data&estimates revisions

[dariaple]

Note on SimPaths UK Updates (4 Nov 2025)
Updates to Initial Populations

1. Partnership status (dcpst)
   o Redefined as a binary dummy distinguishing partnered vs single, removing the separate category for previously partnered individuals.
2. Adult child flag (adultchildflag)
   o Revised definition: an individual is considered an adult child if they:
    have at least one parent in the household (i.e. non-missing parental age);
    are aged 18 or above;
    do not have a partner living in the same household;
    are at least 15 years younger than both their parents; and
    both parents are below state pension age in that year and not retired.
3. Leaving parental home indicator (dlftphm)
   o Reconstructed using the new adultchildflag definition to ensure internal consistency.
4. New variable: liwwh – total months employed since January 2007
   o Captures the cumulative duration of employment over the observation window, providing a consistent measure of accumulated work experience.
   o The timeline could, in principle, be extended backwards using retrospective information from the UKHLS Lifetime History modules (Waves 1 and 5). However, implementing this extension would be a major additional task.
5. Revised variable: dchpd (childbirth indicator)
   o Justin noted that the predicted probabilities of having a child were implausibly high.
   o My investigation revealed this stemmed from how the number of newborns had been computed. Previously, newborns were defined as children aged <1 at interview, not actual birth events.
    In some cases, a baby could be counted twice across waves.
    Adopted or stepchildren under 1 could also have been included.
    At the BHPS–UKHLS transition, this method overcounted legacy BHPS infants already born before the merge but still under one year old in Wave B.
   o The new approach instead uses the UKHLS newborn datasets, where each record directly represents a reported birth event since the previous interview.
    It also ensures that each newborn is counted only once and that the BHPS transition is handled properly.
   o Notably, in Cara’s original SAS version, all BHPS newborns were dropped, which likely underestimated the number of births.
   The new specification corrects this by retaining valid BHPS birth events after Wave B.
   o Importantly, a similar issue exists in the EU models, which also infer births from child age rather than directly recorded birth events.
    This approach will need to be adjusted, for example by accounting for the quarter of birth available in EU-SILC.
    However, in EU-SILC it may not be possible to exclude adopted children.

Updates to Processes

1. Fertility process
   o Re-estimated using the revised dchpd (having a child) and the updated binary dcpst (partnership status) variable.
2. Leaving parental home process
   o Re-estimated using the updated dlftphm variable.
3. Home ownership process
   o Re-estimated at the benefit unit head level.
   o The benefit unit head is defined as an adult aged 18 or above who:
    has the highest personal non-benefit income (ypnbihs_dv);
    or if equal, has the highest age;
    or, if equal, has the lowest idperson.