Impact of tobacco tax increases and industry pricing on smoking behaviours and inequalities: a mixed-methods study

Health and economic impacts of tobacco

Smoking remains the leading preventable cause of death, disease and health inequalities in the UK. 1 Over 100,000 people in the UK are killed by smoking every year. 2 Smokers lose an average of 10 years of life expectancy, and exposure to other people’s smoke also causes death and disease among non-smokers. In England, the annual cost to the economy is estimated to exceed £11B a year, including cost to the NHS, employers and to the wider economy. 3 In 2017, 14.9% of the adult population in England were current smokers, but this figure masks large disparities in smoking across socioeconomic groups (e.g. 26% smoking prevalence among routine and manual vs. 10% among managerial and professional occupations). 4 Any public health intervention that reduces tobacco use or prevents uptake will lead to significant benefits by reducing premature mortality and morbidity and, consequently, lower NHS costs and increased economic output.

The importance of tobacco tax and price

Tobacco tax increases have been shown to be the most effective and cheapest way of reducing tobacco smoking prevalence,5–13 consumption,14,15 initiation16,17 and inequalities in smoking. 18–21 Tobacco tax has been identified by the World Health Organization as a critically important tobacco control intervention22 and the UK has real tobacco prices that are among the highest in the world. 23,24 The public health impact of tobacco taxation can, however, be reduced by the availability of cheap tobacco. The tobacco industry (TI) is aware of the importance of price and its internal documents show that pricing and price promotions are among its most important marketing tools25–28 in order to reduce the effectiveness of tobacco tax increases. 29,30 Hence, research needs to assess the extent to which the effectiveness of tobacco tax increases is influenced by TI pricing strategies and the effect this has on smokers’ behaviour and socioeconomic inequalities in smoking.

The UK tobacco industry

The main tobacco companies involved in the UK market are Imperial Tobacco (now Imperial Brands) and Japan Tobacco International (JTI). 31 In 2008, JTI became a major player in the UK market through purchasing Gallaher. British American Tobacco (BAT) and Philip Morris, also known as Philip Morris International (PMI), have smaller shares. 31 PMI’s main contribution to the UK market is its iconic brand Marlboro (Philip Morris, New York, NY, USA). Scandinavian Tobacco Group (STG) (Copenhagen, Denmark) was previously involved in the UK through its distribution of Natural American Spirit [chiefly roll your own (RYO)], and currently via sales of cigars and cigarillos. The annual profits generated by the UK TI have been estimated to be £1.5B in 2012 and 2013, with profit margins estimated at approximately 50%. 31

Tobacco industry segmentation and pricing strategies

The TI offers a variety of different tobacco products, including factory-made (FM) or manufactured cigarettes, RYO tobacco, cigars, cigarillos and waterpipes. FM cigarettes and RYO tobacco are by far the most popular tobacco products, particularly in the UK. For each tobacco product, the industry offers a number of different brands and it also segments the market into different price categories. Our previous work reviewed the industry literature and sales data on FM segments and identified four price segments: premium, mid-price, economy and ultra-low-price (ULP) brands. 32

When tax increases are announced in the Budget each year, tobacco companies can do one of three things: (1) increase prices on top of tax increases, so that both the price and tax increase are passed on to consumers (known as overshifting), (2) absorb the tax increase so it is not passed on to consumers (undershifting) or (3) pass the tax increase on to consumers in full (fully shifting). 33 The extent of ‘tax pass-through’ to consumers will reflect economic conditions, competition in the market and the extent to which manufacturers are willing to adjust their profits in order to absorb some of the tax increases.

We previously revealed that, between 2000 and 2009 in the UK, the TI undershifted taxes and cut prices on cigarettes in the cheapest segments to keep them cheap, while overshifting taxes on more expensive cigarettes to maximise profits. 32,34,35 Approximately, only half of the price increases were needed to meet tobacco taxation increases, consistent with evidence from Ireland between 2000 and 2010. 36 The range between the cheapest and most expensive cigarettes (the ‘price gap’) therefore increased. Similar TI pricing strategies have now been confirmed in other jurisdictions,37–39 where tobacco companies have also been introducing increasing numbers of lower priced products. 37–39 Smokers are therefore incentivised to down-trade from expensive to cheap cigarettes, rather than quitting smoking.

UK smokers’ behavioural responses to price increases

When faced with a tobacco tax and price increase, smokers may do nothing, quit, reduce consumption or engage in price-minimising behaviours by switching to one of a number of cheaper sources of tobacco. This might include:

• cheaper legal products (e.g. cheaper manufactured/FM and/or RYO cigarettes12,40–44)

• tax avoidance – travellers returning from other European Union (EU) countries (which typically have lower tobacco duties) can legally purchase tobacco for personal use in line with guidelines for reasonable personal use,45 or smokers can purchase tobacco in duty-free shops46

• tax evasion – changing to illicit sources,47 including purchases of a genuine TI product that has entered the illicit market or of counterfeit or ‘illicit whites’

• purchasing larger quantities which afford economies by buying in bulk (e.g. by the carton, rather than the pack). 40,41

The growing availability of legal cheap cigarettes in the UK is therefore a major public health concern, as it provides options to smokers other than changing their smoking behaviour. If this differentially affected smokers in lower socioeconomic groups, it would drive smoking inequalities further. 32,34,35 The use of cheaper tobacco products has indeed increased, particularly by the poorest. 10,12,42–44,48 The same concerns apply to tobacco tax evasion, also used more frequently by lower socioeconomic groups. 49–52 By contrast, tax avoidance is more common among higher socioeconomic groups consistent with their greater propensity to travel. 46 However, self-reported tax avoidance and evasion by UK smokers declined from 2002 to 2011,53 in contrast to TI arguments that they would increase in response to higher tobacco taxes (see Illicit tobacco: tobacco industry complicity and UK government strategies).

It is, however, unclear whether the decline in use of expensive FM cigarettes is due to a decline in uptake, users quitting or trading down to cheaper products. 12 Evidence indicates that the availability of cheap tobacco reduces motivation to quit and quit success. 54–59 Our work, to date, had used repeat cross-sectional data to examine cheap cigarette use in the UK and could therefore not directly determine the impact that tax and industry price changes had on smokers’ behaviour. Two US studies40,60 found mixed results: one60 found smokers using cheaper FM cigarettes were less likely than those using premium brands to quit or reduce smoking between 1988 and 1993, and the second, more recent, study40 found no direct association between use of cheap FM brands or RYO tobacco in 2009 and quitting outcomes in 2010. Evidence from the International Tobacco Control Policy Evaluation Project (ITC), using data prior to 2008, across four countries (Canada, Australia, the USA and the UK), found a lower likelihood of successful quitting among smokers of cheaper compared with more expensive FM cigarettes. 56 However, in comparison with these other countries, the use of cheap tobacco has been found to be significantly higher in the UK. 57

Factors that might influence smokers’ inclinations to switch to cheaper products rather than quitting also remain largely unexplored. For instance, smokers who are brand-loyal,61,62 or loyal to a specific product type (FM cigarettes or RYO tobacco, rather than smoking both63) might be less likely to switch to a cheaper product. To the best of our knowledge, existing research has provided only limited insights into the purchasing patterns among UK smokers of tobacco on which UK tobacco duties have not been paid,46 and no studies have examined trends in tax evasion separately from tax avoidance, or how these purchasing patterns might change in response to tobacco tax increases.

Affordability

The impact of price rises is also modified by changes in incomes and inflation rates, and, hence, it is important to measure affordability (i.e. smokers’ purchasing power for tobacco in relation to their real incomes). There are many different types of ‘aggregate’ tobacco affordability measures: the ‘Big Mac^® index’ representing the number of cigarettes purchased for the price of one McDonald’s Big Mac hamburger (McDonald’s Corporation, San Bernardino, CA, USA);64 the ‘minutes of labour’ needed to purchase a pack of 20 Marlboro cigarettes or an equivalent local brand;65 the ‘relative income price’ (RIP), representing the percentage of per capita gross domestic product (GDP) required to purchase 100 packs of cigarettes (Marlboro or local brand);66 and the cigarette price–daily income ratio, which divides the price of one pack of cigarettes (Marlboro or local brand) by daily income. 67 These measures have different strengths and weaknesses and their merits, particularly when compared across high-, middle- and low-income countries, have been discussed elsewhere. 68

Studies using these methods indicate that cigarettes in the UK have became less affordable since the 1990s, although the rate of decline in affordability appears to be slowing. Between 1991 and 2002 the annual decrease in tobacco affordability was around 5.5%,65 but only around 2–3% between 2003 and 2009. 69 Similarly, estimates using the RIP indicate an annual decrease in affordability of around 2.5–3% between 1990 and 2001,66 but only 1–2% between 2004 and 2010. 70 No studies have examined tobacco affordability in the UK since 2010.

A major limitation of aggregate affordability measures is that they rely on average cigarette prices, usually derived from a small number of brands, and on average national income estimates. However, there are wide income inequalities observed within many countries, including the UK, and, as shown in Tobacco industry segmentation and pricing strategies, a large range of prices between and within different tobacco products. In addition, aggregate measures do not capture the strategies that smokers use to minimise costs, such as buying from cheaper sources or buying in bulk.

Illicit tobacco: tobacco industry complicity and UK government strategies

There is overwhelming evidence of the TI’s historical involvement in the global illicit tobacco trade. 47,71–73 In the UK in the 1990s, tobacco companies were accused of facilitating smuggling by deliberately oversupplying their brands to countries with no demand for them. 74 Although the nature of the illicit tobacco market has since changed substantially and the TI now claims that it has addressed the problem, evidence suggests that the TI has continued to be involved in illicit trade and failed to control their supply chain,73,75,76 despite signing legal agreements to address both these issues. 47 In the UK, Her Majesty’s Revenue and Customs (HMRC) estimated that in 2011 the aggregate supply of certain brands of RYO tobacco to some countries exceeded legitimate demand by 240%. 77 Similarly, massive TI overproduction of cigarettes in the Ukraine has been shown to fuel the illicit market in Europe,78 a finding supported by our pan-European survey, which showed that illicit tobacco use was greatest for those living in countries that shared a land or sea border with the Ukraine, Russia, Moldova or Belarus. 52 Furthermore, PMI’s own data suggest that in 2010 around one-quarter of illicit cigarettes in Europe were genuine PMI brands. 79

Despite this, the TI continues to use the threat of increases in illicit tobacco use to argue against key tobacco control policies, including tax increases. 33,79–81 Although these arguments are intuitive and price is one driver of illicit tobacco use, it is not the main driver; supply-side issues (such as levels of corruption and TI complicity) are more important and are key to controlling illicit trade. 82 Nevertheless, the industry has used these arguments extensively in the UK, despite evidence that much of the price increases is directly attributable to industry price increases rather than tax increases, no observed association between tobacco prices and illicit tobacco use across countries in Europe,52 and an observed overall decline in levels of illicit tobacco in the UK despite continued tax increases. 83

The UK government implemented a number of strategies to reduce illicit tobacco use during the study period. In 2000, the UK launched its first illicit tobacco strategy, including £201M of additional funding, new customs staff, freight scanners, stronger penalties for tobacco smuggling and new ‘UK duty paid’ markings on tobacco packs. Government co-operated with tobacco companies to reduce illicit tobacco and implemented a publicity campaign to raise awareness of illicit tobacco use dangers. 84 In 2006, 2011 and 2015, the strategy was updated and refreshed. A regional programme in the north of England, launched in 2009, also aimed to reduce demand for illicit tobacco. 85

UK government tobacco taxation

In the UK, three types of tax are applied to tobacco: (1) specific tax (a fixed amount per 1000 cigarettes or 1000 grams of RYO tobacco); (2) ad valorem tax (set as a percentage of the retail price and only applied to FM cigarettes); and (3) value-added tax (VAT) (another ad valorem tax applied to most goods and services and applied to both FM cigarettes and RYO tobacco). Compared with ad valorem taxes, specific taxes tend to narrow the price range between expensive and cheaper brands, maximise the impact of tobacco taxes24,86,87 and raise more revenue. 88

Since our previous work, UK tobacco taxes have increased annually, their structure has changed somewhat and the rate of VAT has varied [Table 1, and for further details see the fact sheet published by Action on Smoking and Health (ASH) England89 and UK Budget documents90–98]. Two significant changes occurred in March 2011. For FM cigarettes, there was a shift away from ad valorem taxes,99,100 which were reduced from 24% to 16.5% of the retail price. Instead, specific taxes on FM cigarettes were increased from £119 to £155 per 1000 sticks. Ad valorem taxes generally increase the price range between cheap and expensive products and this move was intended to narrow the price range between different FM brands. For RYO tobacco, there was a sudden increase of 10% above the Retail Price Index (RPI) in tax from £130 to £152 per kilogram. This was intended to narrow the price range between RYO and FM products. A third significant change occurred in March 2012, when the tax escalator (which was otherwise set between 0% and 2% above RPI throughout the study period) increased to 5% above RPI for 1 year. It should be noted that, although the RPI (a measure of private household spending from survey data) is still employed in relation to tobacco taxation changes,101 it was de-designated as a national statistic in March 2013 and replaced with the Consumer Price Index (CPI), which meets international standards and is comparable with other European countries. 90 The RPI, unlike the CPI, includes housing costs, is more volatile than the CPI and rises more quickly. 102 We used the CPI to adjust our FM and RYO prices for inflation (see Chapter 2).

