Log-log panel with fixed effects:
[
\ln Q_{gct} = \alpha + \beta \ln P_{gct} + \gamma \ln Y_{gt} + \theta' X_{gt} + \mu_{gc} + \tau_t + \varepsilon_{gct}
]
| Concept | Derived From | Storage |
|---|---|---|
| Q (Quantity Proxy) | Real expenditures (Mean deflated by CPI or PCE) |
fact_cex.measure='exp_real' |
| P (Price) | CPI/PCE index matched to UCC | fact_price.value |
| Income | Demographic characteristic (CHARACTERISTIC_CODE → income bracket) |
dim_demo.income_class |
| Region/State | Geography mapping (region, division, state) | dim_geo.geo_id |
| Weights | CEX-provided sample weight (if using microdata) or population weights | fact_cex.measure='weight' |
Then elasticities:
Price elasticity = %ΔQ / %ΔP per UCC × SOM × time window
Income elasticity = %ΔQ / %ΔIncome for same segments