22 Million New Smokers Needed: Methodological Appendix

Report Health Care Reform

22 Million New Smokers Needed: Methodological Appendix

July 11, 2007 10 min read

Authors: Andrew Nowobilski and William Beach

PART I: Calculating the Base Year (2005)
We followed these steps in developing our base year estimates:

  1. We calculated the average consumer expenditure (in dollars/year) on tobacco products (aggregate, which includes cigarettes, cigars, smokeless, etc.) for the typical consumer unit in each socioeconomic category (all, age, race, income). Note: There are approximately 2.5 people per consumer unit, although this statistic varies by socioeconomic indicators. Source: U.S. Department of Labor, Bureau of Labor Statistics, "Consumer Expenditures in 2005," Report 998, p. 3, at www.bls.gov/cex/csxann05.pdf.
  1. We found the total number of consumer units in each demographic. Source: Ibid.
  1. We multiplied the number of consumer units in each demographic and the average amount spent on tobacco products by members of each demographic to obtain the total expenditure on tobacco products by demographic.
  1. We calculated the fraction of total tobacco consumption that each demographic group consumes by dividing total expenditure on tobacco products into the demographic's respective subtotals of tobacco product consumption. We assumed the same ratio to hold for cigarette consumption.
  1. After finding the total domestic consumption of cigarettes (in number of cigarettes), we multiplied this fraction by the total to find each demographic's consumption of cigarettes. Source: U.S. Department of Agriculture, "Table 1-Cigarettes: U.S. output, removals, and consumption, 1996-2006," April 2007, at www.ers.usda.gov/Briefing/Tobacco/Data/table01.pdf.
  1. The implicit average price of cigarettes per pack was found by dividing the total expenditures on cigarettes by the total number of cigarettes sold. This gives an average price per cigarette. We divided by 20 to get the implicit average price per pack. Sources: U.S. Department of Agriculture, "Table 21-Expenditures for tobacco products and disposable personal income, 1989/2006," at www.ers.usda.gov/Briefing/Tobacco/Data/table21.pdf, and U.S. Department of Agriculture, "Table 22--Governmental revenues from tobacco products, 1991/92-2004/05," March 2006, at www.ers.usda.gov/Briefing/Tobacco/Data/table22.pdf.
  1. We compared this implicit price to other average prices and found them to be similar.
  1. We found the elasticities for each of the demographic categories. We converted the original categories from which we calculated consumption data to match the socioeconomic categories for which elasticities were calculated. For example, elasticities were given in an "under 25" category. So, we assumed "under 25" to mean 18-25 so it would match the previous demographic. We assumed surveys of children are not often taken or have little impact on the result. Elasticities were also only given in the 25-39 and 40+ ranges. We assumed the original 35-44 category to be evenly distributed among the ages. We divided the number of cigarettes consumed by the 35-44 demographic as well as revenue by two and added it to the already given categories of 25-34 to match it with the 25-39 category of elasticity. The other half of the 35-44 category was added to the remaining categories from the original data. This comprised the 40+ category. We encountered a similar problem with income. The original data was divided by increments of varying degrees whereas elasticities were divided into above or below median income. Median income was given as $56,304 per consumer unit. The income range in the original data was $50,000-$69,999. Then we divided $19,999 (the range of values, namely $69,999-$50,000) by $6,304 ($56,304-$50,000) to get the weighted importance of the values in this original range that were below the median. Then, we multiplied that number by each demographic's consumption of cigarettes and revenue to obtain the consumption and revenue of those people with incomes between $50,000 and $56,304. We added these values to the values of all other incomes below the median. Likewise, the remaining consumption and revenue values (those with incomes between $56,304 and $69,999) were added to the other incomes above $70,000 to get the values for consumers with income above the median. Source: Farrelly et al. "Response by Adults to Increases in Cigarette Prices by Sociodemographic Characteristics." Southern Economic Journal, Vol. 68, No. 1, 2001, pp.156-165.
  1. Next, we used the elasticities for each demographic, the values previously found for quantity consumed by each demographic group, and the implicit price found for a pack of cigarettes to find the change in quantity consumed per demographic for a 61-cent increase in the price of a pack of cigarettes.
  1. We found the new quantity of cigarettes consumed by adding the change in quantity and the previous quantity.
  1. Finally, we calculated the revenue by multiplying the 61-cent tax increase by the quantity of packs of cigarettes now sold per demographic, and then summed all the revenue to obtain the total revenue from the 61-cent tax increase.

