Comprehensive state tobacco control programmes are effective strategies to reduce cigarette consumption and smoking prevalence.1 2 State-wide programmes have contributed to smoking declines in California,3 Massachusetts,4 Florida,5 and nationally.6 Paid media campaigns have contributed to the observed declines in smoking among youth and adults.7^–12

Several state programmes have witnessed dramatic reductions in funding.13 The effects of these budget cuts are not well understood, but available evidence suggests the effects are large and immediate. Youth exposure to anti-tobacco ads declined in the aftermath of budget cuts for several state tobacco control programmes between 2002 and 2003.14 Funding cuts in California in the mid-1990s halted declines in adult smoking rates,3 although some data suggest these declines resumed once funding was partially restored.15 The elimination of paid anti-tobacco media activities targeting youth in Minnesota reduced campaign awareness and may have increased smoking susceptibility.16 The elimination of school-based programmes within Oregon’s comprehensive tobacco control programme led to increases in smoking uptake among teens.17 Little is known about the effect of state tobacco control funding cuts in Massachusetts and Florida, or whether these reductions may be offset by national efforts such as the “truth” campaign. This paper examines consequences of dramatic reductions in the Florida Tobacco Control programme (FTCP), a large state-wide youth prevention programme, between 1998 and 2000.

The Florida Tobacco Control Program (FTCP)

The FTCP launched in April 1998 to prevent youth tobacco use in Florida. The most visible component of the FTCP was “truth”, a media campaign targeting teens with information about the tobacco industry’s deceptive marketing practices.18 The FTCP also established a state-wide youth anti-tobacco organisation, supported local anti-tobacco coalitions, distributed school tobacco prevention curricula and enhanced enforcement of youth tobacco purchase, use and possession laws. Smoking rates declined rapidly among Florida teens within 2 years of the programme’s inception.5 The Florida “truth” media campaign contributed to these declines.9^–11 The FTCP appears to have been successful in reducing youth smoking by increasing anti-tobacco industry attitudes and reducing intentions to smoke among never smokers.9 10 Anti-industry attitudes and smoking intentions are strong predictors of youth smoking, so FTCP effects on these intermediate outcomes are likely responsible for part of the observed reductions in Florida youth smoking prevalence.19^–21

The FTCP’s budget was cut substantially after the first year, from $70.5M in the financial year (FY)1999 to $38.7M in FY2000.22 Enforcement initiatives were eliminated, community- and school-based programmes were reduced, and the Florida “truth” campaign’s budget declined from $11.7M to $7.1M. These reductions produced lower levels of broadcast time for the campaign but did not influence the quality of the ads.23 Lower broadcast time would be expected to reduce campaign awareness and campaign effects on subsequent outcomes. The consequences of this reduction in state tobacco control programme funding, however, remains unknown. Of note, smoking prevalence declined at comparable rates between 1998 and 1999 and between 1999 and 2000, suggesting that effects of the budget cuts were not immediately apparent.5 However, media campaign effects on smoking behaviour often take time. Many published media campaign evaluations include a lag of 1 year or more between programme exposure and subsequent effects on behaviour.8 9 11 Indeed, the downward trend in smoking prevalence among Florida students slowed considerably after 2000.24 Furthermore, there is evidence that the national “truth” media campaign, modelled largely on Florida “truth” and launched in February 2000, contributed to nationwide reductions in youth smoking.25 The national “truth” effort may have likewise contributed to declines in smoking among Florida teens between 1999 and 2000. As a result, it could be misleading to conclude that the FTCP budget cuts did not have immediate negative consequences for Florida tobacco prevention based solely on trends in youth smoking behaviour.5 It would be more instructive to examine changes in intermediate outcomes, including programme exposure, programme-targeted attitudes and non-smoking intentions.

Study objectives

This study analyses survey data between 1998 and 2000 to assess consequences of 1999 FTCP budget cuts on targeted outcomes. First, we examine trends in Florida “truth” recall, beliefs and intentions. Second, we use multivariate models to gauge whether favourable trends in programme exposure, anti-industry beliefs and non-smoking intentions were changed after the 1999 FTCP budget cuts. Third, we estimate whether exposure to the first 4 months of the national “truth” campaign offset otherwise deleterious effects of the cuts to programme funding.

METHODS

Survey data

We analysed data from the Florida Anti-tobacco Media Evaluation (FAME) survey, a repeated cross-sectional telephone survey of Florida teens designed to assess the impact of the Florida “truth” campaign.26 FAME samples were drawn from a commercial vendor list that contained information about gender, race/ethnicity, grade in school and telephone numbers for approximately 50 of the Florida teen population.26 A random sample from this pool was drawn and stratified to be representative of Florida teens by geographic region, gender, age and race/ethnicity for each survey wave. The baseline survey was conducted in April 1998 prior to the launch of the Florida “truth” campaign. Five semi-annual survey waves, each containing approximately 1800 respondents, were conducted between April 1998 and May 2000.

