Sundhedsudvalget 2020-21
L 61 Bilag 1
Offentligt
2259708_0001.png
,
28
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0002.png
29
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0003.png
30
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0004.png
31
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0005.png
32
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0006.png
33
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0007.png
34
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0008.png
35
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0009.png
36
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0010.png
37
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0011.png
38
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0012.png
39
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0013.png
40
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0014.png
41
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0015.png
42
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0016.png
43
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0017.png
44
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0018.png
45
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0019.png
46
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0020.png
47
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0021.png
48
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0022.png
49
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0023.png
50
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0024.png
51
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0025.png
52
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0026.png
53
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0027.png
54
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0028.png
55
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0029.png
56
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0030.png
57
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0031.png
58
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0032.png
59
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0033.png
60
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0034.png
61
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0035.png
62
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0036.png
63
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0037.png
64
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0038.png
65
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0039.png
66
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0040.png
67
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0041.png
68
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0042.png
69
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0043.png
70
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0044.png
71
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0045.png
72
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0046.png
73
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0047.png
74
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0048.png
75
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0049.png
76
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0050.png
77
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0051.png
78
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0052.png
79
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0053.png
80
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0054.png
81
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0055.png
82
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0056.png
83
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0057.png
84
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0058.png
85
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0059.png
1.
Summary description
The research report "Analysis of the impact of Plain Packaging on smoking prevalence and
impact of the introduction of Plain Packaging regulation in Australia on smoking prevalence
and legal tobacco consumption. After conducting a detailed review of the research project and
the analysis, I outline below a final opinion on the appropriateness of the applied methodology
and correctness of the reported results in the report.
2.
Is the methodology appropriate for the analysis of the
problems investigated and correctly applied?
The research report describes the different types of econometric analysis adopted to test the
impact of the introduction of Plain Packaging on both smoking prevalence and tobacco
consumption. The following list summarizes the different methodologies and statistical
assessment, pointing out the appropriateness of their usage.
As regards smoking prevalence, a probit regression model has been applied to
study the smoking status component of the Roy Morgan Single Source ("RMSS") survey
a cross-sectional survey of a nationally representative sample of Australian individuals
aged 14 and over covering the period from 2001 to December 2017. The dependent
variable takes the value of zero if the respondent does not smoke and one if the respondent
smokes. The usage of a probit model is appropriate for this type of survey data and the set
of explanatory variables employed in the model are appropriate for this type of analysis;
and the methodology has been correctly applied.
As regards legal tobacco consumption, a number of methods are employed to
assess the impact of Plain Packaging using retail audit sales data covering the period from
January 2008 to December 2017:
Firstly, a non-parametric approach has been used to carry out a
before-and-after evaluation of the smoking prevalence and tobacco consumption,
86
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0060.png
using New Zealand as a comparative case (control group), which has not been
subject to the introduction of the Plain Packaging (treatment). The methodology
that is applied is useful for providing a preliminary assessment of the separate
effects of the implementation of Plain Packaging from other characteristics that
could affect tobacco consumption, and it has been correctly applied.
Secondly, a structural break analysis has been implemented to
assess the moment in which a break in the cigarette consumption time series
happens. The usage of the Chow test is appropriate and has been correctly applied
to assess whether there is a structural exogenous break in the consumption pattern
at a given date. The results reveals that there has not been any specific change in
the cigarette consumption trend in Australia as a result of the introduction of Plain
Packaging.
Thirdly, a regression analysis has been performed taking the
cigarette consumption as a dependent variable and Plain Packaging with a set of
other explanatory variables as independent variables. The inclusion of the Plain
Packaging as a dummy variable into the structural model is suitable to test the impact
of Plain Packaging, and the set of explanatory variables employed in the model are
appropriate for this type of analysis; and the methodology has been correctly
applied.
Finally a difference-in-differences regression model has been
performed on per capita cigarettes consumption up to December 2017, including
also New Zealand as a comparator. This is in line with the study of Dryden (2017)
and suggests that plain packaging is associated with a statistically significant (to
the 10% level) increase in per capita cigarettes consumption.
3.
Are the results correctly reported?
The research report has been competently written and includes a sufficiently large amount of
technical details, to explain the adopted methodology and obtained results. The results have
been correctly reported and interpreted, providing a clear understanding of the impact of Plain
Packaging regulation in Australia on smoking prevalence and tobacco consumption.
87
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0061.png
4.
Does the analysis meets the standard and rigor needed for
publication in an academic journal?
The empirical research presented in the report constitutes an extensive and rigorous
econometric study to analyse the underlying patterns of smoking prevalence and tobacco
consumption and to assess the impact of the introduction of Plain Packaging.
This research makes use of inclusive data sets, which facilitate the application of a large battery
of statistical approaches. As examined in the previous subsections, the applied methodologies
cover the main aspects of the analysed phenomenon and have been appropriately applied to the
described empirical setting, in line with the state-of-the-art contributions in this field.
The inclusion of supplementary appendixes provides a comprehensive assessment of the
robustness of applied methodologies, and an explanation of the technical aspects which are
relevant to support the correctness of the obtained results.
Professor Stefano Nasini
8 November 2019
88
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0062.png
W. KIP VISCUSI
AN ASSESSMENT OF THE EFFECT OF AUSTRALIAN PLAIN PACKAGING
REGULATION: ANALYSIS OF ROY MORGAN RESEARCH DATA,
CITTS DATA, AND NTPPTS DATA:
2 JANUARY 2018
1
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0063.png
I.
1.
INTRODUCTION
I am the University Distinguished Professor of Law, Economics, and Management at
Vanderbilt University. I hold a Bachelor’s degree in Economics, two master’s degrees,
and a Ph.D. in economics, all from Harvard University. I have published more than 350
articles and 20 books dealing primarily with health and safety risks, and I have been
ranked among the top 25 economists in the world based on citations in economics
journals. I worked extensively with the U.S. Environmental Protection Agency (“EPA”)
on a continuous basis from 1983 to 2012, where much of my work was focused on the
development of guidelines for hazard warnings for dangerous pesticides and chemicals. I
also have extensive professional experience evaluating regulatory impact analyses and
the economic methodology used in benefit-cost analysis, including being the Deputy
Director of the President’s Council on Wage and Price Stability from 1979–1980, which
was responsible for White House oversight of all new federal regulations during that
period as well as executive branch review of all regulatory impact analyses. Further
details of my educational background and professional experience are set out in
Appendix D of this report.
2.
I have been asked to provide a report that examines empirical data from Australia to see
what effect, if any, plain packaging of tobacco products
1
(“Plain Packaging” or “PP”) has
had on smoking behaviors in Australia following its implementation.
2
3.
1
In particular I have been requested to:
Plain packaging, also known as standardized packaging, generally refers to regulation that requires the
removal of all branding (colors, imagery, corporate logos and trademarks) from product packaging,
permitting manufacturers to print only the brand name in a mandated size, font, and place on the pack, in
addition to the health warnings and any other legally mandated information. The appearance of all tobacco
packs is also standardized, including the color of the pack.
Plain Packaging was introduced in Australia under the Tobacco Plain Packaging Act 2011, No. 148, 2011
with all tobacco products sold in Australia required to comply with the requirements from December 1,
2012.
1
2
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0064.png
a.
i.
Review the following datasets:
Roy Morgan Single Source Survey (“RMSS”) data: a nationally
representative, repeated cross-sectional survey of Australians aged 14 and
above covering the period from January 2001 to December 2016.
ii.
The Australian National Tobacco Plain Packaging Tracking Survey
(“NTPPTS”) data: A continuous survey of Australian smokers and recent
ex-smokers commissioned by the Australian Government’s Department of
Health and Ageing to evaluate the impact of changes in the packaging of
tobacco products in Australia for the period from 9 April 2012 to 30
March 2014.
3
iii.
The Cancer Institute New South Wales (NSW) Tobacco Tracking Survey
(“CITTS”) data: a serial, weekly cross-sectional survey of adult smokers
and recent quitters in New South Wales, Australia, for the period from
February 2009 to June 2016.
4
b.
Review and comment on the following publications in relation to the above
datasets:
i.
Analysis of the Australian RMSS data presented in:
3
4
The data collected in the National Tobacco Plain Packaging Tracking Survey is available on request from
the Australian Department of Health, see
http://www.health.gov.au/internet/main/publishing.nsf/content/tobacco-plain-packaging-evaluation,
accessed 29 June 2017.
