Working Paper: CEPR ID: DP17079
Authors: Xulia González; Saul Lach; Daniel Miles
Abstract: We revisit the bias in the estimation of production functions with firm-level data due to the lack of physical quantities on output and inputs. We show that constructing firm-specific prices from available data on firm-specific price changes, and using them to deflate revenues and expenditures, introduces a measurement error into the empirical production function. This error reflects the unobserved base year prices used in the construction of the firm-specific prices. The usual practice of ignoring them generates an omitted variable bias (OVB). Monte Carlo simulations suggest that this bias can be significant. Because of the OVB, the estimates are sensitive to the choice of base year. The OVB disappears in our simulations when firm-specific fixed effects are incorporated into the estimation of the production function.
Keywords: production function estimation; unobserved prices; omitted variable bias
JEL Codes: D22; D24; L10; L60
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Normalization of unobserved base year prices (P22) | Omitted variable bias in estimation of production functions (D24) |
| Omitted variable bias in estimation of production functions (D24) | Accuracy of production function estimates (C51) |
| Incorporating firm-specific fixed effects (C23) | Elimination of omitted variable bias (C20) |
| Normalization of base year prices (P22) | Bias in parameter estimates (C51) |
| Omitted variable bias (C20) | Residual productivity measures bias (D24) |
| Residual productivity measures bias (D24) | Estimates of aggregate productivity growth (O47) |
| Omitted variable bias (C20) | Correlation between input demands and omitted base year prices (E31) |