Rough Draft

Effects of Flexitime on Employee Attendance and Performance: A Field Experiment Author(s): Jay S. Kim and Anthony F. Campagna Source: The Academy of Management Journal, Vol. 24, No. 4 (Dec., 1981), pp. 729-741 Published by: Academy of Management Stable URL: https://www.jstor.org/stable/256172 Accessed: 01-02-2019 18:12 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide

range of content in a trusted digital archive. We use information technology and tools to increase productivity and

facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at

https://about.jstor.org/terms

Academy of Management is collaborating with JSTOR to digitize, preserve and extend access to The Academy of Management Journal

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms

? Academy of Management Journal 1981, Vol. 24, No. 4, 729-741.

Effects of Flexitime on Employee A ttenadance and Performance:

A Field Experiment1 JAY S. KIM

ANTHONY F. CAMPAGNA The Ohio State University

The effects of a flexitime program in a county welfare agency are assessed. The analysis of covariance on the employees' attendance and performance revealed that (1) the flexitime program appears to permit employees to reduce their use of unpaid absences and (2) perfor- mance efficiency tends to be higher among employees under the flexitime program.

Flexitime as an organization development intervention has been increas- ingly popular in American industry. According to one report, approxi- mately 13 percent of the employing organizations in the private sector have adopted a form of flexitime program since it was introduced in the early 1970s. This represents about three million American workers (Nollen & Martin, 1978). The purpose of this study is to investigate the effects of flexitime on employee attendance and performance in a field setting.

Although there are a number of variations in its form, the basic model of flexitime usually consists of five interrelated components: (1) a band width, or the total number of hours in a given workday, (2) a core time, or designated period of time during which all employees are required to be working, (3) a flexible band of hours both before and after the core time that allows employees to vary their starting and quitting times, (4) bank- ing, which allows carry-over of surplus or deficient hours worked, and (5) variability of schedule-the employees' freedom to vary working hours from one period to another without prior approval from their supervisor (Golembiewski & Proehl, 1978). The degree of flexibility and amount of discretion permitted employees depend on the variations in the above five components in the program.

In spite of its popularity in application and numerous success stories, very little is known about the effectiveness of flexitime programs. Most

'An earlier version of this paper was presented at the National Academy of Management Confer- ence, Detroit, Michigan, August, 1980. The authors would like to thank S. D. Nollen and two anony- mous reviewers for their suggestions on an earlier version of the paper.

729

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms

730 Academy of Management Journal December

studies of them are anecdotal, testimonial, and post hoc in nature; few are based on rigorous empirical investigation. Following an extensive review of literature on flexitime, Golembiewski and Proehl (1978) have reported that of the 16 empirical studies examined, only one included any statistical treatment of the data reported. Thus, the results typically attributed to the flexitime program could be artifacts that would cast serious doubt on the internal validity of the findings. Other methodological and measurement weaknesses commonly shared by the flexitime studies include: (1) the ab- sence of a control group, (2) lack of pre-intervention measures, and (3) the use of "soft" data as criteria.

Notwithstanding these weaknesses, a number of investigators have ar- gued that flexitime has a positive impact on employee attitudes among su- pervisory employees (Partridge, 1973) and nonsupervisory employees in a large insurance company (Evans, 1973), among employees in a pharma- ceutical firm (Golembiewski & Hilles, 1977), and in a computer firm (Hopp & Sommerstad, 1977). In addition, flexitime has been reported to help reduce absenteeism and sick leave (Golembiewski & Hilles, 1977; Mueller & Cole, 1977) and to help increase productivity (Martin, 1975; Nollen & Martin, 1978). In general, the proponents of this program have argued that permitting employees to exercise flexibility in their arriving and quitting time will reduce absenteeism, tardiness, overtime, etc., be- cause these would be accounted for under allowed discretionary time. Fur- ther, it has been argued that under the flexitime program the employees would be able to adjust their work activities to their individually more pro- ductive hours, resulting in a more efficient utilization of labor input in a work setting (Nollen, 1979). This increase in performance efficiency would further allow the employees to make positive alterations in their attending and their producing behavior.

