Skip to content | Skip to institutional links

Common menu bar links

Institutional links

Current Subject
- Environmental & Workplace Health
- Reports & Publications
Explore...
Proactive Disclosure

Proactive Disclosure

Environmental and Workplace Health

Print |

Need Larger Text? |

Reducing Work-Life Conflict: What Works? What Doesn't?

Table of Contents

2.3 Statistical Analysis

Data were summarized and analyzed using frequencies and analysis of variance (ANOVA). To assist the reader, key statistical terms are defined briefly in Box Five. Greater details on the use of ANOVA can be found in Duxbury and Higgins, 2005 (see Appendix A).

Box Five: Terms Used in Discussing Statistical Analysis

Analysis of Variance (ANOVA): A technique that can be used to determine if statistically significant differences occur in means between two or more groups.

F-test: Statistic used to evaluate whether the means of groups are statistically different. If two or more means are unequal, we say we have a significant ANOVA.

p-value: Level of statistical significance. Traditionally, p-values of 0.05 or less are considered to be statistically significant.

Dependent variable: The outcome variable of the research (i.e. role overload, work-to-family interference, family-to-work interference, caregiver strain).

Independent variable: A variable that is expected to influence the dependent variable (i.e. coping strategy, work arrangement, management support).

R² (R-squared): The amount of variance in the dependent variable that is explained by the independent variables. This statistic is used to determine the strength of the association between dependent and independent variables and ranges from 0 to 1. The closer R² is to 1, the stronger the association. Researchers often multiply the R² value by 100 and talk about the percent of the variation in the dependent variable (in this case work-life conflict) explained by the independent variable (the various coping strategies examined).

Bonferroni adjustment: This is a more conservative approach to hypothesis testing, which is done to control for what researchers call a type 1 error (i.e. the error of rejecting a null hypothesis when it is true). It is a simple procedure where the p-value of 0.05 (the common rejection level) is divided by the number of dependent variables included in the analysis to get a more conservative rejection level.

2.3.1 Frequencies

As a first step in all analysis, we calculated either (1) the percent of the total sample that used a particular coping strategy, or (2) the availability within the work environment of a particular potential moderator such as perceived flexibility. The methods used to operationalize "use" and "availability" in this analysis are summarized in Box Six.

Box Six: Measurement of Use/Availability of the Different Coping Strategies

Work arrangement: Use was operationalized as follows:

Percent who indicated that they worked a regular work arrangement, flextime, compressed work week (CWW), tele-work, part time (see Chapter Three for operational definitions)
Percent who indicated that they worked a shift schedule (answered yes to working either a fixed shift or a rotating shift)
Percent who performed guerrilla tele-work (spent more than an hour each week working from home during regular work hours)

Perceived flexibility: Availability was operationalized into three categories:

Low flexibility (mean of less than 2.5)
Moderate flexibility (mean of 2.5 to 3.5)
High flexibility (mean of greater than 3.5)

Supportive manager: Availability was operationalized into three categories:

non- supportive manager (mean of less than 2.5)
mixed manager (mean of 2.5 to 3.5)
supportive manager (mean of greater than 3.5)

Non-supportive manager: Availability was operationalized into three categories:

Non-supportive manager (mean of greater than 3.5)
Mixed manager (mean of 2.5 to 3.5)
Not non-supportive manager (mean of less than 2.5)

Supportive benefits: Use was measured as follows:

Used (on survey indicated yes that this benefit was available in their organization, and yes that they used it)
Did not use (on survey either indicated that no, this benefit was not available in their organization, or indicated that yes, the benefit was available but no they did not use it)

Individual coping strategies: Use was measured as follows:

Rarely (combination of never and monthly)
Weekly (weekly)
Daily (combination of several times a week and daily)

Delayed starting family/Had fewer children/Off-shift work hours:

Use defined as percent who indicated yes to this item

Family coping strategies: Use was measured as follows:

Rarely (combination of disagree strongly and disagree that they use strategy)
Sometimes (indicted that they used occasionally)
Often (combination of agree and agree strongly that they use this strategy)

To examine the impact of gender and job type on the use of the different coping techniques, we divided the sample into four groups (male managers/professionals, male other, female managers/professionals,¹¹ female other) and then calculated frequencies for each of these groups using the method outlined above.

To examine the impact of gender and dependent care status on the use of the different coping strategies, we divided the sample into four groups (males with dependent care, males without dependent care, females with dependent care, females without dependent care) and again calculated frequencies by group.

Chi-squared analysis was performed to determine which groups were more or less likely to use the different coping strategies/have access to the different supports within their work environment. Given the very large sample size, between-group differences of 1% or more were statistically significant. In this report, we limit our discussion to differences that were both significant and substantive (between-group difference of 5% or more).

Top of Page

2.3.2 ANOVA

We used a statistical technique called ANOVA to determine how effective the various coping strategies/environmental supports examined in this report are at reducing work-life conflict. A summary of the various ANOVAs calculated in these analyses is given in Box Seven.

