Article Text

A comparison of alternative methods for measuring cigarette prices
1. Frank J Chaloupka1,
2. John A Tauras2,
3. Julia H Strasser3,
4. Gordon Willis3,
5. James T Gibson4,
6. Anne M Hartman3
1. 1Department of Economics, Institute for Health Research and Policy, University of Illinois at Chicago, Chicago, Illinois, USA
2. 2Department of Economics, University of Illinois at Chicago, Chicago, Illinois, USA
3. 3Division of Cancer Control and Population Sciences, National Cancer Institute, Rockville, Maryland, USA
4. 4Information Management Services, Inc., Calverton, Maryland, USA
1. Correspondence to Frank J Chaloupka, Institute for Health Research and Policy, University of Illinois at Chicago, 1747 W. Roosevelt, Room 558, Chicago, IL 60608, USA; Anne M Hartman, National Cancer Institute, 9609 Medical Center Drive, Room 3E-416, Rockville, MD 20850, USA; hartmana{at}mail.nih.gov

## Abstract

Background Government agencies, public health organisations and tobacco control researchers rely on accurate estimates of cigarette prices for a variety of purposes. Since the 1950s, the Tax Burden on Tobacco (TBOT) has served as the most widely used source of this price data despite its limitations.

Purpose This paper compares the prices and collection methods of the TBOT retail-based data and the 2003 and 2006/2007 waves of the population-based Tobacco Use Supplement to the Current Population Survey (TUS-CPS).

Methods From the TUS-CPS, we constructed multiple state-level measures of cigarette prices, including weighted average prices per pack (based on average prices for single-pack purchases and average prices for carton purchases) and compared these with the weighted average price data reported in the TBOT. We also constructed several measures of tax avoidance from the TUS-CPS self-reported data.

Results For the 2003 wave, the average TUS-CPS price was 71 cents per pack less than the average TBOT price; for the 2006/2007 wave, the difference was 47 cents. TUS-CPS and TBOT prices were also significantly different at the state level. However, these differences varied widely by state due to tax avoidance opportunities, such as cross-border purchasing.

Conclusions The TUS-CPS can be used to construct valid measures of cigarette prices. Unlike the TBOT, the TUS-CPS captures the effect of price-reducing marketing strategies, as well as tax avoidance practices and non-traditional types of purchasing. Thus, self-reported data like TUS-CPS appear to have advantages over TBOT in estimating the ‘real’ price that smokers face.

## Introduction

Extensive economic research has demonstrated that changes in cigarette prices lead to significant changes in smoking behaviour.1 ,2 Increases in prices lead current smokers to quit smoking, keep former smokers from restarting, prevent youth from taking up smoking and reduce the number of cigarettes consumed by continuing smokers. Smoking prevalence among young people, the poor and pregnant women is particularly responsive to price changes.1 ,2

Governments considering increases in cigarette excise taxes rely on estimates of the price elasticity of cigarette demand in order to project the impact of tax increases on their revenues and on smoking behaviour. The accuracy of these projections depends on the reliability and validity of the price data on which price elasticity estimates are based. Estimates of the overall price elasticity of cigarette demand in the USA generally range from −0.25 to −0.5, implying that a 10% increase in cigarette prices will reduce cigarette consumption by 2.5–5%.1

US tobacco price studies have primarily used retail-based data collection. Most of these demand studies have relied on state average cigarette prices obtained from the annual Tax Burden on Tobacco report (TBOT).3 A few studies have used other sources, including prices obtained from retail scanner-based databases, prices from the American Chamber of Commerce Researchers Association and prices collected from store observations and mail or phone retail surveys.4

For TBOT, Internal Tobacco Institute documents indicate that average state cigarette prices from 1954 and onward have been collected from a mail survey of retail establishments that sell cigarettes from all 50 states and Washington, DC; the survey is conducted in the fall (generally reporting prices as of November 1) each year.4 The TBOT was originally produced by the Tobacco Institute, which closed as a result of the tobacco industry settlement with the state of Minnesota. Since 1999, Orzechowski and Walker have produced the TBOT. In recent years, approximately 15 000–16 000 retailers have been surveyed, with response rates in the mid-teens and retailer numbers varying from a few dozen to nearly 100 in a given state (personal communication, Orzechowski, 2007).