See also the Australian National Tobacco Plain Packaging Tracking Survey:
Technical Report, available at
http://tobaccocontrol.bmj.com/content/suppl/2015/02/16/tobaccocontrol-
2014-052050.DC1/tobaccocontrol-2014-052050supp.pdf, accessed 29 June 2017.
I have previously addressed the CITTS data and NTPPTS data in reports that I submitted for British
American Tobacco in October 2015 in UK legal proceedings in which PP was being challenged, and in
November 2015 in relation to the Australian Government Department of Health Post-Implementation
Review of the Tobacco Plain Packaging Act 2011 (Aus). My current report provides a more extensive
analysis of these data including detailed multivariate controls, an empirical assessment of additional CITTS
data, and a longer time period for the CITTS data, and also examines aspects of the cohort component of
the NTPPTS data that I did not consider previously. In addition, this report also includes an analysis of the
RMSS data, which I have not considered previously.
2
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0065.png
The report of Dr. Tasneem Chipty entitled “Study of the Impact of
the Tobacco Plain Packaging Measure on Smoking Prevalence in
Australia” (January 24, 2016), which was commissioned by the
Australian Department of Health and is the only econometric
analysis of data that is relied on in the Australian Government’s
Post Implementation Review Report of the Australian Tobacco
Plain Packaging Act 2011 (“TPP Act”) published in February
2016.
5
Diethelm and Farley (2015) “Refuting tobacco-industry funded
research: empirical data shows a decline in smoking prevalence
following the introduction of plain packaging in Australia.”
6
ii.
Analysis of the Australian NTPPTS data presented in several papers
published in
Tobacco Control,
(April 2015), Volume 24, Suppl. 2, titled
“Implementation and evaluation of the Australian tobacco Plain Packaging
policy,”
7
which papers are also relied on in the Australian Post
Implementation Review Report;
8
and
5
6
7
The Australian Post Implementation Review Report and its appendices, including Dr. Chipty's report are
available on the Australian Government Office of Best Practice Regulation website at
http://ris.dpmc.gov.au/2016/02/26/tobacco-plain-packaging/, accessed 29 June 2017.
Some program codes
and data files relating to Dr. Chipty’s report are also available on request from the Australian Department
of Health at
http://www.health.gov.au/internet/main/publishing.nsf/Content/foi-disc-log-2015-16, accessed
29 June 2107.
Pascal A Diethelm, Timothy M Farley, “Refuting tobacco-industry funded research: empirical data shows a
decline in smoking prevalence following the introduction of plain packaging in Australia,” Tob. Prev.
Cessation 2015;1(November):6
http://dx.doi.org/10.18332/tpc/60650.
Available at
http://tobaccocontrol.bmj.com/content/24/Suppl_2.toc.
The papers in this publication that
analyze the Australian NTPPTS data are:
Melanie Wakefield, Kerri Coomber, Meghan Zacher, Sarah Durkin, Emily Brennan, and Michelle
Scollo, “Australian Adult Smokers’ Responses to Plain Packaging with Larger Graphic Health
Warnings 1 Year after Implementation: Results from a National Cross-Sectional Tracking Survey,”
Tobacco Control 2015;24:ii17-ii25. doi:10.1136/tobaccocontrol- 2014-052050;
3
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0066.png
iii.
Analysis of the New South Wales CITTS data presented in Dunlop et al
(2014) “Impact of Australia’s Introduction of Tobacco Plain Packs on
Adult Smokers’ Pack-Related Perceptions and Responses: Results from a
Continuous Tracking Survey,”
9
which is also relied on in the Australian
Post Implementation Review Report.
c.
Review the Australian Government's Post Implementation Review Report of TPP
Act published in February 2016 (the "PIR"),
10
and comment on the validity of the
conclusions expressed in the report regarding the impact of Plain Packaging on
smoking behaviors.
4.
To the best of my knowledge, the analysis of Australian data that I provide in this report
is the most up to date and comprehensive analysis of the data that has been provided to
date. In particular:
a.
My analysis of the RMSS data extends through December 2016, thus providing
15 more months of data in the post-PP period than analyzed in the report of Dr.
8
9
10
Sarah Durkin, Emily Brennan, Kerri Coomber, Meghan Zacher, Michelle Scollo, and Melanie
Wakefield, “Short-Term Changes in Quitting-Related Cognitions and Behaviours after the
Implementation of Plain Packaging with Larger Health Warnings: Findings from a National Cohort
Study with Australian Adult Smokers,” Tobacco Control 2015;24:ii26-ii32.
doi:10.1136/tobaccocontrol-2014-052058;
Emily Brennan, Sarah Durkin, Kerri Coomber, Meghan Zacher, Michelle Scollo, and Melanie
Wakefield, “Are Quitting-Related Cognitions and Behaviours Predicted by Proximal Responses to
Plain Packaging with Larger Health Warnings? Findings from a National Cohort Study with Australian
Adult Smokers,” Tobacco Control 2015;24:ii33-ii41. doi:10. 1136/tobaccocontrol-2014-052057; and
Michelle Scollo, Meghan Zacher, Kerri Coomber, Megan Bayly, and Melanie Wakefield, “Changes in
Use of Types of Tobacco Products by Pack Sizes and Price Segments, Prices Paid and Consumption
Following the Introduction of Plain Packaging in Australia,” Tobacco Control 2015;24:ii66-ii75.
Supra at footnote 5.
Sally M. Dunlop, Timothy Dobbins, Jane M. Young, Donna Perez, and David C. Currow, “Impact of
Australia’s Introduction of Tobacco Plain Packs on Adult Smokers’ Pack-Related Perceptions and
Responses: Results from a Continuous Tracking Survey,"
BMJ Open
2014; 4(12): e005836, Available at
http://bmjopen.bmj.com/content/bmjopen/4/12/e005836.full.pdf.
Supra at footnote 5.
4
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0067.png
Chipty and 3 additional years of data in the post-PP period than in Diethelm and
Farley (2015);
b.
My analysis of the Australian NTPPTS data and the New South Wales CITTS
data includes an analysis of all of the survey outcomes, rather than a selection of
the outcomes as presented in the published papers on these data. Also, unlike the
the Australian Government's Australian Post Implementation Review Report
which simply relies on the conclusions from the published papers on these data, I
also undertake a review of the papers and analyze the underlying data; and
c.
My analysis of the CITTS data includes a longer time period than in any
published study as it extends through June 2016, which includes an additional 37
months of data from that considered by Dunlop et al. (2014) in their analysis of
the CITTS data.
5.
As noted in the report of Dr. Chipty, at the same time that Australia introduced tobacco
Plain Packaging it also introduced updated and enlarged graphic health warnings on
tobacco product packaging under the Competition and Consumer (Tobacco) Information
Standard 2011 (which included expanding the size of the warning on the front of the pack
from 30% to 75%). Given the timing of these changes, it is not possible to separately
identify the effects of tobacco Plain Packaging from those of the updated and enlarged
graphic health warnings without making restrictive assumptions. As such, my discussion
of the effects of Plain Packaging encompasses the effects from both of these changes,
which I refer to collectively as the "2012 Packaging Changes" (as is also the case with Dr
Chipty's analysis in her report).
5
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0068.png
II.
6.
EXECUTIVE SUMMARY
In this report I provide a comprehensive analysis of three key datasets from Australia,
namely: 4 years of post-implementation RMSS data; 3 �½ years of post-implementation
CITTS data; and the Australian Government commissioned NTPPTS data. My use of a
longer post-implementation time frame for my analysis than in any previous study
provides a stronger test of the impact of the 2012 Packaging Changes if, as some have
suggested, the effect of the policy change would increase over time. Each of these
datasets provides somewhat different perspectives, and no single dataset is complete in
terms of addressing both smoking prevalence and various attitudinal responses to plain
packs, or what are sometimes referred to as intermediate metrics. However, despite the
different perspectives provided by these data, the implication of my analysis of the three
datasets yields a consistent conclusion that there is no evidence of the 2012 Packaging
Changes having any impact on reducing smoking prevalence rates or consumption
amongst current smokers. There is also consistent evidence that the policy is associated
with counterproductive effects on some of the intermediate or secondary measures that
are relied upon in the Australian Government's Australian Post Implementation Review
Report and by other proponents to promote Plain Packaging.
