The Consequences of Work–Family Enrichment in Families on the Behaviour of Children
Round 1
Reviewer 1 Report
Overall, I find this work theoretically sound. I also enthusiastically embrace the idea parental WFE can improve kids’ lives.
Nonetheless, I have 2 major reservations regarding this manuscript.
- The selection criteria are not given. The authors report they selected data from 4012 parents, but don’t say how. What fraction of parents of the 10,000 children are represented in the data? What fraction of the children? Who is missing?
- Second, I have a question about the importance even if the results are taken at face value.
The final results are shown in Figure 2 which certainly attests to work-family enrichment contributing to good child behavior via parenting and parent well-being. But if I understood lines 342-344 right, there is no evidence that child behavior is significantly enriched by WFE. “To accurately test the mediating paths of parenting and parental well-being, we fitted the structural model in three steps. First, we considered only the direct path of WFE on children’s behaviour (along with the control variables), which was not significant (β = .284, p = .417).” Doesn’t that mean that despite the statistically significant positive pathways shown in the final model that the net effect on kids is for practical purposes zero? I admit I don’t understand how that might be so (strength of the relationship, effects of control variables, etc.), but it does seem to raise an important question about how much these positive pathways matter in real life.
Similarly, “Moreover, the intervening variables of parental well
395 being and parenting performance fully mediate the relationship between work-family enrichment
396 and children’s behaviour, as this direct relationship is not present.”
But what does it mean to fully mediate a relationship that isn’t there in the first place? If a significant direct effect disappeared when introducing parental well-being and parenting performance then sure, the effect is fully mediated. But how can the change from insignificant to insignificant be described as mediating?
Other comments (less critical).
In the lit review, we learn “Role accumulation theory (Sieber 1974) considers several positive outcomes: (1) role privileges (e.g., rights), (2) buffering of status security, (3) resources from status enhancement and role performance, and (4) personality enrichment”. I think the authors should explicitly say that these aren’t measured in the study (differentiate between a plausible reason for the relationship between WFE and parental well-being and actually measuring the mechanism that connect the two).
There is no questioning of the validity of the Work-Family Gains scale of Marshall and Barnett. We are told that “The composite reliability of this scale was 0.86” which only means that the self-reports of work-family gains were positively and strongly correlated with one another. But is subjective assessment of work-family gains a valid way to measure work-family gains? In the absence of an answer to that question, I think the strongest conclusions that can be drawn are that people who believe themselves to be experiencing work-family gains have children that are better off. The authors say “The results confirm that children’s behaviour is influenced through parental well-being by parents’ experience of work-family enrichment” which is technically correct but, I believe, misleading. It seems to imply that when parents experience WFE their kids have better behavior rather than that when parents report experiencing WFE…” The authors understand the difference—they say in the conclusions “A second limitation in the current study 408 is that the measurements of interest were subjective in nature by self-reporting (work-family 409 enrichment....” I believe this should be mentioned in the methods section to give the reader a heads-up on the issue, and that the phrase “experience enrichment” should not be used because it is ambiguous between experiencing some kind of intervention that increases WFE and reporting more WFE. (The phrase can be easily misinterpreted to imply more causality than is warranted; parents may experience more WFE when their kids are better behaved with the key mechanism being perceptions rather than any change in how much work and family enrich each other).
This issue of validity could could be used to criticize measures of parenting and parent well-being (or at least to suggest the paper be improved by citing evidence of validity for these measures), but because work-family enrichment is the central theoretical construct that this paper tests the implications of, it deserves even more rigorous scrutiny than the mediators do.
Author Response
Dear reviewers,
We thank you for your observations and helpful suggestions. We hope that the reviewed manuscript complies with your suggestions.
Best regards,
The authors
REVIEWER 1
I have 2 major reservations regarding this manuscript.
- The selection criteria are not given. The authors report they selected data from 4012 parents, but don’t say how. What fraction of parents of the 10,000 children are represented in the data? What fraction of the children? Who is missing?
We have included additional details on the selection criteria we used in the study. Partly we lost datalines due to attrition (as we use wave 7) and partly item non response was responsible for a loss of data. All step are extensively documented now with some additional descriptives on the final analytical sample. We also included a warning for the reader that the final sample is not representative for all Australian families due to the focus of the study.
- Second, I have a question about the importance even if the results are taken at face value.
