Introduction

In 2019, tobacco accounted for about 8.17 million deaths globally [1]. Smoking prevalence, which has been declining in past decades, remains significant at approximately 25.0% for men and 5.4% for women [2].

About one in five Filipinos is a smoke [3]–making the Philippines the country with the third highest smoking prevalence in Southeast Asia [4]. Although prevalence has been declining, the annual rate of decline barely changed since 1990 [2]. That average smoking intensity or the number of daily sticks smoked has increased from 18.5 in 1980 to 21.4 in 2012 –a figure higher compared to more than half of the 187 countries studied by Ng et al. [5]–is another cause for concern. This increase in smoking intensity that happened alongside the decline in prevalence lends some support to the “hardening hypothesis,” which posits that tobacco control measures mainly affect less dependent smokers, leaving behind “hardened” smokers who would never quit smoking [6, 7].

The Philippine government significantly increased cigarette excise taxes by passing the Sin Tax Reform Act of 2012 [8]. Since then, a series of tax increases followed (see Table 1 for a summary of excise tax reforms). By 2015, the average price per stick of cigarette was estimated at PhP 2.46 or about 5 US cents [9]. This is about 1.5% of the 2015 average daily individual income in the Philippines of PhP 166.25 (about USD 3.3). The amount is based on authors’ computation using the 2015 average household income [10] and an average household size of 4.4 [11]. There is a convincing body of evidence [12, 13] pointing to tax increases as the “single most effective intervention to reduce demand for tobacco” [14].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Pre- and post- 2012 reform excise tax amounts in Philippine peso (PhP).

https://doi.org/10.1371/journal.pone.0275840.t001

Implementation of the landmark 2012 law was followed by a noticeable decline in smoking prevalence–from 28.3% in 2009 to 23.8% in 2015. However, smoking intensity among daily smokers slightly increased from 10.6 sticks per day in 2009 to 11.0 in 2015, according to the Global Adult Tobacco Survey (GATS) data [3]. This suggests that price increases affect smoking behavior more through the prevention of smoking initiation than the reduction in smoking intensity. One recent study found that while a 10% increase in price would lead to 12.4% reduction in the probability of smoking participation, the same price increase may lead to a 1.4–6.8% increase in sticks smoked [9].

The evidence on which pathway–smoking participation or intensity–excise taxes influence smoking behavior is rather mixed and lacking in the Philippine context. Moreover, not much has been done on smoking cessation determinants in developing countries like the Philippines [12]. These raise the importance of research on varying smokers’ responses to price-based interventions to understand their impact. In this regard, we aim to examine the influence of excise tax increases by uncovering typologies based on smokers’ response to price changes. We also aim to understand the characteristics of smokers in each typology and how they relate to smoking intensity. Using information on how increases in cigarette prices affected individuals’ smoking decisions following the tax reform, we posit that (1) some individuals resort to drastic reduction in smoking either by considering cessation or decreasing sticks smoked, while (2) others find ways to maintain their smoking habits by switching to a cheaper brand, buying in bulk, or asking cigarettes from fellow smokers.

Methods

Results

Our sample has a mean age of 41 years, composed mostly of males (91%), high school graduates (42%), employed individuals (87%), those who smoked within 6 to 30 minutes after waking (30%), rural residents (65%), and has at least one smoker in the household (98%). Our sample smoked about 9 sticks per day and paid about PhP 3 per stick or about 6 cents USD (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Statistics by typology.

https://doi.org/10.1371/journal.pone.0275840.t002

We chose the two-class logit RE LCM since it yielded the lowest BIC and had the lowest proportion of predicted frequencies deviating from the observed frequencies to the total sample (see online S1 and S2 Tables for more information on the LCM).