PART II: Projecting
We followed these steps in creating our projections.

  1. We projected total U.S. consumption of cigarettes per year, per capita, for 2006-2012 simply by extending the linear rate of change for the same figure calculated from 1996-2005 historical data from the USDA and then multiplying this projection by the projected population. Sources: For a description of CBO's methodology for projecting cigarette tax revenues, see Congressional Budget Office, "Description of CBO's Models and Methods for Projecting Federal Revenues," May 2001, p. 27, at www.cbo.gov/ftpdocs/28xx/doc2807/RevenueProjections.pdf. For 1990s population estimates, see U.S. Census Bureau, "Resident Population Estimates of the United States by Age and Sex: April 1, 1990 to July 1, 1999, with Short-Term Projection to November 1, 2000," January 2, 2001, atwww.census.gov/popest/archives/1990s/nat-agesex.txt. For population estimates from years 2000-2006, see U.S. Census Bureau, "Annual Estimates of the Population by Five-Year Age Groups and Sex for the United States: April 1, 2000 to July 1, 2006," at www.census.gov/popest/national/asrh/NC-EST2006/NC-EST2006-01.xls . For 2006+ projects, see U.S. Census Bureau, "U.S. Interim Projections by Age, Sex, Race, and Hispanic Origin," March 18, 2004, at www.census.gov/ipc/www/usinterimproj/.
  1. We projected average cigarette prices for 2006-2012 by extending the linear rate of change in implicit average cigarette prices calculated from 1996-2005 historical data on U.S. expenditures on cigarettes, combined with the above historical data on quantity consumed.
  1. We consolidated the base year (2005) consumer expenditure data into categories compatible with the elasticity statistics we used. First, we summed the number of consumer units in the consumer-expenditure subcategories to get the number of consumer units in each elasticity-compatible sub category. Where the range of an elasticity subcategory split the range of one of the consumer expenditure subcategories, a method was used similar to that described in step 9 of Part I. Next, we generated an average consumer expenditure statistic for each elasticity subcategory by weighting the average consumer expenditure attributable to each consumer expenditure subcategory by the fraction represented by the number of consumer units in that consumer expenditure subcategory, divided by the number of consumer units in the elasticity subcategory in which it falls (if it falls in two elasticity subcategories equally, the weight is divided by two). The products of the average consumer expenditures and the weights were summed over all consumer expenditure subcategories in a given elasticity subcategory.
  1. The number of overall consumer units and the number of consumer units in each elasticity (sociodemographic) subcategory was grown by the rate of growth predicted by the U.S. Census in its interim projections for persons (aged 18+) matching the characteristics of that reference persons in the consumer unit subcategories. Although persons and consumer units are not the same, we had no basis for projecting changes in the number of persons per consumer unit over time, and so we presumed that the ratio of persons per consumer unit was constant across the time, allowing us to apply growth rates derived from population data to consumer units. The number of consumer units in each subcategory was then rescaled by multiplying it by the projected overall number of consumer units for that year divided by the sum of the projected consumer units across all subcategories in a particular sociodemographic dimension. Source: U.S. Census Bureau, "U.S. Interim Projections by Age, Sex, Race, and Hispanic Origin," May 31, 2007, at www.census.gov/ipc/www/usinterimproj/.
  1. The average consumer expenditures for each consumer unit were projected by dividing the figure from the previous year by the previous year's (projected or known) average price per pack to get quantity packs. The elasticity for that subcategory of consumer unit was then multiplied by this quantity, divided by the price for the previous year, and multiplied by the difference between the projected current-year price and the previous-year price. This gave us the change in quantity packs, to which we added the original quantity packs, and multiplied by the projected current-year price to arrive at projected current-year average consumer unit expenditures.
  1. We multiplied the projected average consumer unit expenditures by the number of consumer units in each subcategory, arriving at total expenditures by subcategory on tobacco. This total was then rescaled by multiplying it by the projected overall consumer unit expenditures on tobacco for that year divided by the sum of the projected expenditures across all subcategories in a particular sociodemographic dimension. Finally, the fraction of total consumption attributable to a given sociodemographic subcategory was calculated by dividing the consumption attributable to that subcategory by the overall consumption for that year.
  1. The rest of the process was the same as for the base-year calculation.
  1. We reduced final revenue estimates by 0.25 of the total, to account for the offset attributed to a reduction of income and payroll tax receipts attributable to the increase in excise taxes. See Joint Committee on Taxation, "Overview of Revenue Estimating Procedures and Methodologies Used by the Staff of the Joint Committee on Taxation," February 2, 2005, at www.house.gov/jct/x-1-05.pdf.
  1. Finally, we constructed fiscal year estimates simply by adding one-quarter of revenue projected for the previous year and three-quarters of projected revenue for the current year. 