Also recently, an increasing number of cross-sectional and longitudinal population surveys have collected data on cigarette prices and price-related promotions. These include the Tobacco Use Supplement to the Current Population Survey (TUS-CPS), the National Adult Tobacco Survey, several state surveys, the International Tobacco Control Policy Study, the Global Youth Tobacco Survey and Global Adult Tobacco Survey. The price data collected from these surveys are useful in developing aggregate measures of price for a state or country.1 ,4

Some of these surveys ask for the ‘average’ amount a person spends per pack, but this technique relies on respondents’ ability to perform mathematical computations. For respondents who buy by the pack, this is straightforward; it is somewhat more difficult for those who buy by the carton and even more difficult for those who take advantage of multipack specials (eg, buy-three-get-two-free). Alternatively, some ask how much the respondent paid for their purchase and what quantity was purchased.

This report compares measures of US cigarette prices obtained from two data sources, each with different collection methodologies: the retail-based TBOT3 and the self-report-based TUS-CPS. We compare the collection methods and data for these two systems and analyse their strengths and weaknesses for measuring average cigarette prices in the USA. We predicted that the retail-based TBOT estimates would be higher than the TUS-CPS self-reported population estimates, primarily because the latter reflects the impact of price-reducing marketing tactics and accounts for tax avoidance by smokers, and the former estimate does not.

### Tax Burden on Tobacco

As described in the TBOT,3 prices include state and federal excise taxes at the time they are reported but do not include sales taxes. Because methodological details are scarce, there is some ambiguity about whether or not they include local excise taxes. For many years, TBOT notes that local cigarette taxes are not included in the prices. In more recent years, it indicates that the prices for New York State include prices from New York City, where the local tax is significant, and prices for Illinois include prices from Chicago and Cook County, where the local taxes are also significant. An examination of the correlations between reported state prices and taxes (state only and state plus average local) suggests that local taxes in other states are included in TBOT prices for those states.

The TBOT provides only slight detail on the methods used to collect the price data and on how the average prices are computed. In older TBOT volumes, the prices are described as weighted averages of single pack, carton and vending machine sales, where weights are based on the share of each in overall sales. Data collected in the retail establishment survey include the ‘regular’ prices for full-priced (premium brand) single packs of cigarettes, full-priced cartons, generic (discount brand) single packs and generic cartons.

The sampling frame for the retailer survey is not well described and is likely to miss at least some cigarette retailers. For example, the mail survey does not appear to include retailers located on Native American reservations (where state and local taxes have historically not been applied); excluding these retailers will lead to an overestimate of the true average price of cigarettes faced by residents of a given state, with a greater bias in states where a larger share of the population lives near reservations. If the survey framework has not been regularly updated as new types of cigarette retailers emerged over the past 10–15 years (eg, discount cigarette shops), the resulting price estimates may be biased if the prices charged by emerging retailers differed from those charged by traditional retailers. Finally, these surveys do not appear to include internet cigarette vendors, often located in low-tax states, on reservations, or in other countries. Prices from these vendors can be considerably lower than prices from traditional bricks-and-mortar vendors, and excluding them from the survey may indicate a higher price than average customers pay.

Footnotes to the TBOT price tables indicate that the prices do not reflect temporary price reductions, suggesting that at least some industry marketing efforts that reduce price are not included in the estimates. These might include promotions involving free cigarettes (eg, buy-one-get-one-free promotions), coupons and straight price discounts provided to wholesalers and/or retailers that are passed on to consumers. Industry spending on price-reducing promotions accounts for the vast majority of its marketing spending, at approximately 80% of total advertising and promotional expenditures in the most recent years for which data are available.5

Additionally, the average state prices reported in the TBOT will overstate the price faced by individuals living in jurisdictions that do not apply local taxes and understate the price faced by individuals living in jurisdictions that do apply local taxes. This problem is magnified for individuals living near reservations where no state or local taxes are applied. Similarly, individuals in a high-tax jurisdiction who live near lower-tax jurisdictions will face a lower price for cigarettes than implied by the TBOT price for their state given the potential for cross-border purchases.

