«Patrick Basham IEA Discussion Paper No. 29 July 2010 Institute of Economic Affairs 2 Lord North Street London SW1P 3LB IEA web ...»
Canada's ruinous tobacco display ban:
economic and public health lessons
IEA Discussion Paper No. 29
Institute of Economic Affairs
2 Lord North Street
IEA web publications are designed to promote discussion on economic issues and the role of markets in solving
economic and social problems. Copyright remains with the author. If you would like to contact the author, in the first
instance please contact firstname.lastname@example.org. As with all IEA publications, the views expressed in IEA web publications are those of the author and not those of the Institute (which has no corporate view), its managing trustees, Academic Advisory Council or senior staff.
Canada's ruinous tobacco display ban:
economic and public health lessons Patrick Basham Introduction The critics of tobacco retail displays claim they help to initiate smoking in adolescents and thwart the efforts of smokers to stop smoking. Removing such displays, argue the proponents of display bans, will therefore reduce smoking initiation in the young and increase the likelihood of success of smokers trying to quit, while barely impacting the independent retail sector. This paper assesses the validity of these claims in the context of the international experience, especially in Canada, with display bans.
Accordingly, this paper reviews the empirical evidence about the public health effectiveness and the economic impact of display bans in Canada. While the Canadian public health story is a depressing one, the paper also details what the display ban has done ‘successfully’ in Canada, namely driving the illegal market and decimating the independent retail sector.
The paper then provides a comparative analysis of the display ban data in Iceland, Ireland, and Thailand. With Canada and Iceland, the paper reviews the evidence cited by the UK Department of Health in its 2008 Consultation on the Future of Tobacco Control. With Ireland and Thailand, the paper examines the most recent smoking prevalence data and compares this with prevalence prior to the introduction of their respective display bans.
Public health impact There are three types of evidence in support of the public health claims about display bans. The first type is made up of experimental studies in which individuals are shown simulated tobacco displays and then asked questions about their reactions to the displays. The second type consists of what are called recall and recognition studies in which individuals are asked about their recall of cigarette brands sold in shops and/or their memories of tobacco displays and this information is correlated with their likelihood of being or becoming a smoker. The final type of evidence comes from jurisdictions that have implemented display bans. It typically consists of pre- and post-ban prevalence, consumption, smoking initiation, and smoking cessation data in order to determine what effect, if any, the display has had.
During last year’s UK debate on tobacco displays, each of these three sorts of evidence played a role. The champions of display bans argued that the experimental evidence, the results of recall and recognition studies, and especially the evidence from countries such as Canada, which had implemented display bans, all pointed to the role that tobacco displays played in smoking initiation and cessation, as well as to their effectiveness in reducing smoking initiation among the young, reducing overall prevalence and consumption, and encouraging cessation.
On the other hand, tobacco retailers and the tobacco industry claimed that the experimental, recall, and recognition studies were severely flawed in their design and provided little reliable evidence that tobacco displays played a significant role in the uptake or continuance of smoking. In addition, they argued that the evidence from countries like Canada failed to show that display bans had any effect on smoking initiation, prevalence, consumption, or quitting.
The author has argued previously that the above studies fail to provide any compelling evidence that tobacco displays had a significant role in smoking.1 There is, of course, some room for disagreement about just what, if anything, the experimental and recall and recognition studies reveal. However, the real world evidence – effectively a ‘natural experiment’ – from jurisdictions such as Canada that have implemented display bans is far more definitive in determining whether tobacco displays affect smoking behaviour in the ways their critics claim.
Indeed, as with many tobacco control policies, changes in smoking initiation, prevalence, consumption, and cessation provide a legitimate measure of whether such policies work. If countries that have instituted display bans fail to see statistically significant changes in smoking initiation, prevalence, consumption, and cessation, then the claim that tobacco displays have a causal relationship with these outcomes is significantly weakened. Hence, the results of these natural experiments in policy are crucial for determining whether display bans are effective in changing smoking behaviour and therefore constitute a justifiable instance of tobacco regulation.