European Union tax changes

Since the 1970s, tobacco product regulations have been introduced in European countries. 103 Two pieces of legislation were introduced in the last decade that are relevant to this study: EU Council Directives 2008/118/EC and 2011/64/EU, which define categories for tobacco and tax structures, including minimum tobacco tax rates. Of relevance here are cigarettes (Article 3), cigars and cigarillos (Article 4) and RYO tobacco (Article 5). The regulations require countries to apply a specific excise tax and an ad valorem tax on cigarettes, as well as a minimum excise duty. Countries have more flexibility for other tobacco product categories and can apply specific and/or ad valorem tax. Mandatory minimum tax rates are set out. 104 The European Commission also changed the price benchmark from the most popular price category (MPPC) to a weighted average price (WAP) in each member state in January 2011.

Roll-your-own pricing strategies

To the best of our knowledge, no studies had examined TI pricing of RYO tobacco in the UK, despite its growing market share. Furthermore, the work on FM cigarette pricing dated only to 2009. Up-to-date data on RYO tobacco and cigarette pricing were essential, because TI pricing strategies could evolve over time to reflect market, economic and policy developments, and should be considered when setting tobacco duties.

Tobacco industry pricing strategies and socioeconomic inequalities

There was limited evidence on the impact of TI pricing on levels of cheap tobacco use in the UK. Evidence was also limited on differences in smoking outcomes (including quit attempts and success) among those using expensive compared with cheap cigarettes and the extent to which this explains inequalities in smoking. Further research was therefore needed to examine how UK smokers respond to TI pricing strategies, including the extent to which they engage in price-minimising behaviours, reduce consumption or quit, and whether or not these responses vary by SES.

The prevalence of illicit tobacco use

Although the argument that the TI most frequently uses to prevent tobacco tax increases is the claim that they will lead to an increase in the illicit tobacco trade,105 to the best of our knowledge no study had directly examined links between tax increases and trends in tax evasion or avoidance by smokers in the UK. Furthermore, as the TI uses the illicit trade to argue against tobacco tax increases and other tobacco control measures using its own data,47,81,105 it is important that assessments of the level and nature of illicit use in the UK are conducted independently of the TI. The only recent, independent studies estimating illicit use in the UK covered only England, or regions therein, and did not provide a measure of uncertainty with the estimate. 49,50,52 Some also did not break the data down by FM and RYO. 49 Further independent studies were therefore essential.

Objectives

• Provide up-to-date knowledge of TI pricing and the extent to which this modifies the impact of tobacco taxation on public health, by examining:

  • how the TI segments manufactured cigarettes and RYO tobacco on price

  • the extent to which the TI undershifts, overshifts or fully shifts tobacco tax increases on to consumers, whether or not this varies by product, price segment and over time, and what proportion of price increases by segment is explained by TI price increases versus tax increases.

• To explore the impact of tobacco tax increases, as moderated by TI pricing strategies, on smokers’ behaviour, by examining:

  • the impact of the price range (the difference in price between the most and least expensive products) and price changes (annual tax increases modified via TI pricing) on quitting or switching between products or price segments

  • the impact of the price range and price changes on tobacco consumption

  • whether these behaviours differ by smokers’ previous product or price choices.

• To explore the impact of tobacco tax increases, as moderated by TI pricing strategies, on inequalities in smoking, by examining:

  • the characteristics (socioeconomic, geographic, etc.) of smokers using each product/price segment

  • whether or not behavioural choices (quitting, switching between price segments and reducing consumption) differ by smokers’ SES

  • the proportion of change in smoking inequalities over time attributable to cheap (legal) tobacco use.

• To explore whether or not cheap (legal) products are a means of market entry for the young, by examining:

  • the age of smokers using each product or price segment

  • trends in youth usage of cheap products

  • whether or not young people initiate smoking via cheap products, later upgrading to more expensive products.

• To increase understanding of trends in and the nature of tax avoidance and evasion, by examining:

  • trends in the proportion of smokers engaging in tax avoidance and evasion and their socioeconomic and other characteristics (where they acquire their tobacco and trends therein)

  • whether or not tax/price increases, particularly larger tax increases, are linked to tax avoidance and/or evasion

  • which products are most frequently acquired via tax avoidance and evasion, and from which sources.

• To synthesise findings and develop recommendations to improve the effectiveness of tobacco taxation as a public health intervention.

Once the project was under way, we realised that it was not possible to address the fourth objective [to explore whether or not cheap (legal) products are a means of market entry for the young], as insufficient numbers of young people were enrolled in the ITC. Instead, the ITC data were useful for supplementary analyses on tobacco affordability and weight of RYO cigarettes, which enhanced our understanding of the public health effectiveness of tobacco tax increases.

Sources

The commercial literature included TI documents, industry monitoring reports and trade magazines, from the years 2008 to 2014, with a few articles from early 2015 included when they were reviews of 2014. TI documents included annual reports from BAT, Imperial Tobacco, JTI and PMI, and presentations from BAT, Imperial Tobacco and PMI (JTI does not make presentations available). Industry monitoring reports were from Euromonitor International Ltd (London, UK), Mintel (London, UK) and Keynote (Google Inc., Mountain View, CA, USA), and for 2014 only from Bank of America (Charlotte, NC, USA)/Merrill Lynch (New York, NY, USA) (Imperial Tobacco bought some American brands from Reynolds American at this time). Trade magazines included Retail News (online as Talkingretail.com), The Grocer, Wholesale News and Tobacco Journal International (including the supplement Tobacco Profiles). Online and/or paper copies of trade magazines were searched, depending on availability. Additional trade magazines (2014 editions of Forecourt Trader, Convenience Store and Asian Trader), available from a related literature review for a previous project, were also searched. One article on Imperial Tobacco price segmentation from the Scottish Grocer, found via a post hoc Google (Google Inc., Mountain View, CA, USA) search for segment names, was also included.

Analyses

From articles mentioning tobacco products, we collated the following information in a database: segment name, tobacco company, brand name, pack size, price, price type [recommended retail price (RRP) or price marked], date referred to, date of article, source of article, title of article, reporter, geography and relevant quotations. The number of articles/reports extracted was 517. Each article was searched for mentions of brands and brand variants, alongside segment information and any text relevant to understanding segmentation. An article could mention one or several brands and brand variants and/or could provide one or more quotations regarding segmentation. This resulted in 1892 entries in the final main database, of which 1700 concerned 347 brand and brand variants. The remainder of the entries were articles that discussed price segmentation generally without referring to specific brands.

We also conducted a qualitative analysis to explore how the industry maintained profitability through brand segmentation, in the context of other price-limiting strategies, between 2008 and 2014. For this, 557 quotations about brand segments and tobacco pricing were collected from the commercial literature and analysed. The quotations were imported into NVivo 10 (QSR International, Warrington, UK) and coded at various nodes. There were two super-nodes: ‘market conditions that the TI was experiencing’ and ‘TI actions and responses to these market conditions’. Within each super-node, subnodes were developed iteratively during data collection. Queries for each subnode by year were used to assess patterns over time.

Limitations and strengths

Strengths of the commercial literature are that it can provide unique insights into industry attempts to maintain and enhance profitability. This was also the only data source available which included cigars and cigarillos. However, this literature may also obscure the TI’s real intentions (e.g. to promote sales) and fears (e.g. of declining premium sales and profits), and it will tend to report industry rather than public health viewpoints. The flip side of this is that as this literature is aimed at current and potential investors, rather than government and regulatory bodies that seek to regulate its promotional activities, it often contains illuminating examples of TI efforts to promote sales, sales figures and profitability.

Studies using the commercial literature

The commercial literature was often used as a background to inform our investigations, but three specific studies were also conducted.

1. Study 1 (see Chapter 3): what were the TI actions to maintain profitability?

2. Study 2 (see Chapter 3): how does the TI segment products on price?

3. Study 3 (see Chapter 6): was there any evidence of smokers down-trading to cheaper brands in the commercial literature?

Source

Nielsen’s Scantrak data were based on electronic point-of-sale (EPOS) system data (i.e. information obtained when tobacco products are scanned at the till in sales outlets). 106 Data are collated by Nielsen on grocery sales, which include sales in supermarkets and convenience stores. Nielsen collates 100% of sales data from four supermarkets [Tesco (Welwyn Garden City, UK), Sainsbury’s (London, UK), Morrisons (Bradford, UK) and Asda (Leeds, UK)] and conducts stratified random sampling with replacement from other outlets [e.g. Co-op (Manchester, UK)]. 107 For remaining retail outlets, sales could only be modelled. As an example, in November 2015, Nielsen collected EPOS data from 87% of the UK’s supermarkets, 15% of its convenience stores (including 83% supermarket-owned convenience stores, 59% petrol station shops, 6% convenience store chains and 4% independents) and 17% of Northern Ireland stores with grocery sales (Northern Ireland represents 2.8% of the UK population). 107 The outlets that provide Nielsen data vary over time and changes to the cigarettes being sold by the outlets may also mean that databases need adjusting. Nielsen updates its market coverage and thus its estimation of prices and volumes twice a year. In the UK, Nielsen collates EPOS data on grocery sales from 15,000 shops across all food and drink trade channels, and the total grocery market measured was £145B in 2015. 108 Other estimates of the total UK grocery market vary from £99.2B109 to £179.1B;110 thus, Nielsen’s estimate is consistent with these.

Nielsen data consists of a list of stock-keeping units (SKUs). A SKU is identified through an individual electronic barcode. Each SKU comprises seven different descriptors:

1. product type (e.g. RYO tobacco vs. FM cigarettes)

2. trading company (e.g. BAT)

3. house (e.g. Pall Mall)

4. brand (e.g. Pall Mall Blue king size)

5. pack size (e.g. the number of FM sticks or grams of RYO per pack)

6. sales unit contents – the number of packs sold together (e.g. in a carton)

7. packaging type – whether or not price promoted (i.e. sold in a price-marked pack).

Stock-keeping units are thus the lowest level element within a hierarchical classification system. We compared this Nielsen hierarchical brand typology with other alternatives,111,112 producing one to be used for our study (Table 2). Nielsen provides monthly sales data and for this study we used data from November 2008 to February 2016.

Analyses

Nielsen provided six overlapping data sets of monthly data: 1 (November 2008–October 2011), 2 (January 2009–December 2011), 3 (August 2011–July 2013), 4 (July 2012–December 2014), 5 (January 2013–November 2015) and 6 (March 2013–February 2016). Nielsen updates its estimates (market coverage and thus sales volumes and prices) twice yearly. Hence, the most recent data sets are likely to reflect the market most accurately, so when data for a month were available from multiple data sets, we used the most recent (e.g. we took data for November 2008–December 2008 from data set 1, January 2009–July 2011 from data set 2). We thus created a data set with the most recently provided monthly data, spanning from November 2008 to February 2016.

Up to five types of sales information were provided by Nielsen for each SKU every month:

1. number of units (usually packs) sold

2. price of each unit

3. total value of SKUs sold

4. total volume sold (sticks of FM or kilograms of RYO)

5. percentage of sampled stores to which each SKU was distributed.

Sometimes tobacco companies change the names of brands. We used this sales information, together with the commercial literature and Google searches, to check whether an apparent new brand, family or variant in a data set previously existed under a different name. Thus, SKU behaviour over the entire analysis period was identifiable.

It should be noted that not all five types of sales information were provided for all the data sets. Volumes were not provided for data sets 1 and 2 (used for our estimates between November 2008 and July 2011), so we imputed volume based on number of units sold. When provided, the volume variable showed evidence of modelling by Nielsen, so there may be some discrepancy between earlier and later data sets here. Distribution data were provided only for data sets 5 and 6 (used for our estimates between January 2013 and February 2016).

Limitations and strengths

Nielsen tobacco data are increasingly being used for academic, as well as commercial, research. 14,113–116 They have also been validated for policy evaluation. 117 Only legal grocery store sales are included, but such sales (including supermarkets and convenience stores) represent up to 80% of the total market [see Chapter 6, ITC data study 3: smokers’ tobacco purchasing behaviours by product and store type (licit purchases)] and, as we shall see, illicit and non-grocery store sales are declining (see Chapter 8). 81,118 In this project, we did not differentiate Nielsen sales data in terms of the percentage of tobacco sales that came from supermarkets compared with convenience stores; thus, although Nielsen samples overall grocery sales from around 87% of UK supermarkets and 15% of its convenience stores, we do not have the data to determine what percentage of the tobacco sales came from those supermarkets compared with convenience stores. However, in more recent work (covering the period from May 2015 to April 2018), we have found that about 80% of tobacco sales were from convenience stores (data not shown). This was higher than in this study using the ITC data (41% of legal tobacco sales were from convenience stores; see Chapter 6). Thus, assuming that convenience stores have higher tobacco prices, it might be the case that our Nielsen analyses would provide slightly higher estimates of tobacco prices than our ITC data; however, given that we can triangulate the findings from these two sources, we are confident that we will obtain an accurate estimate of tobacco prices in this project.

Only RYO tobacco prices are included, not papers or filters (which currently cost between £0.002 and £0.006 and between £0.002 and £0.008 each, respectively),119,120 which would add between £0.004 and £0.014 per cigarette.

Wider changes in the competitive landscape may also have contributed to price changes, but this seems unlikely: the e-cigarette market in 2015 was only a twentieth of the size of the tobacco market in the UK121 and four transnational tobacco companies collectively accounted for > 90% of the market, with only relatively small changes in their respective market shares. 122

Studies using the Nielsen data set

• Study 1 (see Chapter 3): allocation of SKUs to segments.

• Study 2 (see Chapter 4): TI pricing strategies.

• Study 3 (see Chapter 4): what proportion of price increases by segment are explained by TI price increases versus tax increases?