Although this proposition has not been explicitly tested, three studies with acceptable methodology and measurements have dealt with the im- pact of flexitime on employees' attendance and performance. Investigat- ing approximately 60 employees in research and development units under a flexitime program, Golembiewski, Hilles, and Kagno (1974) reported findings that suggest that flexitime reduces overtime and absenteeism. They found that the total number of paid absences in the experimental group during a 375-day period was reduced by 35 percent from the imme- diately preceding year, while absenteeism increased 15 percent in the con- trol group. On the other hand, the expected decrease in short term ab- sences (a single day or less) did not occur in the experimental group. In fact, they increased by nearly 13 percent, as compared to an increase of 21 percent in the control group. These findings refute the common belief that flexitime reduces short term absences more than long term absences. In addition, as the authors themselves pointed out, "experimentals and com- parisons started from substantially different bases" (1974, p. 528). This initial difference should have been taken into account when experimental and control groups were compared on criterion variables. In the second

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms

1981 Kim and Campagna 731

study, Harvey and Luthans (1979), investigating the effect of flexitime on employee attendance in a state agency, concluded that flexitime appears to have a relatively favorable effect on absenteeism. However, they did not offer any statistical treatment of the absenteeism data in their study. As is frequently the case, the positive effect of flexitime on employee atten- dance has been claimed with relatively tenuous empirical evidence.

Studies investigating the impact of flexitime on performance are just as limited as are those on absenteeism. One notable exception is a study by Schein, Maurer, and Novak (1977). They investigated the impact of flexi- time on the performance of 246 clerical level employees in five production units within a large financial institution over a period of four months. For two of the production units, in which both an experimentally designated control group and pre- and post-measures were employed, the results revealed no significant differences in performance between the experimen- tal and control groups. In another production unit, however, productivity during the flexitime period was found to be significantly greater than that in the same period of the previous year. Based on these findings, the authors suggested that no clear-cut conclusions with respect to perfor- mance can be offered but that flexitime appears to have no adverse impact on it. In that study, however, the potential changes in the total work hours resulting from the flexitime program were not taken into account in mea- suring productivity. Because the flexitime program has been reported to influence the employees' attending behavior (Golembiewski et al., 1974; Harvey & Luthans, 1979), the results of the Schein et al. study on produc- tivity may have been attributable in part to the confounding of the cri- terion measure with the changes in the total hours worked during the ex- perimental period.

The present study attempts to remedy some of the methodological weaknesses of previous studies in several important ways. First, this study employs a pre-intervention measure of each criterion for both the experi- mental and the control group. This allows the necessary adjustment of cri- terion variables for the initial differences between groups. Second, this study investigates the effectiveness of flexitime using "hard" criterion measures. Especially, it uses the performance efficiency measure after the potential impact of flexitime on attendance behavior is taken into ac- count. This permits testing whether employees under the flexitime pro- gram are performing more efficiently on an hourly basis. Third, the nature of the task under investigation is taken into account when performance measures are analyzed because Nollen (1979) suggested that employees' behavioral responses to flexitime tend to vary with different types of tasks. Fourth, this study attempts to increase the external validity of the findings reported by Schein et al. (1977) by investigating the same research issues in a public sector setting. According to Nollen (1979), there is some evidence suggesting that flexitime has a greater impact on performance in the pri- vate than in the public sector. Given the increasing number of flexitime implementations in the public sector, such information is required in order to guide future action.

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms

732 Academy of Management Journal December

The research issues investigated in this study are: (1) Does the flexitime program affect the employees' absenteeism? (2) Does the flexitime program affect the employees' performance effi-

ciency?

METHOD

Subjects and Procedure

A total of 353 employees in 4 divisions of a county welfare agency par- ticipated in this study. Each division consisted of several administrative units, and each administrative unit was supervised by a common superior. Approximately 78 percent of the participants were female, 53 percent were black, and 90 percent had at least a high school education or some college training. The participants were responsible primarily for determining the eligibility of welfare cases and were all at the same organizational level.