Impact upon Canadian public health
At the time of the 2009 UK parliamentary debate on the display ban legislation, there was considerable evidence presented about whether display bans in Canada (the second country in the world to institute a display ban) and countries such as Iceland (the first country to institute a ban) were effective since bans in these countries had been in place long enough to provide considerable data. Iceland had introduced its display ban in August 2001 and the Canadian province of Saskatchewan in 2002.
In an autumn 2008 report on the effectiveness of tobacco display bans, the BBC’s Peter Jackson claimed that tobacco display bans in respective Canadian provinces made a significant difference in smoking behaviour.2 According to Jackson, ‘smoking rates among 15-19 year olds fell from 29% in 2002 to 19% in 2007’, five years after the first display ban was instituted in Saskatchewan. As of the time of writing this paper, all Patrick Basham and John Luik, Hidden in Plain Sight: Why Tobacco Display Bans Fail, Democracy Institute: London 2009.
Peter Jackson, ‘Smoking ban lessons from abroad’, BBC News, 12 September 2008.
provinces have prohibited tobacco displays with Newfoundland being the most recent to pass legislation in May 2009 taking effect on 1 January 2010.
Similar claims about the positive impact of display bans are regularly made by Action on Smoking and Health (ASH) UK, a lobby group heavily funded by the British taxpayer.
Martin Dockrell of ASH told the BBC that the experience of Saskatchewan showed how
effective display bans were in reducing smoking. The BBC report explains:
As the following sections demonstrate, this account seriously misrepresents, to put it politely, the experience of Saskatchewan’s display ban.
In its 2008 Consultation on the Future of Tobacco Control, the Department of Health was not nearly as certain as ASH about the effectiveness of any Canadian display ban.
At the time, the Department of Health noted:
Table 1 (below) reports the most recent data from Canada based on the Canadian Tobacco Use Monitoring Survey (CTUMS) 1999-2008, produced by Statistics Canada for Health Canada. The data is reported by province.
Department of Health, Consultation on the Future of Tobacco Control, London: 2008: 34.
Department of Health ‘Display Ban Impact Assessment’, Consultation on the Future of Tobacco Control, London: 2008: 76.
Table 1: Canadian smoking prevalence ages 15-19 by province
Source: CTUMS 1999-2008, Statistics Canada, Ottawa.
Prevalence rates vary considerably year by year. For example, Alberta’s prevalence went from 15% in 2006 to 20% in 2007, while British Columbia’s increased from 9% in 2007 to 15% in 2008.
In 2008, all Canadian provinces except for New Brunswick had display bans. The display bans in the provinces of Alberta, British Columbia, Ontario, and Quebec bans came into force in 2008. According to the CTUMS data, there has been a decline, albeit with some reversals, in youth prevalence across Canada from 1999 to 2008.
In 2006, when only two provinces, Manitoba and Saskatchewan, had display bans, there were no statistically significant differences in youth prevalence rates between those provinces with display bans and those without. The provincial rates were all within (+/-) 5% of the national average smoking rate. However, the two provinces with the highest youth prevalence were the two provinces with display bans.
The cross sectional data shows that display bans have no statistically significant effect on youth prevalence. In Prince Edward Island and Nova Scotia, youth prevalence was 13% in 2007 and 14% in 2008, respectively. In Saskatchewan, it was 21% in 2006, 22% in 2007, and 20% in 2008. In Manitoba, it was 20% in 2006, 20% in 2007, and 17% in 2008.
Adult prevalence is equally unaffected by display bans. For example, adult male prevalence in Prince Edward Island was 22% in 2006 and 21.3% in 2008.
As Figure 1 below illustrates, of these provinces with display bans in effect before 2008, the data show that display bans have no statistically significant effect on youth consumption.
Figure 1: Youth daily consumption in three display ban provinces
Source: CTUMS 1999-2008, Statistics Canada, Ottawa.
