5.1 Analysis and Classification of Stayers and Quitters

Many evaluation reports on telecommuting programs make little or no reference to attrition among telecommuters. Very few actually study attrition seriously (Quaid and Lagerberg (1992) is one exception), yet the importance of this factor to estimating the adoption and impacts of telecommuting cannot be emphasized enough. Forecasts of the proportion of the workforce likely to become telecommuters (and derivative forecasts, such as impacts on transportation) implicitly assume "once a telecommuter, always a telecommuter". But if the typical participant only telecommutes six months before terminating the arrangement (or, viewed another way, if only two percent of those who will ever telecommute are doing so at any given time), then impact assessments based on the proportion of the workforce ever expected to telecommute will be wildly overstated. In this chapter, we examine the attrition of telecenter users and explore the reasons for that attrition.

Telecommuting retention is analyzed in the following three ways. In Section 5.2, the attitudes and characteristics of those participants who remain in the program (stayers) are compared with those who left it (quitters). The goal is to determine which work and household variables may affect the decision to stop telecommuting from a center. In this study, work and household data was collected using an attitudinal survey as described in Section 3.2. A subset of the before-wave data is used to compare and contrast the stayers and quitters according to selected factors that are hypothesized to influence the decision to quit.

Another way to determine the causes of attrition is to examine the motivations to quit, as presented in Section 5.3. Some possible reasons for dropping out of a telecommuting program include residential relocation, changes in job duties, technical problems, and discomfort with telecommuting. Particularly important to this project, those who quit using a telecommuting center may, in fact, prefer telecommuting from home. The motivation for leaving the telecenter program was the most important part of the exit interview, which was conducted, where possible, with each telecenter user who left the project.

The comparison of stayers and quitters in terms of the duration and frequency of telecenter use, presented in Section 5.4, is still another way to analyze telecommuting retention. Individual telecommuting duration and frequency can be computed from the attendance log data as described in Section 4.4. A particular pattern of telecenter use may be characteristic of each group. These patterns may suggest a possible motivation for quitting or may be used as an indicator of the likelihood to quit. In addition, questions about the current and ideal distribution of work time were included in the exit interview. A difference between current and ideal work frequency can show whether a preference for telecommuting from a center still exists even though the respondent has ceased to telecommute from the center.

Before the aforesaid analyses can be conducted, an appropriate study group must be chosen. The identification of the sample of stayers and quitters begins with the group of respondents to the before employee survey. The before-wave data are used here since not all participants in the RABO Project (especially those who quit) completed after surveys. The attitudinal data will allow the comparison of attitudes and characteristics between the two groups of interest (see Section 5.2). Of the 61 respondents in the before telecenter user data set, 15 respondents from the sites that ceased operation or withdrew from the RABO Project (Roseville, Anaheim, Concord, Citrus Heights, and Davis) were excluded from further consideration. For these cases, the reason for quitting is likely to be the closing of the center rather than employment requirements or personal preference. In addition, attendance log and exit interview data are scarce or non-existent from these sites (in the final report, telecommuting termination caused by site closing will be investigated further). The remaining 46 respondents were assigned stayer or quitter status according to information collected prior to June 30, 1995 from the attendance logs. The assignations were then cross-checked with current tracking of telecommuters to verify the respondents' project participation status. According to the classification process, 22 of the respondents are stayers and 24 are quitters.

5.2 Attitudes and Characteristics

Unlike the earlier description of telecommuter characteristics (see Section 3.2), the following comparison of the stayers and quitters only uses the data collected in the before surveys. Data from the early wave of surveys are used since the majority of the quitters ceased telecommuting prior to the administration of the later wave of surveys. Thus, information on attitudes and characteristics is only available from the before surveys. The comparison of the survey results for the 22 stayers and 24 quitters covers the sections of the survey in the following order: demographic and travel characteristics, job characteristics, amount of telecommuting, and job satisfaction and work environment characteristics.

The stayers and quitters are relatively similar according to demographic indicators. There are comparable distributions for gender, age, and educational background. Both groups are statistically similar in terms of household size and available vehicles (see Table 5-1). Quitter households have more part-time workers and generally have higher incomes; a total of twelve quitters are in the top two income categories compared to only six stayers. While the stayers and quitters had similarly long commutes to the regular workplace (about 41 miles, on average), quitters had to travel 3.5 miles farther, on average, to the telecenter.