The final results are shown in Figure 2 which certainly attests to work-family enrichment contributing to good child behavior via parenting and parent well-being. But if I understood lines 342-344 right, there is no evidence that child behavior is significantly enriched by WFE. “To accurately test the mediating paths of parenting and parental well-being, we fitted the structural model in three steps. First, we considered only the direct path of WFE on children’s behaviour (along with the control variables), which was not significant (β = .284, p = .417).” Doesn’t that mean that despite the statistically significant positive pathways shown in the final model that the net effect on kids is for practical purposes zero? I admit I don’t understand how that might be so (strength of the relationship, effects of control variables, etc.), but it does seem to raise an important question about how much these positive pathways matter in real life.
Similarly, “Moreover, the intervening variables of parental wellbeing and parenting performance fully mediate the relationship between work-family enrichment and children’s behaviour, as this direct relationship is not present.”
But what does it mean to fully mediate a relationship that isn’t there in the first place? If a significant direct effect disappeared when introducing parental well-being and parenting performance then sure, the effect is fully mediated. But how can the change from insignificant to insignificant be described as mediating?
We have to admit that the results reported were incorrect. When reviewing our output, we have reported the wrong parameters in the manuscript suggesting that the initial path was not significant while it was highly significant. Significance in the second model was still found but to a lesser degree. The conclusion that our final model fully mediated the direct path was correct, but we now have documented this conclusion with the right parameters and tests. We apologize for this error. All results were double checked and we do have the correct story reported now.
Model 1
|
Standardized Effects in Linear Equations |
||||||
|
Variable |
Predictor |
Parameter |
Estimate |
Standard |
t Value |
Pr > |t| |
|
F04 |
F11 |
PF04F11 |
0.16872 |
0.03151 |
5.3552 |
<.0001 |
Model 2
|
Standardized Effects in Linear Equations |
||||||
|
Variable |
Predictor |
Parameter |
Estimate |
Standard |
t Value |
Pr > |t| |
|
F04 |
F11 |
PF04F11 |
0.06736 |
0.02745 |
2.4542 |
0.0141 |
Other comments (less critical).
In the lit review, we learn “Role accumulation theory (Sieber 1974) considers several positive outcomes: (1) role privileges (e.g., rights), (2) buffering of status security, (3) resources from status enhancement and role performance, and (4) personality enrichment”. I think the authors should explicitly say that these aren’t measured in the study (differentiate between a plausible reason for the relationship between WFE and parental well-being and actually measuring the mechanism that connect the two).
It is correct that we do not use role theory as an empirical construct but that is not what we aimed to communicate. At the beginning of the paragraph, we state that WFE “builds upon functional role theory”. We give the reader some background on the origins of the concept without actually claiming that we will use role theory in our models. In order to avoid too much emphasis on the role theory, we now have added some wider explanation on the WFE concept. We hope that the erroneous impression that role theory would be used in a direct way, is now countered.
There is no questioning of the validity of the Work-Family Gains scale of Marshall and Barnett. We are told that “The composite reliability of this scale was 0.86” which only means that the self-reports of work-family gains were positively and strongly correlated with one another. But is subjective assessment of work-family gains a valid way to measure work-family gains? In the absence of an answer to that question, I think the strongest conclusions that can be drawn are that people who believe themselves to be experiencing work-family gains have children that are better off. The authors say “The results confirm that children’s behaviour is influenced through parental well-being by parents’ experience of work-family enrichment” which is technically correct but, I believe, misleading. It seems to imply that when parents experience WFE their kids have better behavior rather than that when parents report experiencing WFE…” The authors understand the difference—they say in the conclusions “A second limitation in the current study 408 is that the measurements of interest were subjective in nature by self-reporting (work-family 409 enrichment....”
I believe this should be mentioned in the methods section to give the reader a heads-up on the issue, and that the phrase “experience enrichment” should not be used because it is ambiguous between experiencing some kind of intervention that increases WFE and reporting more WFE. (The phrase can be easily misinterpreted to imply more causality than is warranted; parents may experience more WFE when their kids are better behaved with the key mechanism being perceptions rather than any change in how much work and family enrich each other).
This issue of validity could be used to criticize measures of parenting and parent well-being (or at least to suggest the paper be improved by citing evidence of validity for these measures), but because work-family enrichment is the central theoretical construct that this paper tests the implications of, it deserves even more rigorous scrutiny than the mediators do.
