Table 5.

Summary of Other Methodologies & Advertising Findings.

Study [ref. no.]; Sample; Outcome; Methods	Innovations & Refinements	Study Findings & Conclusions
Alcohol advertising studies
Calfee & Scheraga [140]; annual time-series data for FR, DE, NL & SE; per capita alcohol use; linear & log regressions for each country.	For Sweden, alcohol advertising has been prohibited since 1979. Models include country prices, income & advertising expenditures.	Advertising coefficients are not significant for any country. The results for Sweden are not different than the other 3 countries, despite the advertising ban. Price is significant for Sweden.
Lariviere et al. [141]; monthly time-series data for Ontario, CN, for 1979–1987 for beer, wine, spirits & soft drinks; demand system model.	Monthly advertising expenditures for four beverages that capture “pulsing” effects across markets; advertising for four beverages, prices, income & demographics.	Advertising for beer & spirits are not significant. Negative sign for wine advertising & positive sign for soft drinks. Study concludes that “advertising is not effective in enlarging markets,” but rather promotes brand-switching.
Markowitz & Grossman [97]; 1976 Physical Violence in American Families survey; overall & severe domestic violence; probit model.	State alcohol tax, availability, illegal drug prices, restrictions on billboard advertising, restrictions on window displays & price advertising.	Restrictions on advertising are ineffective in reducing violence. Violence toward children reduced by higher alcohol taxes.
Markowitz & Grossman [142]; 1976 & 1985 Physical Violence in American Families surveys; physical child abuse by gender; probit model.	State alcohol tax, availability, illegal drug prices, restrictions on billboard advertising, restrictions on window displays & price advertising. State binary variables in some models.	Restrictions on advertising are ineffective for both genders. For females, violence toward children reduced by higher alcohol taxes in 1976 & 1985.
Nelson [143]; state panel data for 1982–1997; per capita pure alcohol use by beverage; panel data model with regional fixed effects & simulations.	Bans of billboard advertising, bans of price advertising & state monopoly control of retail stores. Study considers substitution among beverages due to regulations.	Bans of advertising do not reduce total alcohol consumption, reflecting in part substitution among beverages. Income is always significant and price is generally significant.
Nelson [66]; international panel of 17 OECD countries for 1975–2000; per capita consumption of pure alcohol; panel data model for log levels & growth rates, IV model.	Spirits broadcast advertising bans & bans of broadcast advertising for all beverages, alcohol-control policy index & drinking sentiment. Study adjusts for non-stationary data & endogeneity of the alcohol policy index.	Bans of advertising do not reduce alcohol consumption, regardless of severity. Other alcohol policies and prices have a negative effect on consumption.
Nelson [51]; meta-analysis of 21 longitudinal and panel data studies of alcohol advertising & youth drinking; meta-regression analysis.	Paper examines 23 effect-size estimates for drinking onset & 40 estimates for other drinking behaviors. Meta-regressions account for primary study heterogeneity, heteroskedasticity, omitted variables, publication quality & truncated samples.	Meta-regression results are consistent with publication bias, omitted variable bias in some studies & lack of a genuine effect for advertising, especially mass media. The paper also discusses “dissemination bias” in the use of research results by investigators & health policy interest groups.
Paschall et al. [144]; 2003 ESPAD Alcohol Survey for 26 countries, youth 15–17 years; current drinking & binge drinking; separate trivariate regressions.	Overall alcohol-policy index score, alcohol availability, advertising control rating & country per capita consumption. No other controls for prices, income, drinking sentiment, etc.	Alcohol advertising control rating is not statistically significant at standard 95% confidence level, after controlling for per capita consumption. Policy index is insignificant, but availability rating is significant for current drinking & binge drinking.
Saffer and Dave [145]; 75 media markets, US, 1996–1998 & 1997–1998, youth ages 12–17 & 12–16; past year drinking, past month, binge drinking; probit & OLS regressions.	Composite measure of local advertising expenditures. Significant in 10 of 15 cases for MTF data. Significant in 5 of 6 cases for NLSY data. Log of advertising is significant in 1 of 2 cases for NLSY. T-statistics are 2.3 or less in 14 of 23 cases.	Null effect of advertising on three MTF drinking measures for blacks. Null results for males for MTF for past month & binge drinking. Null results for NLYS for two log models. Concludes that “reduction of advertising can produce a modest decline in adolescent alcohol consumption.”