A nonequivalent control group design was employed in this study. Prior to the flexitime intervention, the administrative units in each division were randomly assigned to the experimental group and to the control group. This randomization procedure, repeated in each division, allowed approx- imately equal numbers of employees in the experimental group and the control group.

The employees participating in the experimental group were allowed to begin their work any time between 6:30 a.m. and 9:30 a.m. and leave be- tween 3:00 p.m. and 6:00 p.m. (the designated pre- and post-work band). All employee participants were required to be at work between the hours of 9:30 a.m. and 3:00 p.m. (the designated core time). In addition, all par- ticipants were required to work an 8-hour day. Those employees assigned to the control group continued to work their normal 8-hour days (8:00 a.m. to 4:30 p.m.). All employees who had overtime had the choice of tak- ing one and a half compensatory time-off or taking one and a half over- time pay. During the four months of the experimental period, there was no change in the supervisory personnel in any of the administrative units under investigation.

Measures

The criterion measures were collected on each participant in the experi- mental and control groups from organizational records for the month prior to the flexitime intervention and for the subsequent four months of the intervention period. Two different types of absenteeism were ob- tained: paid absences and unpaid absences. Each of these was divided fur- ther into short-term and long-term absences. Short-term paid absences for an employee were operationally defined as a monthly total of paid ab- sences (i.e., sick leave) which was taken in blocks of two hours or less a day. Long-term paid absences were operationally defined as the monthly total of paid absences taken in blocks of more than two hours a day.

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms

1981 Kim and Campagna 733

Similarly, short-term unpaid absences for an employee were operation- ally defined as the monthly total of leave time without pay that amounted to two hours or less a day. Long-term unpaid absences were operationally defined as the monthly total of leave time without pay that amounted to over two hours a day. These four measures defined on an a priori basis were available for 161 employees in the experimental group and for 185 employees in the control group.

In regard to performance, it was found that four divisions under the study varied as to the required routine actions, the client contact, the na- ture of the cases to be processed, etc. Consequently, this agency had em- ployed different performance measures in each of these divisions. There- fore, in order to provide a basis for comparison, a performance efficiency measure was computed and was standardized within each division across the 5-month period. The performance efficiency measure for an employee was operationally defined as the rate of output quantity per hour by com- puting the ratio of monthly output to the actual hours worked during a given month. The actual hours worked during the month were obtained by subtracting the total hours of absences, vacation time, and compensatory leave time from the total work hours available plus overtime worked for the month. These performance efficiency measures, analyzed for each sep- arate division, included: (1) the total number of cases approved during the month for Division 1; (2) the total number of routine actions completed during the month for Division 2; (3) the total number of cases processed during the month for Division 3, and (4) the total number of cases pro- cessed during the month for Division 4. To supplement the four division analyses, the performance efficiency measure was standardized within each division to provide the basis for comparison across all four divisions.

It was learned that this agency had been monitoring the employees' at- tendance behavior much more closely than it had the producing behavior such as output quantity. Consequently, when the measures to compute performance efficiency were compiled, it was found that for only 94 em- ployees was there a complete set of necessary data (i.e., output measure, paid and unpaid absences, compensatory leave time, vacation time, over- time, etc.) for each of the five months of the intervention period. All data were compiled by the employees in the personnel department from the centralized standard logs and organizational records, which the unit super- visors were required to report on a monthly basis.

In order to control for the pre-intervention differences between the ex- perimental and the control group, analysis of covariance using the pre- intervention measure as a covariate was employed in analyzing the cri- terion variables for each separate division and for the four divisions com- bined.