In Prince Edward Island, youth consumption averaged 13.8 cigarettes per day in 2006 and remained at an average 13.8 cigarettes per day in 2008. In Manitoba it was 11.7 per day in 2006 and 11.6 in 2008. In Saskatchewan it was 11.5 cigarettes per day in 2006 and 11.4 per day in 2008.
Tellingly, however, as Figure 2 highlights, in Nova Scotia youth consumption actually rose from 12.6 cigarettes per day in 2006 to 14.5 per day in 2008 – a 15% increase.
In a report commissioned by Japan Tobacco International, Dr Andrew Lilico of Europe Economics, a consultancy, argues that a cross sectional analysis of the youth and adult prevalence data and a time series analysis comparing prevalence before and after the implementation of display bans in the relevant provinces shows that, ‘for both age groups, the introduction of a display ban is associated, statistically, with a rise in smoking prevalence…For both age groups, the average number of cigarettes consumed is unchanged’.6 Dr Andrew Lilico, The Impacts of Restrictions on the Display of Tobacco Products: A Supplemental Report, Europe Economics, October 2009: 17-18.
Figure 2: Youth daily consumption in display ban Nova Scotia 14.5 13.5 12.5 11.5
Source: CTUMS 1999-2008, Statistics Canada, Ottawa.
Though it might be objected that the Europe Economics report was produced for a tobacco company and therefore inherently suspect, my own statistical analysis (which was not supported by the tobacco industry) of the same data, essentially confirms Dr Lilico’s analysis. Dr Lilico’s overall statistical results are both robust and replicable.
In other words, the Canadian data do not support the claim that a display ban was effective in ‘cutting smoking’ either by young people or adults. Instead, they show that display bans are associated with increased prevalence for both young people and adults in Canada and no decline in consumption.
In answer to the problems cited by the Department of Health’s Consultation Impact Assessment, namely that the data covered only a small number of time periods and its claim that the ‘evidence on teenage smoking in Canada is largely inconclusive’, the new Canadian data now cover a larger time period and allows for a far more robust conclusion about the effectiveness of display bans.
The Department of Health’s Consultation Impact Assessment noted that the data on youth smoking in Canada ‘does not control for other factors affecting smoking prevalence.’ This is a common problem in many studies that evaluate the effectiveness of tobacco control policies. Since smoking initiation, prevalence, and consumption all have multiple determinants, it is impossible to draw causal conclusions without controlling for these determinants.
Obviously, the cross sectional data reported in the CTUMS survey does not do this. An analysis of the data taking into account all of the factors affecting smoking initiation, prevalence, and consumption in both youth and adult populations in Canada across four years is beyond the scope of this paper. However, Lilico does undertake a limited analysis using the standard statistical estimation technique known as the ‘first difference fixed effects model’, which is widely used to assess the effectiveness of a variety of regulatory measures.
In this analysis, he looks at the effect that cigarette prices, Gross Domestic Product per capita, and the display ban might have on prevalence and consumption for adults and those aged 15-19.7 For Lilico, the data analysis suggests the following three
1. ‘There is no correlation of changes in price, the presence…or introduction of the display ban with changes in the average number of cigarettes consumed by the 15-19 age group.’ 2. ‘The presence of the display ban has no statistical correlation with changes in prevalence…for the general population.’ 3. ‘The display ban is strongly and materially correlated with increased prevalence amongst 15-19 year olds. Specifically, where the display ban is present, smoking prevalence increases by 2 percentage points.’8 Based on a careful reading of Canadian Government data, it is readily apparent display bans have not resulted in a change in prevalence or consumption in young people or adults in Canada to the end of 2008.
Furthermore, based on a more sophisticated statistical analysis, there is no correlation between display bans and changes in prevalence or consumption for 15-19 year olds in Canada and between display bans and changes in prevalence for adults in Canada.
Further the display ban is associated with increased prevalence for 15-19 year olds.