We excluded respondents outside of the campaign’s targeted age range (12 to 18 years) and teens that had tried smoking because the campaign’s primary objective was to prevent initiation. The final analytic sample consisted of 5010 Florida teens. The average response rates for waves 1 (April 1998) to 3 (May 1999) were just under 69; response rates for subsequent waves were not reported and are unavailable.26 The final analytic sample was 52.2 female with a mean (SD) age of 14.92 (1.81) years. Two-thirds of respondents (66.0) identified themselves as non-Hispanic White, with 16.6 African–American, 12.3 Hispanic and 5.0 “other”. Additional details about survey procedures are presented elsewhere.26

Survey measures

Confirmed Florida “truth” ad recall

Each survey (excluding the baseline) asked respondents to indicate whether they had seen three or four specific Florida “truth” ads that aired on television. Items were worded as follows: “Have you recently seen an antismoking ad on television that shows [ad description]?”. Respondents who said yes were asked a follow-up question: “What happens in this ad?”. Respondents who provided additional ad details were deemed to have confirmed ad recall. Confirmed ad recall measures have been shown to be valid indicators of ad exposure.27 We created a dichotomous variable to indicate whether or not a respondent recalled one or more Florida “truth” ads within a survey wave. Baseline survey respondents were assigned a value of 0. On average, 76.7 of Florida teens confirmed recall of at least one Florida “truth” ad across the four survey waves during which the ads were broadcast.

Confirmed national “truth” ad recall

The national “truth” campaign started in February 2000. Procedures identical to those described above were used to gauge recall of the three most prominent national “truth” ads that aired in early 2000. Respondents to the first four surveys were assigned a value of 0 for national “truth” confirmed ad recall. In all, 62.2 of Florida teens confirmed recall of at least one national “truth” ad in May 2000.

Anti-Industry belief index

FAME surveys included a series of statements, with Likert-style response categories (strongly agree to strongly disagree), designed to capture changes in beliefs targeted by Florida “truth” campaign ads. Items were recoded where necessary so that higher values reflected stronger anti-industry sentiment. Four criteria were used to select individual belief items for scale inclusion: the item (1) reflected a belief targeted by the campaign, (2) was included in each of the five FAME survey waves, (3) was positively associated with Florida “truth” recall and (4) was negatively associated with smoking behaviour.28 Six items met these criteria, including the statements, “Tobacco companies try to get young people to smoke because older people quit smoking or die”, and “Most people your age think it is okay to work for tobacco companies”. Cronbach α was not used to assess reliability because the decision to combine the items into an index was not based on the expectation that items would be strongly correlated.29

Non-smoking intentions

Non-smoking intentions were assessed with a single question: “If your best friend offered you a cigarette, would you smoke it?”. Response categories included “definitely”, “probably”, “probably not” and “definitely not”. Youth were coded as having non-smoking intentions if they said “definitely not”. This measure is a strong predictor of resistance to adolescent smoking initiation in longitudinal studies.20 21

Survey wave and budget cuts

Three variables assessed the effects of budget cuts on trends in Florida “truth” recall, anti-industry beliefs and non-smoking intentions. The first variable, “survey wave”, was a linear variable ranging from 1 to 5. The second variable, “budget cut”, was an indicator for whether or not the survey was conducted before (waves 1 to 3; April 1998, September 1998 and May 1999) or after (waves 4 and 5; September 1999 and May 2000) reductions in tobacco control funding. The third variable was an interaction term between survey wave and budget cut.

Potential confounding variables

A number of other variables might influence Florida “truth” recall, anti-industry beliefs and/or non-smoking intentions. These include demographics (age, gender, race/ethnicity) and household characteristics such as the presence of smokers in the home and the degree of parental monitoring. Household smoking was assessed by asking whether various members of the household smoked, including parents, siblings and others. A total of 28.7 of respondents answered yes. Parental monitoring was assessed as follows: “On average, how many days a week do you and all the people in your household eat your evening meal at the same time?”. The mean (SD) response was 3.95 (2.37) days. These variables were associated with Florida “truth” recall, anti-industry beliefs, or non-smoking intentions and were included in all multivariate models.

Analytical approach

We began with logistic regression models to estimate adjusted trends in “truth” recall and non-smoking intentions, and linear regression models to estimate adjusted trends in anti-industry attitudes across the five survey waves. Adjusted trends were required to account for modest differences in the demographic composition of each sample. We hypothesised that progress on each outcome would diminish after the budget cuts.