RESULTS

Table 1 is a summary of the means, adjusted means, and standard de- viations of the short-term and the long-term unpaid absences. The

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms

734 Academy of Management Journal December

TABLE 1

Summary of Means, Adjusted Means, and Standard Deviations of Unpaid Absences

(N=346)

PRE 1-Month 2-Month 3-Month 4-Month Marginal Criterion Measures M SD M SD M SD M SD M SD Meansb

Short-term unpaid absences:

Experimental group (n = 161) .192 .583 .124 .527 .118 .442 .161 .517 .093 .400 .124 (.128)a (.122) (.166) (.097) (.128)

Control group (n = 185) .218 .742 .256 .783 .262 .807 .332 1.062 .267 .794 .279 (.252) (.258) (.328) (.263) (.275)

Long-term unpaid absences:

Experimental group (n = 161) 4.232 19.384 1.335 5.032 2.962 8.913 3.422 11.346 1.465 5.222 2.296 (1.281) (2.909) (3.368) (1.412) (2.242)

Control group (n= 185) 2.005 6.937 2.389 8.420 3.921 11.567 4.662 20.510 4.921 35.921 3.973 (2.435) (3.968) (4.708) (4.968) (4.019)

aThe adjusted means using the monthly measure immediately preceding the intervention as covariate.

bThe average of the criterion measure for the four month intervention period combined.

TABLE 2

Summary of Means, Adjusted Means, and Standard Deviations of Paid Absences

(N= 346)

PRE 1-Month 2-Month 3-Month 4-Month Marginal Criterion Measures M SD M SD M SD M SD M SD Meansb

Short-term paid absences:

Experimental group (n = 161) .503 1.108 .422 .944 .332 .815 .403 .809 .319 .793 .369 (.418)a (.328) (.399) (.315) (.365)

Control group (n = 185) .368 .873 .256 .636 .244 .582 .291 .796 .397 .834 .292 (.260) (.228) (.295) (.400) (.295)

Long-term paid absences:

Experimental group (n= 161) 9.875 10.619 6.732 7.725 6.807 7.583 8.416 12.060 7.484 7.635 7.359 (6.688) (6.762) (8.371) (7.439) (7.315) Control group (n =185) 7.737 9.552 7.718 14.602 8.762 11.828 8.172 8.912 7.659 9.746 8.077 (7.757) (8.800) (8.211) (7.698) (8.116)

aThe adjusted means using the monthly measure immediately preceding the intervention as covariate.

bThe average of the criterion measure for the four month intervention period combined.

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms

1981 Kim and Campagna 735

adjusted mean of short-term unpaid absences during the 4-month inter- vention period was .128 for the experimental group and .275 for the con- trol group. A 2 (experimental vs. control) x 4 (time) analysis of covariance using the pre-intervention measure as a covariate revealed a statistically significant difference between the experimental and the control group (F1,343= 9.15, p < .002). This result represents approximately a 35 percent decrease in the experimental group's short-term unpaid absences from the pre-flexitime period to the flexitime period, and there was about a 28 per- cent increase in the short-term unpaid absences among the employees in the control group. Neither the time effect nor the interaction effect was statistically significant.

Long-term unpaid absences in the experimental group decreased by 45 percent from the pre-flexitime period to the flexitime period. On the other hand, there was about a 98 percent increase in the same measure among the employees in the control group. This increase appears to be notable, but it did not reach a level of statistical significance when the initial differ- ence between the experimental and the control group was accounted for (F1,343 = 2.24; N.S.). As with short-term unpaid absences, neither the time effect nor the interaction effect was noted. Yet, the trend of the data on unpaid absences seems to be clear. Long-term unpaid absences were lower in the experimental group than in the control group in every month during the 4-month flexitime intervention period. In fact, the long-term absen- teeism among the employees in the control group was more than three times greater than that in the experimental group for the last 30 days of the flexitime intervention.

Table 2 shows the findings on the short-term and the long-term paid ab- sences taken by the employees in the experimental and the control groups. Although the short-term paid absences among the employees under the flexitime program were slightly greater than the same measure among the employees in the control group, no statistically significant difference was observed during the 4-month intervention period (M'=.365 vs. M'= .295; F1,343 = 2.27, N.S.). Neither time effect nor the group by time interac- tion effect was statistically significant. For the long-term paid absences, the findings were reversed with the similar statistical results. The em- ployees under the flexitime program tended to show a lower amount of paid absenteeism than those in the control group, but this difference did not reach a statistical level of significance. Again, neither time effect nor the group by time interaction effect was statistically significant.