We use scales in our research knowing the strengths and limitations of these techniques in social sciences. All survey research is essentially a subjective measurement from people’s experience. That is exactly what scales are developed to do: measure a latent trait in people’s lives unmeasurable by behavior or hard facts. Reliability measures are a well-established way of showing the internal consistency of a scale. As our study is not a validation study of the WFE scale, we cannot go into depth in all psychometric qualities of the scale. We use the scale to test our hypotheses. However, we did add more information of validation studies done with this scale so that the reader who likes to know more about the technicalities of the scale is pointed in the right direction.
The question whether or not to use “experiencing WFE” is a difficult one as the scale items do refer to the experience of WFE. It is also explicitly included in the definition of the concept “the extent to which experiences in one role improve the quality of life in the other role”. In that respect, removing all references to “experience” would be a conceptual error. Off course we do not want to give any impression on causality but we hoped that this was made clear in our limitations at the end of the article. We hope you can agree with our decision to opt for conceptual purity relying on the reader not to see “causal language” when we use “experience”. Additionally, we more clearly refer to our found results as being relationships in the discussion.
Reviewer 2 Report
The paper deals with a very interesting topic, especially because the positive spillover between family and work and its connection with children well-being and behavior is a very neglected one.
I think the state of the art and the methodology are well illustrated, the results are very interesting and contribute to the scientific debate, despite the well-documented limitations described in the concluding section.
I have only one suggestion: I would recommend the Authors to move the hypothesis of their work in the methodological part. I appreciated the effort to link the hypothesis with the existing literature, but in the present form the presentation, in my view, is confusing for the reader. Moreover, at page 5 (197-198) the Authors talk about three hypothesis, but only two are explicitly mentioned and described in the previous section, as well as in the Discussion section. I would suggest to add a (H3) near to the third hypothesis as they did for the other two and to discuss it in a more articulated way.
Author Response
Dear reviewers,
We thank you for your observations and helpful suggestions. We hope that the reviewed manuscript complies with your suggestions.
Best regards,
The authors
REVIEWER 2
I think the state of the art and the methodology are well illustrated, the results are very interesting and contribute to the scientific debate, despite the well-documented limitations described in the concluding section.
I have only one suggestion: I would recommend the Authors to move the hypothesis of their work in the methodological part. I appreciated the effort to link the hypothesis with the existing literature, but in the present form the presentation, in my view, is confusing for the reader. Moreover, at page 5 (197-198) the Authors talk about three hypothesis, but only two are explicitly mentioned and described in the previous section, as well as in the Discussion section. I would suggest to add a (H3) near to the third hypothesis as they did for the other two and to discuss it in a more articulated way.
We agree with the reviewer that there are many ways to present hypotheses in an empirical paper. We do indeed have three hypotheses of which the third was not written out in a forma way. We have corrected the text in that respect. We have added the third hypothesis in the text and presented it more clearly as was suggested. We also refer to our hypotheses more clearly in both the results and the discussion section.
We did not move the hypotheses to the methods section as we believe the integration with the literature makes it easier for the reader to follow our story.
Round 2
Reviewer 1 Report
Sample selection:
203 A subsample was selected to perform the current research. As the Family Enrichment Scale is 204 adopted in each wave, we could have performed a longitudinal (fixed or random) regression. 205 However, this choice would have been hampered by the dynamic nature of families.
Not really: a fixed effects regression would have looked at the effects of change (dynamics) within families holding constant things constant. Unmeasured dynamic things are a problem whether or not a longitudinal analysis were performed. I think it was a mistake to forfeit potential benefits of the longitudinal data.
As we aim to 206 look at the relationship between WFE and child’s behaviour, we decided to use only the latest 207 available wave (wave 7) in this study. Family dynamics are taken into account as we used the panel 208 data to reconstruct family decomposition (divorce and separation) and recomposition (repartnering) 209 into a variable determining whether or not the parent currently has a partner.
This description is not of taking family dynamics into account. It sounds much more like parents in the household were measured in an earlier wave and the authors constructed that variable for the most recent wave using family composition changes in between.
Wave 7 had a response 210 rate of 78.3 % for cohort B and 73.9 % for cohort K resulting in 3381 and 3089 interviews respectively. 211 As our independent variable, the WEF scale, is crucial to the study, non-response on the scale further 212 decreased our analytical sample to 2936 (B-cohort) and 2704 (K-cohort) respondents (5640 in total). A 213 final reduction in the data were the datalines where we did not have information on the child’s 214 behaviour or where the parental variables were missing for both parents.
