Multiple pweights in svyset, diff-in-diff, R-squared zero

[Oscar Saraya]

Hello everybody, I hope someone could help me.
I am using a panel data survey (MxFLS), household level, for 2 rounds: 2002 and 2005. Every round has moreless10 books of data, divided by topics. Each book has its own pweight related to the household folio. About my work: I want to run a Diff in Diff regression to assess the impact of a governmnent health insurance program on household savings. I constructed the savings variable using the difference between income and consumption of households. The problem relies here, since I used different data books to construct the savings variable, I don't know which pweight to use since every book has a different pweight. Is there a way to input different pweights in svyset or should I only decide for one pweight?

The problem is that I ran the DID regression and I got an R-squared=0.00 which is very disappointing. I was wondering this could be because of not weighting my data. My DIDs are the following (without svy :

regress savings after sp04 health_ins after*sp04 after*health_ins sp04*health_ins after*sp04*health_ins

where:

• savings: savings dependent variable=income-consumption
• after: dummy variable, 1 if 2005 and 0 if 2002 (after and before the program)
• health_ins: dummy variable indicating whether the household was eligible to the program, 1 if eligible. (also i tried using a continuous variable between 0 and 1)
• sp04: municipalities where the program was already available in 2004 (2004, between before=2002 and after=2005)

also I did a fourth differences regression using a dummy for positive savings since many households were really indebted:

regress savings after sp04 health_ins pos_savngs after*sp04 after*health_ins sp04*health_ins pos_savngs*after pos_savngs*sp04 pos_savngs*health_ins pos_savngs*after*sp04 pos_savngs*after*health_ins pos_savngs*sp04*health_ins after*sp04*health_ins after*sp04*health_ins*pos_savngs

where:

• pos_savngs: dummy variable, 1 if households had savings>0

I also run them using xtset and xtreg controlling for state fixed effects, but the results weren´t better. I dont know what else to do. Am I establishing the regressions in a wrong way? Does anyone reccomend me to add any new variable in the regression? a friend told me to add a variable more related to the savings dependent variable because all were dummies, for example minimum wage per municipality, states GDP or inflation in 2002 and 2005. But shouldn't this complicate more the regressions?

thank you very much!!