As suggested earlier, two different analyses were conducted for perfor- mance efficiency: (1) four separate division analyses to account for the variations in the nature of the task; and (2) a combined analysis across the four divisions to increase the degree of generalizability. Table 3 shows the summary of means, adjusted means, and standard deviations of perfor- mance efficiency for each of the four divisions under study.

For division 1, the adjusted mean for the experimental group was .803 as compared to .583 for the control group during the 4-month intervention

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms

736 Academy of Management Journal December

TABLE 3

Summary of Means, Adjusted Means, and Standard Deviations of Performance Efficiency in Four Divisions

(N= 94)

PRE 1-Month 2-Month 3-Month 4-Month Marginal Criterion Measuresa M SD M SD M SD M SD M SD Meansc

Division 1:

Experimental group (n=6) .507 .586 .768 .559 .802 .671 .671 .598 .590 .157 .707 (.793)b (.833) (.699) (.889) (.803) Control group (n = 5) .568 .032 .763 .190 .587 .082 .553 .082 .804 .746 .676 (.733) (.550) (.500) (.549) (.583)

Division 2:

Experimental group (n =14) .725 .314 .948 .332 .745 .332 .873 .279 .794 .260 .840 (.969) (.758) (.887) (.808) (.855) Control group (n= 16) .769 .288 .732 .342 .681 .408 .827 .375 .741 .450 .745 (.714) (.671) (.814) (.728) (.731)

Division 3:

Experimental group (n =16) .286 .114 .301 .120 .300 .131 .246 .069 .263 .075 .277 (.285) (.281) (.238) (.250) (.263) Control group (n = 26) .237 .075 .262 .085 .259 .097 .248 .089 .224 .107 .248 (.271) (.271) (.253) (.232) (.256)

Division 4:

Experimental group (n 6) .165 .067 .177 .078 .247 .120 .240 .130 .172 .083 .209 (.178) (.249) (.240) (.172) (.209)

Control group (n = 5) .168 .162 .164 .104 .395 .472 .194 .079 .234 .055 .246 (.164) (.393) (.193) (.234) (.246) aDivision 1: Hourly rate of cases approved during the month; Division 2: Hourly rate of routine actions completed during the month; Division 3: Hourly

rate of cases processed during the month; Division 4: Hourly rate of cases processed during the month.

bThe adjusted means using the monthly measures immediately preceding the intervention as covariate.

CThe average of the criterion measure for the four month intervention period combined.

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms

1981 Kim and Campagna 737

period. A 2 (experimental vs. control) x 4 (time) analysis of covariance using the pre-intervention measure as a covariate revealed a statistically significant difference between the experimental and the control group (F1,8 = 12.642; p < .007). Although the pattern of performance efficiency between the two groups appeared to be different during the experimental period, the group by time interaction effect failed to reach a statistical level of significance (F3,24 = 1.948; N.S.). The results for the remaining three divisions showed that neither the main nor the interaction effect was statistically significant. Yet, in two (divisions 2 and 3) of these three divi- sions, performance efficiency tended to be higher for the experimental group than for the control group. For example, when the initial difference in performance efficiency prior to the intervention was accounted for, it was found that the average hourly rate of routine actions completed for the experimental group in division 2 was about 17 percent higher than that for the control group during the intervention period. In division 4, a re- versed trend was shown. The average hourly rate of cases processed for the experimental group was about 16 percent lower than that for the control group during the intervention period.

Table 4 shows the findings on the standardized performance efficiency for all four divisions combined. The standardized average performance ef- ficiency adjusted for the initial difference was higher for the experimental group (M' = .109) than for the control group (M' = -.076). However, this

difference did not reach a statistical level of significance (F1,91 = 1.590; N.S.). Neither the time effect nor the group by time interaction effect was statistically significant.