Empirically, the previous reports in this series determined that the four dependent variables included in this analysis, as well as some of the independent variables (i.e. perceived flexibility, decision to have children), are significantly associated with gender, job type and dependent care status (see Duxbury & Higgins, 2001, 2003; Higgins & Duxbury, 2002 for a review of the relevant theory and findings). To minimize the impact of uncontrolled confounds on our findings, we did each of the ANOVAs in Box Seven twice. In the first set of analysis, we controlled for gender and job type by including an independent variable in the ANOVA (gender by job type), which was operationalized as follows:

Male managers/professionals
Male other
Female managers/professionals
Female other

In the second set of analysis, we controlled for gender and dependent care status by including an independent variable in the ANOVA (gender by dependent care), which was operationalized as follows:

Male - no dependents
Male - dependents
Female - no dependents
Female - dependents

This data analysis strategy will give us a greater appreciation of how gender, job type and dependent care status are related to the various moderators of work-life conflict explored in this study. It will also allow us to target our recommendations about how employees in these different demographic groups can best cope with the four different forms of work-life conflict.

Box Seven: ANOVAs

Overall impact of alternative work arrangements examined using work arrangement as the independent variable and each form of work-life conflict as the dependent variable (i.e. 4 ANOVAs)
Overall impact of perceived flexibility examined using total perceived flexibility measure as the independent variable and each form of work-life conflict as the dependent variable (i.e. 4 ANOVAs)
Impact of the different types of flexibility examined individually using each of the 10 items making up the measure as the independent variable and each form of work-life conflict as the dependent variable (i.e. 40 ANOVAs)
Overall impact of supportive manager determined using total supportive manager measure as the independent variable and each form of work-life conflict as the dependent variable (i.e. 4 ANOVAs)
Overall impact of non-supportive manager determined using total non-supportive manager measure as the independent variable and each form of work-life conflict as the dependent variable (i.e. 4 ANOVAs)
Impact of the 16 different supportive and non-supportive behaviours examined individually using each of the behaviours making up the measure as the independent variable and each form of work-life conflict as the dependent variable (i.e. 64 ANOVAs)
Impact of the different organizational supports examined using each of the 13 supports making up the measure as the independent variable and each form of work-life conflict as the dependent variable (i.e. 52 ANOVAs)

ANOVAs Done to Look at Effectiveness of Individual Coping Strategies

Impact of personal coping strategies examined using each of the 13 items making up the measure as the independent variable and each form of work-life conflict as the dependent variable (i.e. 52 ANOVAs)
Impact of decision to have fewer children examined using this strategy as the independent variable and each form of work-life conflict as the dependent variable (i.e. 4 ANOVAs)
Impact of decision to delay or not have children examined using this strategy as the independent variable and each form of work-life conflict as the dependent variable (i.e. 4 ANOVAs)
Impact of off-shifting examined using this strategy as the independent variable and each form of work-life conflict as the dependent variable (i.e. 4 ANOVAs)

ANOVAs Done to Look at Effectiveness of Family Coping Strategies

Impact of family coping strategies examined using each of the 18 items making up the measure as the independent variable and each form of work-life conflict as the dependent variable (i.e. 72 ANOVAs)

There are three statistics of interest in the ANOVA run:

The interaction term: If this is significant, it means that the ability of the moderator to reduce work-life conflict varied depending on either the gender by job type or gender by dependent care of the respondent.
The main effect for the moderator: If this is significant, it means that the level of this form of work-life conflict (i.e. overload) varies depending on the level of the moderator (i.e. supportive manager, mixed manager, non-supportive manager). In other words, the moderator has a significant impact on work-life conflict.
The main effect for either gender by job type or for gender by dependent care: If this is significant, it means that the level of this form of work-life conflict (i.e. overload) varies depending on the level of the gender by job type or gender by dependent care.

In each of these cases (i.e. interaction term, main effect) there are two statistics of interest: the significance level of the F statistic, and the R² (i.e. the amount of the variation in work-life conflict explained by the independent variables). The following conventions were employed during data analysis:

We used an R² of 4% or greater (i.e. .04) as an arbitrary determination of substantiveness (i.e. the moderator is linked to work-life conflict in a meaningful way and/or worthy of note).
To compensate for the large number of tests being performed, significance levels for the F test were set at p = .01 or less.

While a complete set of findings is given in the Appendices, only those results that met both these criteria are discussed in detail.

ANOVA analysis was done as follows. The interaction term was examined first. If the interaction was significant and substantive, then we examined the mean work-life score reported for the different demographic groups at low, medium and high use (or low, medium and high availability) of the moderator. If the interaction term was not significant, we looked at the main effect for the moderator to determine the relationship between the moderator and work-life conflict. If the main effect for the moderator was significant and substantive, then we looked at the mean work-life score for the three levels of the moderator variable. If the main effect for the moderator was not significant and substantive, we concluded that this coping strategy did not reduce/increase work-life conflict. Finally, it should be noted that we do not include an in-depth analysis of the gender by job type or gender by dependent care main effects. Such analysis can be found in Duxbury and Higgins, 2003.

¹¹ In this report, those in the "other" job group include those who work in technical, administrative, clerical and production positions. Those in the manager/professional group self identified themselves as working in these types of positions. A complete discussion of the operationalization of job type in this study can be found in Higgins and Duxbury, 2002.