Since the authors did have single parent families in the sample, this sounds like dual parent children were omitted if one parent's data were missing and single parent children were included with the same amount of data.
The final analytical sample 215 consisted of 4163 respondents. There were slightly more mother (2406, 57.8%) than father (1757, 216 42.2%). Most parents (91%) were partnered (whether or not after a break-up with the biological parent 217 of the child). On average, included families had 2.49 children with a mean age of the youngest child 218 of 8.4 years (SD = 2.4) for those in the B cohort and 12.0 years (SD = 2.9) for the sample members in 219 the K cohort. The data from both cohorts were used, accounting for the age differences by controlling 220 for it in the analyses. Both partner 1 and partner 2 were used in the analysis. When only one parent 221 had answered the WFE scale, they were adopted as analytical unit. When both parents had answered 222 the scale, the analytical unit was chosen at random and only one of these parents was used in the 223 analysis.
This seems a good way of including both dual and single parent families.
The authors said in their response that they had emphasized that the sample was not representative, but this is how: 224 Even though the LSAC aims to be representative for the cohorts under study, we need to 225 underline that the study is not representative for the whole population of Australian (working) 226 parents and their children. Because we use data from cohort B and K, children’s ages are limited to 227 12/13 and 16/17 which implies that parents are older. A consequence is that most families have a 228 complete fertility history at that age which also explains why the families under study are, on 229 average, larger. Two issues here:
1) there is no attention to non-representativeness in terms of panel attrition or missing data, and
2) I'm not buying that the families are larger on average because fertility is more likely to be complete. The period TFR in Australia hasn't been over 1.9 since 1990. One of the reasons this sample has higher average family size is that no zero child families are included.
The model seems to still have an interpretation error:
"First, we considered only the direct path of WFE on children’s 366 behaviour (along with the control variables), which was highly significant (β = .0169, p = .000). In the 367 second model, we added the direct path from parental well-being and parenting to children’s 368 behaviour. This addition maintained the significant direct path from the first step (β = .067, p = .01). 369 The effect decreased and significance dropped substantially." Decreased? The coefficient was more the 4x the size.
Plus it seems a little odd to conclude that the relationship was fully explained by the mediating variables when the effect size in the final model was greater than when estimated without the mediators. Dropping from statistical significance is important, but it isn't everything.
Having the data to do a study is not a contribution: it should not be listed as #3.
Overall, the presentation could be less wordy.
E.g., Generally, it appears that the experience of work-family enrichment 161 has positive relations with parent-child interactions, family role performance, and family satisfaction 162 through the increased availability of resources, positive affect, and feelings of commitment (Cooklin 163 et al., 2015, 2016; Vieira et al., 2016). Cooklin et al. (2016) focused only on fathers in their study and 164 reported enhanced effective parenting techniques in terms of more warm affective parenting and 165 buffers against hostile parenting when fathers were interacting their children. Another study by 166 Cooklin et al. (2015) investigated the same associations, but for mothers. Their findings suggest a 167 clear positive association between work-family enrichment and warm and consistent parenting.
Consider: Generally, it appears that the experience of work-family enrichment 161 has positive relations with parent-child interactions, family role performance, and family satisfaction 162 through the increased availability of resources, positive affect, and feelings of commitment (Cooklin 163 et al., 2015, 2016; Vieira et al., 2016). Cooklin et al. (2015, 2016) reported enhanced effective parenting techniques in terms of more warm affective parenting and buffers against hostile parenting when interacting children among both mothers and fathers.
E.g., Two alternatives have been suggested in the literature: 234 higher-order factor models and the bifactor model (Mulaik & Quartetti, 1997). Given that the bifactor 235 model assumes orthogonal associations between factors (an assumption that is too strict for our 236 conceptual model), we opted to use a second-order hierarchical SEM model (Chen et al., 2005).
Consider: Of the models suggested in the literature (Mulaik & Quartetti, 1997; Chen et al., 2005), we opted for a second-order hierarchical SEM model because it does not assume an orthogonal associations between factors, as we do not.
Other examples of wordiness are the use of phrases like: "The interaction of family members within their contexts being acknowledged, the relationship 35 between the experience of work-family enrichment as a parent and the behaviour of a child is not 36 direct." And "Therefore, our third, overarching hypothesis goes as follows: both parental well187 being and parenting fully mediate the relationship between work-family enrichment and the behaviour of the 188 child (H3).
Comments for author File: 
Comments.pdf
Author Response
See document attached.
Author Response File: Author Response.pdf