Tobacco advertising studies
Bardsley & Olekalns [146]; 1962–1996 time-series data for AU; per capita tobacco consumption; rational addiction model & dynamic simulations.	Aggregate consumption in Australia peaked in the late 1960s. Real ad expenditures per capita declined after a peak in late-1960s. Most tobacco advertising banned in 1992.	Effect of pro-smoking advertising & policy interventions are small relative to economic variables for taxes, income & demographics. Evidence of forward-looking behavior; virtually all reductions in smoking due to tax increases.
Czart et al. [98]; 1997 Harvard Alcohol Study survey, students at 140 US colleges; current smoking & ave. daily number; probit & logistic models.	State, local and school variables for smoking policies, availability & school-level advertising bans (newspapers, bulletin boards).	Bans of cigarette advertising on campus and bans of sales of cigarettes on campus have no significant effect on smoking behavior. Price is significant for smoking participation & level of smoking.
Hammar & Martinsson [147]; 2000 county-based survey in northern SE; smoking initiation age (9–25 years); duration analysis.	Anti-smoking policies enacted in Sweden from 1955 to 1986, including 1979 laws on marketing.	Public policies do not show a significant effect on the age of smoking initiation. Age of initiation depends on gender, parental smoking & time trend.
Hublet et al. [148]; 2006 Health Behaviour (HBSC) survey for 29 European countries, youth 11–15 years; regular smoking by gender; multilevel model.	Country-level variables for price, public bans, advertising bans, sales to minors, vending machines, adult smoking, affluence, etc.	Bans of advertising & public smoking bans are insignificant. For regular smoking, price is significant for boys, but not for girls.
Lewit et al. [149]; 1990 & 1992 surveys of 9th grade students in 21 CN & US cities; current smoking & smoking intentions by gender; logistic model.	Site-specific smoking control variables. Includes prices, minimum age, access to vending machines, and anti- & pro-smoking media exposure. Media exposure is self-reported index for 5 media for pro-smoking & 10 media for anti-smoking.	For current smoking, pro-tobacco media significant for boys, but not for girls. For smoking intentions, pro-tobacco not significant for either gender. Concludes that “only very modest support to the notion that media-focused policy interventions will be effective.” Price significant for boys’ current smoking & girls’ intentions.
McLeod [150]; 1953–1983 time-series data for AU; tobacco & cigarette consumption; double-log model with intervention binaries.	Australia banned cigarette & tobacco broadcast advertising in 1976.	Ban of broadcast advertising has a short-run effect on tobacco use, but no effect on cigarette use. Price is significant, but income is insignificant.
Nelson [151]; international panel model for 20 OECD countries for 1970–1995; per capita cigarette & tobacco use for levels & growth rates; OLS panel model with time & country fixed-effects, IV model.	Strong bans (print + all broadcast), moderate bans (3–4 media), weak bans (TV-radio only), no. of banned media & warning labels. Study adjusts for endogeneity of advertising bans, non-stationary data & structural change.	Bans of advertising have no effect on cigarette consumption, regardless of the time period considered or the severity of the bans. Price & income are significant, but evidence of structural change beginning around 1985.
Nelson [152]; Global Youth Tobacco Survey for 42 developing countries for 1999–2001, youth 13–15 years; current smoking & ever smoked prevalence; linear probability models by gender & combined with interaction terms.	Countries with complete bans (all major media), moderate (TV or other media) bans & no media banned; warning labels & minor sales prohibited. Other covariates for availability, education, peer smoking, income, Muslin faith, former Soviet-bloc countries, etc.	Bans of advertising have no effect on youth smoking prevalence in developing countries for either gender or combined. Higher income levels reduce smoking in developing countries & smoking by peers is important. Youth in Muslin countries have lower predicted prevalence & Soviet-block countries have higher prevalence.
Nelson [153]; meta-analysis of 33 advertising elasticities for US and 16 elasticities for other countries; 19 studies of four major regulatory effects; meta-regressions.	Study adjusts for heterogeneity of estimates, heteroskedasticity & non-independence of observations. The study also reviews 50 years of advertising regulation by the FTC.	Advertising elasticities are very small and not statistically significant regardless of the time period. The 1971 ban of broadcast advertising did not affect cigarette consumption.