• Study 4 (see Chapter 6): what are sales volumes overall and by tobacco type over time?

• Study 5 (see Chapter 6): what are sales volumes by differently priced segments over time?

Data source and sampling characteristics

This research predominantly used the UK data from the ITC. This includes Canada, the USA, the UK and Australia. It is a longitudinal cohort study of adult smokers (aged ≥ 18 years) who are initially recruited as smokers (at least 100 cigarettes smoked in their lifetime and smoking at least monthly at the time of the survey). They are subsequently retained in the study for as long as possible, even if they quit smoking. At each survey, a new ‘replenishment’ sample of smokers are also recruited to compensate for those lost to follow-up (ranging for the UK between 22% and 34%), maintaining sample sizes at around 1500 for each country at each survey. This research mostly used the first 10 surveys or ‘waves’ of the UK ITC, conducted approximately annually between 2002 and 2014 (Table 4) (for some analyses, a subset of these surveys was used). More recent data were excluded because the methodology, sampling and questionnaires changed.

The ITC surveys were administered via computer-aided telephone interviewing or online (the online version was piloted at survey 7 in 2008 and then introduced gradually from survey 8 in 2010 onwards). Stratified random sampling was used to obtain samples broadly representative of national distributions of age, sex and geographical region for each of the four countries at the time of each survey. At each survey, cross-sectional sampling weights were also calculated to be representative of these characteristics, as well as longitudinal weights that were adjusted for attrition. Detailed methodological information about the ITC has been published elsewhere. 123,124

Measures

Not all studies used all measures and details will be provided in subsequent chapters when appropriate.

Tobacco use and purchase measures

Usual tobacco product type (product loyalty)

This was measured by asking ‘Do you now smoke . . . (packet/FM cigarettes only, RYO cigarettes only, or both)?’. For the measure of product loyalty, those who responded ‘both’ were classified as non-product-loyal ‘mixed’ users; and those who responded as ‘FM cigarettes only’ or ‘RYO tobacco only’ were classified as product-loyal ‘non-mixed’ users.

Last tobacco purchase type

This was coded as factory-made (cigarettes sold in) packs (FM-P); factory-made (cigarettes sold in) cartons (FM-C) or RYO (Figure 1). For participants who reported having a usual brand of tobacco, they were asked if this was the same as their last purchase. Those purchasing other products may be less brand-loyal and potentially more influenced by price promotions or discounts.

FIGURE 1.

International Tobacco Control Policy Evaluation Project survey questions and response categorisation for tobacco product purchases.

Brand loyalty

Participants were asked ‘Do you have a regular brand and variety of cigarettes (yes/no)?’.

Analyses

The ITC data analyses were conducted using the Stata^® software package, Stata SE version 12.1 (StataCorp LP, College Station, TX, USA), unless otherwise indicated. When relevant, more detailed analytical plans are discussed for each study in the relevant chapters. In general, for analyses in which we were interested in simple overall population trends (e.g. in prices paid for different products over time, or changes in the proportions of participants purchasing tobacco from different sources), we used generalised estimating equations (GEEs) with population-weighted data. However, the estimates provided by GEE analyses are not well suited for predictive modelling (e.g. for simulating or predicting responses of individuals or different groups of individuals),130 so for analyses in which we were interested in examining the role of other factors (e.g. nicotine dependence, sociodemographics or taxation rates) on our outcomes from our particular sample, we instead used random-effects regression analyses with unweighted data.

The ITC data are hierarchical, in that the same individual can provide multiple responses by participating in more than one survey, so the data are responses clustered within individuals, with a large number of small clusters (an average of around two responses per individual) and even some ‘clusters’ of size 1. GEEs require that the nature of the correlation between responses in the same cluster (in the ITC data that is the correlation between responses made by the same individual in different surveys) be specified. Stata allows for seven different types of correlation structure (independent, exchangeable, unstructured, autoregressive, stationary, non-stationary and fixed). From the outset, the autoregressive, stationary and non-stationary options were ruled out for our data, as these rely on equally spaced timing of observations with no gaps, which was not the case for our data. For the remaining options, we made our selection using a theory-driven, rather than a data-driven, approach. The fixed option allows for the user to specify a correlation matrix; however, we had no theoretical basis for producing such a matrix, so this option was also ruled out. The independent option assumes that responses provided by the same individual at different time points are independent (uncorrelated). We felt that this was highly unlikely and thus ruled out the independent correlation structure also. The likeliest scenario for our data, we felt, was that participants’ responses at different time points would be related in such a way that responses made closer together would be more closely related than those made further apart. The exchangeable correlation structure imposes the same relationship at all time points and does not allow for this scenario. This left only the unstructured correlation option, in which correlations are estimated for each time point from the data themselves. We thought that, theoretically, this was the best option to fit our data and therefore used this as the default option (the exchangeable structure was the next best option and, thus, we used this when the unstructured method failed to converge). In addition, the standard errors (SEs) provided by GEE are quite robust to mis-specification of the correlation structure if robust SEs are specified, which we did routinely. However, we did also use a data-driven method to compare model fit of the unstructured compared with exchangeable options, using the quasi-likelihood under the independence criterion (QIC), and the results of this are presented in Table S1 in Report Supplementary Material 2. The QIC statistic indicated an equivalent or better fit using the unstructured correlation matrix in all cases, with one exception, and the implications of this are discussed in Chapter 8.

Our reasons for choosing random- (and not fixed-) effects regression were both practical and theoretical. Our objective was to make inferences at the group level, controlling for the correlation of responses provided by the same individual (e.g. exploring the effect of time or age on tobacco affordability). Fixed-effects regression focuses on cluster-level analysis, in which the clusters themselves are of interest (e.g. in data in which pupils are clustered in schools, looking at the effect of different schools on academic outcomes), and clusters of large sizes are desirable. Random-effects regression is more focused on the group-level effects and can handle large numbers of small clusters and clusters of 1 (as in our data). Fixed-effects regression therefore also requires a higher degree of within-cluster variability than was present in our data. Furthermore, we were also often interested in the effects of time-invariant variables (e.g. sex, ethnicity), and fixed-effects regression does not allow for this. 130

We note also that we have not, in any of our analyses, adjusted our alpha levels for multiple comparisons, as we agree with others that this is not a meaningful or useful strategy. 131 We have, however, provided p-values throughout the report, enabling interested readers to make their own adjustments if so desired.

Limitations

The ITC data are based on self-reported details of smoking behaviour, so the results are dependent on the accuracy of that reporting. In general, several steps are taken to minimise bias and maximise response rates (e.g. through using different times of the day and days of the week for recruitment calls). Multiple contact information is collected to minimise attrition, with up to 25 attempts at follow-up, using different methods. Demographic profiles of the unweighted samples are compared with benchmark surveys and weights calculated, so that weighted data are consistent with corresponding national benchmarks. Full details of methods and weighting are given elsewhere. 123,124 Data from single countries can be used to examine how smoking behaviour changes over time in response to specific policies. Given the commonality of methods, study design and questionnaires across different countries, comparisons across countries can also be made using other countries as controls. The ITC conceptual model123 focuses on how single policies affect behaviour, rather than how combinations of policies and other factors interact to influence behaviour.

Similarly to the limitations discussed with the Nielsen data, the ITC survey did not ascertain whether or not RYO pouches were purchased as part of a bundle, including filters and/or rolling papers. However, the impact of this on our price estimates is likely to be minimal, with papers and filters contributing only between £0.004 to £0.014 to the prices per RYO cigarette, and any resulting biases would be consistently applicable over the study period, thus not affecting time-related trends.

A particular strength of the ITC is that it has been subject to consideration by ethics committees in multiple organisations and the hundreds of outputs from ITC data have been subjected to extensive peer review. As an international collaboration, all the processes and practices are subject to ongoing quality improvement. Specific limitations pertaining to particular analyses and subsamples are described in the relevant chapters.

Studies using ITC data

• Study 1 (see Chapters 4 and 8): prices paid for tobacco by smokers over time from both licit and tax avoidance/tax evasion sources.

• Study 2 (see Chapters 5–9): changes in tobacco affordability over time among smokers.

• Study 3 (see Chapters 6 and 8): smokers’ tobacco purchasing behaviours by product and store type, for both licit and tax avoidance/evasion sources.

• Study 4 (see Chapters 6, 8 and 9): impact of tobacco pricing changes on smokers’ purchasing patterns and product choices (analysis 4a) and smoking behaviour (analysis 4b).

• Study 5 (see Chapters 7–9): impact of tobacco pricing changes and product choices on smokers’ quit attempts (analysis 5a) and sustained abstinence (analysis 5b).

• Study 6 (see Chapter 7): weight of RYO tobacco cigarettes, analysis over time and international comparison.

Background: UK market conditions reported to be facing tobacco companies in the commercial literature

The TI strategies undertaken to maintain profitability during the review period (2008–14) were reported within the context of two main price-impacting tax changes that were also reported in the commercial literature. These were (1) the EU-wide plans to increase taxes and (2) to begin to equalise tobacco taxes (and thus prices) between different tobacco products to reduce the price differential between cigarettes and other forms of tobacco. 132 These changes were often discussed in emotive terms:

The government is mounting an assault on the downtrading trend in tobacco sales . . .

Hayward133

The commercial literature reported detailed implications of proposed changes, such as the removal of certain loopholes:

The proposal would also remove loopholes which allow some cigarettes made from fine cut tobacco to be sold as cigars, cigarillos or pipe tobacco and therefore to benefit from a lower tax rate.

Euromonitor International Ltd132

The banning of small pack sizes as a result of the revised EU Tobacco Products Directive (TPD) in 2016 was also reported. 134 Additionally, analysts noted that EU tax policy aimed to reduce the range in prices between different EU countries, although its ability to do this was limited:

. . . the mainly ad valorem (i.e. as tax as a percentage of price) mechanism of setting tax levels in the tobacco market means that equalising taxes and duties would not, by any means, make for equal prices, though the cross-border gaps would be narrowed.

Euromonitor International Ltd132

The commercial literature therefore revealed that tobacco analysts and vendors were expecting tax increases in this period and that tax structure changes were intended to reduce the price range between different products.

Findings: tobacco industry actions undertaken to maintain profitability

Commercial literature study 1 conclusions

During the review period (2008–14), the commercial literature reported that smokers were continuing to down-trade to lower priced tobacco options for two main reasons: (1) increased prices (generally blamed in the commercial literature on tax increases) and (2) the recession reducing smokers’ disposable incomes (see Chapter 5).

The TI responded in a number of ways. First, innovations (capsules, flavours, etc.) were used to promote premium products and signal their greater attractiveness. However, as the trend to down-trade grew, tobacco companies appeared to blur the boundaries between price segments and innovations, in that both product and packaging were increasingly used in lower price segments. Growing numbers of products were launched in cheaper price segments, often within the same brand/brand family (e.g. cheaper variants or RYO tobacco versions of known FM cigarette brands). Additionally, a variety of techniques were used to promote cheaper products and help maintain their lower prices, most notably price marking and smaller pack sizes (e.g. packs with < 20 cigarettes or low-weight RYO packs).

The compensation mechanism of premium brands offsetting the lowered profit margins of cheaper products was forecast, by Euromonitor International Ltd, to stop working in the UK in the 2020s. Indeed, many of the promotional approaches noted during the review period are now restricted in the UK (see Chapter 10).

Segmentation of stock-keeping units for factory-made cigarettes

For the first month of data, using the list of SKUs for each price segment which was developed in our PPACTE study32 (premium, mid-price, economy and ULP), we allocated FM SKUs to segments (except that the PPACTE economy segment was split into ‘subpremium’ and ‘value’ segments based on common segment naming of two subgroups of brands in this segment; see Table 5). All SKUs sold in the first month were listed in order of increasing price per stick, alongside their PPACTE segment. If there was no PPACTE segment, we allocated the SKU to the segment within which its price fell. There were a few brand families, generally that arose towards the end of the PPACTE data series, where the segment once the brand was established, as shown in the new data set, had moved. Segments for such brand families were revised. We then could calculate a price range for each segment and establish price boundaries for each segment that did not overlap.

For subsequent months, we again listed all SKUs sold by price per stick, alongside their segment from the previous month. New brand families were added to the segment within which their price fell. Problem SKUs, which appeared to cross segment boundaries, were flagged and addressed separately. These problem SKUs were not included in the boundary designation for that and subsequent months until their price had stabilised within the designated boundary. If a new SKU emerged or reached valid market share and its price was in between two segments, it was classified with the closest segment, unless in the future it was persistently in the other segment, and the commercial literature suggested that segment or the emergence of a new segment.

Our PPACTE work, the updated commercial literature review and an initial analysis of the data, showed that all SKUs within a FM brand family tended to be within the same price segment and that brand families tended to stay in the same price segment over time. Originally, we had believed that brand variants in the same brand family would be located at different price segments because the commercial literature (study 1) had suggested that brand variants were being introduced at different price points to their main brand family. When this occurred, however, Nielsen generally split the lower priced variants into a different brand family. For example, L&B Blue was presented as a different brand family to L&B. The final allocation of FM SKUs was thus always consistent with their brand family.