11
Chief among these
potentially counterproductive effects is that there has been an increase in the belief that
the warnings exaggerate the risks. Other impacts, such as the effects on quitting behavior
and smoking rates, are more mixed. On balance, these results undermine any conclusion
that the 2012 Packaging Changes have had a net beneficial effect. In particular:
11
See e.g. World Health Organization, Regional Office for Europe, (2014) "Evidence Brief Plain packaging
of tobacco products: measures to decrease smoking initiation and increase cessation"; and McNeill A,
Gravely S , Hitchman SC, Bauld L, Hammond D, Hartmann-Boyce J. "Tobacco packaging design for
reducing tobacco use". Cochrane Database of Systematic Reviews 2017, Issue 4. Art. No.: CD011244.
DOI: 10.1002/14651858.CD011244.pub2 .
6
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0069.png
a.
My analysis of the RMSS data, which includes 15 additional months of data in the
post-2012 Packaging Changes period than was considered in the report of Dr.
Chipty, and 3 years of additional data in the post-2012 Packaging Changes period
than was addressed in Diethelm and Farley (2015), found that the estimated
statistical association of the 2012 Packaging Changes with smoking prevalence
rates is zero. Instead, my analysis of the RMSS data found that the decline in
smoking prevalence rates in Australia is a continuation of past nonlinear time
trends, overall economic trends such as the general Australian consumer price
index, and influences such as rising cigarette prices, and is not significantly
related to the adoption of the 2012 Packaging Changes. The only sound
conclusion based on this evidence is that the 2012 Packaging Changes are not
associated with any change in smoking prevalence rates.
b.
An evaluation of the CITTS and NTPPTS data relating to actual cigarette
consumption behavior in Australia indicates that the 2012 Packaging Changes
have not been associated with a decrease in smoking behaviors amongst current
smokers. The results for the CITTS sample are mixed, with no clear cut evidence
of efficacy. The number of cigarettes smoked per day experienced a statistically
significant increase of about one cigarette. There has also been a change in the
distribution of smoking activity. More respondents report that they smoke daily,
fewer report that they smoke at least weekly (not daily), fewer report that they
smoke less often than weekly, and a statistically insignificant larger number report
that they currently smoke not at all, though they did smoke in the last year.
Within the NTPPTS sample, there is no statistically significant change in the
7
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0070.png
number of cigarettes smoked per day.
12
My analysis of the CITTS data includes a
longer time period than in any published study as it extends through June 2016,
which includes an additional 37 months of data beyond that considered in Dunlop
et al. (2014).
13
c.
There is also consistent evidence from the CITTS and NTPPTS data indicating an
unfavorable association of the 2012 Packaging Changes with a number of so
called intermediary metrics (e.g., increasing the efficacy of health warnings) even
setting aside issues pertaining to the efficacy of these intermediate variables in
predicting actual smoking behaviors. For example, my analysis of the CITTS
data shows that after the implementation of 2012 Packaging Changes in Australia:
i.
respondents rate it significantly more difficult to quit both in terms of how
difficult it would be to quit and how difficult they thought it would be to
quit, and respondents are significantly less confident that they can quit,
which is an impact that could arise if the policy made consumers think that
quitting would be a more formidable challenge; and
ii. there is a statistically significant 16% increase in whether respondents
believe that the graphic warning labels policy exaggerate the risk of
smoking, a statistically significant 7% increase in beliefs that the
government pesters people too much about smoking risks, a statistically
12
13
I note that consistent with my analysis, Scollo et al. (2015), which is the only published study of the
NTPPTS data that discusses the data on actual consumption behavior, also found that the 2012 Packaging
Changes had no impact on consumption: see Michelle Scollo, Meghan Zacher, Kerri Coomber, Megan
Bayly, and Melanie Wakefield, "Changes in Use of Types of Tobacco Products by Pack Sizes and Price
Segments, Prices Paid and Consumption Following the Introduction of Plain Packaging in Australia,"
Tobacco Control
2015;24:ii66-ii75.
Sally M. Dunlop, Timothy Dobbins, Jane M. Young, Donna Perez, and David C. Currow, “Impact of
Australia’s Introduction of Tobacco Plain
Responses: Results from a Continuous Tracking Survey",
BMJ Open
2014; 4(12): e005836;
8
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0071.png
significant 5% increase in beliefs that the health effects are exaggerated,
and a statistically significant 3% increase in the belief that smoking is only
harmful to heavy smokers. These result could arise from defensive
processes and increased reactance activated by the the 2012 Packaging
Changes, leading to an increased degree of rejection of the graphic
warnings message.
d.
My analysis of the NTPPTS data also shows that after the implementation of the
2012 Packaging Changes in Australia:
i.
respondents were less likely to think about quitting either once or once
every few days over the previous week, less likely to stub out many times
after the policy, and were less likely to stop many times upon having the
urge to smoke;
ii.
there was a decrease in the number of respondents who intend to quit
smoking in the next month, and a decrease in the number of respondents
who stub out their cigarette many times after thinking about the harms of
smoking; and
iii.
there is no statistically significant impact on beliefs regarding the
harmfulness of cigarettes; zero effects with respect to all categories of
responses regarding whether the person thinks about the money spent on
cigarettes; and an increase in the agreement that the dangers are
exaggerated.
7.
The evidence of a lack of impact of the of the 2012 Packaging Changes in Australia and
of a number of potentially counterproductive effects is not unexpected given that
9
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0072.png
consumers are informed of the risks of smoking and the 2012 Packaging Changes do not
provide any new information to consumers. An assumption that making the warnings
larger and more prominent will increase their effectiveness is misplaced. There is no
empirical evidence that “shouting” works in increasing behavioral compliance in this
context where no new information is being provided. The evidence of negative outcomes
is also consistent with research that demonstrates that fear-based warnings may in fact
elicit responses that are the opposite of their intended effect.
8.
In this report, I also evaluate and provide a critique of previous analysis of the Australian
RMSS data presented in the report by Dr. Tasneem Chipty and in Diethelm and Farley
(2015):
a.
The report by Dr. Tasneem Chipty was commissioned by the Australian
Department of Health to assess the impact of Plain Packaging on smoking
prevalence in Australia, and is the only econometric analysis of data that seeks to
identify the actual effect of the 2012 Packaging Changes on smoking that is relied
on in the Australian Post Implementation Report. As explained below, I have
identified several flaws in in Dr. Chipty's approach that render it unreliable,
namely:
i.
the use of overlapping indicator variables which create confounding
effects, meaning that any conclusions drawn from Dr. Chipty’s analysis
are highly speculative;
ii.
the use of a linear time trend when the time trend is nonlinear. Dr.
Chipty’s procedure violates basic principles of statistical analysis since
she reported no statistical tests of the use of a linear trend as opposed to a
10
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0073.png
nonlinear trend. Capturing any nonlinear relationship with a linear trend
line, as Dr. Chipty does, leads to an unexplained “policy impact” that is
spurious, as it is not an effect of the policy but instead reflects an
underlying nonlinear trend; and
iii.
the use of indicator variables only for the major tax increases that occurred
in Australia, which fails to recognize the continuous nature of excise tax
levels and generates a source of error in the treatment of taxes; and
iv.
the failure to include a cigarette price variable in her model which is the
most important variable in models of the economic demand for any
consumer product.
As a result of these shortcomings, the report of Dr Chipty provides no sound
evidence in support of the efficacy of plain packs policies.
The two most
important flaws in her study were the failure to consider the nonlinearity of
the temporal trend in smoking prevalence rates and the omission of cigarette
prices from the model.
v.
Three principal results from the RMSS data analysis are apparent in both
the analysis of the extended dataset that I used, as well as in my analysis
of the shorter time period considered in Dr. Chipty’s report:
Properly recognizing that the temporal trend is nonlinear rather
than linear (as Dr. Chipty wrongly assumes) alone accounts for the
downward trend in smoking rates;
Even with only a linear trend, reasonable specifications of the
model using either overall consumer prices or continuous measures
11
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0074.png
of the recommended retail price of cigarettes rather than the crude
excise tax indicator variables approach used by Chipty eliminates
the statistical significance of the 2012 Packaging Changes variable.
It is only by ignoring both the nonlinearity of the smoking
prevalence time trend and the role of prices, as Dr. Chipty does,
that it is possible to generate non-zero statistically significant
estimates of the 2012 Packaging Changes variable; and
Third, even if there were a purported association of the 2012
Packaging Changes with smoking prevalence based on Dr.