Table 5-1: Demographic and Travel Characteristics for Stayers and Quitters

Question	Mean1		t-statistic	p-value
	Stayers (N=22)	Quitters (N=24)
Household size	2.91	2.71	-0.54	0.594
Full-time workers	1.45	1.21	-1.14	0.259
Part-time workers	0.18	0.54	2.22	0.031
Vehicles per household	2.18	1.96	-1.07	0.293
Miles to regular workplace (one-way)	41.402	41.69	0.04	0.969
Miles to telecenter (one-way)	4.112	7.72	1.50	0.140

¹ Bolded means are significant at a .LE. 0.10.     ² N = 21

Job characteristics also did not vary much between stayers and quitters. Similar numbers of the two groups had manager/administrator and professional/technical positions, but five of the seven administrative and only one of the six sales workers were quitters. Perhaps administrative support duties were less amenable to telecommuting from a center than sales tasks since the latter usually involves site visits while the former requires office equipment and files. Quitters had generally longer experience (about two more years with the firm and in the occupation) which suggests a possible difficulty in adapting to a new work arrangement (see Table 5-2). This hypothesis is supported by the fact that more quitters (9) had conventional eight-hour schedules than stayers (4) and more stayers (13) had flextime schedules than quitters (5). However, both groups had worked for their respective supervisors about the same length of time on average, and none of the differences in the experience variables were statistically significant. Surprisingly, quitters spent significantly more time working independently and remotely (tasks well-suited for telecommuting) than stayers and, consequently, spent less time working face-to-face or at a specific location (see Table 5-2). Stayers had significantly more work-related travel, probably because more of this group were sales people.

Table 5-2: Work Characteristics for Stayers and Quitters

Question	Mean1		t-statistic	p-value
	Stayers (N=22)	Quitters (N=24)
Years worked for supervisor	2.582	2.853	0.27	0.792
Years worked for employer	6.72	8.563	0.83	0.412
Years worked in occupation	8.062	10.154	0.91	0.367
Independent and remote time	58.27%	71.13%	1.91	0.062
Face-to-face and location time	28.36%	22.69%	-1.16	0.253
Work-related travel time	12.91%	6.19%	-1.85	0.071

¹ Bolded means are significant at a .LE. 0.10.
² N = 20     ³ N = 21     ⁴ N = 22

The two groups also had comparable experience with telecommuting. While a similar proportion of each group (about one-quarter) had prior experience with telecommuting (most likely the home-based form), the quitters have had the longer average duration of past telecommuting experience (see Table 5-3). However, this difference is not statistically significant and is based on a small sample (seven stayers and six quitters).

The average ideal distributions of time among the possible work locations were relatively similar (see Table 5-3); however, the quitters did assign a slightly larger percentage to the regular workplace and a slightly smaller percentage to the telecommuting center. Quitters and stayers have similar average center-based telecommuting frequencies for job suitability, manager support, preference, and expectation (see Table D-1, Appendix D). However, there are larger differences for home-based telecommuting for the same categories, with quitters favoring this option more than stayers. In fact, quitters preferred to telecommute from home for 24% of the time compared to only 12% for stayers. However, quitters preferred to telecommute from a center more than twice as frequently as from home (50% of the time). So, for some of the quitters, their experience with and preference for home-based telecommuting may have led them to choose to telecommute from home rather than from a center.

Table 5-3: Telecommuting Experience and Preference for Stayers and Quitters

Question	Mean1		t-statistic	p-value
	Stayers (N=22)	Quitters (N=24)
Years of telecommute experience	1.602	5.003	1.18	0.263
Ideal % at the regular workplace	41.32%	47.97%	0.91	0.367
Ideal % at the telecenter	42.05%	36.82%	-0.70	0.489
Ideal % at home	10.05%	11.25%	0.23	0.816
Preferred % from a center	55.90%	50.21%	-0.64	0.525
Preferred % from home	12.16%4	23.54%5	1.72	0.094

¹ Bolded means are significant at a .LE. 0.10.
² N = 7     ³ N = 6     ⁴ N = 21     ⁵ N = 19

There were few meaningful differences between stayers and quitters in attitudes about work and the workplace. In fact, none of the average responses to job performance and satisfaction questions had significant differences (see Table D-2, Appendix D). A few of the work environment characteristics were different, however (see Table D-3, Appendix D). Stayers expected distractions at the telecommuting center to be less of a problem than quitters (p = 0.012), so perhaps the quitters are likely to be more sensitive to distractions than the stayers. In addition, quitters were less worried about over-indulging at the regular workplace (p = 0.038) or at the center (p = 0.011) and felt they had fewer opportunities for professional interaction at the regular workplace (p = 0.047). T-tests on the mean importance rating of the work characteristics showed only two significant differences between the groups (see Table D-4, Appendix D). Quitters rated having equipment and services and dressing comfortably more important than stayers did (p = 0.031 and 0.050, respectively). The need for equipment may show a preference for returning to the regular workplace, but the desire to work comfortably likely reflects a desire for working at home where formal business attire is not necessary.

