Model: | I | II | III | IV |
Panel A: all-cause fire mortality (any location) | ||||
Effect estimate (IRR, 95% CI) | 0.78 (0.75 to 0.81) | 0.95 (0.90 to 1.01) | 0.95 (0.90 to 1.01) | 0.95 (0.90 to 1.00) |
State effects | Yes | Yes | Yes | Yes |
Year effects | Yes | Yes | Yes | |
Time-varying covariates | Yes | Yes | ||
State-specific trends | Yes | |||
N states | 49 | 49 | 49 | 49 |
Average observations per state | 14 | 14 | 14 | 14 |
Valid state-years | 685 | 685 | 685 | 685 |
Panel B: all-cause fire mortality (residential fires) | ||||
Effect estimate (IRR, 95% CI) | 0.80 (0.76 to 0.84) | 0.94 (0.86 to 1.03) | 0.94 (0.86 to 1.02) | 0.94 (0.86 to 1.02) |
State effects | Yes | Yes | Yes | Yes |
Year effects | Yes | Yes | Yes | |
Time-varying covariates | Yes | Yes | ||
State-specific trends | Yes | |||
N states | 46 | 46 | 46 | 46 |
Average observations per state | 11.9 | 11.9 | 11.9 | 11.9 |
Valid state-years | 547 | 547 | 547 | 547 |
Poisson regression models with clustered-robust SEs were used to estimate the effects while accounting for correlated errors within states. The dependent variable is offset by population size (per state-year) in all models. The time-varying covariates included in models III–IV are per capita disposable income, cigarette sales, ethanol consumption and the per cent of the population aged ≥65 years.