Chipty's analysis, one should be skeptical of the import of these
results given that her statistical analysis includes four overlapping
indictor variables for the 2010 to 2015 period. Given her statistical
format, it is difficult to disentangle the effect of the multiple policy
shifts that occurred around the 2012 period. Attributing the lower
smoking prevalence rates to the 2012 Packaging Changes as
opposed to the excise tax increases both before and after the advent
of the 2012 Packaging Changes policy is not warranted.
vi. Properly specified multivariate regression analyses that corrects for the
flaws in Dr. Chipty’s analysis demonstrates that the estimated effect of the
2012 Packaging Changes on the smoking prevalence rate cannot be
distinguished statistically from zero. As noted above, I also extended Dr.
Chipty’s data period with an additional 15 months of data through to
December 2016 and found that the impact of the 2012 Packaging Changes
12
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0075.png
on smoking prevalence rates cannot be distinguished statistically from
zero for this longer time period as well.
b. The article by Diethelm and Farley (2015) is identified as the only published
study that assesses the impact of Plain Packaging on smoking prevalence, in a
recent Cochrane review of the plain packing literature that was published in April
2017 (the "Cochrane Review").
14
However, as I explain below, the article lacks
scientific validity for several reasons:
i.
The authors had no original data, but instead relied on estimates of
monthly averages inferred from a figure in a working paper by Kaul and
Wolf.
15
As a result, their sample size for their analysis only included 156
imputed monthly average figures, not the more than 700.000 individual
observations in the Kaul & Wolf sample.
16
While the authors express
concern about the possible error in imputing data based on the chart in
Kaul and Wolf, the more important limitation is that the aggregation of the
data by month reduces the informational content of the data and prevents
the ability to match any data to particular respondents;
ii.
Because of this reliance on monthly average data the authors have no
information by individual respondent and consequently their analysis
includes no controls in the model for individual characteristics such as
14
15
16
McNeill, A., Gravely, S., Hitchman, S.C., Bauld, L., Hammond, D., and Hartmann-Boyce, J., "Tobacco
Packaging Design for Reducing Tobacco Use," Cochrane Database of Systematic Reviews 2017, Issue 4,
Art. No.: CD011244.
Kaul, A. and Wolf, M. "The (Possible) Effect of Plain Packaging on Smoking Prevalence in Australia: A
Trend Analysis," University of Zurich Department of Economics Working Paper, June 2014.
I note that the Diethelm and Farley (2015) sample size is wrongly listed as 700,000 on p. 4 of the Cochrane
Review.
13
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0076.png
age, gender, education, income level, and region, and changes in the
sample composition that may have occurred over time.
iii.
In addition to ignoring all demographic variables, the Diethelm and Farley
(2015) article also omitted other key determinants of smoking prevalence
rates. Cigarette prices are not included in the model. Excise tax rates are
ignored except in terms of a single tax shift. Also, the nonlinear nature of
the smoking prevalence rates before the advent of plain packaging is not
taken into account.
17
iv.
The net impact of these flaws is that this study lacks any scientific
credibility. The deficiencies I cited are not minor limitations nor matters
of a difference of opinion, but are fundamental problems that make it
inappropriate to rely on their study.
v.
I note that the Cochrane Review graded the quality of the Diethelm and
Farley (2015) paper as “low” (p.4), which is an assessment I believe
nevertheless actually overstates the quality of the paper for the reasons
stated above.
Had the Cochrane Review also taken into account the
factors I cite above, the article’s value would be below the rating of “low,”
as it should not be regarded as having any scientific merit at all.
9.
My examination of the outputs of each of the CITTS and NTPPTS datasets also indicates
that the published articles analyzing these data are disturbing from the standpoint of
academic integrity and are highly misleading. Rather than provide an unbiased
assessment of the survey results, the studies present selected findings that purport to
17
Diethelm and Farley (2015) faulted the study by Kaul and Wolf, which discarded the first 42 months of
observations because they were not consistent with a linear trend. However, Diethelm and Farley (2015)
do not present any statistical tests supporting the validity of a linear trend.
14
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0077.png
demonstrate the efficacy of the 2012 Packaging Changes policy which a more thorough
analysis of the data shows is misleading.
In particular, these studies have ignored
substantial evidence from the same Australian datasets they purport to analyze; which
evidence is consistent with a lack of efficacy of the 2012 Packaging Changes and
underscores the point that the post-implementation publications analyzing these data have
selectively and misleadingly presented the results they do present. Viewed in their
entirety, the datasets consistently indicate that the 2012 Packaging Changes are not
associated with any reduction in smoking behaviors. Empirical assessments are biased if
a researcher only cites the results that portray a policy in a positive light and fails to
report the evidence that indicates not only a lack of a favorable impact, but also rather
important counterproductive effects.
A comprehensive analysis of a broader set of
questions in the CITTS and NTPPTS datasets leads to the conclusion that on balance the
2012 Packaging Changes policy is not working.
10.
The fact that these articles are peer reviewed does not provide any assurance that the
analyses and conclusions of the papers are valid. I have served for three decades as the
founding editor of a peer-reviewed journal and have been on the editorial boards of 20
other peer-reviewed journals. Peer review only means that one or more persons in the
field has reviewed the article and has recommended publication of it. The peer reviewers
do not generally have access to the data used in the article to replicate the study.
Reviewers typically only read the article to assess whether the methodology and findings
appear to be sound and novel contributions. Other researchers who have access to the
original data often can undertake a more thorough analysis than in a peer review, as I
have done with the data that I have reviewed.
15
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0078.png
11.
Similar to my review of the data, the only consistent evidence that the recent Cochrane
Review of the Plain Packaging literature found was that Plain Packaging was associated
with a decrease in the appeal of the pack. However, the responses to these questions
could be a result of the increase in the size of the warning to 75% that result in the packs
being dominated by graphic health warnings, rather than Plain Packaging. Indeed the
Cochrane Review highlights the high risk of confounding in these studies given that Plain
Packaging was introduced alongside enhanced health warnings in Australia making it
difficult to isolate the effects of Plain Packaging. For studies that focused on the effects
of Plain Packaging on actual behavioral outcomes, including smoking prevalence and
consumption, the Cochrane Review concluded that the confidence in the findings was
“limited, due to the nature of the evidence available” and that the evidence was “mixed.”
The Cochrane Review also noted that: "[n]o studies assessed uptake, cessation, or relapse
prevention"(p. 2). The results of many of the studies on other intermediate outcomes were
also limited and mixed so that there is no consistent evidence of Plain Packaging being
effective across a large number of variables. The Cochrane Review concludes that "[t]he
available evidence suggests that standardised packaging may reduce smoking prevalence"
(p. 2), which itself doesn't demonstrate the efficacy of Plain Packaging. In addition,
based on my review of the actual data emanating from Australia and critical analysis of
the published papers on this data, which the Cochrane Review did not undertake, the
overwhelming evidence across all the datasets is not consistent with the 2012 Packaging
Changes being effective in reducing smoking, while there is also evidence consistent with
the policy being counterproductive on many of the intermediate or secondary measures
that are relied upon to promote Plain Packaging.
16
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0079.png
12.
The Australian Government Post-Implementation Review Tobacco Plain Packaging 2016
report (the "PIR") provided an inadequate and incorrect assessment of the effect of the
2012 Packaging Changes on smoking. The PIR’s only statistical evidence of the effect of
the 2012 Packaging Changes is based on the flawed report by Dr. Chipty. What is also
striking is that the PIR did not review the implications of the NTPPTS and CITTS data
with respect to smoking prevalence and consumption.
The PIR merely relies on
published papers without any critique or review of those papers. Based on my review of
the papers and the underlying NTPPTS and CITTS datasets, I conclude that they cannot
be relied upon. There is not a sound basis for the PIR’s conclusion (p. 4): “[i]n light of all
this evidence, the PIR concludes that tobacco plain packaging is achieving its aim of
improving public health in Australia and is expected to have substantial public health
outcomes in the future.” The PIR’s reference to “all this evidence” is especially
inappropriate because the cited studies did not report all the evidence from the NTPPTS
and CITTS datasets, but only the selected results that provide the most favorable
perspective on the performance of plain packs. In addition, as noted above my extended
analysis of the RMSS data and the CITTS data which is the most extensive data analysis
undertaken to date (and includes 15 months of additional data to the analysis undertaken
by Dr Chipty) confirms that Plain Packaging has not been effective, as the statistical
association of the 2012 Packaging Changes with smoking prevalence rates cannot be
distinguished from zero. This result further demonstrates that the conclusion reached in
the PIR is unjustified.
