F(9, 193) � 3.07, p � .01. Effects for family supportwere diminished (� � –0.12), t(193) � –1.84, p �

Table 3Results of Analysis of Unique Contributions of Each Type of Supportto Burnout, Job Satisfaction, and Productivity

Support type

Burnouta Job satisfactiona Productivityb

B SE t � B SE t � B SE t

Family �.50 .18 �0.27** �.18 .37 .17 2.13* .15 .15 .09 1.65Coworker �.34 .21 �1.65 �.11 .51 .20 2.62** .18 .14 .10 1.34Immediate supervisor �.26 .20 �1.26 �.09 .85 .18 4.60*** .31 .29 .10 3.07**Unit supervisor �.49 .17 �2.88** �.20 .60 .16 3.79** .26 .07 .09 0.78

Note. Controlling for age, gender, work unit size, and number of supervisors. df � 194 for all t tests.a Hierarchical multiple regression. b Mixed-models regression (does not provide �).* p � .05. ** p � .01. *** p � .001.

89SOCIAL SUPPORT AND WORK PERFORMANCE

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

Page 7: Sources of social support and burnout, job satisfaction, and productivity

.07, whereas the effects for unit supervisor supportwere essentially unchanged (� � –0.18), t(193) �–2.36, p � .02.

Examined separately, each measure of support wassignificantly and positively associated with job satis-faction (all ps � .05). The overall R2 for the equationassessing the association of the covariates and thecluster of four social support variables to job satis-faction was 0.18, and the adjusted R2 was 0.15, p �.0001. The cluster of support variables contributed11.8% of the variance, F(8, 194) � 7.00, p � .0001,above that associated with the control variables.Within the cluster, only immediate supervisor sup-port (� � 0.21), t(194) � 2.5, p � .02, was signifi-cantly positively associated with the level of jobsatisfaction.

When trait anger was included as a covariate inanalyses of job satisfaction, the partial R2 associatedwith the support cluster decreased slightly (partial R2

� 0.10), F(9, 193) � 6.19, p � .0001, but was stillsignificant. The effects of immediate supervisor sup-port remained significant (� � 0.20), t(193) � 2.42,p � .02.

Four separate PROC Mixed analyses revealed thatimmediate supervisor support was significantly pos-itively associated with productivity, whereas familysupport (p � .11), coworker support (p � .16), andunit supervisor support (p � .40) were not. With allfour indices of perceived support serving as predictorvariables, immediate supervisor support (B � 0.35,SE � 0.12), t(194) � 2.82, p � .01, was positivelyassociated with productivity, and the effects for allother sources of support were not significant (0.03 �t � 1.2). When trait anger was added as a controlvariable, perceived immediate supervisor support re-mained significantly associated with summons issu-ance rate (B � 0.33, SE � 0.12), t(194) � 2.75, p �.01.

To determine the size of the relationship betweenimmediate supervisor support and productivity, aftercontrolling for enforcement group and the remainingcontrol variables, we divided the estimated variancefor the between-persons effects (the intercept) ob-tained with age, gender, number of agents, number ofsupervisors, enforcement group, and perceived sup-port in the equation by the estimate obtained when allof the variables except support were included in theequation. The resulting proportion was subtractedfrom 1 to obtain a measure of the proportion ofvariance explained by the immediate supervisor sup-port variable. Immediate supervisor support ac-counted for almost 7% of the between-persons vari-ance in summons issuance rate that remains after

controlling for enforcement group and the controlvariables. Examination of the untransformed rates ofsummon issuance adjusted for all control variablesand enforcement group revealed that agents with highlevels of perceived supervisor support (i.e., hadscores above or below the median) wrote an averageof 31.44 (SE � 0.77) summonses per day, whereasthe agents with low levels of perceived supervisorsupport wrote an average of 26.87 (SE � 0.73) sum-monses per day.

Squad and Unit-Level Perceptions of Support

The following analyses include average squad orunit ratings of perceived support. These analysespermit examination of variations among supervisorsin the degree to which their agents perceived them assupportive. For immediate supervisors, an average ofthe perceived ratings provided by each member of thesquad was obtained. There was an average of 4(range � 1–8) members of each squad who com-pleted these surveys. Agents could switch squadsduring the course of a month. If the agent switchedsquads, the average value for the support level of thenew squad was used for the days the agents workedin that squad. To obtain average ratings of unit su-pervisor support, we averaged ratings across the unitas a whole. To evaluate the degree to which individ-ual differences in perceptions of supervisor supportreflected between-squads differences in supervisorsupportiveness, we performed analyses using PROCNested from SAS Institute. The results suggest that38% of the variation in individual rating of perceivedimmediate supervisor support was attributable to dif-ferences among squads, and this implies that thecorrelation between an individual’s rating and themean for the entire unit or squad is about .62 (i.e., thesquare root of .38). To assess intraunit agreement, weused the Spearman–Brown formula, and it indicatedthat for a typical squad with four members, the esti-mated reliability of the mean is 0.71.