Seven FM segments were originally identified from the Nielsen data and literature: (1) premium; (2) vestigial mid-price (brands identified as being in the mid-price segment in the PPACTE study, which mostly either disappeared or were repriced during the data series); (3) subpremium; (4) value; (5) original ULP (brand families in this segment at, or near to, the beginning of the data series); (6) new ULP (brand families that appeared from 2011 onwards that were priced generally within the original ULP bracket); and (7) sub-ULP (brand families that appeared from 2012 onwards and were consistently priced below the ULP brand families). Two researchers analysed in detail all flagged SKUs, to ascertain why they were not in the expected segment. When possible, rules were then developed for allocating flagged SKUs. Reasons for SKUs not being in the expected segment (based both on data observations and on the literature review) included (1) delayed price increases (e.g. Dunhill); (2) high-priced SKUs within one segment overlapping with low-priced brand families within another segment (e.g. JPS Black, Sovereign Black and Vogue Perle were priced between main subpremium brand families and value brand families for long periods); (3) low introductory pricing (e.g. Marlboro Bright Leaf Platinum); (4) very low exit pricing, particularly in 2014 when the number of sticks per pack was changing from 20 sticks to 19 sticks (e.g. Richmond, Mayfair) and the old 20-stick packs were sold off very cheaply; (5) repricing a brand family into a different segment (e.g. Superkings moved from vestigial mid-price to premium, or Berkeley, where the segment declined several times); and (6) gimmicks (e.g. click on or menthol, which often led to higher priced SKUs).

We produced graphs of SKUs within brand families and segments over time to understand the segmentation better. Two examples are provided to illustrate the utility to our allocation process of visualising the data. Figure 2 shows the atypical brand Berkeley for which repricing changed the segment several times. The three Berkeley superking SKUs (SKUs 1–3; black lines in Figure 2) present at the beginning of the data series dropped from vestigial mid-price to value around May 2010 and from value pricing to original ULP pricing around December 2011. Berkeley king size 19-stick packs (SKUs 4 and 5; grey lines in Figure 2) showed the opposite pattern and increased from subULP in late 2014 (when they arrived on the market) to original ULP pricing by the end of the data series.

FIGURE 2.

Nielsen data study 1: Berkeley FM cigarette SKUs alongside preliminary price segments, illustrating how anomalous SKUs were used to understand segmentation. All SKUs are single packs with no price marking. Prices are not weighted for volumes.

Mayfair was a key brand in the value segment. However, there were some anomalous SKUs (Figure 3). Mayfair Menthol SKUs could sometimes be priced higher than expected and for many months the green SKU illustrated (SKU 5 in Figure 3) was priced in the subpremium segment. Dramatic falls in price of 20-stick Mayfair packs in 2014 when the 19-stick packs were introduced are also shown. Note, that one of the replacement 19-stick packs was actually priced higher than the 20-stick pack, at the border between value and subpremium pricing.

FIGURE 3.

Nielsen data study 1: Mayfair FM cigarette SKUs alongside preliminary price segments, illustrating the dramatic discounting of some SKUs being phased out of production, as new SKUs were introduced. All SKUs are single packs with no price marking. Prices are not weighted for volumes.

Our visual analysis of flagged brands showed that although some SKUs might temporarily be priced within a neighbouring segment, this generally occurred only for short periods and in the margins of the unexpected segment. As such, it was simple to allocate each SKU to the segment where it spent most time. There were two exceptions: (1) there was considerable overlap in price between the original ULP and new ULP segments (these were therefore merged, see Figure 2); and (2) it was impossible to allocate a clear segment to Berkeley brand families due to several marked price changes. Berkeley was therefore excluded (see Generalisation and final segments).

Segmentation of stock-keeping units for roll-your-own tobacco

Stock-keeping unit prices were listed and graphed separately by pack size, as these were priced differently, with heavier packs cheaper per gram. Initial graphical analysis of the RYO SKUs revealed only three main price segments and that these were priced consistently above, in line with or below one of the main traditional RYO brands (Amber Leaf) (Figure 4). RYO SKUs were therefore listed by price per gram together with their segment, based on their pricing relative to Amber Leaf. If SKUs appeared to move segments, flags were used in the same way as for the FM SKUs.

FIGURE 4.

Nielsen data study 1: RYO SKUs compared with Amber Leaf SKUs, showing the pattern that was observed when other RYO SKUs were priced consistently above (e.g. Golden Virginia SKUs; blue lines), in-line with (e.g. Golden Virginia Smooth SKUs; light green lines) or below (e.g. Gold Leaf SKUs; dark green lines) the more traditional RYO brand Amber Leaf (black lines). Note, all SKUs displayed were single, non-price-marked packs.

Generalisation and final segments

Nielsen did not always differentiate FM and RYO SKUs at the same point in the brand classification hierarchy (see Chapter 2 and Table 2). For FM SKUs, Nielsen differentiated brand families (their ‘houses’) when they were introduced at different price points, for example Marlboro Red and Marlboro Gold were in the same brand family, whereas Marlboro Bright Leaf, which had lower prices, was differentiated. For RYO SKUs, this was not always the case. For example, Amber Leaf Signature Blend SKUs were in the same brand family as the other Amber Leaf SKUs, despite being more expensive. This difference may flow from there being fewer RYO SKUs (e.g. in 2015 there were 934 FM SKUs and 226 RYO SKUs). Thus, for FM, SKUs in different brand families were reliably likely to have different price points, so we allocated FM SKUs to price segments at the brand family level. As this was not clearly the case for RYO SKUs, these were allocated to price segments at the more specific brand variant level.

The segment allocations, which were initially determined using the restricted pack types and sizes described in Segmentation of stock-keeping units for factory-made cigarettes and Segmentation of stock-keeping units for roll-your-own tobacco, were then extended to all other relevant pack sizes, cartons and price-marked SKUs. For simplicity and for consistency between FM and RYO, the vestigial mid-price, subpremium and value segments were merged into a ‘mid-price’ segment, the original and new ULP segments were merged into a ‘value’ segment and the subULP became the ‘subvalue segment’. Thus, for analysis, the segments for FM were premium, mid-price, value and subvalue; and the segments for RYO were premium, mid-price and value. The final list of FM brand families and RYO brand variants allocated to these prices segments is provided in Table 6.

To explore convergent validity, FM brand families and RYO brand variants were listed within their allocated price segments, together with associated price-related terms used in the commercial literature (see Tables S2 and S3 in Report Supplementary Material 2). In general, there was good consistency between segments and literature terms; this was especially true for FM families in the premium segment, which were clearly differentiated. For FM families in the mid-price segment, the literature terms used could depend on whether the family was positioned as desirable because of cheaper pricing (resulting in some overlap with terms used for brand families in the value and subvalue segments), or for being more sophisticated than cheaper brands (resulting in the use of the term ‘premium’ in the literature). The greatest variation in literature terms was seen for the FM brand families in the value and subvalue segments (although there was no overlap with terms used for the premium segment). For RYO tobacco, brand variants allocated to the premium and value price segments were clearly differentiated by price-related descriptors in the commercial literature. RYO variants in the mid-price segment, however, could be described by low- or high-price terms, or a mixture. This often occurred when they were being compared with a differently priced brand variant in the same brand family (e.g. Golden Virginia Smooth is ‘low priced’ compared with the original Golden Virginia brand variant). By contrast, Marlboro RYO was described with high-priced descriptors, perhaps as a reflection of the premium nature of the Marlboro brand overall.

Segmentation allocation of extra stock-keeping units based on industry pricing analyses

Some additional SKUs that did not meet the inclusion criteria for the original segmentation analysis (see Table 3), due to low market share, became valid when the Nielsen data sets were merged for the industry pricing analyses described in the following chapter (see also Chapter 2). These extra SKUs were allocated to either the same price segment as their brand family (FM) or their brand variant (RYO). If this was not possible, they were allocated to the segment that their price was closest to for the majority of the time that the SKU was valid (e.g. FM Lucky Strike and RYO Turner SKUs). One brand, Swan, was excluded due to extremely low pricing. It is possible that there was some confusion by retailers, given that some Swan RYO accessories are sold in packaging similar to cigarette packs; thus, Swan filter tips or rolling papers, for example, could have incorrectly been given a tobacco brand SKU (see Figures S1–S3 in Report Supplementary Material 3 for illustrations of this allocation process).

Allocation of factory-made cigarette brands in the ITC data to price segments

Analysis of the ITC data set included exploring how smokers moved between FM cigarette price segments (see Chapter 6). Thus, we needed to know the price segment of the brands that ITC respondents reported smoking. For ITC participants surveyed prior to November 2008 (see Table 4), when Nielsen data from this current study became available, we used PPACTE classifications to allocate brands to price segments. Some brands recorded by ITC respondents did not appear in the Nielsen data and were allocated to the segment in which their price fell (i.e. allocated by price not brand). Price boundaries for each segment had to be recalculated slightly for this purpose, because ITC does not differentiate between price-marked and non-price-marked packs. It was necessary to develop some new exclusion criteria to recalculate the price boundaries. In addition to the usual exclusions (SKUs with market share ≤ 0.05%, packs containing < 17 sticks or > 20 sticks, retailer and wholesaler brands, cartons and all Berkeley SKUs), we excluded all Chesterfield, Marlboro Touch and Mayfair Green superking 20-stick pack SKUs when recalculating the price boundaries. Some further SKUs were partially excluded in some months only, due to anomalous pricing in visual analysis of the boundaries (see Table S4 in Report Supplementary Material 2). Unlike the original Nielsen segment analyses, we included price-marked packs. Wholesaler and retailer brands could not be included in the Nielsen analysis because they were distributed only to a small selection of stores. Such brands appeared in the ITC data, so tentative boundaries from the Nielsen data were also provided for these (they tended to be priced lower than other products). The final price segment boundaries used for ITC brand allocation are provided in Figure 5.

FIGURE 5.

Nielsen price boundaries used for ITC brand allocation of FM cigarettes to price segments (used in ITC studies 3–5). Note, that for some ITC analyses the value and subvalue segments were combined.

Introduction

Here we assess UK smokers’ store-based tobacco purchase costs over time. As described in Chapter 2 (see also Chapter 6), these are assumed to be legal, fully taxed purchases. During the study period, 2002–14, ITC data suggested that most (≥ 80%) tobacco purchases were from UK store-based sources. The most common source of tobacco purchases changed from convenience stores to supermarkets over the study period, which presumably conferred price savings. We compare purchase costs of FM-P, FM-C and RYO tobacco in order to understand possibilities for price savings, which could be undertaken as an alternative to quitting.

Methods

Results

The real price (base year 2014) of all product types purchased increased significantly during the study period: prices remained relatively unchanged up until 2010, but increased thereafter (Figure 6). The real price per stick for both FM-P and FM-C increased by only £0.10 from 2002 to 2014. For FM-P, median prices per stick increased from £0.27 to £0.37. FM-C were typically £0.01 to £0.02 cheaper per stick. The median price per stick of RYO tobacco increased by only £0.05, from £0.12 in 2005 to £0.17 in 2014.

FIGURE 6.

International Tobacco Control Policy Evaluation Project data study 1: weighted median prices (black lines) and 95% ranges (shaded areas: top and bottom 2.5% excluded) of (a) FM-P and RYO tobacco; and (b) FM-C purchased from UK store-based sources only. Blue lines indicate 25th and 75th percentiles. All prices are adjusted to 2014 values (base year = 2014) and based on the most recent reported tobacco purchase. a, No data collected for this year.

The range of prices differed by product type, but FM-P/FM-C and RYO smokers in 2014 were able to buy the same type of product at real prices similar to 2002 and 2005, respectively. The price range for FM-P was relatively narrow and changed little over time: £0.12 per stick in 2002 to £0.13 in 2010; then £0.16 in 2013 to £0.23 in 2014, when an increase in the highest price and a decrease in the lowest price was evident. The price range for FM-C was wider and more variable, despite median prices being similar to FM-P, ranging between £0.19 and £0.27 per stick in most years, but £0.37 in 2007 and £0.45 in 2014. The price range for RYO tobacco changed little over time, from £0.10 per stick in 2005 to £0.12 in 2014. FM-P prices were evenly distributed over the range, whereas RYO prices were negatively skewed, with greater variation at the cheaper (below median) end.

Limitations

No data were collected for most of 2011 and all of 2012. Hence, we are unable to comment on changes specific to this period and this might overestimate the linear nature of the trends observed over time. However, the large size of the data set does give some confidence in the observations reported. Some store-based purchases might be ‘under the counter’ and hence tax evasion, but we minimised this possibility by excluding very low prices paid from UK store-based sources. To be included in the study, participants had to be current smokers, so our procedure meant that quitters were progressively excluded from the analysis. However, the missing data analysis also suggested that users of cheap tobacco (who other research has shown may be less likely to quit54–56,59) were more likely to be excluded. Thus, it is possible that these two effects balance each other to some extent. The ITC survey is replenished at each wave with a representative sample of current smokers, to minimise attrition effects. We therefore think it is unlikely that the observed trends are due to attrition.

Implications

Changing between product types would have saved money (e.g. purchasing FM-C rather than FM-P would typically confer a saving of £0.01–0.02 per stick on the median price). Switching to RYO conferred the largest savings, a 0.5-g stick being typically less than half the price (around £0.18 cheaper) of one FM-P cigarette. This means that a saving of about £750 per year could be made for the average UK smoker smoking around 11.4 CPD,217 by switching from FM-P to RYO.

Changing within product types (to different brands) also saved money. The price range between the cheapest and most expensive FM-C products purchased was consistently wide and, from around 2010 onwards, the price range of FM-P products also widened markedly. These findings suggest that the TI strategy of overshifting prices on premium and undershifting prices on cheaper products was becoming more aggressive in the UK. In contrast, the observed price range for RYO tobacco was disproportionate, due to more variation in the cheap prices, suggesting that TI undershifting may be particularly relevant within the RYO tobacco market.