Although the results of the attitudinal survey serve to characterize the stayers and quitters, they do not give much insight into why some participants chose to quit. Even the statistically significant results reported above do not provide compelling reasons for wanting to stop telecommuting. In fact, the quitters do not view the telecenter negatively for any characteristic which might be expected of those who want to quit (such as feeling motivated at work, distractions from others, having necessary equipment, etc.). However, these surveys were administered prior to the start of telecenter use, so attitudes about working from the telecommuting center may have changed over time. As a result, there is little basis for predicting who will quit telecommuting from the center using the data collected with the before wave of attitudinal surveys.

5.3 Reason for Quitting

For the 24 participants who were identified as quitters (see Section 5.1), an attempt was made to conduct an exit interview with each one, primarily in order to identify the reason for quitting. However, contacting and eliciting information from all quitters proved to be difficult since they may no longer have felt an obligation to participate or may have changed phone numbers. Despite this, at least some data (some of it second-hand) is available for all but four of the quitters.

The reason(s) given for leaving the telecenter program are shown in Table 5-4 (the frequencies do not sum to 24 because one respondent gave more than one reason for quitting). The most frequently given explanation (25% of the quitters) was that they changed positions within the company or the tasks assigned to their position changed. This was mostly the result of corporate downsizing which required employees to take over some tasks that formerly belonged to other positions. Thus, they needed to be at the main office more or in some other way could not complete their new tasks at the telecommuting center. Similarly, on a previous study of home-based telecommuting retention, office problems and job changes accounted for 29% of the reasons for dropping out of the program (Quaid and Lagerberg, 1992).

Table 5-4: Reasons for Quitting

The next most important reasons relate specifically to the supervisor and/or employer. Supervisors' and/or employers' attitudes led five of the participants to quit (21%). Of these five employees, two employees were required to quit by their current employer and/or supervisor, two had new supervisors which required the change, and one was encouraged to quit by the employer and/or supervisor. Other reasons for quitting were leaving the company for a new job (12.5%) and residential relocation (8.3%). For four of the 24 quitters (16.7%), the reason for leaving the program is unknown. It appears that these telecenter users were generally forced to quit because of outside circumstances (change in job duties or supervisor decision). This contrasts with the previous retention study where lack of equipment (17%) and personal dissatisfaction with telecommuting (8%) were important reasons for quitting (Quaid and Lagerberg, 1992).

The exit interview also contained a pair of questions about the prospects for future telecommuting by the respondent. Of the nine quitters (37.5%) who completed this part of the interview, only one respondent categorically refused to consider telecommuting from a center again, while two would not consider telecommuting from home. For this subgroup of the quitters, their preference is to continue to use telecommuting centers. The preference for more telecenter use in the future, as well as the reasons given for quitting, suggest that external constraints often prevent the adoption and the continuation of telecommuting programs.

5.4 Telecommuting Duration and Frequency

The attendance logs are the primary data source for telecommuting duration and frequency (see Sections 4.4.1 and 4.4.2 for duration and frequency for the overall sample of project participants). Although both the attitudinal survey and the exit interview ask these questions about telecenter use, these responses are considered to be only rough estimates. While there are undoubtedly a few missing telecommuting occasions in the attendance logs, they likely provide the best measurement of both duration and frequency. Importantly, some of the quitters had zero duration and frequency: these four (16.7%) did not use the center at all. In addition, four more quitters used the center only once, and hence both duration and frequency measures will be zero for them. As a result, statistics for both the total set and the "non-zero" set of quitters are reported below.

The average duration of telecommuting can be used to estimate how long telecommuters will stay in the program before dropping out. For stayers, this measure is not final since they are still telecommuting. However, for comparison purposes the current duration of telecommuting for stayers is 10.7 months, on average, as of June 30, 1995. From the attendance log data, the average duration for all quitters is 5.1 months (assuming 30-day months). For those quitters who telecommuted at least two occasions, the average duration was about 2 months longer, 7.2 months (see Table 5-5). The median duration for both groups is shorter (4.3 and 6.5 months, respectively), which suggests that the majority of the quitters leave the telecommuting program in less than 7 months. The distribution of telecommuting duration by the quitters, shown in Figure 5-1, has a peak between 5 and 8 months and a long tail.