To test whether average ratings of support (reflect-ing differences among squads or units) contributed tothe prediction of work-related variables above andbeyond the variance associated with individual-levelperceptions, we entered average ratings into the mod-els after controlling for individual ratings of super-visor support and demographic and work-unit vari-ables. Average unit ratings of supervisor support didnot contribute significantly to the prediction of burn-out. Similarly, average squad ratings of immediatesupervisor support were not associated with job sat-isfaction. However, PROC Mixed analyses revealed

90 BARUCH-FELDMAN, BRONDOLO, BEN-DAYAN, AND SCHWARTZ

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

Page 8: Sources of social support and burnout, job satisfaction, and productivity

that average squad ratings of immediate supervisorsupport were positively associated with summonsissuance rate (B � 0.43, SE � 0.19), t(3524) � 2.3,p � .02, even after controlling for individual percep-tions of supervisor support.

Gender Differences

Some studies have suggested that women are morelikely than men to use social support to cope withwork stress (Hurst & Hurst, 1997; Thompson, Kirk-Brown, & Brown, 2001). To evaluate potential gen-der differences in the relationship of support to burn-out, satisfaction, or productivity, we first enteredgender into the model as a main effect and thenincluded terms representing the interaction of genderand each dimension of support. Separate tests wereperformed for each dimension of support. There wereno significant gender differences in the relationshipof any source of support to burnout, satisfaction, orproductivity.

Discussion

The purpose of this study was to examine thework-related correlates of different sources of socialsupport. These effects were examined in New YorkCity Traffic Agents who perform a structured andrepetitive task and experience frequent interpersonalconflict, making them a reasonable sample in whichto investigate the correlates of support. The datarevealed negative relationships of support to burnoutand positive relationships of support to job satisfac-tion and productivity, with variations in the strengthof the relationships depending on the source ofsupport.

Family support was more closely correlated withstress responses to the job (i.e., burnout) than withproductivity. These findings of a stronger associationof family versus coworker or immediate supervisorsupport to burnout are consistent with studies ofburnout among teachers (Greenglass et al., 1994).Immediate supervisor support was associated withproductivity and satisfaction but was not significantlyassociated with burnout. The association of supportto productivity was significant and meaningful whena salient and objective measure was used, contrary tosuggestions that support is not closely associatedwith objective measures of productivity (Gerstner &Day, 1997).

These findings suggest that there may be multiplepathways through which support can affect work-related outcomes. Family members or other intimate

associates may have a vested interest in supportingthe individual’s ability to maintain employment byreducing stress. They may also have the requisitepersonal knowledge and authority to provide effec-tive support for emotional distress. In contrast, pro-ductivity may be a more salient concern for supervi-sors, and they may receive more training or havemore skill in addressing work-related behaviors ver-sus stress-related responses.

The effects of perceived support on work out-comes are moderate in size, accounting for about 7%of the variance in burnout and productivity and al-most 12% of the variance in job satisfaction. Analystsfor the agency have estimated that each summonsgenerates about $29 in revenue after costs are con-sidered. Therefore, even small differences in produc-tivity enhancements can have significant differencesin traffic flow and revenue generation. Other factors,such as location, availability of vehicles, and localdifferences in parking regulations, may account foradditional variance.

Trait anger was negatively associated with percep-tions of support, satisfaction, and productivity andpositively associated with burnout. However, the as-sociation of support to each work outcome variableremained significant, even after controlling for traitanger. This suggests that the effects of the supportrelationship add unique variance above that associ-ated with the characteristics of the recipient ofsupport.

In this sample, men and women have the sameresponsibilities and pay, making them a useful sam-ple in which to assess gender differences. There wereno gender differences found for any source of supportand for any work-related outcome. Both men andwomen may be able to use support when it is avail-able, and there is a significant need for support.