In relation to the influence of tobacco taxation, from 2011 onwards, when tobacco duty was higher than in previous years (at 2–5% above inflation; see Table 1), there was a greater increase in the prices of purchased tobacco products. However, even when assuming that this was a causal relationship, these tax increases did not substantially impact real prices, did not prevent a widening price range between cheapest and most expensive FM-P products, and made little impact on the lowest price paid for RYO tobacco. Hence, although there was a statistically significant increase in real price paid over the study period, this was not substantial.

Summary of ITC data on tobacco prices

The real price paid by smokers changed little over time, despite taxation changes. For some smokers, it was possible that the prices they were paying for their tobacco from licit sources in 2014 had not changed in real terms compared with 2002. Tobacco taxation had not prevented the success of the TI strategy of widening the price range between expensive and cheap products.

Although the ITC data were able to tell us about the range of available prices, there were too few cases to break down tobacco products into detailed price segments for both FM and RYO. In addition, insufficient detail was available on products purchased to understand fully how the wide variation in prices was achieved. Nielsen tobacco sales data were used for these purposes.

Analysis

Results

Price per stick

When comparing prices per stick, RYO sticks were considerably cheaper than FM sticks (Figure 7, and see also Table S5 in Report Supplementary Material 2). Over the study period, the price range between cheapest and most expensive price segments widened. For RYO sticks, from £0.02 to £0.03 per stick and for FM sticks from £0.07 to £0.10 per stick. This reflected smaller price increases in the cheaper segments. However, price increases were seen in every segment, suggesting that the fall in price of subvalue packs (see Price per pack) was attributable to the declining number of sticks per pack. Prices per stick also increased much faster among FM cigarettes than RYO tobacco, with the exception of the subvalue FM segment.

FIGURE 7.

Nielsen data study 2: weighted average real prices (base year = 2014). (a) Per stick; and (b) per pack for FM cigarettes and RYO tobacco in different price segments, over time. Data are for all pack sizes combined (cartons excluded). Note, the price of FM subvalue packs appears unstable initially as different pack sizes were introduced.

Price marking

Cheaper segments were more likely to be price marked than expensive segments (Figure 8 and see also Table S6 in Report Supplementary Material 2). For example, between 60% and 100% of subvalue FM products were price marked compared with a negligible proportion of premium FM products. Patterns for RYO products were similar, although a higher proportion of premium RYO products were price marked than premium FM products (36% in 2012 and about 25% thereafter). Half of RYO value products were price marked.

FIGURE 8.

Nielsen data study 2: percentage of packs price marked by (a) segment for FM cigarettes and RYO tobacco; and (b) pack size for RYO tobacco. Nielsen data on price marking was only available from August 2011.

For RYO tobacco, but not FM cigarettes, smaller packs were also more likely to be price marked than larger packs (see Figure 8 and see also Table S7 in Report Supplementary Material 2). For most segments and pack sizes, price marking appeared to increase gradually over time. However, there were notable exceptions to this pattern. When the new subvalue FM brands were introduced in early 2012, 100% were price marked but this then fell rapidly to between 60% and 70%. There was substantial price marking of 10-g RYO packs prior to growth in the mid-price segment (2011) and when 10-g packs were introduced into the value segment (2013). In addition, price marking on premium brands in both FM and RYO segments fell.

Differences between the prices of price-marked and standard packs varied over time and between segments; however, price-marked packs were typically cheaper (see Table S8 in Report Supplementary Material 2). Overall, the increasing availability of price-marked products enabled the TI to sell products at lower prices, albeit compromising convenience store profit margins.

Tobacco industry pricing conclusions

Tobacco sales and ITC data indicated that, although overall tobacco prices increased, the cheapest products did not increase in price in real terms. The data sets also revealed types of tobacco products with cheaper pricing: RYO tobacco (all data sets); cartons (ITC data); and mid-price, value and subvalue FM cigarettes, mid-price and value RYO tobacco, price-marked packs and small packs (Nielsen data). The availability of cheap tobacco products that did not increase in price in real terms helps explain why smokers could still purchase their tobacco at prices similar to those at the beginning of our study period in 2002 and also the role of RYO tobacco in keeping the cost of smoking low. In the following section, we explore how the TI uses its pricing strategies to respond to tobacco taxation changes.

Methods

Results

Net real revenues were considerably greater for higher than for lower priced segments, although the gap was more marked for FM than RYO segments (Figure 9; see also Table S9 in Report Supplementary Material 2). A cyclical pattern of a drop in revenue immediately post Budget, followed by an increase thereafter, emerged with the 2011 Budget and from this point there was progressively more differentiation in revenue between segments. Within the premium segment, revenue was greater on FM than RYO products, whereas in the value segment this pattern reversed and net revenues in the cheapest FM segment changed little throughout the whole period.

FIGURE 9.

Nielsen data study 3: net real (base year = 2014) TI revenue per pack (most popular pack sizes) for FM cigarettes and RYO tobacco, by price segment. Vertical reference lines indicate tax changes. Net revenue refers to revenue after tax only (other costs, e.g. distribution, production, etc., have not been deducted).

Declines in net revenue at times of tax (VAT or tobacco tax) increases can be seen more clearly using changes in net revenue post Budget (Figure 10 and see also Table S10 in Report Supplementary Material 2). Every year and in every segment, in the month after the Budget, net revenues fell. Thus, tobacco tax changes were not passed straight on to consumers, but were initially absorbed by tobacco companies (indicating slightly lower profits per pack). In general, the extent and duration of the undershifting was greatest in the lowest segments, whereas in higher priced segments the undershifting was less marked and often short-lived with profits recovering to and then exceeding pre-Budget levels (indicating overshifting) within 1–3 months each year. For example, in 2014, revenue on premium FM products recovered to pre-Budget levels by May and by the end of the Budget year was up by £0.23. By contrast, revenues on subvalue FM products fell in 2014 until August and, although they increased thereafter, never recovered to pre-Budget levels (see Figure 10).

FIGURE 10.

Nielsen data study 3: change in net real (base year = 2014) revenue per pack (£) post Budget [difference in revenue per pack in each post-Budget month compared with Budget month (last month of previous tax year)] for popular pack sizes of FM cigarettes and RYO tobacco, by price segment: (a) 2009 tax year; (b) 2010 tax year; (c) 2011 tax year; (d) 2012 tax year; (e) 2013 tax year; (f) 2014 tax year; and (g) 2015 tax year. Budgets were held in April 2019 and March in subsequent years. A negative change indicates undershifting and a positive change indicates overshifting: when values cross the £0.00 reference line in any tax year, it indicates that revenue has reached the same level (in real terms) as in the last month of the previous tax year, just prior to when the Budget was enacted. Note that, in 2009, FM value packs are 20 sticks, not 19 sticks, as 19 stick packs were not available at this time. Note that the FM subvalue segment was introduced in 2012, so this is shown from 2013. Net revenue refers to revenue after tax only (other costs, e.g. distribution, production, etc., have not been deducted).

Patterns varied somewhat by year, reflecting the different tax changes. Unlike for other tax years, in January 2010 and January 2011 there were VAT increases. The declines in revenue at both these points indicate these VAT increases were absorbed by tobacco companies. The years 2011 and 2012 saw the most marked tax changes (see Table 1). In line with this, after the 2011 tax system changes, a more marked pattern of initial undershifting and greater differentiation in revenue change by segment emerged, consistent with the widening revenue gap between segments.

Background: affordability in the commercial literature

The ‘Great Recession’ was alluded to in the commercial literature, the effect of which was seen as producing down-trading via reduced incomes and welfare freezes, affecting the disposable income of (particularly low-income) smokers. In 2009, and still in 2012, the recession was seen as a driver pushing smokers to switch to RYO tobacco, at least partially:

‘Dualling’ is [. . .] a trend identified by Imperial Tobacco towards smoking both roll-ups and factory made cigarettes [. . .] partial down-trading [. . .] More and more UK smokers, and particularly, younger women, are turning to roll-ups. Part of the reason is the recession.

Euromonitor International Ltd219

The RYO market continues to expand as people feel the pinch from government cuts and wage freezes. 220

In 2010, Watkins from Imperial Tobacco did not think that ‘convincing smokers to trade back up will be an issue’ after the recession,169 but down-trading was still apparent in 2014 despite the end of the recession:

. . . while there appear to be the early signs of a wider economic recovery, many smokers are still feeling the pinch and so growth remains the preserve of the value for money subsector of cigarettes, RYO and even value for money RYO.

Hegarty221

Thus, in 2014, there were still stated to be affordability problems. Wholesaler Paul Ford was quoted in Wholesale News, stating ‘consumers cannot afford to smoke, drink, drive a car’. 211

Hence, the commercial literature suggested that tobacco products were becoming less affordable partly due to reduced incomes via the government response to the recession.

Methods

Results

The characteristics of the included sample are given in Table 11. There were slightly more females (58%) than males, the mean age was 49 years (SD 14.2 years) and the majority were white with low to moderate levels of education. Most smoked their first cigarette within 30 minutes of waking and average cigarette consumption was 16.8 CPD (SD 8.2 CPD).

There was an average annual decrease of 1.6% in incomes for our sample of smokers over the study period (Figure 11). This deviated considerably from national annual per capita GDP, which increased from £26,206 to £30,299 between 2002 and 2007, when mean income in our sample decreased from £32,202 to £29,423. From 2007, incomes in our sample continued to decrease at a more marked rate than GDP, to a low in 2012 of £24,976 (GDP declined to £27,196). From 2012, both income in our sample and annual per capita GDP modestly increased (see Figure S6 in Report Supplementary Material 3).

FIGURE 11.

International Tobacco Control Policy Evaluation Project data study 2: measures of affordability (individualised and aggregate versions) and their constituent components (income and tobacco prices) over time. (a) Annual gross income and tobacco prices for FM cigarettes and RYO tobacco (values are adjusted for inflation using CPI with 2014 as the base year); and (b) individualised measures of affordability (for FM cigarettes and RYO tobacco) and the aggregate version of affordability (FM only). Individualised measures are unadjusted and inclusive of all purchase sources. The aggregate version of affordability is based on annual per capita GDP and mean cigarette prices from annual sales in the MPPC (prior to 2011) or WAP (2011 onwards) for FM cigarettes from fully taxed UK sources only. Note, shaded areas indicate 95% confidence intervals. a, No ITC data collected for this year.

There was an average annual increase of 2.6% in the price of FM cigarettes (see Figure 11), which, combined with the decrease in income, contributed to an average annual decrease in the affordability of FM cigarettes of 0.24%, from 91.5% [± 95% confidence interval (CI) 91.0% to 91.9%] in 2002 to 87.8% (± 95% CI 87.0% to 88.5%) in 2014 (see Figure 11, model 1 unadjusted). Similarly, the average annual increase of 4.5% in the price of RYO tobacco (see Figure 11) and the decrease in income contributed to an average annual decrease in the affordability of RYO tobacco of 0.31%, from 96.3% (± 95% CI 95.7% to 96.9%) in 2006 to 93.7% (± 95% CI 93.0% to 94.4%) in 2014 (see Figure 11, model 1 unadjusted). RYO tobacco was significantly more affordable than FM cigarettes.

The aggregate version of affordability (fully taxed FM cigarettes only) also decreased (see Figure 11). There was an increase in aggregate affordability between 2010 and 2011, which coincided with the change in calculation of cigarette prices from the MPPC to the WAP and, hence, marks the switch between these two data series. When MPPC was used, aggregate affordability decreased at an average annual rate of 0.13% between 2002 (93.4%) and 2010 (92.5%); when using the WAP, an average annual rate of 0.40% between 2011 (93.7%) and 2014 (92.5%) was observed.

For FM cigarettes, focusing on the fully adjusted regression (model 2), there was a significant linear decrease, χ²(1) = 59.8 (p < 0.0001), in individualised affordability over time (Table 12). However, as indicated by the reverse adjacent contrasts, none of the year-to-year decreases were significant, except for the two instances when there was a 2-year interval between surveys: 2008–10, χ²(1) = 5.4 (p < 0.05) and 2010–12 χ²(1) = 4.3 (p < 0.05), a period that also coincided with tax increases greater than the rate of inflation. The same pattern of results was obtained for the fully adjusted model 2. For model 3, when non-UK/non-store sources were excluded, a significant linear decrease in affordability was still observed over time, χ²(1) = 54.3 (p < 0.0001). However, the adjacent contrasts indicated that none of year-to-year decreases remained statistically significant. Model 3 was the most similar to the generalised affordability measure, but was around 3–6% lower each year and showed a more marked decline. In both models 2 and 3, 2010 was the first year that FM cigarettes became significantly less affordable than they had been in 2002. Full results for FM cigarettes for all three models and including all covariates can be found in Table S11 in Report Supplementary Material 2.

For RYO tobacco, the changes in individualised affordability over time were similar to FM cigarettes (see Table 12). The fully adjusted model 2 indicated a significant linear decreasing trend, χ²(1) = 39.45 (p < 0.0001) in affordability over time; however, there were no significant year-to-year changes indicated in the reverse adjacent contrasts. Findings were similar for the unadjusted model 1 [except the decrease from 2008 to 2010, χ²(1) = 5.3 (p < 0.05), was statistically significant] and model 3, excluding purchases from non-UK/non-store sources. Full results for RYO tobacco for all three models and including all covariates can be found in Table S12 in Report Supplementary Material 2.

