Articles

Effects of comprehensive smoke-free legislation on smoking

behaviours and macroeconomic outcomes in Shanghai,
China: a difference-in-differences analysis and modelling
study
Hongqiao Fu, Sian Tsuei, Yunting Zheng, Simiao Chen, Shirui Zhu, Duo Xu, Winnie Yip

Summary
Background China has one of the highest levels of tobacco consumption globally, and there is no national smoke-free Lancet Public Health 2024
legislation. Although more than 20 Chinese cities have passed local smoke-free laws since 2008, evidence on their Published Online
effectiveness in reducing smoking behaviours and their economic benefits is scarce. By exploiting a natural quasi- November 19, 2024
https://doi.org/10.1016/
experiment, whereby a comprehensive public smoking ban was implemented in Shanghai in March, 2017, this study
S2468-2667(24)00262-7
aims to assess the impact of the policy on individual smoking behaviours and quantify its effect on macroeconomic
See Online/Comment
outcomes. https://doi.org/10.1016/
S2468-2667(24)00272-X
Methods In this difference-in-differences analysis and modelling study, we used data on smoking behaviours from School of Public Health, Peking
the 2012, 2014, 2016, and 2018 waves of the China Family Panel Studies. We used a difference-in-differences approach University Health Science
to investigate trends in smoking prevalence in respondents in Shanghai, relative to respondents from other direct- Center, Beijing, China
(H Fu PhD, S Zhu MSc);
administered municipalities, provincial capital cities, and subprovincial municipalities (control group), after the Department of Family Practice,
implementation of a smoking ban in 2017. All respondents aged 18 years or older were included, with the exception University of British Columbia,
of people who lived in Beijing and rural areas. The primary variable of interest in the difference-in-differences analysis Vancouver, BC, Canada
(S Tsuei PhD); Faculty of Health
was self-reported smoking status. Based on the difference-in-differences estimation of reduction in smoking
Sciences, Simon Fraser
prevalence, we then used a health-augmented macroeconomic model to estimate the potential macroeconomic gains University, Burnaby, BC,
if such a ban was implemented across China for the period 2017–35. Canada (S Tsuei); School of
Health Management, Fujian
Medical University, Fuzhou,
Findings 14 688 respondents were included in the analysis: 5766 from Shanghai and 8922 from the control group.
China (Y Zheng PhD);
After the implementation of the smoking ban in Shanghai in 2017, smoking prevalence decreased by 2·2 percentage Heidelberg Institute of Global
points (95% CI 2·1–2·3), equivalent to an 8·4% reduction in the number of current smokers. The smoking ban had Health, Heidelberg University,
a larger effect on men, people with a higher level of education, unmarried people, and younger people when compared Heidelberg, Germany
(S Chen ScD); Chinese Academy
with their respective counterparts. The modelling analysis showed that implementing a nationwide comprehensive
of Medical Sciences and Peking
public smoking ban similar to that in Shanghai would result in a 0·04–0·07% increase in the national gross domestic Union Medical College, Beijing,
product in China between 2017 and 2035, outweighing the economic costs of smoking ban enforcement. China (S Chen); Institute of
Population and Labor
Economics, Chinese Academy
Interpretation The smoking ban in Shanghai shows that a comprehensive public smoking ban with strict enforcement
of Social Sciences, Beijing,
is effective in curbing smoking behaviours. Moreover, the implementation of a comprehensive public smoking ban China (D Xu PhD); Department
across China would be cost-effective. of Global Health and
Population, Harvard T H Chan
School of Public Health,
Funding National Social Science Fund of China. Boston, MA, USA
(Prof W Yip PhD)
Copyright © 2024 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 Correspondence to:
license. Dr Duo Xu, Institute of
Population and Labor
Introduction such policies protect non-smokers from second-hand Economics, Chinese Academy of
Social Sciences, Beijing 100006,
Tobacco use is the leading cause of preventable deaths in smoke and influence the behaviour of people who do China
the world,1 accounting for approximately 8 million deaths smoke.5–7 Many studies have investigated the effects of [email protected]
and 200 million disability-adjusted life-years (DALYs) public smoking bans on smoking prevalence, smoking
globally in 2019.2 A growing awareness of adverse health intensity, second-hand tobacco smoke exposure, and
consequences of smoking has led to the worldwide health outcomes,8–16 and have shown that the effects are
implementation of various tobacco control policies, country-specific and subject to differences in policy
including increased taxation, smoke-free legislation (ie, enforcement.6,17 However, most of these studies have been
smoking bans in public places), restrictions on done in high-income countries. Evidence on the effects of
advertising, health warning labels, and cessation support public smoking bans in low-income and middle-income
for current smokers.3,4 Among these policies, smoke-free countries (LMICs) is relatively scarce, whereas most of
legislation has gained considerable attention because the tobacco-related disease burden occurs in LMICs.2 It is

www.thelancet.com/public-health Published online November 19, 2024 https://doi.org/10.1016/S2468-2667(24)00262-7 1