Despite the limitations of the TBOT price data, these are the most widely used cigarette price measure employed in studies of US cigarette demand, in large part because they are readily available and have been consistently collected and reported for nearly 60 years. Many studies apply different approaches to controlling for some of the limitations of the TBOT price data, particularly those related to the incentives for cross-border cigarette shopping and reservation purchases.1 ,6 ,7

### Tobacco use supplement to the current population survey

The TUS-CPS8 ,9 is a national household address-based probability sample surveying approximately 240 000 persons per wave every 3–4 years since 1992. The overall response rate is 82%; of these, 80% of responses are self-reports. Among the 185 000 self-reports, there are about 30 000 current smokers providing cigarette price information.

The TUS-CPS asks current smokers to report their most recent purchase information either by the pack or by the carton according to their typical purchasing habits. In collecting cigarette cost information, the TUS-CPS eliminates respondent calculation burden and simultaneously captures the use of any additional price-reducing promotions (eg, coupons) at most recent purchase. Some other surveys ask about ‘usual’ price, which has the advantage of capturing typical behaviour but will not pick up any changes in behaviour that may be relevant (eg, a buy-one-get-one-free promotion for a brand other than the respondent's usual brand). However, focusing on the most recent purchase minimises recall error and helps provide a consistent measure of current prices.

The TUS-CPS8 ,9 asks the following:

• What price did you pay for the LAST pack of cigarettes you bought? Please report the cost after using discounts or coupons.

• What price did you pay for the LAST carton of cigarettes you bought? Please report the cost after using discounts or coupons.

In addition to framing price questions in the most efficient way, the TUS-CPS also collects information on the location of cigarette purchases, which is particularly helpful for assessing tax avoidance behaviour among smokers. In the TUS-CPS, the survey asks whether the purchase was in an individual's home state, another state or elsewhere.

## Methods

We compared estimates of cigarette prices based on data from the 2003 and 2006/2007 TUS-CPS with those from the TBOT for those periods. Multiple, state-level measures of cigarette prices were constructed based on responses to the questions, including measures of average prices for single-pack purchases, average prices for carton purchases and weighted average prices for a single pack (using the single-pack price derived from the carton price and weighting by the shares of smokers who report buying single packs and buying cartons). Prices were constructed for all purchases, purchases made in-state only, purchases made out-of-state, purchases made from other sources and purchases made out-of-state or from other sources.

The calculation methods are outlined below:

TUS-CPS Average Cigarette Price per Pack Purchases from All/In-State/Out-of-State/Other Sources:

[(fraction of current smokers who buy single packs) ×  (their average reported price per pack)]+[(fraction of current smokers who buy cartons) × (their average reported price per carton converted* to price per pack)]

*To convert price per carton to price per pack, we divided price per carton by 10 (the number of packs in a carton).

• All: Reported prices from all smokers for national estimates or for all smokers who are residents of the given state for state estimates regardless of where they purchased their last pack or carton of cigarettes.

• In-state: Reported prices only from smokers who reported purchasing last pack/carton in the given state of residence.

• Out-of-state: Reported prices only from smokers reported purchasing out of their state of residence.

• Other sources: Reported prices only from smokers reported purchasing by internet, mail, phone, duty-free, Indian reservation or from foreign country.

Additionally, several measures of tax avoidance were constructed from the TUS-CPS data, including the percentage of smokers reporting buying out-of-state, the percentage of smokers buying from other sources and the percentage of smokers buying either out-of-state or from other sources. These measures will capture some tax avoidance behaviour by smokers—specifically interstate cross-border shopping, direct purchases (internet, mail and phone), some Indian reservation purchases and duty-free and other foreign purchases. Other in-state tax avoidance behaviour, such as buying from street sellers likely to be selling bootlegged cigarettes, cannot be captured given the limited questions included in the survey. Also, to the extent that a respondent includes purchases on Indian reservations as ‘in-state’ or ‘out-of-state’ purchases rather than ‘from elsewhere’, the purchases on Indian reservations specifically (and thus ‘other sources’ more generally) will be underestimated.