PART III: Descriptive Statistics

  1. The final revenue projections for the years 2008-2012 were recorded. Then the average annual rate of change in revenues was calculated for these years and used to extend linearly the trend to arrive at figures for years beyond 2012.
  1. The baseline projections for number of packs consumed were recorded. The current tax rate of 39 cents per pack was multiplied by these figures to arrive at projected tax revenues under the status quo. Then the average annual rate of change in revenues was calculated for these years and used to extend linearly the trend to arrive at figures for years beyond 2012. These numbers were then compared to the numbers calculated above in graph form. The distance between the two steadily declined in time.
  1. The number of smokers per year was estimated by taking 0.21 times the Census projected total populations for each future year. Sources: U.S. Census Bureau, "U.S. Interim Projections by Age, Sex, Race, and Hispanic Origin," March 18, 2004, at www.census.gov/ipc/www/usinterimproj/, and U.S. Department of Health and Human Services, "Health Data for All Ages," Smoking among adults: US, 2000-2005, at http://209.217.72.34/hdaa/ReportFolders/
    ReportFolders.aspx?IF_ActivePathName=P
    /Health%20Conditions%20and%20Risk%20Factors
    .
  1. We recorded the number of cigarette packs consumed per year, per capita, projected for 2008 (see Part II). Next, we calculated the proportion of the predicted 2008 U.S. population aged 18+. Finally, we divided the average number of packs per capita by the above proportion and then divided the result by 0.21 to get the average cigarettes per smoker projected for 2008, holding the proportion of the adult population that smokes cigarettes constant.
  1. We developed revenue targets for each year. To do this, we recorded a historical average of annual percentage changes in medical prices for the years 2000-2006 (4.29%). We next calculate the amount of expenditures needed per year to provide the same amount of medical care each year. Source: Bureau of Labor Statistics, "Consumer Price Index Detailed Report Tables," 2006, at www.bls.gov/cpi/cpi_dr.htm.
  1. We held the number of smokers as a constant fraction (0.21) of the projected population for each of the years examined and calculated how many additional packs would have to be consumed by each in order to achieve the annual revenue target. We subtracted the projected revenue for each year in question from the revenue target for that year. We then divided by 0.61 (the proposed increase in the federal excise tax rate per pack) to obtain the packs needed to make up this deficit. Finally, we divided by the number of projected smokers for that year.
  1. We held the number of packs per smoker constant using the numbers estimated in step 4 above. Again, we subtracted the projected revenue for each year from the revenue target for that year and divided by 0.61 to obtain the packs needed to make up this deficit. Finally, we divided by the average number of packs per smoker, as estimated in step 4 above.

William W. Beach is Director of, and Andrew Nowobilski is an Intern in, the Center for Data Analysis at The Heritage Foundation.

Authors

Andrew Nowobilski

William Beach

Senior Associate Fellow