Comments to readings

Comments to readings

by Kate Chirikova -

Here we are, witnessing the application of sequence analysis once again in two other contexts. It appears to be a commonly used method in life-course studies!

Sabbath et al. 2015

1. Authors mention that initially there were 6,489 possible work-family-kids trajectories from which common prototypes were derived with sequence analysis, and then they were compiled in 7 clusters. In the context of the magnitude of data reduction, I'm wondering how many prototypes were derived from such a huge number of possible trajectories. I'm also interested what variables were fed to the clustering algorithm -- e.g. was it based on work-family-kids prototypes only or were there other factors that determined clustering of individuals? I'm also generally interested in this 2-step method of data reduction. Particularly, what is the conceptual difference between sequencing and clustering steps? Both seem to have a goal of data reduction.

2. I wonder how generalizable these findings are to more recent age cohorts. The study captures the trajectories of women born in 1935-1956. Given that the norms and culture change over time (e.g. individuals are having fewer children, couples tend to cohabitate rather than marry etc.), we might expect seeing different clusters of work-family-kids trajectories vs. the ones discovered in this study. Because some other trajectories are more common now, and others are more "extinct". This could potentially affect the discovered relationship between work-family-kids trajectories and health outcomes.

3. Due to various adjustments performed in the data analysis stage it's quite clear what contributes to increased mortality in single non-working mothers or single working mothers (e.g. health-related behaviors, SES, education). But it is less clear what drives the increased mortality in married non-working mothers. After adjustment for various factors, the magnitude of association between this trajectory and death remained large. Authors make the argument that work may be very beneficial for health. It would be interesting to investigate this issue further.


Harrati et al. 2019

1. Interesting that in this case the results of sequence analysis (resulting typology) were used as a dependent variable rather than an independent.

2. Last cluster (representing long-term disability leaves) comprises only 1.65% of the sample. I'm wondering if this raises any issues in terms of how confident we are of the results for this particular level of the "outcome". Is this small number of individuals in a cluster or imbalance between cluster sizes problematic?

3. Together with the authors, I'm also very interested why females have more disruptions and short-term disability leaves vs. males, given that potential differences in air pollution exposure or occupation do not explain these disparities. As the authors state, this could be due to selection -- i.e. women working in this manufacturing setting comprise a unique population whose distinctive features may have contributed to these disparities. Would be interesting to investigate this further.

Re: Comments to readings

by Charles Fleischmann -
Re: both of your #3's

Sabbath et al- I think drives/contributes might be the wrong words. Correct me if I am wrong, but all of the associations in this paper are correlations. I think the paper's main innovation is observational- i.e. to show that these work-family forms have variable and in some cases statistically significant mortality outcomes.


Harrati et al- I also find this observation really interesting. I can think of a few additional possibilities. 1) there is some other work place exposure risk, apart from particulate matter, that women could be disproportionally exposed to or more susceptible to compared to men. 2) there is a difference in cultural expectations between men and women with regards to seeking disability leave (i.e it could be seen as taboo/ not stoic for men to do so). 3) Psychological differences between men and women may result in women seeking PTO more often.

Comments to readings

by Carolyn Hughes -
Really great organization of your observations and questions, Kate!

Sabbath 1) I was wondering the same thing about the process for data reduction and the creation of the 7 trajectories. I'm also curious how divorce/disrupted marital status was handled. I tried to go back to see if I missed a mention of this (they mention marital status at first birth), though I didn't have time to fully scour the paper considering what we read about last week regarding the effect of disrupted marital status for women and poorer health (when combined with high fertility). This then relates to your point 2) what would this study look like in more recent cohorts and how would the clusters/trajectories look different?

