COMMENTS ON FINDINGS BY THE TOBACCO CONTROL PROJECT THAT ADVERTISING PROMOTES AGGREGATE CIGARETTEE CONSUMPTION
Professor Daniel F. Leach
AB (Penn) Ph.D.(UCLA)
Department of Business Economics, University of the Witwatersrand
Portfolio Committee on Health Hearings on the Tobacco Products Control Amendment Bill
19-20 October 1998, Cape Town
1. In this report I will comment on the "findings" of the Economics of Tobacco Control Project (ETCP, Control Project) that restricting advertising by cigarette manufacturers and anti-tobacco advertising are both effective in reducing aggregate cigarette consumption.
Deficiencies in the Control Project's Cigarette consumption Data
2. The cigarette consumption data for 1990-1994 used by the Control Project has been-and continues to be-seriously flawed. The table and figure below compare two sets of figures used by the Control Project with a set of figures publicly available from the South African Revenue Service (SARS), adjusted for imports.
2.1 in its first papers (Project Update #1 Aug.1996 and Project Update #2, Sept.1996) the Control Project reports obtaining cigarette consumption data for 1970-1989 from an article by Reekie (1994), (which Reekie had obtained from publicly available SARS figures), but, inexplicably, the Control Project estimates consumption figures for 1990-1994.
[Ed. note: The graph of SARS vs. ETCP Data has not been included.]
2.2 As shown by the table and figure, the consequence was that the Control Project's figures show a large 42.5% increase in consumption from 1989 to 1994, in sharp contrast to the SARS figures that show consumption to have peaked in 1991 and to have declined by 1994 to below the consumption level of 1989. It should be noted that consumption has continued to decline to the point where 1997 consumption was below 1984 consumption.
2.3 The Control Project's wildly inaccurate estimates were pointed out to Professor Abedian, yet, in a Control Project Media Release (20 January 1997), he initially continued to insist that his figures 'show an increase in cigarette consumption in South Africa".
2.4 Eventually, however, the Control Project released a revised and corrected version of its Project Update # 1 now indeed showing that consumption had peaked in 1991 and declined thereafter. This is the series labeled "ETCP2" in the table and figure. However. the revised figures still deviate considerably from the SARS figures, showing a much higher peak in 1991-almost 4 billion cigarettes higher-than shown by the SARS data and declining more rapidly than the SARS figures.
2.5 Even if there were some doubt about what the exact consumption figures are, the revised Control Project figures just don't look right. The 1991 peak looks too sharp, with a 4 billion cigarette rise in consumption from 1990-1991, and a subsequent 7 billion decline the following year.
Deficiencies in the Control Projectís Advertising Data
2.6 The Control Project's advertising data are even worse than its consumption data. The basic source of advertising data in South Africa is the "Adindex" published by Market Research Africa which is owned by Nielsen Media Services, but the Control Project states (Project Update # 2, p.14) that before 1990 reliable data were available only for 1980, 1986 and 1989. The Control Project therefore estimated figures for the remaining 17 years between 1970 and 1989(1970-1979, 1981-1985, 1987-1988)and the procedure they used was to assume that tobacco advertising was 5 percent of total South African advertising spending. Thus 17 of the 25 data points for the sample period 1970-1994 were basically made up by the Control Project.
2.6.1 The publishers of Adindex categorically reject the Control Project's contentions about the reliability of the advertising data prior to 990. Yet in the Control Project's Media Release already referred to, Professor Abedian stated that "Our data on advertising is likewise the best time series data available in the country. We have checked with all relevant institutions and agencies to confirm data in this respect." One can only ask if the publishers of Adindex were among the "relevant institutions and agencies" that confirmed the Control Project's advertising data.
The Control Projectís Estimation of the Effect of Advertising on Aggregate Cigarette Consumption
3. In its Project Update # 3 the Control Project reports estimates of the demand for cigarettes that include both industry advertising and "anti-tobacco" advertising as explanatory variables.
3.1 Using "simultaneous equation" estimation the Control Project finds no relationship between advertising and cigarette consumption for their full sample period, 1970-1994, but a (barely) statistically significant relationship for the shorter period, 1970-1990. Using "single equation" estimation they find no relationship for 1970-1993 (1994 is excluded for technical reasons) but, again, a statistically significant relationship for the shorter period, 1970-1990.
3.1.1 No reason is given for selecting and reporting the particular shorter period, 1970-1990. What about 1970-1991 or 1970-i 992? If the reason is to exclude the period of "anti-tobacco advertising" (1990-1992), why not 1970-1989?
3.1.2 Initially, on page 3 of its document, the Control Project is at least cautious: "Based on the results, there is some ambiguity as to the effect of cigarette advertising on demand for cigarettes."