Although excluded from the overall experimental design of this study, additional data were obtained by an evaluation committee of seven agency members on the employees' reactions on the flexitime time. The em- ployees' responses to the flexitime program, obtained by the questionnaire after the intervention period, were very favorable. For example, among those in the experimental group, approximately 84 percent reported that flexitime had a strong positive influence on their morale; 92 percent re- ported that the flexitime program had a strong influence on arranging time for personal matters outside of agency hours; 62 percent reported that the flexitime program improved their job satisfaction; 88 percent reported that the flexitime program had a strong positive effect on the reduction of morning tension; 91 percent reported that flexitime had a positive influ- ence on the reduction of commuting time and traffic congestion problems, and 61 percent reported that the flexitime program reduced their parking problem.

CONCLUSION AND DISCUSSION

In spite of a general lack of empirical supports, it has been often sug- gested that the employees under a flexitime program would reduce absen- teeism and increase their productivity. The results of this study suggest

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms

738 Academy of Management Journal December

TABLE 4

Summary of Means, Adjusted Means, and Standard Devilations of Standardized Performance Efficiency in Four Divisions Combilned

(N = 94)

PRE .1-Month 2-Month 3-Month 4-Month Marginal Criterion Measures M SD M SD M SD M SD M SD MeanSb Experimental group (n =42) .095 1.192 .246 1.118 4154 1.065 .121 1.092 .098 .972 .154 (.195)a (.107) (.080) (.057) (.109) Control group (n =52) -.055 .895 -.148 .884 -.053 .959 -.044 1.017 -.100 1.049 -.086 (-.107) (-.015) (-.01 1) (-.067) (-.076)

aThe adjusted means using the monthly measure immediately preceding the intervention as covariate.

bThe average of the criterion measure for the four month intervention period combined.

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms

1981 Kim and Campagna 739

that flexitime significantly reduces employees' unpaid absences in general. However, the positive impact of flexitime on short-term unpaid absen- teeism (leave without pay for two hours or less a day) appears to be much stronger than on the long-term unpaid absenteeism (leave without pay for more than two hours a day). This finding suggests that the flexible work- ing hours allowed under the flexitime program have served as a substitute for short-term leave without pay. This substitutability was not apparent for the long-term unpaid absences. However, no substantial increase or decrease in paid absences was noted. Employees under the flexitime pro- gram tend to take just as many paid absences (i.e., sick leave) as the em- ployee with regular work hours.

As stated earlier, empirical studies investigating the effect of flexitime on performance are virtually nonexistent. One study by Schein et al. (1977) suggested that a flexitime program has no adverse impact on per- formance in a private sector organization. The results of the present study, conducted in a public sector agency, concurs with that finding. In three out of four divisions, the performance during the 4-month flexitime period was higher in the experimental group than in the control group, al- though in only one of these divisions did it reach a statistical level of sig- nificance. Thus, one may conclude that flexitime had either a positive im- pact on performance or did, at a minimum, encourage an increase in per- formance among public employees.

The results of this study imply that the employees under flexitime tend to take it as an alternative benefit to be used for nonwork activities with- out affecting their economic benefits. This would imply that, although one may expect the reduction of employees' leave without pay under the flexitime program, the costs associated with absenteeism would not neces- sarily be reduced. Yet, to the extent to which flexitime increases efficiency of labor input, the overall organizational effectiveness can be enhanced.

However, the findings also imply that the potential positive impact of flexitime on performance efficiency may not be constant across different tasks. For example, Partridge (1973) reported findings that suggest that flexitime tended to hinder managers in supervising their staffs, in fitting in with unpredictable work times, and in setting up formal meetings in the office. He concluded that flexitime can generate difficulties in coordinat- ing subordinates' activities that arise from uneven work flow. One can argue that the degree of difficulty caused by flexitime would vary accord- ing to the kind of coordination needed for the completion of a certain task. If the job requires an extensive communication and information ex- change among supervisors, co-workers, clients, etc., the potential payoff in performance under flexitime may be low. On the positive side, if the nature of the task and task interdependence is such that each employee can perform an entire module in a relatively autonomous manner (i.e., pooled interdependence), performance can be increased under the flexitime pro- gram. Because the nature of interdependence and coordination needs for task completion was not measured in this study, one can only speculate on

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms

740 Academy of Management Journal December

the validity of this argument. However, if it is valid, it implies that the na- ture of the task should be considered carefully before a decision to imple- ment flexitime is made. For example, if the nature of the task is sequential in nature (i.e., assembly line), the implementation of flexitime would not be feasible or would incur a considerable increase in coordination costs in production.