We call classes 1 and 2 as “unlikely to quit” (UTQ) and “potential quitters” (PQ), respectively, because the two are clearly distinguished by their marginal outcome probability on Stop, Thought, and Decrease–the more drastic measures compared to the price-minimizing Switch, Bulk, and Ask tactics. Compared to those in the UTQ class, smokers classified under PQ have significantly higher marginal outcome probabilities for Stop and Thought, meaning, they are more likely to answer affirmatively to questions closely related to smoking cessation. PQs were also slightly more likely to Decrease than UTQ smokers. In terms of the marginal outcome probabilities of price-minimizing adaptations, those in the PQ and UTQ classes are almost similar (Fig 1). The UTQ individuals make up 37.38% (N = 624) of the sample while PQs make up 62.62% (N = 1,027). The UTQ smokers smoke three more sticks than PQs. As per the bivariate analyses, sex, employment status, nicotine dependence, presence of other smokers at home, and secondhand smoke exposure have statistical association with class membership (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Marginal outcome probabilities by type of smoker.

https://doi.org/10.1371/journal.pone.0275840.g001

Table 3 reports the multivariate analyses results. The odds of females being in the UTQ group is about 51–59% less than males, while students have 55–63% lower odds of being in the UTQ class when compared to the employed. Adding nicotine dependence to the model slightly increased the odds of being in the UTQ group for females vs males and students vs employed. Those in the middle of the nicotine dependence scale have about 39–46% lower odds of being in the UTQ group versus the most-dependent group. Meanwhile, non-daily smokers or the least nicotine-dependent have 73% lower likelihood of being in the UTQ class than those who smoke within 5 minutes of waking (Columns 1 and 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Regression results.

https://doi.org/10.1371/journal.pone.0275840.t003

Robust OLS model results for both measures of smoking intensity is almost similar, so we only show coefficients for sticks smoked. Individuals in the UTQ class smoke about three more sticks than PQs but the difference declines to about 1.6 sticks when nicotine dependence is accounted for. Females smoke about three sticks less than males but the sex-gap in sticks smoked drops to 1.2 sticks when nicotine dependence is considered.

Compared to college graduates, those with less than elementary education smoke about 1.6 sticks less while high school graduates smoke about one stick less. Student smokers consume about 3 sticks less than the employed, although this figure drops to two sticks when nicotine dependence is added. Those who are not in the labor force smoke about two sticks less than the employed. Meanwhile, those in the lowest wealth tertile smoke about one stick less than those in the highest wealth tertile. Individuals who smoke within 6 to 30 min and within 31 to 60 min of waking are smoking four sticks less than those who smoke within 5 min of waking. Moreover, compared to the most nicotine-dependent group, those who smoke after an hour from waking consume six sticks less while the non-daily smokers consume 12 sticks less (Columns 3 and 4).

Discussion

Since 2012, the Philippines made cigarettes significantly more expensive through a series of excise tax reforms. Only a few studies have been done to assess how smoking behavior was affected by the series of tax increases.

The tax could have stopped non-smokers from initiating smoking, made smokers quit, or reduced sticks smoked, but it could have also inadvertently pushed smokers to find ways to pay less to maintain current smoking levels. Our study tried to uncover the types of adaptations in response to the tax increase. Doing so may help understand why, despite the tax reform, average sticks smoked among smokers increased by almost one cigarette per day.

Two types of Filipino smokers emerged: one who made more drastic changes by thinking about quitting or attempting to quit, and another who did not drastically change smoking behavior and instead opted for price minimization. We called the groups “potential quitters (PQ)” and “unlikely to quit (UTQ),” respectively.

PQs had significantly higher probabilities of reporting that they tried to stop or thought about quitting smoking compared to UTQs. UTQs were less likely to resort to decreased consumption. The two classes did not differ much on the probability of price-minimization (i.e., switching to cheaper brands, buying in bulk, asking for cigarettes from others).