3.1.3 This caution is apparently forgotten by page 5, where the Control Project boldly reports that "Estimates show that a 1 percent increase in the growth in advertising expenditures will increase growth in demand for cigarettes by between 0.10 and 0.24 percent." In any case, even if these estimates were all statistically significant- which they are not-such small estimates are hardly economically significant, representing as they do a decline of between 1 and 2.4 cigarettes out of 1000. Therefore the Control Project has no justification to conclude that 'limiting advertising by cigarette manufacturers . . . [is] an effective weapon in discouraging growth in demand for cigarettes".
The Control Projectís Estimation of the Effect of Anti-Tobacco Advertising on Aggregate Cigarette Consumption
4. The Control Project's estimation of the effect of anti-tobacco advertising on cigarette consumption is based on the fact that from 1990-1992 there were a few flightings of anti-tobacco advertising in cinemas.
4.1 Although the Control Project finds a statistically significant negative relationship between their measure of anti-tobacco advertising and cigarette consumption, the economic effect is trivial-A decline in consumption of 0.04%, i.e. 4 cigarettes out of 10 000. The Control Project's researchers don't appear to understand the basic distinction between statistical and economic significance.
4.2 Again, they have no justification to conclude (p.5) that anti-tobacco advertising is an "effective weapon in discouraging growth in demand for cigarettes."
The Effect of Advertising and Anti-Tobacco Advertising Found in International Studies
5. It is a result of longstanding, well-known in the fields of marketing and economics, that advertising has little or no effect on aggregate consumption for any product. The major effect of advertising is rather on market shares.
5.1 These findings apply to the most heavily advertised products, including cigarettes, and were confirmed for cigarette advertising by a 1996 survey article by Martyn Duffy published in the International Journal of Advertising. Duffy's survey of studies done over the world finds that advertising and advertising restrictions, including bans, "have had little or no effect upon aggregate consumption of cigarettes."
6. In the Control Project's Media Release cited above, Professor Abedian, in contrast, states that
"our findings concur with the mounting international evidence that advertising does have a positive impact on consumption of cigarettes. . . . we refer. South African policy makers to professional papers such as [Hu], Sung & Keeler's 1995 paper published in American Economic Review. . . or Trembaly & Tremblay's 1995 paper published in Contemporary Economic Policy".
6.1 If South African policy makers had followed Professor Abedian's suggestion and had examined these papers carefully, they would have found the following:
6.1.1 The Tremblay and Trembaly paper found "one would expect a 1 percent increase in [U.S. industry] advertising expenditures to generate an increase [of] 3.67 cigarettes per person.. . per year". The average annual per capita U.S. consumption during the period of the study was 3865 cigarettes.
6.1.2 The Hu, Sung and Keeler paper found that a one million dollar increase in anti-cigarette advertising by the State of California reduced per capita consumption by 5.3 cigarettes per year (1995. p.188, table 1, equation iii).
7. Professor Abedian is being disingenuous in suggesting that this recent "mounting international evidence" represents a change from Dufly's finding that advertising has little or no effect on aggregate cigarette consumption.
SUMMARY AND CONCLUSlON
1. The data used by the Control Project is seriously deficient. The consumption data is suspect for the period 1990-1994 and the advertising data is very bad, with 17 of 25 data points estimated by assuming that tobacco advertising was a fixed percentage (5%) of a much larger aggregate-total S.A. advertising spending. The Control Project basically made up their advertising data.
1.1 Even if this data were, as Professor Abedian asserts, 'the best time series available in the country" it hardly warrants the strong conclusions reached by the Control Project.
2. The Control Project reports a statistically significant effect of advertising on cigarette consumption only for only the subperiod, 1970-1990. For the full sample period there is no effect of advertising on consumption. No reason is given why we should attach special significance to this one subperiod. in any case the economic effect is trivial: At most 2.4 cigarettes out of 1000.
3. The Control Project reports a statistically significant effect of anti-tobacco advertising on cigarette consumption. The economic significance is even more trivial than the industry advertising effect: 4 cigarettes out of 10000. The Control Project researchers don't appear to understand the basic distinction between statistical and economic significance.
4. On the concluding page of Project Update # 3 the Control Project concludes that "limiting advertising by cigarette manufacturers and promoting anti-cigarette awareness are both effective weapons in discouraging growth in demand for cigarettes."
4.1 This strong conclusion misrepresents its true findings, which are more consistent with the cautious earlier statement that "Based on the results, there is some ambiguity as to the effect of cigarette advertising on demand for cigarettes."
5. It is a result of longstanding, well-known in the fields of marketing and economics, that advertising has little or no effect on aggregate consumption. This result was confirmed for tobacco in a 1996 survey article in the International/ Journal of Advertising.
5.1 Professor Abedian cites two studies in support of his assertion of "mounting international evidence that advertising does have a positive impact on consumption of cigarettes". One study shows that a 1% increase in industry advertising increased per capita cigarette consumption by 3.67 cigarettes per year. The other study shows that a one million dollar increase in anti-cigarette advertising reduced per capita consumption by 5.3 cigarettes per year. The "mounting international evidence" continues to show little or no effect of advertising on aggregate cigarette consumption.