Corrigendum: The role of tobacco taxes in starting and quitting smoking: duration analysis of British data

Authors


Andrew M. Jones, Department of Economics and Related Studies, University of York, Heslington, York, YO10 5DD, UK.
E-mail: amj1@york.ac.uk

We estimated hazard rates for starting and quitting smoking in which the tax rate on cigarettes was measured as a time-varying covariate. Since completing the analysis an error has come to light in the procedure that was used to map the real tax rate on cigarettes to the year at risk for the analysis of the hazard of starting smoking. This error affects mainly the discussion of the starting models in Section 5.1. It does not affect any of the analysis of the hazard of quitting smoking that is reported in Section 5.2.

The effect of correcting this error is to change the estimates of the tax elasticity of starting, as reported in the summary (page 517), Section 1 (page 518) and Section 5.1 (Table 2, pages 533 and 534 and Table 3). In the paper it was reported that the tax elasticities were relatively small but that they were statistically significant (page 533). The corrected results show that the estimates are even closer to zero in magnitude and are not statistically significant: the reported point estimate of 0.16 for men is replaced by 0.006 and the estimate of 0.08 for women is replaced by −0.028.

Table 2.  Split population log-logistic–probit results for starting†
ParameterResults for menResults for women
DurationParticipationDurationParticipation
  1. t-statistics are given in parentheses.

NO_QUALIFICATION0.04200.3748−0.03850.1552
(−2.94)(5.56)(−2.80)(2.88)
A_LEVEL0.05560.16950.0828−0.1274
(2.25)(1.25)(3.11)(−1.25)
HND0.0318−0.16560.0198−0.1932
(1.69)(−1.96)(1.00)(−2.63)
DEGREE0.0790−0.15170.0608−0.2243
(4.36)(−1.85)(3.17)(−3.16)
OTHER−0.01060.26330.05510.0193
(−0.49)(2.46)(1.81)(0.16)
ASIAN0.1542−0.20670.1369−0.9862
(2.97)(−1.08)(1.63)(−4.71)
AFRO-CARIBBEAN0.1369−0.09530.0734−0.4508
(2.75)(−0.42)(1.13)(−2.28)
OTHER_NON-WHITE0.07750.16600.0151−0.0214
(1.57)(0.64)(0.23)(−0.09)
MOTHER_SMOKED−0.04550.4784−0.03910.5191
(−1.95)(4.21)(−1.69)(5.92)
FATHER_SMOKED−0.05310.4910−0.02530.2744
(−3.21)(6.96)(−1.51)(4.86)
PARENTS_SMOKED−0.08230.4828−0.05620.5493
(−4.68)(6.45)(−3.26)(9.19)
YEAR−0.0187−0.0609
(−4.52)(−6.24)
YEAR2/1000.16970.3579
(5.07)(5.48)
YEAR3/1000−0.0523−0.0844
(−6.04)(−5.22)
YEAR4/100000.00500.0066
(7.12)(4.96)
In(TAX)0.0062−0.0281
(0.21)(−0.86)
CONS2.8783−0.06813.3244−0.2827
(102.96)(−0.81)(63.22)(−4.24)
γ0.12950.1350
Log-likelihood−8812.18−10301.79
χ2(16)261.90426.96
Predicted proportion of starters0.690.53
Observed proportion of starters0.660.52
Number of observations37374861
Number of failures24602508
Table 3.  Sensitivity analysis of tax elasticity in models of starting†
ModelResults for menResults for women
  1. t-statistics are given in parentheses.

1920–1985 data
Split population0.006 (0.210)−0.028 (−0.86)
Log-logistic−0.021 (−0.750)−0.034 (−1.12)
1905–1985 data
Split population0.060 (2.13)0.009 (0.29)
Log-logistic0.003 (0.10)0.003 (0.10)

The interpretation of the findings reported for the effect of other variables, such as education and parental education, on the hazard of starting, is unchanged by the correction.

The correction makes no discernible difference to the figures reported in Section 5.1 or their interpretation; nor are changes required to any part of Section 6 of the paper on ‘policy implications’. Corrected versions of Tables 2 and 3 are given here.

Ancillary