We continued by estimating a series of linear and logistic regression models using an interrupted time series technique that allows the slope and intercept of the model to change at a given point of time. Specifically, we tested for significant differences in the rates of change in Florida “truth” recall, anti-industry beliefs and non-smoking intentions (a) prior to the budget cuts (FAME waves 1 to 3, from April 1998 to May 1999) and (b) after the budget cuts (FAME waves 4 and 5, from October 1999 to May 2001). This procedure used three variables (plus controls): survey wave, budget cuts and their interaction term. The survey wave variable gauges the direction of trends in the outcome variable prior to the budget cuts. We hypothesised that this variable would be positive and significant for all three outcomes. The interaction term tests whether the positive trend in the outcome was significantly reduced after the budget cuts. The budget cuts coefficient is not interpretable because it represents the effect of the budget cut if survey wave was equal to zero (logically impossible because the budget cuts occurred between wave 3 and 4). As a result, the budget cuts coefficient is not discussed in the present work.

We conducted the regression models in two steps for anti-industry beliefs and non-smoking intentions. We first estimated a model (linear regression for anti-industry beliefs, logistic regression for non-smoking intentions) that excluded national “truth” recall as an independent variable. This model assessed whether differences were observed in the rates of change in each outcome prior to the budget cuts and after the budget cuts. In support of a deleterious effect of the budget cuts, we expected the interaction term to be negative and statistically significant for each outcome. Next, we estimated a model that included national “truth” recall. This model was designed to test the extent to which national “truth” recall offset otherwise deleterious effects of the budget cuts. We hypothesised that the interaction between survey wave and budget cuts would become larger when we included national “truth” recall in the model.

RESULTS

Trends in confirmed Florida “truth” recall, anti-industry beliefs and non-smoking intentions

Confirmed Florida “truth” recall increased sharply between April 1998 and May 1999, followed by gradual reductions in October 1999 and May 2000 (table 1). Anti-industry beliefs also increased sharply between April 1998 and May 1999. No change was observed between May 1999 and October 1999, while a non-significant increase occurred between October 1999 and May 2000. The percentage of Florida teens classified as closed to future smoking increased rapidly between April 1998 and May 1999. This upward trend continued until October 1999, but experienced a non-significant decrease between October 1999 and May 2000.

Changes in trends for confirmed “truth” recall after budget cuts

Upward trends in confirmed Florida “truth” ad recall were significantly reduced after the 1999 FTCP budget cuts (table 2). The survey wave coefficient was positive and highly significant, indicating that there was a substantial upward trend in confirmed Florida “truth” recall prior to the budget cuts. The interaction between survey wave and the budget cut indicator was negative and statistically significant, suggesting that the 1999 FTCP budget cuts reduced the upward trend in confirmed “truth” recall.

Changes in trends for anti-industry beliefs after budget cuts

Upward trends in anti-industry beliefs were not significantly reduced after the 1999 FTCP budget cuts (table 3, model 1). The survey wave coefficient was positive and statistically significant, showing a substantial upward trend in anti-industry beliefs prior to the budget cuts. The interaction between survey wave and the budget cut indicator was negative but not statistically significant, indicating that the 1999 FTCP budget cuts did not reduce the upward trend anti-industry beliefs. However, the interaction term more than doubled in magnitude and became statistically significant when controlling for confirmed recall of national “truth” ads (table 3, model 2). Therefore the multivariate model suggests that upward trends in anti-industry beliefs would have been reduced after the 1999 FTCP budget cuts in the absence of the February 2000 launch of the national “truth” campaign.

Changes in trends for non-smoking intentions after budget cuts

Upward trends in non-smoking intentions were significantly reduced after the 1999 FTCP budget cuts (table 4, model 1). The survey wave coefficient was positive and statistically significant, while the interaction between survey wave and the budget cut indicator was negative and significant. The interaction term remained similar in magnitude and significant when controlling for confirmed recall of national “truth” ads (table 4, model 2), indicating that reductions in upward trends in non-smoking intentions were not offset by national “truth”.

DISCUSSION

This study provides evidence that reductions in tobacco control funding have immediate effects on programme exposure and cognitive precursors to smoking initiation. Dramatic reductions in funding for the FTCP in 1999 reversed upward trends in recall of the Florida “truth” anti-smoking campaign, which contributed to large declines in Florida youth smoking prevalence.9^–11 The launch of the national “truth” campaign in February 2000 may have offset otherwise deleterious effects of the budget cuts on trends in anti-industry beliefs, which are also linked to reduced smoking initiation.9^–12 However, national “truth” did not offset the negative effects on trends in non-smoking intentions among Florida teens. This may be attributable to the fact that the national “truth” campaign had only been in existence for 4 months. Campaign effects on smoking intentions may take longer to manifest themselves than campaign effects on targeted attitudes.30 It is nevertheless noteworthy that national “truth” appeared to influence anti-industry beliefs in such a short amount of time. This finding is not implausible; recall of the Florida “truth” campaign reached 75 in the first 2 months of the campaign,26 and national “truth” campaign reached comparable levels of recall and increased anti-industry beliefs nation-wide within the campaign’s first 10 months, if not sooner.31