17
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0080.png
III.
13.
ROY MORGAN RESEARCH DATA
The most extensive set of individual survey data on smoking prevalence in Australia are
the RMSS data from Roy Morgan Research. This independent firm employs a large
nationally representative Australian sample using cross-sectional surveys to generate
survey data on individuals aged 14 and over regarding their smoking status, where
interviewers are dispatched weekly and all areas are covered monthly.
18
The data have
been used in previous peer-reviewed published research to obtain estimates of smoking
prevalence in Australia.
19
14.
Previous reports have provided analyses of the effect of Plain Packaging on smoking
prevalence in Australia using the RMSS data, including:
a.
A non-peer-reviewed report prepared for the Australian Department of Health by
Dr. Tasneem Chipty, which analyzed the RMSS data from January 2001 to
September 2015, and concluded based on a regression analysis that during the
period after the implementation of the 2012 Packaging Changes in Australia
through to September 2015, smoking prevalence rates declined by 0.55
percentage points relative to what the prevalence would have been without the
2012 Packaging Changes;
20
and
b.
Diethelm and Farley (2015) which assessed the effect of Plain Packaging on
smoking prevalence in Australia using the RMSS data for adults (aged 18+) for
18
19
20
For details of how Roy Morgan Research collects the RMSS data see Roy Morgan Research. "How we
collect and process Single Source data in Australia". Available from:
http://www.roymorgan.com/products/single-source/single-source-fact-sheets,
accessed 2 May 2017.
See Wakefield MA, Durkin S, Spittal MJ, Siahpush M, Scollo M, Simpson JA, et al. "Impact of tobacco
control policies and mass media campaigns on monthly adult smoking prevalence" Am J Public Health.
2008;98:1443-50. doi: 10.2105/AJPH.2007.128991; and Wakefield MA, Coomber K, Durkin SJ, Scollo M,
et al. "Time series analysis of the impact of tobacco control policies on smoking prevalence among
Australian adults," 2001-2011. Bull World Health Organ 2014; 92:413-422 doi: 10.2471/BLT.13.118448.
Dr. Tasneem Chipty, "Study of the Impact of the Tobacco Plain Packaging Measure on Smoking
Prevalence in Australia," Report of Dr. Tasneem Chipty, January 24, 2016, supra at footnote 5.
18
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0081.png
the period from January 2001 to December 2013 (one year after mandatory full
implementation of PP in Australia) and found a statistically significant reduction
of smoking prevalence of 3.7% following the introduction of Plain Packaging in
Australia.
I note that the recently published Cochrane Review of the Plain
Packaging literature notes that this paper is the only peer-reviewed published
paper that assesses the impact of Plain Packaging on smoking prevalence in
Australia.
21
15.
In this report I present an analysis of a larger Roy Morgan Research sample that includes
a longer post-2012 Packaging Changes time period than that provided in Diethelm and
Farley (2015), and in the report of Dr. Chipty. I also present analysis of the time period
considered in Dr. Chipty’s report to facilitate a comparison of the results. The starting
date for my Roy Morgan Research data is January 2001, which is the same as that of
Diethelm and Farley (2015) and Dr. Chipty's report. However, the data I analyze extend
through December 2016, thus providing 3 years of additional data in the post-2012
Packaging Changes period than was addressed in Diethelm and Farley (2015) and 15
additional months of data in the post-2012 Packaging Changes period than was
considered in the report of Dr. Chipty, making my analysis the most up to date available.
According to Dr. Chipty, the inclusion of the new data should lead to even larger
estimates of the effect of plain packs as she hypothesizes, “the benefits of the Packaging
Changes will likely grow over time.”
22
A.
Analysis of RMSS Data
21
22
McNeill A, Gravely S , Hitchman SC, Bauld L, Hammond D, Hartmann-Boyce J. "Tobacco packaging
design for reducing tobacco use". Cochrane Database of Systematic Reviews 2017, Issue 4. Art. No.:
CD011244. DOI: 10.1002/14651858.CD011244.pub2.
Chipty Report, p. 3.
19
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0082.png
16.
Here I provide a detailed analysis of the RMSS data on smoking prevalence rates. The
principal matter of interest is the effect of the 2012 Packaging Changes in Australia on
smoking prevalence rates based on a regression analysis of whether the respondent is a
smoker as a function of pertinent demographic and policy variables. My assessment of
the RMSS data is patterned generally after that in the report of Dr Chipty. The table
below lists the extensive set of demographic and locational variables from the RMSS data
that I include in my multivariate regression analysis for which the detailed estimates
appear in Appendix A. My variables address the same range of demographic influences
as in Dr. Chipty’s report, though there are a few minor differences. For example, I use
continuous measures of age and income rather than a large series of categorical variables
for different age and income categories.
Explanatory Variables Used in Regressions
Variable
Female
Marital status, single
Marital status, divorced
Marital status, widowed
Marital status, separated
Student
Years of education
Age
Non-adults (14-17)
Employed full time
Retired
Income (thousands)
Income, multiple household members
Bread winner
Mean
0.5181
0.2422
0.0818
0.0781
0.0370
0.0238
12.34
47.46
0.0575
0.5524
0.0111
47.63
0.4162
0.6502
Std. Dev.
0.4997
0.4284
0.2741
0.2683
0.1887
0.1525
3.18
19.30
0.2328
0.4972
0.1048
40.93
0.4929
0.4769
20
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0083.png
Household size
Home owner
Victoria
Queensland
South Australia
Western Australia
Tasmania
Darwin-Alice Springs
Lives in capital city
17.
2.7263
0.6812
0.2300
0.1985
0.0780
0.0950
0.0481
0.0101
0.5812
1.3806
0.4660
0.4208
0.3988
0.2681
0.2932
0.2139
0.1001
0.4934
The principal differences between my analysis and Dr. Chipty’s, as I discuss below, are
with respect to the following matters. First, I include different measures of cigarette
prices in the equation to account for the important economic dependence of smoking
behavior on the cost of cigarettes. Second, my analysis accounts for the nonlinear trend
in smoking prevalence rates and includes a statistical test of the importance of
nonlinearity, whereas Dr Chipty assumes without any testing that the trend is linear.
Thirdly, my analysis also accounts for the continuous changes in cigarette excise tax rates
rather than focusing on the major increases alone. As I discuss below, if the analysis
correctly includes either a measure of cigarette prices or a nonlinear trend, or both of
these influences, then the estimated statistical association of the 2012 Packaging Changes
with smoking prevalence rates is zero. The only sound conclusion based on this evidence
is that the 2012 Packaging Changes are not associated with any change in smoking
prevalence rates.
18.
I begin with an analysis of the RMSS data time period used in Dr. Chipty’s report and
excluding October and November 2012 from the sample. As noted above, my variables
address the same range of demographic influences as do Dr. Chipty’s, with only a few
21
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0084.png
minor differences. In addition to these variables, I have constructed a series of policy-
related variables based on the different time periods relating to the policies noted above.
These include indicator variables for the 2006 graphic warnings policy and the 2012
Packaging Changes as well as measures of consumer prices and the recommended retail
price of cigarettes. I capture the effect of the cost of cigarettes in several separate ways.
First, I use indicator variables for the major excise tax eras indicated in the table above.
However, as I discuss further below in my critique of the report of Dr. Chipty, this
formulation ignores the continuous nature of excise tax changes and also ignores the level
of the taxes. Second, instead of these indicator variables I include a variable for the level
of excise taxes per pack, in real inflation-adjusted terms. This measure accounts for both
the excise tax level embodied in the major excise tax increases and also recognizes the
periodic updates of the excise taxes during the year. Third, as a measure of the cost of
cigarettes I have used two different measures of the total cigarette prices, not simply the
excise tax component. The first cost variable is the overall consumer price index (CPI),
which is a measure of general price trends in the economy, not just the cost of cigarettes.
I also use a more cigarette-specific price measure, which is the recommended retail price
per pack for Craven Cork Tip 20s cigarettes. This data is provided in Scollo, and
Winstanley, "Tobacco in Australia: Facts and Issues," where the authors explain that
Craven is a longstanding brand in Australia and one of a handful of brands available in
1940 that is still available in 2016.
23
Scollo, and Winstanley also provide recommended
23
See table 13.3.1 in Section 13.3 in Scollo, MM and Winstanley, MH. "Tobacco in Australia: Facts and
issues". Melbourne: Cancer Council Victoria; 2016. Available from
www.TobaccoInAustralia.org.au.