Two methodological limitations of this study should be noted. First, the employees in both the control group and the experimental group knew that they were participating in a field experiment and that the results of this experiment would have potential influence on policy decisions on flex- itime program implementation for the entire agency. Because the em- ployees were not "blind" to what was taking place, their responses might have been influenced by the "Hawthorne effect" and/or by a "Rosenthal effect." This possibility could not be totally ruled out from the current re- search context: the policy of this agency was such that the employees were kept informed on this flexitime experiment. However, the employees, in- cluding supervisors, did not have knowledge as to which criteria would be employed in evaluating flexitime effectiveness. Further, because (1) the employees were led to believe the questionnaire responses to be the pri- mary data source for the evaluation of the flexitime program and (2) the criterion measures were compiled from the centralized data pool by a third party, the possible influence of the expectations on the experimental group would have been minimal. Second, subjects were not selected completely at random. However, the intact groups (i.e., administrative units) were chosen randomly and assigned to the experimental and the control group for each division. They thus shared all practical bases such as common tasks, common supervision, and common level. Thus, meaningful com- parisons could be made in each division.

Notwithstanding these limitations, the results of this study appear to justify the argument that the flexitime program has a positive impact on attendance and performance among public sector -employees. It should be noted that this study investigated only one of several forms of the flexitime program. Additional research on various other forms of the flexitime pro- gram in different organizational contexts is clearly warranted. Future studies might fruitfully investigate the potential moderating effect of task on the effectiveness of flexitime as an organization development interven- tion.

REFERENCES

1. Evans, M. G. Notes on the impact of flexitime in a large insurance company: 1. Reactions of non- supervisory employees. Occupational Psychology, 1973, 47, 237-240.

2. Golembiewski, R. T., & Hilles, R. J. Drug company workers like new schedules. Monthly Labor Review, 1977, 100, 69-71.

3. Golembiewski, R. T., & Proehl, C. W. A survey of the empirical literature on flexible work- hours: Character and consequences of a major innovation. Academy of Management Review, 1978, 3, 837-853.

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms

1981 Kim and Campagna 741

4. Golembiewski, R. T., Hilles, R., & Kagno, M. S. A longitudinal study of flexitime effects: Some consequences of an OD structural intervention. Journal of Applied Behavioral Science, 1974, 4, 503-532.

5. Harvey, B. H., & Luthans, F. Flexitime: An empirical analysis of its real meaning and impact. MSU Business Topics, Summer 1979, 31-36.

6. Hopp, M. A., & Sommerstad, C. R. Reaction at computer firm: More pluses than minuses. Monthly Labor Review, 1977, 100, 69-71.

7. Martin, V. H. Hours of work. Washington, D.C.: Business and Professional Women's Founda- tion, 1975.

8. Mueller, 0., & Cole, M. Concept wins converts at federal agency. Monthly Labor Review, 1977, 100, 71-74.

9. Nollen, S. C. Does flexitime improve productivity. Harvard Business Review, 1979, 57 (5), 16-18, 76, 80.

10. Nollen, S. D., & Martin, V. H. Alternative work schedules, part I: Flexitime. New York: AMACOM, 1978.

11. Partridge, B. D. Notes on the impact of flexitime in a large insurance company: II. Reactions of supervisors and managers. Occupational Psychology, 1973, 47, 241-242.

12. Schein, V., Maurer, E., & Novak, J. Impact of flexible working hours on productivity. Journal of Applied Psychology, 1977, 62, 463-465.

This content downloaded from 205.146.48.6 on Fri, 01 Feb 2019 18:12:41 UTC All use subject to https://about.jstor.org/terms