 Articles

Research in context
Evidence before this study the effectiveness of the comprehensive smoking ban with strict
We searched PubMed, Web of Science, Google Scholar, and enforcement in curbing smoking behaviours relative to partial
China National Knowledge Infrastructure databases for studies smoking bans. Furthermore, this study assessed the effect of
published between Jan 1, 2000, and June 1, 2024, using the changes in smoking prevalence on macroeconomic outcomes.
terms “smoke-free legislation”, “smoke-free policy”, or “public Implementing a nationwide comprehensive public smoking
smoking ban”, combined with “smoking behaviors”, ban would result in a 0·04–0·07% increase in China’s gross
“macroeconomic benefits”, or “health outcomes”. Many studies domestic product between 2017 and 2035. A preliminary
had demonstrated the effectiveness of smoke-free legislation in calculation suggests that implementing such a comprehensive
curbing tobacco use in high-income countries. However, robust public smoking ban would be highly cost-effective because the
evidence on the effects of smoke-free legislation on smoking macroeconomic gains far outweigh the costs of enforcing the
behaviours and macroeconomic impacts in low-income and ban. Our analysis provides a clear economic interpretation of
middle-income countries (LMICs) was scarce. No previous the impact of the policy in addition to improvements in health
studies had rigorously evaluated the effects of the behaviours.
implementation of the comprehensive public smoking ban in
Implications of all the available evidence
Shanghai, China, in 2017.
Comprehensive public smoking bans are an effective strategy to
Added value of this study reduce smoking prevalence and alleviate the tobacco-related
To the best of our knowledge, this study is the first to evaluate economic burden. LMICs, including China, should accelerate the
the effectiveness of comprehensive public smoking bans in implementation of comprehensive public smoking bans. Strong
China and their macroeconomic outcomes. The comprehensive political leadership, adequate resources for enforcement, and a
public smoking ban in Shanghai reduced smoking prevalence in monitoring and evaluation system are needed to ensure
Shanghai by 2·2 percentage points, equivalent to a reduction of effective implementation.
8·4% in the number of current smokers. This finding highlights

unclear whether the effects of smoking bans are similar in Shanghai stands out for its extensive coverage and
in LMICs, since LMICs have different demographic stringent enforcement. A partial public smoking ban was
structures, enforcement mechanisms, and cultural first launched in the city in 2010, which was expanded to
contexts. a comprehensive public smoking ban in March, 2017.
The highest tobacco consumption globally is in China, This expansive ban is comprehensive in terms of
where more than 300 million people are smokers, coverage, prohibiting smoking in all indoor public
accounting for more than a third of global tobacco venues, workplaces, and public transportation without
consumption in 2019.2 In 2018, the prevalence of smoking exemptions or designated smoking rooms. The ban also
was higher than 25% among Chinese adults aged extends to some outdoor public venues, including
18–69 years,18,19 and was the leading cause of DALYs and children’s playgrounds, open areas of health-care
the second leading cause of premature death in the facilities, and bus stops, in alignment with Article 8 of
country.20 Smoking was estimated to cause more than the WHO Framework Convention on Tobacco Control,
1 million deaths in China in 2010, and it is estimated that which recommends smoking bans in indoor workplaces,
this number will increase to 3 million by 2050, unless public transport, indoor public places and, as appropriate,
there is widespread smoking cessation.21 Although there other public places. To ensure effective enforcement, the
has been laudable progress in China to reduce smoking municipal government of Shanghai strengthened
prevalence since 2005, when the Chinese Government administrative oversights and introduced fines for
For more on the WHO ratified the WHO Framework Convention on Tobacco individuals and entities who violate the ban. For each
Framework Convention on Control, considerable gaps remain.22 According to the offence, individuals who violate the ban could receive
Tobacco Control see
https://fctc.who.int
2023 WHO report on the global tobacco epidemic,1 China fines of up to Chinese Yuan Renminbi (¥)200 (US$29·6,
scored lowest in protecting people from tobacco smoke 1$ was equal to ¥6·75 in 2017), and entities could be
via smoking bans, compared with its efforts on fined up to ¥30 000 (approximately $4444). By contrast,
monitoring, mass media campaigns, and taxation. There in many Chinese cities where local smoke-free laws have
is no unified smoke-free legislation at the national level. been passed, the bans only cover a small number of
Although more than 20 Chinese cities have passed local indoor venues, such as schools, libraries, public transport
smoke-free laws since 2008, regulations only affected vehicles, and health-care facilities. Smoking remains
approximately 13·8% of the Chinese population in permitted in many public places, such as hotels and
2021.23,24 restaurants, and office work spaces were not considered.24
Among the Chinese cities that have passed local Moreover, most of these cities faced challenges in
smoke-free laws, the comprehensive public smoking ban enforcement due to resource constraints, and violations

