Did Real-Estate Collateral Drive Corporate Investment??
(This short summary omits a lot of detail in the paper.)
| Key Regression Spec Aspect | Coefficient | Interpretation |
|---|---|---|
| capex/lagppe ← redol/lagppe + FE | +0.074 | Result in Chaney-Sraer-Thesmar (AER 2012). |
| capex/lagppe ← 1/lagppe + FE | +0.127 | Absurd variable. (capex trends up with lagppe.) |
| 1.0/lagppe ← redol/lagppe + FE | +0.204 | Absurd variable. (redol trends up with lagppe.) |
| Δ(capex/lagppe) ← Δ(redol/lagppe) | +0.185 | Year-by-Year changes in both num and denom. |
| (Δcapex)/lagppe ← (Δredol)/lagppe | -0.102 | Numerator-only association is not positive! |
where
-
capex = capital expenditures
-
redol = real-estate dollar vallue
-
lagppe = lagged property, plant, and equipment
-
FE = fixed effects (and more).
-
Δ is the year-to-year change in the variable.
Ergo: Most CST regressions suffered from spurious mechanical correlation due to shared year-to-year (steady) increases in ppe in the dependent and independent variables. (A few regressions suffered from spurious non-mechanical trend-induced correlations for firms with vs firms without real-estate.)