Understanding Society User Support

Hello,

I am currently doing analysis for my dissertation studying how the experience of intergenerational and intragenerational social mobility is associated with subjective wellbeing using Understanding Society data. I am using Stata to do the analysis, using waves 2-14 of the survey. When I use the following survey weight- indscus_lw, it really reduces the sample size. For intergenerational mobility for example the raw sample (using a dummy variable for mobility using father's occupation at 14 (panssec8_dv) and current occupation (jbnssec8_dv) has 9140 non-missing values without any weights (for men and women), which decreases to 53 non-missing values when the weight is applied. The sample for intragenerational mobility also decreases but it begins from a larger sample given that more people answered their occupation when first joining the survey vs their father's occupation at 14.

Please could you let me know if there is an appropriate remedy for this? I understand the importance of using survey weights but this is a significant reduction in the sample which is likely to also bias results. My dissertation is due on 3rd of September and any guidance would be really appreciated.

For reference some of my code is also included here (I have used missing dummy analysis, with fixed effects and pooled OLS):

areg scghq1_dv mobdum2 mobdum3 mobdum4 marrieddum2 marrieddum3 marrieddum4 hiqual_dummies2 hiqual_dummies3 hiqual_dummies4 chdum2 chdum3 chdum4 regiondum2 regiondum3 emp_dummies2 emp_dummies3 i.wave healthstatdum2 healthstatdum3 healthstatdum4 healthstatdum5 healthstatdum6 agedum2 agedum3 agedum4 agedum5 agedum6 agedum36 ///
if sex == 1 ///
[pweight= indscus_lw], absorb(pidp)

reg scghq1_dv mobdum2 mobdum3 mobdum4 marrieddum2 marrieddum3 marrieddum4 racedum2 racedum3 racedum4 hiqual_dummies2 hiqual_dummies3 hiqual_dummies4 chdum2 chdum3 chdum4 regiondum2 regiondum3 emp_dummies2 emp_dummies3 i.wave healthstatdum2 healthstatdum3 healthstatdum4 healthstatdum5 healthstatdum6 agedum2 agedum3 agedum4 agedum5 agedum6 agedum36 if sex ==2 [pweight= indscus_lw], vce(cluster pidp)