Adaptations varied by smokers’ sex, employment status, and nicotine dependence. However, we did not find any difference in class affinity on the basis of educational attainment or wealth, contrary to the existing body of evidence [33–36]. The absence of a significant link for educational attainment is likely due to small sample size and not necessarily the absence of any relationship. Related work done in Brazil [37] found that those with less than nine years of education are twice more likely to try quitting and switching to cheaper brands than those with nine or more years of schooling. In contrast, for a handful of developed western countries, the better educated had lower probabilities of having intentions to quit and adopting price-minimization strategies in response to a hypothetical cigarette price increase [38].

This study is an improvement of previous related work given that we considered possible adaptations as interrelated using LCM rather than looking at them in silos. One study [38] came up with non-mutually exclusive categories to derive a quit intention scale that runs from no change in behavior, adapting in ways other than quitting, adapting both other methods and quitting, and having quitting intentions alone albeit more arbitrarily than data driven. A handful of studies using the LCM to classify smokers in general exist but to the best of our knowledge, only one has been done specifically to classify smokers’ adaptive behavior towards price increases. Crespi et al. [39] uncovered five types of Italian smokers’ adaptations to price increases–typologies differed mainly by the level of sensitivity the smoker has towards risks and whether the smoker chose not to change habits through price-minimization tactics. Unlike our study, they were unable to observe quitting tendencies.

Our study has other limitations: (1) the presence of omitted variable bias despite best efforts to account for confounders, (2) the use of cross-sectional data which prevents us from testing whether our findings will hold over time, and (3) the use of self-reported behavior which can be tainted with social desirability bias.

Our study finds some evidence to suggest that the sin tax reform is associated with both smoking participation and smoking intensity because (1) 63% of smokers who were affected by the price increase considered stopping or quitting, and (2) these PQs smoked one to three sticks less than the UTQs. One strong hindering factor to potential cessation for the 37% in the UTQ class is nicotine dependence. Nicotine dependence significantly increased the odds of being in the UTQ class, consistent with previous studies [38, 40]. In sum, nicotine dependence hindered the impact of cigarette price increases not only on smoking cessation intentions but also smoking intensity.

We emphasize the need to take a comprehensive look at not only price but also non-price-based tobacco control interventions in the Philippines. First, that PQs are more likely to quit due price increases warrants a continuing monitoring and evaluation of the sin tax reform laws, notably, RA 11346 passed in 2019. The results of this paper builds on previous findings [40] by showing that smokers are not merely responding through independent decisions on cessation, demand reduction, or price minimization. Certain characteristics determine whether they are PQs or UTQs, and are therefore predisposed to respond in specific ways. Having a better understanding of the population of current smokers and those most likely to quit given tax-induced price increases can better guide policymaking.

Second, it is time to evaluate whether non-price-based interventions complement price-based measures. The Philippine Regulation Act of 2003 (RA 9211), passed almost two decades ago, provides the basis for most non-priced-based regulation, including smokefree places and prohibitions on advertisements and sale to minors. The law provides for the development of a National Smoking Cessation Program and creation of withdrawal clinics. In relation, the health department has instituted mandatory trainings on the policies and guidelines on smoking cessation since 2003 but only for government healthcare workers. Expanding its coverage should be seriously considered. Moreover, because nicotine dependence remains a strong predictor of adaptations to price increases, we suggest that the DOH provide services to address nicotine dependence. To the best of our knowledge, the closest program the DOH currently has for nicotine dependence is the National Quitline introduced in 2017, which had received only 983 calls [41]–a disproportionately small figure compared to the number of smokers in the country.

That the unlikely to quit smokers are less inclined to quit or reduce consumption points to the importance of complementing tax increases with targeted measures. Studies to better understand the UTQ group can help shed light on the best approach to address this group’s needs. Finally, research towards a better understanding of the smoking epidemic in the country is needed.

Supporting information

Acknowledgments

We are grateful for the substantial comments provided by Shannon Monnat on an earlier version of this paper. We thank the attendees of the 2020 Population Association of America, and the members of Syracuse University’s Policy, Place, and Population Health Lab (P3H Lab) and Lerner Center for Public Health Promotion for their feedback. Errors remain the authors’ alone.