Results of the sensitivity analysis looking at changes in tobacco consumption showed that FM cigarette smokers made a slight reduction in their CPD from 17.5 (95% CI 17.2 to 17.9) cigarettes in 2002 to 16.1 (95% CI 15.6 to 16.7) cigarettes in 2014, with a statistically significant linear trend: χ²(1) = 40.9 (p < 0.0001). The consumption of RYO tobacco smokers changed from 16.8 g (95% CI 16.0 g to 17.6 g) in 2006 to 17.3 g (95% CI 16.4 g to 18.3 g) in 2014, but the linear trend was not statistically significant: χ²(1) = 1.7 (p = 0.20). Given that tobacco became less affordable during this period, this decrease cannot be attributed to these observed decreases in CPD (an increase in CPD would result in a decrease in affordability). However, it is possible that the decreases in consumption (especially for FM cigarettes) attenuated the magnitude of the observed decrease in affordability and we might have observed a greater reduction in affordability had smokers not reduced their consumption to compensate.

Limitations

As we included only current smokers, some recent ex-smokers may have quit because they could not afford tobacco, which would mean that our data somewhat underestimated the decrease in affordability. We also excluded non-daily smokers and those who usually smoked both FM cigarettes and RYO tobacco because of insufficient data, which may have slightly underestimated affordability, as these groups might be particularly adept at controlling their tobacco expenditure by smoking less or switching as necessary. We also used repeat cross-sectional, rather than longitudinal data. Finally, we had data only on gross (before tax) rather than net (after tax) income. Individuals in the UK with higher incomes are taxed at progressively higher rates, so their net income (what is actually available to spend on tobacco) will be reduced by relatively more than those on lower incomes. For the high-income groups, affordability will therefore be slightly overestimated; we do not expect this to cause a large bias in our estimates, as just over two-thirds of our sample (68%) had gross incomes < £30,000, which was below the threshold for moving beyond the lowest tax rate in all years analysed, except the 2002–3 tax year in which the threshold was £29,000. Income differences between our sample and national figures are probably reflective of smokers being more disadvantaged overall.

Implications

Our new individualised measure of tobacco affordability provided a more sensitive and nuanced insight into the impact of tobacco taxes compared with aggregate measures based on national estimates of income and average tobacco prices. In the UK, tobacco was significantly less affordable in 2014 than in 2002, although the rate of decrease (0.24% annually) was low and year-to-year declines were not significant, and potentially attenuated by slight reductions in consumption. Affordability was modified by larger tax increases and RYO tobacco was considerably more affordable than FM cigarettes.

We have developed a novel measure of affordability which, beyond using individual reported tobacco prices and incomes, departs from traditional price/income conceptualisations of affordability by also incorporating tobacco consumption. The two other studies223,224 of which we are aware that have adopted individualised224 or partly individualised223 measures of tobacco affordability, although they did not incorporate consumption directly into their calculation of affordability, weighted their estimates of tobacco prices by consumption. It is therefore necessary to discuss the implications of this.

Owing to the inclusion of (or weighting for) consumption, it is possible for these analyses to indicate that tobacco has become more affordable, even if prices have increased relative to incomes. This would only occur only if smokers reduced their consumption by more than what is necessary to compensate for the price increase. We think this is unlikely. The more likely scenario is that they have reduced consumption just enough to compensate for the price increase (we have observed exactly this with users of RYO tobacco; see Chapter 7), in which case our measure would indicate no change in affordability. It is also possible, as we observed in this case, that smokers reduce their consumption, but not enough to compensate for the price increases, which would lead to smaller than expected (based solely on prices relative to incomes) reductions in affordability. It is important to examine this interrelationship among price, income and consumption, because the health gains of reducing consumption without quitting are limited,226 but small reductions in consumption (e.g. by 1.4 FM cigarettes over the 12-year study period we observed here) might be just enough to offset tobacco price increases to a satisfactory level to deter smokers from quitting, thereby undermining the effectiveness of tobacco tax increases. For researchers using our measure, we therefore recommend that, as we have done here, that our measure be examined alongside the separate trends in prices, incomes and consumption, to gain a more comprehensive understanding of how smokers are adjusting their behaviour to keep tobacco affordable.

Background: evidence of down-trading in the commercial literature

For industry, down-trading from expensive products to cheaper products was interpreted as a consequence of tax increases136 and the recession:227

The one concession [rather than quitting] smokers are likely to make to tougher times [Watkins from Imperial Tobacco claims] is to save money where possible: there is already a trend to down trade and this is likely to be accelerated over the coming months.

Hegarty169

In 2008, PMI, with a limited share of the UK market and a predominance of premium brands (notably Marlboro), claimed to be unaware of down-trading in the UK:

The down-trading that everyone is referring to or afraid of, we have just not seen.

Hermann Waldemar, Chief Financial Officer from PMI, quoted in The Grocer228

However, spokespeople from both JTI and Imperial Tobacco described evidence of smokers down-trading. 229

Imperial Tobacco had noticed similar trends:

Increasing numbers of adult smokers are moving down through the cigarette price sectors into the economy sector.

Iain Watkins, trade communications manager at Imperial Tobacco quoted in Wholesale News230

Dual use of manufactured cigarettes and RYO was later reported to be a growing trend. 139 Furthermore, even RYO tobacco users were noted to be trading down from more expensive to cheaper RYO tobacco brands. 169

In 2014, 90% of smokers switching brands were now switching to a lower priced brand,231 causing a ‘full blown subprime crisis in the heart of the global cigarette market’. 232 Down-trading became increasingly essential because ‘consumers cannot afford to smoke, drink, drive a car – something has to give; although tobacco is the last thing some people will give up, they will down-trade’ (wholesaler Paul Ford, quoted in Wholesale News)211 and possible because ‘Smokers used to be embarrassed to buy the cheaper brands, but the culture has changed radically’ (retailer quoted in Retail Newsagent). 233

Thus, there was evidence from the commercial literature to suggest that down-trading was smokers’ behavioural response to the reduced affordability discussed in the previous chapter. Down-trading was discussed as an issue in the commercial literature throughout the time period studied (2008–14) and was seen as a more likely response to financial pressures than smoking cessation.

Methods

The population-weighted proportion of smokers using cheap tobacco (types and sources) at each survey was calculated. To test for linear trends in these proportions, clustered logistic regression analyses using GEEs were employed. These analyses used a binomial distribution with a logit link function and an unstructured correlation matrix (or exchangeable when the unstructured failed to converge). In these analyses, covariates are not controlled for.

Results

Results

Changes in volumes of FM and RYO sticks sold varied by segment (Figure 13 and see also Table S14 in Report Supplementary Material 2). Among FM cigarettes, annual volumes of premium and mid-price products declined markedly throughout the period (54% and 61% declines, respectively). Volumes of FM value cigarettes grew by 126% over the study period, although growth slowed from mid-2011. From their introduction in 2012, volumes of FM subvalue cigarettes increased to 4.3 billion sticks in 2015.

FIGURE 13.

Nielsen data study 5: monthly tobacco sales volumes (millions of sticks sold) for FM cigarettes and RYO tobacco, by price segments.

Within the RYO tobacco segment, from 2009 to 2015, premium sales declined by 43%, whereas mid-price volumes grew by 78% overall, although they have stagnated since 2013. RYO value tobacco grew to 2.3 billion stick equivalents, but growth slowed after 2012.

Implications

The market share of price segments varied considerably over the study period. FM mid-price cigarettes and RYO premium tobacco were the most popular of each type of product at the beginning, whereas at the end of the study period FM value cigarettes and RYO mid-price tobacco were most popular. Tobacco sales data therefore provide further evidence of down-trading, as sales of expensive FM and RYO products declined, but sales of cheap FM and RYO products grew.

Methods

Measures

The variable of primary interest was tobacco product type. Participants provided the specific brand variant of FM cigarettes they usually (if they indicated having a regular brand) or currently smoked. These were classified into one of three price segments: premium, mid-price or value, as described in Chapter 3. The two cheapest FM segments (value and subvalue) were combined because of the low prevalence of subvalue brands in our ITC sample, a segment that only appeared on the market from 2012. RYO tobacco was a fourth category.

We also examined the tobacco taxation increase rate, brand and product loyalty, when appropriate, to assess degree of adherence to particular brands/products, the HSI to assess level of tobacco dependence and frequency of purchases from cheap (probably illicit) sources. Sociodemographic variables (sex, age, education, income, region of residence, relationship status and ethnicity) were also included, as were the sample design control variables of recruitment type (replenishment vs. recontacted samples) and survey mode (telephone vs. online). More details relating to these variables are provided in Chapter 2. The influence of tobacco dependence and sociodemographics is discussed in Chapter 9, and of frequency of purchases from cheap sources in Chapter 8; however, these variables were all controlled for in the present analyses.

For the longitudinal analysis (analysis 4b), we assessed the prevalence of product switching by examining the relationship between tobacco product type at baseline and product type at outcome. Baseline RYO tobacco smokers were excluded as they were not able to trade down. This resulted in the baseline–outcome matrix presented in Table 14, which we used to classify participants to one of four outcomes: (1) traded up/no change, that is participants who remained smoking FM cigarettes from the same price segment or traded up to a more expensive segment (trading up was combined with no change because of its low prevalence of n = 74, comprising 5.4% of the sample); (2) traded down FM cigarettes, that is participants changed to a cheaper FM cigarette; (3) switch to RYO tobacco, that is participants switched from FM cigarettes to RYO tobacco; and (4) stop purchasing (having quit for at least 6 months with no relapse recorded at the time of the outcome survey). As the length of their quit attempt could not be determined, participants recorded as quit for < 6 months at outcome were excluded from analysis 4b.

Analyses

Initially, patterns in the prevalence of cheap tobacco product use over time were assessed descriptively, then two analyses performed.

Results

Sample characteristics for the cross-sectional sample (analysis 4a) are presented in Table 15 and are broadly similar to the characteristics of the longitudinal subsample in analysis 4b (data not shown). Overall, 25.0% of participants smoked premium FM cigarettes, 25.6% smoked mid-price FM cigarettes, 19.0% smoked value FM cigarettes and 30.4% smoked RYO tobacco. Only 4.2% of observations were made in an interval in which the tobacco tax increase rate was 5%. The vast majority (93.2%) were brand and product loyal (87.6%).

Patterns in the prevalence of cheap tobacco product use over time

Overall, the use of cheap tobacco (mid-price FM cigarettes, value FM cigarettes and RYO tobacco) in our sample increased modestly, but significantly, from 72.4% (95% CI 70.1% to 74.7%) in 2008 to 77.6% (95% CI 75.0% to 80.2%) in 2014, with a corresponding drop in premium FM cigarettes from 27.6% (95% CI 25.3% to 29.9%) in 2008 to 22.4% (95% CI 19.8% to 25.0%) in 2014 (Figure 14). The greatest increase (17 percentage points) was observed for value FM cigarettes and there was actually a decrease of 15 percentage points in mid-price FM cigarette use. These time-related trends were interrupted in the 2012 tax year, when the rate of tobacco tax increase was highest (5% above inflation) and there was an apparent marked dip in premium and mid-price FM cigarettes and a spike in value FM cigarette and RYO tobacco use. These patterns mirror those in the Nielsen tobacco sales data.

FIGURE 14.

International Tobacco Control Policy Evaluation Project data study 4: percentage of participants using each product type over time (tax year) with tobacco tax increase rates (per cent above inflation) in brackets. Note that these data are from the cross-sectional sample (analysis 4a, n = 2418, observations = 4339). Error bars represent 95% CIs.

Limitations

Owing to survey timings, only 4.2% of observations across the data set experienced the highest tobacco tax increase rate (of 5%), which could be allocated only to participants surveyed during February 2013 and, hence, there is the possibility that this is a seasonal effect only. Our analysis of Nielsen data indicated that price increases in response to taxation changes occurred throughout the year (see Chapter 4), rather than all at once, which would make it less likely that there were seasonal patterns for switching and down-trading.

Conclusions

Cheaper tobacco products (mid-price FM cigarettes, value FM cigarettes and RYO tobacco) were used by 72% of our sample in 2008, consistent with the 71% reported in our earlier work. 12 This increased by six percentage points by 2014; a modest increase in comparison to prior years (15 percentage points increase from 2001 to 200812), although use of the cheapest type of FM cigarettes (value FM cigarettes) rose by 17 percentage points during the period. Continuing with the same products and down-trading to a cheaper FM product were more common outcomes for smokers than stopping purchasing tobacco. Switching to RYO tobacco was a similar proportion to stopping.

Methods

Results

In analysis 5a (Table 18), a quit attempt was made by 39.4% of smokers. Overall, baseline tobacco product type significantly predicted making a quit attempt. However, those using cheaper tobacco (mid-price FM cigarettes, value FM cigarettes or RYO tobacco) did not significantly differ from users of premium FM cigarettes in their likelihood of trying to quit. To help understand the overall significant finding better, when RYO tobacco was used as the reference category, those smoking mid-price FM cigarettes (OR 1.54, 95% CI 1.08 to 2.18) and value FM cigarettes (OR 1.57, 95% CI 1.07 to 2.31) were significantly more likely to try to quit (data not shown). Quit attempts were more likely at the higher tax increase rate of 2% rather than 1% above inflation (OR for 5% was consistent with the 2% tax increase rate, but non-significant). No significant associations were observed between quit attempts and brand or product loyalty.

In analysis 5b, sustaining a quit attempt of at least 6 months was achieved by 9.7% of participants and was significantly more likely among smokers of mid-price FM cigarettes, value FM cigarettes and RYO tobacco, compared with premium FM cigarettes (see Table 18). Quitting was more likely the higher the tobacco tax increase rate at the outcome survey. Neither brand nor product loyalty was significantly associated with sustained quitting.