I
recognize that the actual cost per pack may be different due to the influence of discounting. However, if
discounting policies are consistent across time, the retail price will differ from the discounted price by a
multiplicative constant, leaving the statistical significance of the estimated impact of prices unaffected.
Even if discounting policies change over time, recommended retail prices will be strongly correlated with
actual prices as evidenced by the negative effect of prices on smoking prevalence rates.
22
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0085.png
retail price data for Winfield 25s which they state is a popular Australian brand, and I
obtained similar results using these data in results which are not reported here.
19.
For simplicity I report here only the estimates for the 2012 Packaging Changes policy
variable. Representative regression results for my full sample appear in Appendix A. The
table below summarizes 10 different ways in which the model could be formulated—
whether the model includes a linear or nonlinear trend and the formulation of the
cigarette cost variables using indicator variables, the excise tax level, the overall
consumer price index, the retail price per pack for Craven 20 cigarettes, and an
instrumental variables (IV) version of the Craven 20 measure to account for the possible
mutual dependence of cigarette prices and smoking prevalence.
24
In 8 of the 10 estimates
reported below, it is not possible to reject the hypothesis that the coefficient for the 2012
Packaging Changes in the smoking prevalence rate equation is zero. The estimated
coefficient is statistically indistinguishable from zero for all models including a nonlinear
time trend or either a linear or nonlinear time trend but also including the consumer price
index, the Craven 20 price level, or the IV version of the Craven 20 prices. It is only by
ignoring both the nonlinearity of the smoking prevalence trend and the role of prices, as
Dr. Chipty does, that it is possible to generate non-zero statistically significant 2012
Packaging Changes coefficient. Given the strong correlation of the nonlinear trend
variable and the nonlinear trend in cigarette prices, including both these variables is not
needed to eliminate the statistical significance of the 2012 Packaging Changes. Thus, the
only two estimates below that can be distinguished from zero assume a linear time trend
and use either indicator variables or the cigarette excise tax as a proxy for cigarette
prices, which as explained below is unjustified.
24
The instruments used to predict the Craven 20 price are the consumer price index and the excise tax levels.
23
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0086.png
Estimates of the 2012 Packaging Changes Coefficient for Equations Using
the Chipty Sample
2012 Packaging Changes Variable
Equation characteristics
Tax policy indicators and linear time trend
Tax policy indicators and nonlinear time trend
Cigarette tax levels and linear time trend
Cigarette tax levels and nonlinear time trend
Consumer price index and linear time trend
Consumer price index and nonlinear time trend
Cost per pack and linear time trend
Cost per pack and nonlinear time trend
IV cost per pack and linear time trend
IV cost per pack and nonlinear time trend
Notes:
Significance levels: *0.10, **0.05, ***0.01
Coefficient
–0.0062 ***
–0.0029
–0.0050 ***
–0.0012
–0.0026
–0.0004
–0.0027
–0.0006
–0.0024
–0.0006
Standard Error
0.0021
0.0026
0.0015
0.0024
0.0022
0.0023
0.0021
0.0023
0.0022
0.0023
20.
Similar results are found for the entire RMSS data time period extending through
December 2016. The Roy Morgan Research sample that I use includes 857,355
observations from January 2001 through December 2016. Here I report results not
excluding October and November 2016 from the sample and using December 2012 as the
2012 Packaging Changes starting date. As indicated in Appendix A, the results are
similar when using October 1 as the starting date, December 1 as the starting date, or
December 1 as the starting date but discarding the October and November 2012 data. As
with the results above, the 2012 Packaging Changes variable is negative and statistically
significant in only 2 of the 10 equations. However, in the other specifications, the 2012
Packaging Changes coefficient is substantially reduced and is never statistically
significant. The estimated effect of the 2012 Packaging Changes is always
indistinguishable from zero if the model includes a nonlinear trend term or includes a cost
24
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0087.png
measure based on the overall CPI, the recommended retail price of Craven 20 cigarettes,
or an IV version of the Craven 20 variable.
Estimate of the 2012 Packaging Changes Coefficient for Equations Using
the Full Sample
2012 Packaging Changes Variable
Equation characteristics
Tax policy indicators and linear time trend
Tax policy indicators and nonlinear time trend
Cigarette tax levels and linear time trend
Cigarette tax levels and nonlinear time trend
Consumer price index and linear time trend
Consumer price index and nonlinear time trend
Cost per pack and linear time trend
Cost per pack and nonlinear time trend
IV cost per pack and linear time trend
IV cost per pack and nonlinear time trend
Notes:
Significance levels: *0.10, **0.05, ***0.01
Coefficient
–0.0061***
–0.0030
–0.0058***
–0.0019
–0.0029
–0.0013
–0.0032
–0.0015
–0.0029
–0.0015
Standard Error
(0.0021)
(0.0026)
(0.0018)
(0.0023)
(0.0021)
(0.0022)
(0.0021)
(0.0022)
(0.0021)
(0.0022)
21.
Using a Roy Morgan Research RMSS dataset that includes an additional 15 months of
data not included in Dr. Chipty's report should have led to larger estimates of the effect of
2012 Packaging Changes if the impact of the policy is increasing over time, as Dr. Chipty
hypothesizes. What I find instead is that the estimated effect is not distinguishable from
zero if one correctly recognizes either the nonlinear nature of the time trend or the impact
of cigarette prices on smoking prevalence rates.
25
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0088.png
B.
Critique of Previous Analyses of RMSS Data
i.
Report of Dr. Tasneem Chipty entitled “Study of the Impact of the Tobacco Plain
Packaging Measure on Smoking Prevalence in Australia” (24 January 2016).
22.
The report of Dr. Chipty was commissioned by the Australian Department of Health and
is the only econometric analysis of data that is relied on in the Australian Government's
Post Implementation Review Report. I also understand that it has been cited by a number
of other regulators and proponents of Plain Packaging to support claims that the policy
has been successful in reducing smoking.
25
23.
The report focuses on the effect of policy changes in Australia on the probability that
members of the RMSS sample report are smokers. As noted above, Dr. Chipty
acknowledges that it is not possible to separately identify the effects of tobacco plain
packaging from those of the updated and enlarged graphic health warnings which
Australia implemented at the same time. As such, Dr. Chipty's analysis, as well as the
analysis in my report, encompasses the estimated effects of Plain Packaging and the
updated and enlarged graphic health warnings (which Dr. Chipty also refers to as the
2012 Packaging Changes). Dr. Chipty’s multivariate regression analysis controls for
demographic factors, a linear time trend, and various tax increases and other policy shifts.
24.
The principal matter of interest for this analysis is the estimated effect of the 2012
Packaging Changes. Dr. Chipty’s model captures this influence with an indicator variable
that takes on a value of 1 from December 1, 2012 through 2015, and a value of zero
otherwise. Her particular 2012 Packaging Changes variable excludes the transition period
25
See e.g. U.S. National Cancer Institute and World Health Organization.
The Economics of Tobacco and
Tobacco Control.
National Cancer Institute Tobacco Control Monograph 21. NIH Publication No. 16-CA-
8029A. Bethesda, MD: U.S. Department of Health and Human Services, National Institutes of Health,
National Cancer Institute; and Geneva, CH: World Health Organization; 2016.
26
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0089.png
of October and November 2012 during which plain packs began to appear in the
Australian market, so that the variable has a value of 0 through September 2012, with a
value of 1 starting in December and the months of October and November excluded from
the analysis. In Appendix A, I report representative parallel results using an October 1
start date, a December 1 start date, or a December 1 start date but excluding the October
and November transition period from the analysis altogether, as does Dr. Chipty. The
results are similar in all these cases. Dr. Chipty’s indicator variable for the 2012
Packaging Changes is intended to capture shifts in smoking rates with the advent of the
2012 Packaging Changes. Smoking behavior will, of course, also be affected by other
factors such as the cost of cigarettes. Dr. Chipty's analysis seeks to isolate the effect of
the 2012 Packaging Changes on smoking prevalence rates by including indicator
variables for different excise tax eras, notably the increases in excise taxes in 2010, 2013,
and 2014. Similar to the indicator variable for the 2012 Packaging Changes, these
variables take on a value of 0 in the years before the excise tax increase and 1 in those
years and thereafter. In my analysis above, I also present models that avoid this
undesirable variable overlap by using continuous measures of prices and taxes.
25.