Squad-level ratings of supervisor support were sig-nificantly associated with an objective measure ofproductivity but did not add variance above thataccounted for by individual ratings of support whensubjective measures (i.e., burnout and satisfaction)were the dependent variables. Supervisors who aregenerally perceived to be supportive may engage inbehaviors that strengthen the agent’s performance.This may include providing instrumental support inthe form of advice and materials, although we did notdirectly test this hypothesis. In contrast, the degree towhich an agent personally felt supported by the su-pervisor may be sufficient to influence jobsatisfaction.

There were several limitations to the present study.First, the study evaluated the correlates of emotional

91SOCIAL SUPPORT AND WORK PERFORMANCE

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

Page 9: Sources of social support and burnout, job satisfaction, and productivity

but not instrumental support. Second, the data arecross-sectional, not prospective. It may be the casethat supervisors are more supportive of agents whoperform well, rather than the reverse. In this case,supervisors would simply be responding to the qual-ity of personnel assigned to them. However, the dataon average squad perceptions of supervisor behaviorand the link between these perceptions and produc-tivity argue against this notion. It is unlikely that asupervisor would be assigned all productive or allunproductive workers.

Third, although we examined the correlates ofsupport in a sample generally facing a high level ofwork-related stress, the data examined only the maineffects of support. No tests were made of the mod-erating or buffering effects of support, because wedid not determine if the correlates of support differedamong agents who experienced few versus manyconflicts with the public. Our longitudinal studiesinvolving repeated examinations of newly hiredagents as they move from a training program toindependent patrol will permit a more rigorous test ofthe hypothesis that social support influences the ef-fects of increased work stress on work-relatedoutcomes.

In sum, the data confirmed hypotheses about therelationship of emotional social support to burnout,satisfaction, and productivity but highlight the com-plexity of the relationships. The findings varied de-pending on the source of support and the nature of theoutcome. These data reinforce the notion that supportis an important feature of the work environment andsuggest that there are several pathways throughwhich support may influence work-related outcomes.

References

Brondolo, E., Jellife, T., Quinn, C., Tunick, W., & Melhado,E. (1996). Correlates of risk for conflict in NYC TrafficAgents. In G. Vandenbos & E. Bulato (Eds.), Violence onthe job: Identifying risks and developing solutions (pp.217–228). Washington, DC: American PsychologicalAssociation.

Brondolo, E., Masheb, R., Stores, J., Stockhammer, T.,Tunick, W., Melhado, E., Karlin, W., Schwartz, J., Har-burg, E., & Contrada, R. (1998). Anger-related traits andresponse to interpersonal conflict among NYC TrafficAgents. Journal of Applied Social Psychology, 28, 2089–2118.

Brown, C., & O’Brien, K. M. (1998). Understanding stressand burnout in shelter workers. Professional Psychology:Research and Practice, 29, 383–385.

Cherniss, C. (1980). Professional burnout in human serviceorganizations. New York: Praeger.

Dansereau, F., Graen, G., & Haga, W. J. (1975). A verticaldyad linkage approach to leadership with formal organi-

zations. Organizational Behavior and Human Perfor-mance, 13, 46–78.

Dignam, J. T., Barrera, M., & West, S. G. (1986). Occupa-tional stress, social burnout, and burnout among correc-tional officers. American Journal of Community Psychol-ogy, 14, 177–193.

Downey, H. K., Sheridan, J. E., & Slocum, J. J. W. (1975).Analysis of relationships among leader behavior, subor-dinate job performance and satisfaction: A path-goalapproach. Academy of Management Journal, 18, 253–262.

Duarte, N. T., Goodson, J. R., & Klich, N. R. (1993). Howdo I like thee? Let me appraise the ways. Journal ofOrganizational Behavior, 14, 239–249.

Dubinsky, A. J., Yammarino, F. J., & Jolson, M. A. (1994).Closeness of supervision and salesperson work out-comes: An alternate perspective. Journal of BusinessResearch, 29, 225–237.

Eisenberger, R., Cummings, J., Armeli, S., & Lynch, P.(1997). Perceived organizational support, discretionarytreatment, and job satisfaction. Journal of Applied Psy-chology, 82, 812–820.

Eisenberger, R., Fasolo, P., & Davis-LaMastro, V. (1990).Perceived organizational support and employee dili-gence, commitment, and innovation. Journal of AppliedPsychology, 75, 51–59.

Eisenberger, R., Huntington, R., Hutchinson, S., & Sowa,D. (1986). Perceived organizational support. Journal ofApplied Psychology, 71, 500–507.

Gerstner, C. R., & Day, D. V. (1997). Meta-analytic reviewof leader–member exchange theory: Correlates and con-struction issues. Journal of Applied Psychology, 82, 827–844.