Harrati 3) This is fascinating, and I have very little to add beyond what you and Charles mention. That said, I'm curious about the effect of childbearing. I saw the footnote that maternity leave and family leave were excluded from the analysis and therefore were unlikely to contribute to the differences by sex. This was really helpful to see, because an initial reaction for the difference in STD and higher disruptions among women would be around the effect of childbearing on work interruptions. I'm curious if the states included in the analysis had different rules/laws on how much leave women can take for pregnancy and maternity leave, especially for the dataset (and generally what those policies are). In some states, pregnancy leave is non-existent and maternity leave is very short; this can cause women to take leave in other forms that may not get attributed to pregnancy/maternity leave, or it may force them back into the workforce when really they medically may not be ready, and could leave them more susceptible to other morbidities. In other studies, there is an association between duration of maternity leave and other health effects, like depression. It would be interesting to see if the higher risk of STD among women varied by non-mothers and mothers. It would also be cool to see an analysis looking at patterns that were inclusive of pregnancy and family leave to see if there’s any interesting pattern in terms of proximity of medically indicated STD or LTD to maternity or family leave (though I’m sure it’s extremely complicated!).

Re: Comments to readings

by Richard Hu -
Harrati: Thinking about generalizability, this study looked at clusters in which many of the populations were around 80% white. I wonder how the results would compare to similar factories in different cities around the country with different racial make-ups. I also wonder what the rates of disabilities look like in different careers/industries, especially those that tend to see racial discrepancies in employment.

Sabbath: I'm glad that you point out the range of birth years, Kate, because it's so interesting to speculate on all the different historical factors and events that may have contributed to womens' classification in these clusters over their lifetimes. For instance, I think of wars that may have left them widowed, the Civil Rights movement, Women's Rights movements, improvements in health (smoking was obviously much bigger back then). I'm also curious to see the other side of the story: mortality rates in married and single fathers given what we know about sex differences in suicides.

Re: Comments to readings

by Leah Koenig -
Carolyn,

I had a similar question about how divorce and separations were handled. I think marital status was taken at each year's interview, which would have allowed them to use a binary measure of married/unmarried at each time. I'm not sure whether they assigned each woman to the marital status where they spent most of their time during the study period? I'd be curious if others have a better grasp on this than I do.

Re: Comments to readings

by Lufan Wang -
Sabbath 2) I am also very interested in the generalizability of these findings to more recent cohorts. As you mentioned, the norms and culture changed over time, the meaning of such work-family-kids trajectories may change as well. For example, a normative work-family trajectory for women born in 1930 was to marry in their early 20s and exit the labor force at the birth of their first child. While in 2020, among the families with children, more than half of them (59.8%) had both parents employed. Working mothers might be the new normative.

Harrati 3) I am also curious about the relationship between these trajectories and job characteristics, health and demographic characteristics in other industries? Do females still have more disruptions and short-term disability leave than males in industries other than manufacturing?

Re: Comments to readings

by Priz Espinosa Tamez -
Sabbath, 2015
1. (Kate's comment 1) I was also wondering about the methods they used to have this final prototypical work–family trajectories. They sound familiar and compatible with what we would think the trajectories could be but I wondered if inside some categories there could be a lot of variability, such as women intermittently going between working and not working. I was wondering if the methods explained in this document https://www.statmodel.com/download/JungWickramaLCGALGMM.pdf are related to the ones used in the paper, or if this is something completely different.
2. I was wondering the same thing as Carolyn and Kate, how they handled cohabitation and divorce. I think it would be especially important to be able to think of the interpretation of these results in other contexts (countries), where cohabitation might reflect very similar behaviors as marriage (instead of being single). I wonder if inside of "single" categories they also included widowed/divorced, since that would add another source of stress/strain in that population compared to single women.

Harrati, 2019
1. I also found it very interesting that female workers were more likely to experience at least one disability episode. I thought about pregnancy and maternity leave and wondered if even after excluding those periods, there could be residual events associated with maternity. I would also be concerned about selection bias, since women working in manufacturing might have a different distribution of other covariates, and therefore agree with their statement in the discussion that it is important to explore this issue further in this specific population of women. This difference could also be due to less barriers (or perceived acceptability) for women to express pain or seek medical attention, compared to men.