The following table lists the different policy events affecting smoking. In Dr. Chipty's
analysis each of the events leads to an indicator variable with a value of 0 before the
event and 1 after the policy event. For Dr. Chipty's report, that ends with data from
September 2015, the analysis includes four overlapping 0-1 indicator variables beginning
in 2010: from 2010 through 2015 for excise taxes, from 2012 through 2015 for the 2012
Packaging Changes, from 2013 through 2015 for excise taxes, and from 2014 through
2015 for excise taxes. Given the overlapping nature of the construction of her variables
27
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0090.png
that fall just short of having an indicator for every year, any conclusions drawn from her
analysis are highly speculative. Dr. Chipty’s procedure is not wrong from a statistical
analysis standpoint, but the use of multiple time period indicators provides very limited
insight into the separate effects of the 2012 Packaging Changes policy. In particular,
there is only a single year in which the 2012 Packaging Changes indicator variable’s
effect is not also confounded with other policy shifts.
Moreover, any lag time in
behavioral responses to policy changes due, for example, to the difficulty of quitting
cigarettes, will tend to lead to smoking prevalence shifts from the earlier excise tax
increases that extend over multiple periods, thus contaminating the purported effect
associated with subsequent time period indicators.
Relevant Policy Changes During Period of Dr. Chipty's Analysis
Year
2006
2010
2012
2013
2014
26.
Policy
Graphic warning labels on cigarette packages (Jan. 1)
Tax increase, 25% per pack (May)
Plain packaging of cigarette packs and increase in graphic health warnings
from 30% to 75% of the front of pack (Oct. 1 begin / Dec. 1 full)
Tax increase, 12.5% per pack (Dec.)
Tax increase, 12.5% per pack (Sept.)
The policy change summary above and the chart below indicate the year overlaps and the
difficulties they create. Excise tax increases in 2010 may have an impact in reducing
smoking cessation rates in 2012, creating a confounded effect along with any impact of
the 2012 Packaging Changes and the possibility of attributing an association with the
2012 Packaging Changes where there isn't one. Similarly, the influence of the 2012
Packaging Changes that is captured with an indicator variable starting in December 2012
28
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0091.png
spans a period that includes subsequent cigarette excise tax increases. The only year in
which the 2012 Packaging Changes are introduced but no new excise tax measure is
introduced is 2012.
26
The overlapping nature of Dr Chipty's indicator variables leads to
results that suggest statistically that the analysis may be capturing general time trends due
to the impact of excise taxes rather than effects correlated with the role of the 2012
Packaging Changes.
Dr. Chipty’s Indicator Variables Specification
Year
Policy
Excise Tax 2010
2012 Packaging
Changes
Excise Tax 2013
Excise Tax 2014
2009
0
0
0
0
2010
1
0
0
0
2011
1
0
0
0
2012
1
1
0
0
2013
1
1
1
0
2014
1
1
1
1
2015
1
1
1
1
27.
Dr. Chipty’s use of indicator variables for the major tax increases is also a crude
empirical approach that generates a source of error in the treatment of taxes. The reliance
on the indicator variables fails to recognize the continuous nature of excise tax levels,
which are updated periodically for inflation. From 2001 through September 2015 (the
period of Dr Chipty's analysis), cigarette excise tax levels in Australia had 32 different
values.
27
My excise tax variable accounts for the level of excise taxes throughout my
26
27
More specifically, if the full implementation of the 2012 Packaging Changes was in December 2012, and
the 2013 excise tax was introduced in December 2013, the time period in which the 2012 Packaging
Changes alone is the incremental change is from December 2012 to November 2013. My analysis accounts
for the monthly policy changes, but for simplicity, the chart above focuses on years.
See table 13.2.3 of Scollo, M, Bayly, M. 13.2 "Tobacco taxes in Australia". In Scollo, MM and Winstanley,
MH [editors]. "Tobacco in Australia: Facts and issues". Melbourne: Cancer Council Victoria; 2016.
29
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0092.png
estimation period, more accurately characterizes the tax rate than simply identifying the
major tax increase periods, and avoids the use of overlapping indicator variables.
28.
The chart below also indicates the nature of Dr. Chipty's analysis and its fundamental
shortcomings with respect to her analysis of temporal factors. As illustrated, there is a
pronounced decline over time in smoking prevalence rates. Smoking prevalence rates
over time may be affected by factors other than those explicitly captured by variables in a
regression model. Progressive changes in public space smoking restrictions, differences
across different population cohorts in attitudes toward smoking, and the rising role of
vaping as an alternative to smoking are among the time-related variables that may not be
accounted for in a regression model. To incorporate the role of omitted temporal factors,
researchers may include variables reflecting the time period. However, there is no
theoretical basis for assuming a particular temporal relationship as Dr. Chipty has done,
as it might be linear or nonlinear. The proper form is an empirical question which Dr.
Chipty doesn't consider. In estimates reported in Appendix A, I report multivariate
regression equations including both time and time squared in the analysis.
28
The
statistically significant coefficient on the time squared variable is the statistical test that
shows that the temporal trend in smoking prevalence rates is consistent with the time
trend being nonlinear. Contrary to Dr. Chipty’s analysis, one can reject the hypothesis
that the trend is linear.
29.
Using data from the pre-2012 Packaging Changes period, I have fitted a nonlinear
temporal relationship to the data indicated by the gray curve, focusing solely on the
relationship between smoking propensities and time as well as time squared. The curve
28
Thus, for example, for the first month of data in the RMSS sample the value of Time would be 1, in the
second month it would be 2, the third month would be 3, etc. The value of Time squared is just the square
of this value, or 1, 4, 9, etc.
30
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0093.png
fitting the pre-2012 Packaging Changes data generates the illustrated post-2012
Packaging Changes projections that track the overall trend nicely.
29
Thus, there is no
discontinuous shift in the trajectory of smoking prevalence rates once one takes into
account the nonlinear trend in smoking prevalence rates that already existed before the
advent of the plain packs policy. My conclusion that there is no discontinuous shift in
smoking prevalence rates that took place in 2012 is also consistent with findings by other
researchers using a different statistical approach.
30
30.
A graphical analysis indicates the mistaken conclusions that will be generated by
assuming that an underlying nonlinear trend is linear. Imposition of a linearity
assumption when the underlying trend is nonlinear will always result in the projected
values of smoking rates exceeding the actual future levels. This phenomenon can be
illustrated using a variety of different time periods which, according to Dr. Chipty’s
logic, would indicate a shift in smoking prevalence rates, whereas in fact no shift has
occurred, only a continuation of the underlying nonlinear trend.
29
30
The equation is a regression of the 0-1 smoking prevalence variable on a constant term, time in months, and
time squared in months, where the smoking probability = 0.24 – 1.32E-4 time – 1.76E-6 time squared,
where all coefficients are statistically significant at the 0.05 level or better.
Lilico, A. (2016). "Analysis of the Chipty Report’s Conclusions Regarding Packaging Changes and
Smoking Prevalence in Australia", August 30, 2016, available at
http://www.jti.com/about-tobacco/key-
regulatory-submissions/.
31
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0094.png
25%
24%
23%
2012
Smoking Prevalence
22%
21%
20%
19%
18%
17%
16%
15%
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
Date
31.
Dr. Chipty's analysis assumes a linear trend indicated by the blue line in the chart, which
she estimated for the time period up to the 2012 Packaging Changes. As her analysis
points out, this linear trend line does not predict the post- 2012 Packaging Changes trend,
which exhibits an additional drop after the advent of the 2012 Packaging Changes. She
then attributes the unexplained drop in smoking prevalence rates to the impact of the
2012 Packaging Changes. That there is a shift if one assumes that trends must be linear is
illustrated by the green line, which is Dr. Chipty’s linear trend line using only post-2012
Packaging Changes data. The blue line and the green line have clearly different slopes,
indicating a purported drop in smoking prevalence rates associated with the 2012
Packaging Changes. However, one can generate other possible purported policy effects
32
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0095.png
for any time period in the chart since the underlying trend is nonlinear. Similar linear
trend lines such as the red trend line based on the first 3 years of data indicate an
unpredicted drop in smoking prevalence rates thereafter, as does the purple linear trend
line based on the first 6 years of data, and the 9-year trend line in yellow. The
“unexplained” departures from the trend lines occur because a linear trend line does not
properly capture the nonlinear trend. Capturing any nonlinear relationship with a linear
trend line as Dr. Chipty does, will lead to an unexplained subsequent apparent “policy
impact” that is spurious, as it is not an effect associated with the policy but instead
reflects an underlying nonlinear trend.