Green, S. G., Anderson, S. E., & Shivers, S. L. (1996).Demographic and organizational influences on leader–member exchange and related work attitudes. Organiza-tional Behavior and Human Decision Processes, 66,203–214.

Greenglass, E. R., Burke, R. J., & Konarski, R. (1998).Components of burnout, resources, and gender-relateddifferences. Journal of Applied Social Psychology, 28,1088–1106.

Greenglass, E. R., Fiksenbaum, L., & Burke, R. J. (1994).The relationship between social support and burnout overtime in teachers. Journal of Social Behavior and person-ality, 9, 219–230.

Houston, B. K., & Kelly, K. E. (1989). Hostility in em-ployed women: Relation to work and marital experience,social support, stress, and anger expression. Personalityand Social Psychology Bulletin, 15, 175–182.

Hurst, T. E., & Hurst, M. M. (1997). Gender differences inmediation of severe occupational stress among correc-tional officers. American Journal of Criminal Justice, 22,121–137.

Karasek, R., Gordon, G., Pietrokovsky, C., Frese, M.,Pieper, C., Schwartz, J., Fry, L., & Schirer, D. (1985).Job content instrument: Questionnaire and user’s guide.Los Angeles: University of Southern California, Depart-ment of Industrial and Systems Engineering.

Landy, F. J., & Farr, J. L. (1983). The measurement of workperformance: Methods, theory, and applications. NewYork: Academic Press.

Lepore, S. J. (1998). Problems and prospects for the social

92 BARUCH-FELDMAN, BRONDOLO, BEN-DAYAN, AND SCHWARTZ

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

Page 10: Sources of social support and burnout, job satisfaction, and productivity

support–reactivity hypothesis. Annals of BehavioralMedicine, 20, 257–269.

Lepore, S. J., Allen, K. M. A., & Evans, G. W. (1993).Social support lowers cardiovascular reactivity to anacute stressor. Psychosomatic Medicine, 55, 518–524.

Littell, R., Milliken, G., Stroup, W., & Wolfinger, R. D.(1996). SAS systems for mixed models. Cary, NC: SASInstitute.

Maslach, C., & Jackson, S. (1981). Maslach Burnout Inven-tory manual. Palo Alto, CA: Consulting PsychologistsPress.

Schriesheim, C. A., Castro, S. L., & Cogliser, C. C. (1999).Leader–member exchange (LMX) research: A compre-hensive review of theory, measurement, and data-ana-lytic practices. Leadership Quarterly, 10, 63–113.

Schwartz, J., & Stone, A. (1998). Strategies for analyzingecological momentary assessment data. Health Psychol-ogy, 17, 6–16.

Smith, D., & Tziner, A. (1998). Moderating effects ofaffective disposition and social support on the relation-ship between person environment fit and strain. Psycho-logical Reports, 82(3, Pt. 1), 963–983.

Spielberger, C. D., Johnson, E., Russell, S., Crane, R.,Jacobs, G., & Worden, T. (1985). The experience andexpression of anger: Construction and validation of ananger expression scale. In M. A. Chesney & R. H.Rosenman (Eds.), Anger and hostility in cardiovascular

and behavioral disorders (pp. 5–30). Washington, DC:Hemisphere.

Steers, R. M., & Rhodes, S. R. (1978). Major influences onemployee attendance: A process model. Journal of Ap-plied Psychology, 63, 391–407.

Thompson, B., Kirk-Brown, A., & Brown, D. (2001).Women police: The impact of work stress on familymembers. In P. A. Hancock & P. A. Desmond(Eds.), Stress, workload, and fatigue: Human factorsin transportation (pp. 200 –210). Mahwah, NJ: Erl-baum.

Vecchio, R. P. (1998). Leader–member exchange, objectiveperformance, employment duration, and supervisor rat-ings: Testing for moderation and mediation. Journal ofBusiness and Psychology, 12, 327–341.

Wolfinger, R. D. (1993). Covariance structure selection ingeneral mixed models. Communication in Statistics, Sim-ulation and Computation, 22, 1079–1106.

Wolfinger, R. D., & Chang, M. (1995). Comparing SASGLM and MIXED procedures for repeated measures.SAS User’s Group International Conference ProceedingSUGI, 20, 1172–1182.

Received August 8, 2000Revision received July 3, 2001

Accepted July 7, 2001 y

93SOCIAL SUPPORT AND WORK PERFORMANCE

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.