Limitations

As described in Chapter 6, only 4.2% of observations across the data set experienced the highest tobacco tax increase rate (of 5%) in February 2013 and, hence, it is possible that this is a seasonal effect only. However, smokers surveyed in February 2013 could have begun their quit attempt any time between July 2010 (if they were interviewed at the beginning of the second survey) and July 2012 in order to have achieved a minimum of 6 months quit by the time they were reinterviewed for the third survey in February 2013; this would mean that there were a number of months over which the March 2012 5% tax increase could have had an impact on the findings (March 2012–July 2012), thus reducing the chances of a seasonal effect being responsible. Our assessment of quitting for at least 6 months meant that those who had quit for < 6 months at the follow-up surveys were excluded. As most smokers relapse soon after quitting, our interest was in those who were able to sustain a quit attempt for a longer period of time.

Implications

Overall, baseline tobacco product type significantly predicted making a quit attempt and the likelihood of achieving at least 6 months’ abstinence at follow-up. Those smoking mid-price or value FM cigarettes were more likely to make a quit attempt than those smoking RYO tobacco. Six months’ abstinence was significantly more likely among smokers of mid-price FM cigarettes, value FM cigarettes and RYO tobacco than among premium FM cigarette smokers. Quit attempts and success were also more likely during periods of higher tobacco tax increase rates. Whether smokers had a usual brand, were exclusive or were mixed users was also not associated with the quitting outcomes.

Methods

Analyses

Each observation was allocated to the calendar year of the survey (see Table S15 in Report Supplementary Material 2) and time was treated as a continuous variable in the analyses. TTFC was used as the indicator of nicotine dependence and treated as a continuous variable (CPD and HSI were used to calculate tobacco weights so could not be used).

There were too few observations for Canada and the USA (see Table 19) to enable reliable statistical analyses of trends. To compare the weight per cigarette across countries, we therefore pooled the data across all survey waves. A clustered linear regression analysis was carried out, with weight as outcome and country as predictor. We controlled for TTFC and the sociodemographic measures, and used maximum likelihood estimation with a robust variance estimator.

To assess changes in weight of RYO tobacco over time, and the effects of sociodemographic differences, only the UK and Australia (n = 1349, observations = 2705) were included. We conducted multilevel linear regression analyses clustered over individuals, with maximum likelihood estimation and robust variance estimators. The outcome variable was weight of tobacco per cigarette and the predictor variables were time (calendar year), country (UK vs. Australia), TTFC and the sociodemographic variables. We first examined the unadjusted univariate effects of each predictor variable and then looked at the fully adjusted multivariate effects in which all predictors were included simultaneously. We report the sociodemographic findings in Chapter 9.

Results

Sample characteristics are presented in Table 19. A missing data comparison between the 1924 individuals suitable for the RYO tobacco weight study and the final sample of 1639 individuals indicated that the final sample featured participants who were significantly more likely to be from Australia or the UK, be male, younger, with a high income, and to be white with an English-speaking background and likely to have higher nicotine dependence (shorter TTFC). These variables were controlled for in the statistical analysis. No significant differences between the samples were found on education level (see Table S16 in Report Supplementary Material 2).

Just over one-quarter (25.8%) of UK smokers from 2006 to 2015 (six ITC surveys) used RYO tobacco exclusively; this proportion was 3.5% in the USA, 6.0% in Canada and 13.8% in Australia. Pooled data across the survey period indicated that the mean weight of tobacco per RYO cigarette was lowest in the UK (mean 0.51 g, SD 0.32 g) and Australia (mean 0.53 g, SD 0.28 g), intermediate in Canada (mean 0.76 g, SD 0.45) and highest in the USA (mean 1.07 g, SD 0.51). All pairwise country differences (clustered linear regression results adjusted for TTFC and SES) were statistically significant, except for the UK and Australia, as the weight of tobacco per RYO cigarette was comparable (see Table S17 in Report Supplementary Material 2).

For the UK and Australia, there was a statistically significant decrease of approximately 0.01 g per RYO cigarette per year using clustered linear regression analysis of tobacco weights over time [β= –0.011 (SE 0.0021); p < 0.001 for the unadjusted model]. This is roughly equivalent to a 2% decrease of the average weight per cigarette per year and persisted even after adjusting for TTFC and SES [β = –0.014 (SE 0.0021); p < 0.001]. As may be seen in Figure 15, in the UK, the mean tobacco weight per cigarette decreased from 0.59 g (SD 0.42 g) in 2006 to 0.46 g (SD 0.25 g) in 2014. In Australia, the comparable change was 0.59 g (SD 0.29 g) to 0.48 g (SD 0.26 g). The differences in weights between the two countries was not statistically significant [β = –0.026 (SE 0.014); p = 0.057 for the fully adjusted model].

FIGURE 15.

International Tobacco Control Policy Evaluation Project data study 6: changes in the mean weight of tobacco (in grams) per RYO cigarette over time for the UK and Australia (AU). Also shown are the model-predicted values from a fully adjusted clustered linear regression, controlling for TTFC, sex, age, ethnicity, level of education and income. a, No data were collected from the UK in 2011 or from Australia in 2012. Sample sizes from Australia for 2009 and 2011 were low (n = 26 and n = 25, respectively, see text and Table 19 for more details).

Limitations

International Tobacco Control Policy Evaluation Project (ITC) data are self-reported and did not directly report on the weight of tobacco used in hand-rolled cigarettes, so that had to be calculated using several other points of information. The resulting estimates could not be verified by independently ascertaining the weight through a direct measurement process. It is possible that some participants might have shared their pouches, although sharing between individuals might balance out over time, thus mitigating any biasing effect.

Although the ITC data set is larger than any other data set previously used to investigate the weight of tobacco in hand-rolled cigarettes, there were still not enough observations to consider how the weight of tobacco changed over time in Canada and the USA. This analysis was also limited because ITC survey waves did not cover all countries on a yearly basis.

Implications

UK smokers reduced the weight of RYO tobacco over time, by about 2% per year. This will have offset, to some extent, the annual tax increases designed to make tobacco less affordable. Smokers in higher tax/tobacco price countries, such as the UK and Australia, used less tobacco per RYO cigarette than those in the USA and Canada, where taxes and prices were generally lower. This suggests, again, that the weight of tobacco is linked to price changes. Although the number of RYO cigarettes smoked per day was similar across the four countries, the UK had the lowest at 16.9 cigarettes and Canada had the highest at 20.4 cigarettes, suggesting that smokers in high tobacco price countries also smoke fewer cigarettes.

Any reduction in the weight of tobacco used in RYO cigarettes is unlikely to benefit the health of smokers, as evidence to date indicates that thinner cigarettes do very similar damage to fatter ones. 235–239

Applying the most recent weight of RYO tobacco in the UK we observed here (0.46 g per cigarette), would mean about 2174 cigarettes per kilogram of RYO tobacco. Using 2017 taxation rates, this means that on a per cigarette basis, a tax of £0.28 per FM cigarette would apply, compared with just £0.10 per RYO cigarette, just over one-third of the level.

Methods

First, we assess the proportion of smokers purchasing from non-UK/non-store sources and, of these, which sources were the most common over the study period [see ITC data study 3: smokers’ tobacco purchasing behaviours by product and store type (licit purchases), Methods and Limitations in Chapter 6]. Then we compare the price of products purchased through these sources with the prices of products purchased through UK store-based sources (see ITC data study 1: prices paid for tobacco by smokers over time, Methods and Limitations in Chapter 4). Sample characteristics are shown in Table 7.

Results

Non-UK/non-store purchase prices (ITC study 1)

Overall, the most common purchase bought from non-UK/non-store sources was FM-C (≈10%), followed by RYO tobacco (≈6%); very few were FM-P (< 2%). Given the small proportion of tobacco purchases that were non-UK/non-store FM-P and, hence, limited data to calculate median prices, trend analyses were not conducted. Median prices for FM-P purchased from non-UK/non-store sources were consistently at least £0.02 below that of FM-P from UK store-based sources from 2002 to 2014, although with large fluctuations (Figure 16). For FM-C and RYO tobacco purchases, more consistent price estimates could be obtained. Median prices for FM-C increased significantly from £0.11 per stick in 2002 to £0.16 in 2014 and were often less than half the price of FM-C from UK store-based sources. Median prices for RYO tobacco purchases from non-UK/non-store significantly increased from £0.05 per stick in 2005 to £0.08 in 2014, considerably cheaper than RYO tobacco from UK store-based sources. For all product types, median prices from non-UK/non-store sources were often cheaper than the cheapest products available from UK store-based sources.

FIGURE 16.

International Tobacco Control Policy Evaluation Project data study 1: weighted median prices of different product types purchased from UK store-based and non-UK/non-store-based sources. (a) FM-P and RYO tobacco; and (b) FM-C. Medians for non-UK/non-store sources are superimposed on the UK store-based purchases originally shown in Figure 6. For UK store-based sources, black lines represent medians, with shading for 95% ranges (top and bottom 2.5% excluded, blue lines are 25th and 75th percentiles). Dashed lines are medians for non-UK/non-store sources. All prices are adjusted to 2014 values and based on the most recent reported tobacco purchase. a, No data collected for these years.

As already discussed, it is difficult to ascertain with any certainty which purchases were truly illicit and which were legal untaxed sources. When we examined the trends in prices from only the sources arguably most likely to represent illicit purchases (i.e. from informal sellers, friends/relatives and non-UK/non-store ‘other’ responses), median values were comparable to those presented in Figure 16 from all non-UK/non-store sources combined (see Figure S7 in Report Supplementary Material 3). However, sample sizes were low and these findings should be interpreted with caution.

Implications

Overall, there was no evidence of an increase in the proportion of non-UK/non-store purchases, even when using the most liberal assumptions. The majority of non-UK/non-store purchases appeared to be tax avoidance, constituting duty free (which significantly increased during the study) or sources outside the UK (which significantly decreased, most likely influenced by the UK ‘Great Recession’ that commenced at the start of 2008). Typically, less than one-quarter of non-UK/non-store purchases and < 5% of all tobacco purchases were from sources likely to constitute tax evasion (e.g. informal sellers or from friends or relatives). Purchasing tobacco from friends or relatives significantly increased from 2002 to 2014, peaking in 2010 at just under one-quarter of all non-UK/non-store purchases. More research is required to understand whether this constitutes a form of tax avoidance and/or evasion, or an effort to pool resources and buy in bulk. Purchases from informal sellers significantly decreased. Purchases from the internet and from toll-free telephone lines together accounted for < 3% of all non-UK/non-store purchases. FM-C and RYO were the products most commonly purchased from non-UK/non-store sources, but purchase of both decreased during the period of study.

These data indicate that UK strategies aimed at reducing illicit tobacco supplies and use77,84,85 appear to have been successful and hence we found no evidence to support TI arguments that illicit tobacco usage has been increasing. The price analysis presented in Chapter 4 supports the theory that smokers have little need to turn to illicit products to obtain cheap tobacco, because large price differences exist between different types of products purchased legally (from UK store-based sources). Thus, although the median prices for all product types (FM-P, FM-C and RYO tobacco) purchased from non-UK/non-store sources were typically lower than the cheapest products from UK store-based sources, all three product types could be purchased legally in 2014 below their median prices in 2002.

Methods

This study uses the same sample, methods and limitations as described in Chapter 5, and sample characteristics are shown in Table 11. In brief, our new measure of tobacco affordability was regressed onto purchase sources, separately for FM cigarette and RYO tobacco smokers. The fully adjusted model controlled for time (tobacco tax year), sex, age, region of residence, ethnicity, education and TTFC (see Chapters 2 and 5 for more details).

Results

Sixteen per cent of smokers included in this sample purchased from non-UK/non-store sources. RYO tobacco smokers were more likely to purchase tobacco from non-UK/non-store sources (28.1%) than FM cigarette smokers (13.9%) (see Table 11). FM cigarettes were significantly more affordable from non-UK/non-store sources [β = 4.15 (SE 0.25); p < 0.001 for the unadjusted and β = 4.10 (SE 0.25); p < 0.001 for the fully adjusted model]. Similarly, RYO tobacco was also significantly more affordable from non-UK/non-store sources, although the effect was not as strong [β = 2.50 (SE 0.31); p < 0.001 for the unadjusted and β = 2.35 (SE 0.30); p < 0.001 for the fully adjusted model].

Purchasing FM cigarettes from non-UK/non-store sources rather than UK stores conferred a saving of around 5% of smokers’ annual incomes (Figure 17), which was approximately £1300 in 2014. A similar saving was observed from purchasing RYO cigarettes from non-UK/non-store sources rather than UK stores.

FIGURE 17.

International Tobacco Control Policy Evaluation Project study 2: individualised tobacco affordability over time for (a) FM cigarette smokers; and (b) RYO tobacco smokers, for purchases made from all sources and separately for non-UK/non-store sources only. Affordability estimates are from the fully adjusted model 2 (see Chapter 5), controlling for all covariates. a, No ITC data collected for this year.

Methods

Analysis 4a: product type

Overall, 7.6% of our sample reported frequently purchasing from illicit sources. This figure was 8.4% among users of premium FM cigarettes, 3.6% for mid-price FM cigarettes, 3.6% for value FM cigarettes and 12.7% for RYO tobacco.

Smokers who reported a high frequency of illicit purchases were significantly less likely than those with a low frequency to use mid-price FM cigarettes [relative risk ratio (RRR) 0.45, 95% CI 0.29 to 0.67; p < 0.001] or value FM cigarettes (RRR 0.70, 95% CI 0.35 to 0.93; p = 0.023), and significantly more likely to use RYO tobacco (RRR 1.68, 95% CI 1.22 to 2.33; p = 0.002) than to use premium FM cigarettes (see Table S18 in Report Supplementary Material 2 for full results).

