Support #2102
open
Partnership
50%
Description
Hi, I'm doing some analysis on partnerships using the USOC data and I had a couple of questions I was hoping you could help with.
1) I've just done a very simple tabulation using the marstat_dv variable for different waves and age groups using the indall table, but I'm seeing some strange trends in the percentage of people living with/married to a partner for different waves.
For example, in wave A 32% of 16-29 year olds have a partner (weighted by a_psnenus_xw), but this drops to 16% for wave M (weighted by m_psnenui_xw). By contrast the ONS figures for 16-29 is pretty steady at around 30% over this time period. Is there an explanation for this? I've just used marstat_dv in (1,2) to denote whether a person is living with a partner, and the age_dv variable for age. Attached a file of what I'm seeing (both weighted and unweighted).
2) I'd like to derive a variable for the length of time (in years) that someone has been with their current partner for, and was wondering what the best way to get this info was? For wave A there is the Cohab table and the Marriage table. So it's possible to take the latest live episode of cohabitation/marriage for each person in a partnership and merge this info on. However, this information is not necessarily consistent between the people in a relationship, and there are a lot of people who do not have any information. One way to fill the gap appears to be by using xwavedat.lmarly_dv, which is the first marriage start year. This fills in a lot of the gaps, but the methodogy feels a little flimsy.
So I was wondering if there was a standard way to do this (i.e. calculate length of time a relationship has been on-going for)? My main focus at this stage is on getting the info for wave A, as I think if that is available subsequent waves should be easier to derive.
Thanks for your help
Michael
Files
Updated by Understanding Society User Support Team 8 months ago
- Status changed from New to Feedback
- % Done changed from 0 to 50
- Private changed from Yes to No
Hello Michael,
UKHLS has recently released a partnership history file, which includes details about the start and end dates of all partnerships, such as marriages, civil partnerships, and cohabitations (living together as a couple) reported by adult respondents in all Understanding Society (UKHLS) and British Household Panel Survey (BHPS) samples. This dataset could be helpful for your analysis. You can download the dataset from the following link SN8473. Understanding Society: Marital and Cohabitation Histories, 1991-2022
Regarding your initial query, I'll consult with the weighting team and get back to you as soon as possible. They may reach out to you directly for further clarification.
I hope this information is helpful.
Best wishes,
Roberto Cavazos
Understanding Society User Support Team
Updated by Olena Kaminska 8 months ago
Michael,
Thank you for your question and sending us the graph.
Our dataset is not directly comparable to ONS, as excludes recent immigrants. To make it comparable, I suggest you exclude recent immigrants from ONS to match UKHLS and directly compare it. Nevertheless, the difference you mention for youngest group is large, and in general we are looking into this issue and conducting research on its causes at the moment. In such comparisons it is also important to check meaning of questions, as much of such difference is often explained by the questionnaire design and question definitions.
I am not an expert on the topic but the fact that weighting makes the proportion of those in relationship lower is likely to be due to downweighting the oversampled ethnic minority and recent immigrant groups, and possibly Northern Ireland, Scotland and Wales. This would be the case if the above listed groups are more likely to be married in 16-29 than others.
Finally, for your specific analysis I would be very careful in determining whether you can use TSMs in your analysis as this would influence your results very much. Check if you have the history of marital information you need for all TSMs. If you do, you can use them and therefore you can use cross-sectional weight. If not, and you need previous waves to determine your variables you should only use longitudinal weight (and therefore only OSMs). Your results will be influenced very much in case of error, basically the proportions can differ two times from the population, especially for recent relationships.
Hope this helps,
Olena