Table 2

Effects of fire safe cigarette laws on the incidence of cigarette-caused fires

Model:IIIIIIIVV
Panel A: cigarette-caused fires
Effect estimate (IRR, 95% CI)0.83 (0.71 to 0.96)1.01 (0.83 to 1.23)1.02 (0.85 to 1.22)1.07 (0.87 to 1.31)1.00 (0.91 to 1.11)
State effectsYesYesYesYesYes
Year effectsYesYesYesYes
Time-varying covariatesYesYesYes
State-specific trendsYesYes
DDDYes
N states5151515151
Average observations per state15.415.415.415.430.9
Valid state-years7877877877871574
Panel B: cigarette-caused fire fatalities
Effect estimate (IRR, 95% CI)0.66 (0.55 to 0.79)0.70 (0.55 to 0.88)0.74 (0.60 to 0.92)0.82 (0.66 to 1.02)0.92 (0.74 to 1.14)
State effectsYesYesYesYesYes
Year effectsYesYesYesYes
Time-varying covariatesYesYesYes
State-specific trendsYesYes
DDDYes
N states5151515151
Average observations per state14.914.914.914.929.8
Valid state-years7597597597591518

  • 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 DDD model (column V) includes two observations per state-year; one for cigarette-related events and one for (cause-determined) events unrelated to smoking materials. This model includes main effects for all states, years, time-varying covariates and state-specific trends, as well as their interactions with event type (a dummy for cigarette-caused event or event unrelated to smoking). The time-varying covariates included in models III–V are per capita disposable income, cigarette sales, ethanol consumption and the per cent of the population aged ≥65 years, as well as the proportion of events with unknown ignition source.

  • DDD, difference-in-difference-in-differences model.