We estimated monthly prices from the TBOT data by computing prices before taxes are added (net-of-tax) as of November 1 in relevant years, assuming a linear trend in the net-of-tax price between Novembers, estimating the net-of-tax price for relevant months based on this trend, and then adding the state and federal tax in effect for each of the relevant months that the TUS-CPS was fielded to the net-of-tax price. National TBOT price measures were constructed as weighted averages of state prices, with weights based on the number of smokers who report prices in each state for each month.

## Results

Table 1 provides summary data for the TBOT and TUS-CPS prices and the measures of tax avoidance constructed from the TUS-CPS data. As expected, the prices constructed from the TUS-CPS were significantly lower than the TBOT prices. For the 2003 wave, the average TUS-CPS price for all respondents regardless of where they purchased was 71 cents per pack less than the average TBOT price; for the 2006/2007 wave, the difference was 47 cents. At the state level, the TUS-CPS price for all purchases and the TBOT price were significantly different in all states for each of the 3 months included in the 2003 wave of the TUS-CPS, and in nearly all states in each of the 3 months included in the 2006/2007 wave. We determined significance by constructing 95% CIs for the TUS-CPS estimates and determined whether TBOT results were outside of that interval.

Table 1

Comparison of TBOT and TUS-CPS summary measures for 2003 and 2006/207

When the out-of-state and other purchases are excluded from the TUS-CPS price measure, the average TUS-CPS in-state price rises slightly (by 3 cents per pack in the 2003 wave and 2 cents per pack in the 2006/2007 wave). The small change reflects the fact that relatively few smokers in the TUS-CPS report engaging in these types of tax avoidance—5.6% in the 2003 wave and 5.2% in the 2006/2007 wave. For those who do engage in tax avoidance, the savings are considerable, with average prices 58 cents per pack lower than those buying in-state for those engaging in tax avoidance in the 2003 wave and 52 cents per pack lower in the 2006/2007 wave.

## Discussion

The differences between the TUS-CPS and TBOT prices likely reflect several factors. In particular, the TUS-CPS prices are more likely to reflect the effects of price-reducing promotions than are the TBOT prices, which are not supposed to include any temporary price reductions. As described above, cigarette companies spent billions of dollars on price-reducing promotions in the relevant years; only part of the effect of these expenditures is likely to be picked up in the TBOT prices.

Similarly, the TBOT prices will not reflect tax avoidance, including purchase of cigarettes in nearby lower tax states, purchases on Native American reservations and purchases from other sources with lower or no taxes. The extent of tax avoidance can only be partially sorted out with the 2003 and 2006/2007 TUS-CPS data. Specifically, out-of-state cross-border shopping and purchases from other sources can be identified in the TUS-CPS data. However, the TUS-CPS in-state purchase data will reflect some in-state tax avoidance, including some purchases on reservations (to the extent the respondent did not select purchased ‘elsewhere’ for in-state Indian reservation purchases), intrastate cross-border shopping (eg, New York City smokers buying elsewhere in New York State in order to avoid the city tax) and purchases of bootlegged or smuggled cigarettes from informal vendors.

The narrowing of the difference between the TUS-CPS and TBOT prices from the 2003 wave to the 2006/2007 wave is likely due to a combination of factors. During this period, states began implementing efforts targeting direct purchases of cigarettes, particularly from internet vendors, by adopting new laws targeting these purchases, obtaining customer lists from internet vendors and then collecting unpaid taxes from state residents on these lists, and working with credit card and shipping companies to prohibit the use of their services for online cigarette purchases. These measures all reduce the incidence of tax avoidance behaviour. The drop in the percentage of ‘other’ tax avoiders from the 2003 wave (0.49%) to the 2006/2007 wave (0.34%) likely reflects the impact of these efforts on internet purchases.

