The GWPS weighting strategy was designed to reduce under coverage and nonresponse bias and maintain representativeness of the target population. Three types of household weights are provided:
A. Longitudinal household weights (weight_r1, weight_r2, weight_r3, weight_r4): Available for all rounds and vary by round. Intended for comparing the full interviewed sample in each round relative to the 2023 PECS baseline. These weights are not calibrated to updated population projections.
B. Cross-sectional population-calibrated household weights (weightx_r3, weightx_r4): Available for rounds 3 and 4. Intended for estimating population totals using updated population projections.
C. Panel household weights (weight_panel): Available for all rounds and constant across waves. Intended for analysis of the balanced panel of 777 households interviewed in all four rounds.
Weight construction involved: (i) defining the relevant analysis sample (round-specific or balanced panel); (ii) assigning initial weights based on the most recent available survey weight (or 2023 PECS weights for the balanced panel); (iii) modeling response propensity using pre-conflict poverty status, governorate, and location type to assess attrition bias; and (iv) calibrating weights through raking to match 2023 PECS household totals by governorate and other characteristics associated with nonresponse.
The resulting longitudinal and panel weights are appropriate for longitudinal analysis but are not intended for estimating current population totals. For rounds 3 and 4, an additional calibration step incorporated updated population projections for six age-sex population groups (males and females aged 0-14, 15-64, and 65+). The resulting weights (weightx_r3 and weightx_r4) support cross-sectional population estimates consistent with these updated projections.