Analysis 4b: changes in smoking behaviour

Trading up (5.5%) or not changing (66.5%) between the baseline and outcome surveys was most common (72.0% combined); 13.5% traded down to cheaper FM cigarettes, 6.2% switched from FM cigarettes to RYO tobacco and 8.3% achieved a current sustained quit attempt of at least 6 months. Frequency of illicit purchases had no significant association with any of these measures, even after controlling for all included covariates (p-values for all RRRs > 0.05) (see Table S19 in Report Supplementary Material 2 for full results).

Implications

Factory-made cigarette smokers who have easy access to cheaper premium FM products, through sources likely to be illicit, appear to have less incentive to smoke cheaper, non-premium, FM brands. However, this analysis indicated that those who frequently purchased from potentially illicit sources were most likely to buy RYO tobacco. In the longitudinal analysis, the frequency of purchasing illicit products did not affect subsequent purchasing patterns (i.e. trading down, switching or sustained quitting).

Methods

Here we report the results of longitudinal analyses 5a and 5b of ITC data study 5 (see Chapter 7), pertaining only to associations with illicit purchases. Analysis 5a (1204 participants providing 2202 observations) examined making quit attempts and analysis 5b (1194 participants providing 2017 observations) examined 6 months’ sustained quitting. Both analyses were longitudinal clustered (aggregated over three survey pairs) logistic regressions, controlling for product type (premium FM cigarettes, mid-price FM cigarettes, value FM cigarettes or RYO tobacco), tobacco tax increase rates at outcome, HSI, brand loyalty, product loyalty, sex, age, education, income, region of residence, relationship status, ethnicity and survey design variables. Detailed methodology and limitations are provided in Chapter 7.

Implications

Frequent purchasing from sources that were potentially used to avoid or evade tax did not appear to influence subsequent quit attempts or success.

Methods

This study uses the same sample and analyses as in ITC data study 2, described in Chapter 5, of daily FM cigarette smokers from surveys 1–10 (2002–14) and RYO tobacco smokers from surveys 6–10 (2006–10). We focus on the fully adjusted clustered linear regression analyses (model 2), regressing affordability onto all predictor variables (tobacco tax year, smoking dependence, purchase source and sociodemographic variables) concurrently, to identify any significant independent predictors. Here we report the findings pertaining only to the sociodemographic variables: sex, age (and age squared, to test for non-linear associations with age), ethnicity, region, education and smoking dependence (TTFC). Detailed methodology and limitations are provided in Chapter 5.

Results

In the overall sample, FM cigarette and RYO tobacco smokers were broadly comparable in age and education, but differed on other characteristics (see Table 11). FM cigarette smokers were less likely to be white (95% vs. 98% RYO tobacco smokers) and male (38% FM vs. 63% RYO tobacco smokers) and there were regional differences (e.g. more RYO tobacco smokers from the South West region). Cigarette consumption was similar (16.7 FM cigarettes vs. 17.0 RYO tobacco), but a higher proportion of RYO tobacco smokers were most addicted (19.8% of RYO tobacco smokers smoked within 5 minutes of waking vs. 15.4% of FM cigarette smokers).

Factory-made cigarettes

In the multilevel linear regression analyses, the following subpopulations were estimated to find smoking FM cigarettes less affordable (Table 21): women (vs. men); older smokers (but please note that there were very few under 25s in the sample); socioeconomically disadvantaged smokers, as measured by education and region (smokers living in Scotland, Northern Ireland and the north-east of England, areas which include some of the poorest localities in the UK). Smokers who were more dependent (as assessed by TTFC) also found smoking less affordable. These sociodemographic and smoking dependence patterns were the same even when excluding non-UK/non-store purchases (see Table S11 in Report Supplementary Material 2).

TABLE 21 - International Tobacco Control Policy Evaluation Project data study 2: linear random-effects clustered regression analyses of individualised affordability regressed on sociodemographic and other covariates (fully adjusted), separately for FM cigarette and RYO tobacco smokers

Ref, reference.

Notes

Analyses are fully adjusted for tobacco tax year, sex, age, age squared, region, ethnicity, education, TTFC and purchase source.

Chi-squared statistics are for overall effects of linear trends (time) and any variables with three or more categories.

Values in bold are significant.

Sample characteristics	FM cigarettes (n = 3420, observations = 7475)			RYO tobacco (n = 734, observations = 1468)
	β	SE	p-value	β	SE	p-value
Sex
Female	Ref			Ref
Male	1.93	0.31	< 0.001	0.90	0.39	0.021
Age (continuous)
Age	0.23	0.05	< 0.001	0.16	0.08	0.047
Age squared	–0.0033	0.00060	< 0.001	–0.0023	0.00083	0.006
Region	χ2(11) = 31.12; p = 0.0011			χ2(11) = 18.87; p = 0.064
London	Ref			Ref
Yorkshire and The Humber	–0.45	0.64	0.481	–0.84	0.97	0.385
East Midlands	–0.46	0.69	0.505	–0.64	0.88	0.465
Eastern	–0.29	0.64	0.648	0.67	0.83	0.421
North East	–1.82	0.81	0.024	–0.77	1.12	0.492
South East	0.19	0.55	0.729	0.41	0.79	0.600
South West	–0.71	0.75	0.322	0.57	0.79	0.473
West Midlands	–0.82	0.65	0.210	–0.82	0.88	0.350
North West	–0.85	0.61	0.163	0.89	0.86	0.300
Wales	–0.33	0.80	0.681	–1.94	0.93	0.037
Scotland	–1.50	0.61	0.013	0.18	0.92	0.844
Northern Ireland	–4.09	0.93	0.000	–0.38	1.32	0.771
Ethnicity
White	Ref			Ref
Not white	–0.26	0.68	0.698	–0.08	1.13	0.947
Education	χ2(2) = 95.14; p < 0.0001			χ2(2) = 17.15; p = 0.0002
Low	Ref			Ref
Moderate	1.45	0.35	< 0.001	1.49	0.43	< 0.001
High	4.23	0.44	< 0.001	1.75	0.54	0.001
TTFC (minutes)	χ2(3) = 116.41; p < 0.0001			χ2(3) = 23.89; p < 0.0001
> 60	Ref			Ref
31–60	–1.25	0.31	< 0.001	–0.05	0.48	0.913
6–30	–2.33	0.31	< 0.001	–1.10	0.40	0.019
Within 5	–3.78	0.38	< 0.001	–2.00	0.53	< 0.001

Figure 18 illustrates individualised affordability for FM cigarette smokers, showing female smokers, those with high levels of education and those who are more dependent (i.e. smoke their first cigarette within 5 minutes of waking) compared with overall affordability. It is evident that these demographic and smoking groups are spending relatively more of their income on tobacco.

FIGURE 18.

ITC data study 2: individualised affordability for FM cigarette smokers, showing female smokers, those with high levels of education and more dependent smokers, compared with overall affordability. Shaded areas represent 95% CIs. Affordability estimates are from the fully adjusted model 2 for FM cigarette smokers only. More dependent smokers are those who smoke their first cigarette within 5 minutes of waking (TTFC < 5 minutes). a, No ITC data collected for this year.

Roll-your-own tobacco

The same subpopulations were estimated to find RYO tobacco smoking less affordable (see Table 21 and see also Table S12 in Report Supplementary Material 2), with one exception (Wales was the only area to find RYO tobacco smoking less affordable). Again, Wales has areas of particular socioeconomic disadvantage. However, the overlapping CIs in Figure 19 illustrate that RYO tobacco is uniformly cheap and, as yet, there is not as much differentiation by sociodemographic and dependence factors, compared with FM cigarette smokers.

FIGURE 19.

International Tobacco Control Policy Evaluation Project data study 2: individualised affordability for RYO tobacco smokers, showing female smokers, those with high levels of education and more dependent smokers, compared with overall affordability. Shaded areas represent 95% CIs. Affordability estimates are from the fully adjusted model 2 for RYO smokers only (see text for details). More dependent smokers are those who smoke their first cigarette within 5 minutes of waking (TTFC < 5 minutes). a, No ITC data collected for this year.

To illustrate the differences in affordability across sociodemographic and smoking dependence factors, we used the fully adjusted regression coefficients from model 2 (see Tables S11 and S12 in Report Supplementary Material 2 for a complete list of regression coefficients) to plot the affordability curves for two hypothetical groups of smokers by age in 2014, one group with low affordability and one with high affordability. These plots were based on the fact that model-based estimates can be obtained for any theoretical values of the included variables, using the regression equation (y = b₀ + b₁x₁ + b₂x₂ + . . . + b_nx_n), where y is the outcome (in this case affordability), x_n represents each predictor variable and b_n is its corresponding standardised regression coefficient (b₀ is the constant or y-intercept). As we wanted to plot affordability over different ages, we did not estimate values for age or age squared, so these were left as x and x² in our equations, respectively. For the remainder of the variables (all categorical), we chose the values corresponding to the highest or lowest affordability. Categorical variables are dummy-coded, so x is set to 0 for all non-selected categories and set to 1 for the selected category (b for the reference category = 0). For example, for our low-affordability curve (based on FM cigarettes, as they are generally more expensive than RYO tobacco), we selected TTFC to be within 5 minutes, as this was associated with the lowest regression coefficient (–3.78) (see Table 21), so the part of our equation pertaining to TTFC was [(0 × 0) + (–1.25 × 0) + (–2.33 × 0) + (–3.78 × 1)], which solves to –3.78. Thus, we selected the high-affordability group to be white, male, smoking RYO tobacco, residing in London, with high levels of education, smoking their first cigarette after 60 minutes from waking and purchasing from non-UK/non-store sources. The low-affordability group was selected to be female, non-white, smoking FM cigarettes, living in Northern Ireland, with low levels of education, highly addicted (smoking their first cigarette within 5 minutes of waking) and purchasing from UK store-based sources. Time was set to 2014 for both groups. The curves are displayed in Figure 20, which shows how affordability decreases with age, but illustrates the large differences in affordability across all ages, with some groups, depending on their characteristics and product choices, potentially spending > 15% more of their incomes on tobacco than others.

FIGURE 20.

International Tobacco Control Policy Evaluation Project data study 2: individualised affordability in 2014, for two theoretical smokers by age [based on regression coefficients for the fully adjusted model 2 for RYO tobacco (high) and FM cigarette (low) smokers; see Report Supplementary Material 2]. The high affordability smoker is a white, male, RYO smoker, residing in London, with high levels of education, a TTFC > 60 minutes, and purchasing from a non-UK/non-store source. The low-affordability smoker is female, non-white, smokes FM cigarettes, resides in Northern Ireland, has low levels of education, is highly addicted (TTFC within 5 minutes of waking) and purchases from a UK store-based source.

Implications

Socioeconomically disadvantaged groups, women and dependent smokers were more likely to struggle to afford smoking as tobacco prices increased. In theory, we would expect that this would make them more likely to quit. However, from Chapters 6–8 we know that there were other options available: smoking cheaper products and smoking fewer CPD, or using less tobacco in RYO cigarettes. We now assess whether or not there were differences in quitting or reducing behaviours related to sociodemographics and dependence.

Methods

Methodological details and limitations are provided in Chapter 6, with basic sample characteristics in Table 15. Analyses were clustered, multinomial logistic regressions: analysis 4a was cross-sectional, whereas analysis 4b was longitudinal and included only exclusive FM cigarette smokers at baseline.

Results

Notable characteristics of the cross-sectional aggregate data sample (see Table 15) include < 8% classified as being highly dependent and an under-representation of younger smokers (with over three-quarters of the sample being aged ≥ 40 years).

Implications

The findings that men, but also younger smokers, were more likely to switch to RYO tobacco may be the result of shifting perceptions among younger groups of the desirability of RYO tobacco smoking, which has been traditionally more common among men. 12 Furthermore, RYO tobacco was marketed as a natural product that may have particularly appealed to young environmentally conscious smokers. 241

Low-income smokers were more likely to switch to RYO tobacco than high-income smokers; and those with low levels of education more likely than those with moderate levels of education to trade down to cheaper FM products. However, there were no other clear indications that down-trading was more common among socioeconomically disadvantaged smokers overall. This might be because socioeconomically disadvantaged smokers were already smoking cheaper products and thus did not have down-trading as an option. However, the data also suggested that there was no socioeconomic difference in quitting. This explains why the over-representation of socioeconomic disadvantage among smokers has been perpetuated, despite high tobacco taxes. If equal proportions of low-SES and high-SES smokers quit, there will still be more low SES smokers because there was a larger pool to start with.

Limitations

We note that for this study, the first outcome survey was in 2010/11 and the largest tax increases were in 2011 and 2012, so it is possible that had the study started post 2012 then more socioeconomic differentiation would be apparent. We could not isolate the post-2012 period due to an insufficient number of cases. Another limitation was that the numbers for the highest dependence (HSI) group were relatively low (see Table 15). This may have contributed to our lack of significant findings relating to dependence, although we note that the ORs were in the expected direction, with trends for the highest HSI group to be more likely to use value FM cigarettes or RYO tobacco (than premium FM cigarettes) and less likely to achieve a sustained quit attempt. Future research needs to examine this further.

Methods

Here we report the results of longitudinal analyses 5a and 5b of ITC data study 5, pertaining only to sociodemographics and dependence (HSI). Both analyses were longitudinal, clustered, logistic regressions, with analysis 5a predicting making quit attempts and analysis 5b predicting achieving at least 6 months’ sustained abstinence from smoking. We focus our reporting on model 2 of these analyses, which fully controlled for all included predictors.

Results