Similarly, many states implemented efforts aimed at reducing incentives for smokers to avoid taxes by buying on Native American reservations. These efforts included entering into compacts with tribes under which a tax equivalent to the state tax was imposed on cigarettes sold on the reservation, with all or most revenues from the ‘tribal tax’ remaining with the tribe (eg, Washington); applying the state tax to all cigarettes sold on the reservation, with tax rebates for purchases made by reservation residents (eg, Wisconsin); and applying state taxes to cigarettes sold on reservations to non-residents (eg, Arizona).10

In addition, cigarette company total marketing expenditures, the vast majority of which have gone into price-reducing promotions in recent years, declined from $15.1 billion in 2003 to$12.5 billion in 2006.11 Price-related marketing expenditures (for price discounts, coupons, free samples and retail-value-added promotions involving cigarettes) fell from $12.2 billion (66 cents per pack) in 2003 to$10.7 billion (61 cents per pack) in 2006.11

Finally, the TUS-CPS data will likely reflect purchases from a wider range of retail sources, including more non-traditional cigarette vendors, than will the TBOT data. To the extent that these sources are more likely to carry brands produced by companies that emerged following the Master Settlement Agreement (MSA) and that did not sign on to the agreement (non-participating manufacturers (NPMs)), the difference in prices between the TUS-CPS and TBOT will partly reflect the greater prevalence of NPM brands (typically sold at prices well below the brands of participating manufacturers (PMs)) in the TUS-CPS data. Over time, loopholes in the MSA that had been exploited by NPMs and that allowed them to sell at relatively low prices were closed, reducing the differences in prices between the NPM and PM brands.12 Part of the narrowing of the gap between the TUS-CPS and TBOT prices may be accounted for by this trend.

These hypotheses are largely supported by an examination of the state-level data from the TUS-CPS all purchases measure and TBOT. The largest difference in price is in New York: \$1.00 in 2006/2007 (see figure 1). This is not surprising given the potential for intrastate and interstate tax avoidance in New York, given the large share of the state population facing the local tax in New York City and having relatively easy access to reservations throughout the state. Other states with relatively large differences are generally those with significant populations residing near reservations (eg, WA, MT, OK). The one exception appears to be Kentucky, which has the second largest difference in prices, no Native American reservations, and a relatively low cigarette tax (30 cents per pack). One possible explanation for this is the presence of numerous NPMs in Kentucky, suggesting that NPM brands may account for a relatively large share of cigarettes consumed in the state. In fact, six (OK, AR, KY, TN, KS, WV) of the top 10 states in our 2006–2007 ranking of the price differences are also among the top 10 states for NPM share in 2006. Overall, access to reservations and/or to NPM brands most likely account for larger differences in prices in some states (note that 2003 TUS-CPS results showed similar patterns; data not shown).

Figure 1

Difference in Tax Burden on Tobacco and Tobacco Use Supplement to the Current Population Survey (TUS-CPS)—all purchases’ prices by state, 2006/207. Note: TUS-CPS price measure by state is based on the smoker's state of residence, regardless of where the last pack/carton was purchased.

Figure 2 shows the percent of smokers in each state who report some types of tax avoidance for 2006–2007. As this figure illustrates, there is very little tax avoidance in most states. Further, data from table 1 suggest that little tax avoidance consists of buying from sources other than in the smoker's own state or other states (most likely various direct and duty-free purchases). In general, we found that there is a positive correlation between the percent of smokers in a given state who report the types of tax avoidance captured in the TUS-CPS survey and average in-state cigarette prices. Also, as expected, the extent of tax avoidance tends to be largest in relatively small states with high taxes and sizable populations near the border of lower tax states, and tax avoidance tends to be lowest in both low-tax states and states with taxes that are relatively low compared with neighbouring states.

Figure 2

Percentage of Tobacco Use Supplement to the Current Population Survey tax avoiders by state, 2006/2007.

If one considers buying by the carton a type of price minimisation strategy, as well as the tax avoidance strategies we have outlined above, Pesko et al13 indicate that US adult price minimisation strategies are likely as high as 28.8%. Xu et al14 find even higher rates—as much as 55.4%—when considering five types of price minimisation strategies, including purchasing by the carton, using coupons, purchasing discount brands, buying on reservations and buying on the internet.