The negative effects of the 1999 FTCP budget cuts on the trajectory of several targeted outcomes likely slowed progress in reducing smoking rates among Florida teens. The effects of anti-tobacco media campaign exposure, anti-industry beliefs and non-smoking intentions often take time to influence population trends in smoking behaviour8 9 11 As a result, the negative effects of the 1999 budget cuts were not apparent in examining Florida teen smoking prevalence in 2000.5 However, the downward trend slowed considerably after 2000. Smoking rates among middle school students declined from 18.5 to 11.1 between 1998 and 2000, a reduction of 7.4 percentage points.24 The rate of decline was much slower between 2000 and 2002, from 11.5 to 9.2 (a reduction of 2.3 percentage points). These data suggest that the influence of large tobacco control programme budget cuts on smoking behaviour may take time to become apparent, but do so over time. This underscores the importance of monitoring intermediate outcomes in assessing the impact of reductions to programme funding.

It is possible that the impact of the budget cuts would have been larger in the absence of the national “truth” campaign. Nevertheless, future reductions in state tobacco control funding are unlikely to receive such a benefit from the national campaign. The first 2 years of the national “truth” campaign were funded at an unprecedented $100M a year. The campaign’s current annual funding has been reduced to less than $20M. As a result, today’s teens are unlikely to receive the same level of exposure to the national “truth” campaign that previous youth cohorts received. Reductions in exposure are likely to inhibit the campaign’s efforts to achieve similar levels of success in decreasing smoking rates.24 The national “truth” campaign is unlikely to compensate for reductions in state tobacco control funding today or tomorrow.

Study limitations

The study’s reliance on a commercial vendor marketing list raises questions about the degree to which the sample represents Florida teens. While the sample had contact information for nearly 50 of Florida teens, it is possible that youth on the vendor list were systematically different than teens excluded from the list. We were unable to test for this possibility. There were also modest differences in the distribution of age, gender and race/ethnicity across survey waves. We adjusted for demographics to ensure that changes in trends over time were not a function of differences in sample composition. Nevertheless, the lack of consistency in sample demographics raises questions about the generalisability of the marketing list sample.

There are also alternative explanations for changes in trends over time. Analyses focused on changes in trajectories for programme-targeted outcomes before and after the 1999 budget cuts. While we attributed any changes to the budget cuts, it is also possible that trends are attributable to a “ceiling effect”. For example, the percentage of Florida non-smoking teens with non-smoking intentions rose from 82.7 in April 1998 to 90.8 in May 1999. At this rate, one would predict that 100 of Florida non-smoking teens would have non-smoking intentions by early 2001. This scenario is implausible; the rate of change would inevitably decline as the percentage of teens with non-smoking intentions approached 100. Nevertheless, two factors suggest that the observed changes in trends are attributable to the budget cuts. The shift in trends mirrored the timing of the 1999 FTCP budget cuts, an unlikely coincidence. The anti-industry belief index also showed a similar pattern when accounting for national “truth” recall, but values on the index were far below their maximum. Thus, ceiling effects do not likely explain the change in trends observed for this outcome.

It is possible that changes in cigarette prices could explain part of the observed changes in outcomes between 1998 and 1999. There was a $0.45 (30) increase in Florida cigarette prices in November 1998 and a 7 increase in the second year of the programme, which econometric models predict a 10–20 decline in youth smoking prevalence for 1998 and a 2–5 decline for 1999.32 However, price changes cannot explain why programme exposure declined between 1999 and 2000, or why increases in anti-industry beliefs slowed between September 1998 and May 1999 and stalled completely between May 1999 and October 1999. The timing of the price increases does not match changes in programme-targeted outcomes, suggesting that price effects on prevalence and programme effects upstream outcomes are largely independent.

The repeated cross-sectional data used for this study limits our ability to determine whether national “truth” ad recall influenced industry manipulation beliefs, or vice versa. However, other studies found that recall of specific “truth” ads predicted changes in subsequent anti-industry beliefs among Florida teens,9 providing some evidence that a positive association between national “truth” ad recall and anti-industry beliefs reflects a cause-effect relationship.

CONCLUSIONS

Reductions in tobacco control funding have immediate effects on programme exposure and cognitive precursors to smoking initiation. While the national “truth” campaign may have mitigated some of these negative effects in Florida between 1998 and 2000, reductions in funding for the campaign make it unlikely that national “truth” would have similar compensatory effects today. There is a critical need to maintain and enhance funding for state tobacco control programmes to continue nationwide progress in preventing youth from initiating cigarette smoking.