Unlike telecommuting duration, the frequency of telecommuting occasions for both stayers and quitters can be compared (see Table 5-6). The average frequency of telecommuting among stayers is 29.6%, or about 1½ days per week. For all quitters, the average frequency is 16.0%, but the average of those who telecommuted more than once is 24.0% (1¼ days per week). Although the stayers telecommute more frequently on average, the difference between the two groups is small. As above for duration, the median is less than the mean for telecommuting frequency for all three groups. The distribution of telecommuting frequency is shown in Figure 5-2. Over half of the non-zero quitters (9 out of 16) telecommuted one to three days per month, while only a little less than one-third (7 of 22) of the stayers telecommuted this amount. In contrast, the majority of the stayers (13 of 22) telecommuted more than one day per week, while less than half of the quitters telecommuted this much (7 of 16). Although the differences in frequency are not that great, it appears that the high-frequency telecommuters are more likely to continue using the telecenters. This is not to say that higher frequency of telecommuting is a cause of longer retention; rather, it is likely that a stronger motivation to telecommute (for some undetermined reasons) is the cause of both outcomes.

Table 5-5: Telecommuting Duration (Months)

Table 5-6: Telecommuting Frequency of Stayers and Quitters

As mentioned earlier, questions concerning telecommuting duration and frequency were included in the exit interview. On average, the quitters reported using the telecenter for 7.4 months (standard deviation = 8.5, N = 18) although there is a wide range of responses (from none to three years). The average telecommuting frequency given by this group was about six days per month, or 27% (std. dev. = 4.8, N = 19). Only 3 (15.8%) of the quitters said they telecommuted from the center for more than two days per week. Although there is missing data from the exit interviews, the self-reported telecommuting duration and frequency are much higher than the averages calculated according to the attendance log data. This difference is either caused by misperception on the part of the participants, missing data in the attendance log, or for frequency, variation in the pattern of telecenter use.

Only nine respondents (37.5%) provided responses to the last part of the exit interview, regarding the ideal distribution of work time. At the time of the interview, these quitters worked primarily at the regular workplace (for 82.2% of their time) or at another location (13.3%, on average). Only one respondent telecommuted from home and none were utilizing another telecenter. However, according to the respondents, the average ideal distribution of their work time at the regular workplace, the telecommuting center, and home would be 50.9%, 38.7%, and 8.1%, respectively. So, many of the workers still preferred to use the telecenter but were not able to due to certain circumstances.

5.5 Summary of Retention Analysis

This chapter explored the attrition of telecommuters in the RABO Project. First, the characteristics of the respondents who quit were compared with the characteristics of those who stayed with the program. Second, the motivations of the quitters to leave the program were described. Third, the duration and frequency of telecommuting among both the stayers and quitters was investigated. In order to conduct this analysis, a particular sample of project participants was identified.

The responses to the before-wave surveys were used to find differences between 22 stayers and 24 quitters. According to employment type, administrative workers were more likely to quit (5 of the 7) and sales workers were more likely to remain in the program (5 of the 6). Surprisingly, quitters spent more time working remotely (more suited for telecommuting) than stayers. However, this result may explain why quitters preferred to spend more time working from home (23.5%) than stayers did (12.2%). Although job performance and satisfaction did not differ significantly between the two groups, certain work environment characteristics were significantly different. Quitters were more likely to worry about distractions at the telecenter and to consider the need for essential equipment to be important. Unfortunately, the findings from the survey data cannot point to the motivation for quitting, especially since the survey was administered prior to the experience with telecenter use.

The exit interview captured the reasons for quitting. The most important reason given was that respondents changed position within the company or their assigned tasks changed (25%). Thus, external corporate downsizing and reorganization were the most likely causes of quitting. Other important reasons include the supervisor requiring or encouraging the respondent to quit (21%) and leaving the company (13%). The reason for quitting is unknown for four individuals (17%).

The attendance log data showed that telecommuting duration and frequency varied widely within each group. The quitters who actually used the telecenter at least twice (16 of the 24 quitters) telecommuted an average of 7.2 months before leaving the program. For comparison, the stayers had telecommuted for an average of 10.7 months at the time of the final data entry. Quitters with some telecenter experience generally telecommute less often than stayers (24% vs. 30%, or 1¼ vs. 1½ days per week). This difference is reflected in the distribution of telecommuting frequency, where 9 quitters (56%) telecommuted less than one day per week while 13 stayers (54%) telecommuted one day per week or more. Perhaps quitters did not telecommute often enough to make the changes to their work schedule worthwhile.

Previous
Next
Table of Contents

Return to the Abstracts of Non-Refereed Publications page.
Return to the Telecommunications and Travel Main page.
Return to the Teleworking Reports page.