## Limitations

In addition to the limitations of TBOT already described, it is important to elaborate on the primary limitation of this analysis. We have not used an independent criterion measure of the true price of cigarettes. While the TUS-CPS appears more valid than the TBOT through constructive validation, we cannot conclusively say how close the self-reported TUS-CPS price estimates are to true prices. Future studies may benefit from comparing self-reported price data, such as from the TUS-CPS, to observed prices and scanner price estimates. Although periodic data from the TUS-CPS can be useful for research purposes and monitoring long-term trends, one limitation of the TUS-CPS itself is that it cannot be used as an annual source of price data due to its tri-quadrennial fielding. This problem could potentially be addressed through more frequent fielding of a brief survey consisting of only the cigarette smoking status and price set of questions, which could also be more rapidly released to the public given its brevity.

## Conclusions

In summary, the self-reported prices from the population-based TUS-CPS appear to be valid measures of price that are highly correlated with the prices measured in the well-established TBOT. Furthermore, construct validity is demonstrated by differences between the TUS-CPS and TBOT prices that are consistent with what is known about the limitations of the TBOT prices, most notably their exclusion of marketing activities that result in temporary price reductions, their omission of prices related to opportunities for smokers to avoid taxes and also other differing state policies for NPMs versus PMs.

The comparisons of the TUS-CPS and TBOT price measures provided in this report suggest that self-reported price data collected in population surveys can be used to construct valid measures that capture the prices smokers actually face in the market, reflecting not just the list prices for cigarettes but also temporary price reductions resulting from cigarette companies’ price-related marketing efforts as well as prices from sources providing opportunities for tax avoidance. Given the pervasiveness of these behaviours, population-based surveys may not only prove useful but may also hold significant advantages over more traditionally used measures that do not account for tax avoidance or local variability in taxes. It may therefore be advantageous for other tobacco use surveys to include similar question types to the TUS-CPS, thus allowing them to serve as alternatives to the TBOT; this may be particularly useful for surveys with state-representative samples that are conducted more frequently than the TUS-CPS.

The future use of self-report surveys of tobacco cost data is likely to be particularly important in countries where other systems for collecting prices are not in place. Moreover, these efforts will be most valuable in low-income and middle-income countries where relatively few studies have estimated price elasticity given the lack of good data on prices (and often on tobacco use).

• Gathering reliable and valid data on tobacco prices is important for a number of reasons, from providing government agencies accurate data when considering cigarette excise taxes to encouraging changes in smoking behaviour based on increased prices. This paper provides a unique comparison between the widely used retail-based Tax Burden on Tobacco report and the self-report, population-based Tobacco Use Supplement to the Current Population Survey. The results discussed in this paper suggest an advantage of self-report population-based data over the retail-based data, as self-report data may reflect tax avoidance strategies such as promotional sales. In addition, this paper can serve as a validity citation for current and future research using similar self-reported populated-based price data in tobacco control analyses. These implications will not only affect research in the USA, where the surveys were conducted, but also globally, particularly in countries where self-reported price data are the most practical data to collect and often the only data available.

## Acknowledgments

All the authors extend appreciation to Anne Rodgers for her comments on a previous version of the paper.

## Footnotes

• Contributors AMH initiated the study concept, designed the study and participated in writing manuscript drafts, editing and preparation of final manuscript for submission to Tobacco Control. In addition, AMH directed the 2003 and 2006–2007 TUS-CPS project, which provided some of the data that were used in analysis. JTG served as system manager for the 2003 and 2006–2007 TUS-CPS project data used in this paper. JTG provided the TUS-CPS data files used in the analysis, participated in writing the methods sections and providing details of sample size for the TUS-CPS data used in analysis and helped in editing the final draft for submission of this paper. FJC led the contract, designed the study conducted the analysis, wrote a detailed study report from which a shorter summary was developed for journal publication and worked on drafts, edits and final revisions. JAT designed the study, conducted the analysis, contributed to a detailed study report from which a shorter summary was developed for journal publication and worked on drafts, edits and final revisions. JHS drafted a shorter manuscript for journal publication from detailed study report and worked on drafts, edits and final revisions. GW worked on drafts, edits and final revisions of this paper. GW developed and tested 2003 and 2006–2007 TUS-CPS data collection questionnaires, which were used to collect data used in analysis.

• Funding Funding support provided by the National Cancer Institute Contract #: HHSN21200800697P.

• Competing interests None.

• Provenance and peer review Not commissioned; externally peer reviewed.

## Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.