32.
While there are multiple factors that could contribute to such a nonlinear trend such as
progressive restrictions on public smoking and cohort effects as the population changes
over time, an additional influence is the rising cost of cigarettes. The figure below
indicates the recommended retail price trajectory for Craven 20 cigarettes, which has
undergone a steep nonlinear increase, reflecting in part the influence of multiple boosts in
the excise tax rate. The nonlinear nature of the price hikes is consistent with the nonlinear
decline of smoking prevalence rates. Any increases in the non-monetary costs of
smoking, such as the convenience costs arising from smoking restrictions, would
reinforce such influences.
33
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0096.png
Price of a Pack of Craven 20 Cigarettes Over Time
33.
Three principal results from the RMSS data analysis are apparent in both the analysis of
the extended dataset that I used as well as in my analysis of the shorter time period
considered in Dr. Chipty’s report. First, if one properly recognizes that the temporal trend
is nonlinear rather than linear (i.e., by including a quadratic time trend term), that
reformulation alone accounts for the downward trend in smoking rates without there
being any additional downward shift associated with the 2012 Packaging Changes.
Second, even with only a linear trend, reasonable specifications of the model using either
overall consumer prices or continuous measures of the recommended retail price of
cigarettes rather than the excise tax indicator variables approach used by Dr. Chipty
eliminates the statistical significance of the 2012 Packaging Changes variable. Third,
even if there were a purported association of the 2012 Packaging Changes with the
34
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0097.png
decline in smoking prevalence rates based on Dr. Chipty's analysis, one should be
skeptical of the import of these results. Her statistical analysis ignores the multiple
changes in excise tax rates and includes four overlapping indicator variables for the 2010
to 2015 period in which there is only a single year in which the 2012 Packaging Changes
are introduced without any other new smoking policies also being introduced. Attributing
the decline in smoking prevalence rates to the 2012 Packaging Changes as opposed to the
excise tax increases, both before and after the advent of the 2012 Packaging Changes
policy, is not warranted. While the final concern may reflect a difference in statistical
approaches, the first two shortcomings are fundamental. Dr. Chipty’s failure to consider
the pivotal role of prices on smoking behavior and the underlying nonlinear trend in
smoking prevalence rates have generated the mistaken conclusion that the 2012
Packaging Changes are associated with an unexplained drop in smoking prevalence rates.
ii.
34.
Diethelm and Farley (2015)
31
The article by Diethelm and Farley (2015) analyzed the average monthly smoking
prevalence rates in the RMSS data which they derive from a research paper by Kaul and
Wolf.
32
They did not use any original RMSS data but instead relied on visual inspection
of the figure in Kaul and Wolf to impute 156 monthly smoking prevalence rates. Because
they have no individual respondent data, their analysis includes no demographic
variables. As a result, changes in sample composition over time, such as the age and
income level of respondents, are ignored. Similarly, there are no controls for different
31
32
Pascal A Diethelm, Timothy M Farley, “Refuting tobacco-industry funded research: empirical data shows a
decline in smoking prevalence following the introduction of plain packaging in Australia,” Tob. Prev.
Cessation 2015;1(November):6
http://dx.doi.org/10.18332/tpc/60650.
Kaul, A. and Wolf, M. "The (Possible) Effect of Plain Packaging on Smoking Prevalence in Australia: A
Trend Analysis" University of Zurich Department of Economics Working Paper, June 2014.
35
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0098.png
Australian state territories so that the mix of the sample across states and any state-
specific smoking-related policies are not taken into account and may be incorrectly
attributed to the 2012 Packaging Changes. The only explanatory variables included in
the regression analysis reported in the paper are a linear time trend, an indicator variable
for the 2010 excise tax increase, an indicator variable for smoke-free policies, and a plain
packaging indicator variable. Even setting aside the omitted demographic and regional
variables, the equation is a very 'bare-bones' specification. The many other excise tax
changes, cigarette prices, and the nonlinear smoking prevalence trend are all omitted.
The recent Cochrane Review notes the nonlinearity in the smoking prevalence trend but
claims that the “additional covariates” in Diethelm and Farley (2015) address the
nonlinearity. This claim is simply not true since their specification includes only two
policy indicator variables and no variables that capture the evident nonlinearity in the
trend apart from these shifts.
35.
The impact of the limitations and flaws in the Diethelm and Farley (2015) paper is that
this study lacks any scientific credibility. The deficiencies I cite above are not minor
limitations or matters of a difference of opinion, but are fundamental problems that make
it inappropriate to rely on their study. The paper has no scientific merit at all.
36
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0099.png
IV.
A.
36.
IMPLICATIONS OF THE CITTS DATA
Overview of the CITTS Data
The CITTS data that I analyze consists of cross-sectional telephone data utilizing a
sample of 17,468 adult smokers and recent quitters (people who quit smoking in the past
12 months). The CITTS data (and the NTPPTS data) only include smokers and recent
quitters, so they cannot be used to assess the effect of the 2012 Packaging Changes on
smoking prevalence. However, these data provide valuable insights into smoking-related
behaviors of smokers and recent quitters, which is clearly an important target group of
the policy. The survey also includes a number of variables relating to the perceptions of
the 2012 Packaging Changes by this group even setting aside issues pertaining to the
efficacy of these ‘intermediate’ variables in predicting actual smoking behaviors. The
CITTS is undertaken in New South Wales, which is the most populous state in Australia.
Further details of this survey are provided in the paper by Dunlop et al. (2014).
33
The
sample analyzed by Dunlop et al. focused on smokers interviewed between April 2006
and May 2013, which includes only six months of data post the implementation of PP in
Australia. My sample through June 2016 adds an additional 37 months and 2,045
observations to the sample, which takes it through June 2016 (i.e., 3 1/2 years of data post
the implementation of the 2012 Packaging Changes in Australia).
34
Accordingly, the
analysis I present here is the most up to date analysis of this dataset.
33
Sally M. Dunlop, Timothy Dobbins, Jane M. Young, Donna Perez, and David C. Currow, “Impact of
Responses: Results from a Continuous Tracking Survey,"
BMJ Open
2014; 4(12): e005836;
As noted above, I have previously addressed the CITTS data in reports that I submitted for British
American Tobacco in October 2015 in UK legal proceedings in which the 2012 packaging change
regulation was being challenged, and in November 2015 in relation to the Australian Government
Department of Health Post-Implementation Review (PIR) of the Tobacco Plain Packaging Act 2011 (TPP).
In this report, in addition to using a larger sample, I also undertake new analyses using more detailed
37
34
L 61 - 2020-21 - Bilag 1: Høringssvar og høringsnotat, fra sundheds- og ældreministeren
2259708_0100.png
37.
The 2012 Packaging Changes took full effect in Australia on December 1, 2012. October
and November of that year were a transition period in which some packs began to
conform to the 2012 packaging change format. To streamline the exposition below, I only
report results in the main body of the report using the December 1, 2012 starting date for
the 2012 Packaging Changes. Results reported in Appendix B indicate that the results are
stable using other policy starting dates.
38.
The sampling procedure for the CITTS survey changed in 2013, as recruitment of
respondents changed to include mobile phone users instead of only landline users.
35
As a
result, my analysis of the CITTS data draws on the findings using multiple regression
analyses in which there is a statistical control for the mobile phone recruitments in the
sample as well as detailed set of demographic variables.
39.
Many of the CITTS data questions are in the form of qualitative Likert rating scales in
which some measures are rated on a scale from 1 to 5 (strongly disagree, somewhat
disagree, neither, somewhat agree, strongly agree), where 1 equals strongly disagree and
5 equals strongly agree. These qualitative scales do not provide a basis that permits a
comparison either within or across people since the cut-off between these categories will
vary across individuals and across questions so that the distinctions are not very
meaningful. For example, there is no way of knowing whether person A’s score of a 4 for
pack attractiveness implies a lower or higher level of attractiveness than person B’s score
of 3. Similarly, we cannot tell if a drop of a score from 4 to 2 is twice the size as a
decrease from 4 to 3. Accordingly, focusing on only one set of extreme responses, such
as shifts in the “strongly agree” category, will distort the assessment of the implications
controls than in my previous assessments. In my previous reports I controlled for cell phone usage and
time trends, but the current report also includes controls for a detailed demographic variables.
Cancer Institute NSW’s Tobacco Tracking Survey (CITTS) Research Plan 2013-2014 at page 11.
38
35