2 www.thelancet.com/public-health Published online November 19, 2024 https://doi.org/10.1016/S2468-2667(24)00262-7

 Articles

against public smoking bans were common.25 collects data on participants’ demographic characteristics,
Considering this context and data availability, we focused socioeconomic status, health status, and health-care
on Shanghai to evaluate the impact of comprehensive utilisation. The original target sample size was
smoke bans in a middle-income country. More details of 16 000 households in the initial wave of data collection.
the comprehensive public smoking ban in Shanghai are Five selected provincial-level administrative regions
in the appendix (pp 1–5). A previous study showed that (Liaoning, Shanghai, Henan, Guangdong, and Gansu) See Online for appendix
in Shanghai smoking prevalence decreased by were oversampled in the CFPS to provide provincially
3·4 percentage points between 2014 and 2018.23 However, representative samples, and each of the oversampled
the extent to which this reduction can be attributed to the provinces included approximately 1600 house­holds. The
comprehensive public smoking ban and the other 8000 households were from an independent
macroeconomic outcomes of such legislation remain sampling frame composed of 20 provinces. The response
unclear. rate at the individual level in the initial wave was 84·1%.
The difference in timing with regard to the With second-stage sampling, the five oversampled
implementation of public smoking bans between provinces, together with the other 20 provinces, made up
Shanghai and other Chinese cities has created a natural the overall sampling frame representative of the national
experiment that provides an opportunity to evaluate the population. The follow-up surveys tracked all family
effectiveness of comprehensive public smoking bans members of CFPS respondents from the 2010 wave, and
implemented in LMICs such as China. In this study, we their children, and the individual participant response
aimed to use data from a nationally representative rates were 74·1% for the 2012 wave, 72·8% for the 2014
longitudinal survey and a difference-in-differences wave, and 67·4% for the 2018 wave. Details on the
approach to examine the effects of the comprehensive sampling and questionnaire designs in the CFPS have
public smoking ban on smoking behaviours in Shanghai. been published previously.32,33
Furthermore, based on changes in smoking prevalence We used data from the 2012, 2014, 2016, and 2018 waves
derived from the difference-in-differences model, we of CFPS for the difference-in-differences analysis,
applied a health-augmented macroeconomic model to because they covered the period before and after
estimate the macroeconomic gains from implementing Shanghai implemented the comprehensive smoking ban
such a comprehensive public smoking ban across China. in 2017. We excluded data from the 2010 wave due to the
partial ban implemented in Shanghai that year, and we
Methods omitted the 2020 wave to minimise COVID-19 pandemic-
Data sources, measures, and outcomes related impacts.34 Respondents in Shanghai were defined
This study used a difference-in-differences approach to as the treatment group, and subsamples from other
investigate the impact of a comprehensive public direct-administered municipalities, provincial capital
smoking ban on individual smoking behaviours in cities, and subprovincial municipalities were defined as
Shanghai. The difference-in-differences model is widely controls (appendix p 24). We selected these samples as
used in impact evaluation studies, as it can account for the controls, because they had a higher level of
observed factors, unobserved time-invariant differences, socioeconomic development than other prefecture-level
and secular trends.26–29 The key identifying assumption of cities and, therefore, were more likely to be at a similar
the difference-in-differences method is that the average stage of the tobacco epidemic,35,36 and followed the
outcome among the treated and comparison populations parallel trends assumption of the difference-in-
would have followed so-called parallel trends in the differences method. Most cities in the control group
absence of treatment.30,31 implemented partial public smoking bans with no
We used the China Family Panel Studies (CFPS) as our substantial changes in implementation reported during
primary dataset for the difference-in-differences analysis. the study period (appendix p 24), with the exception of
CFPS was approved by the institutional review board at Lanzhou and Xi’an. Lanzhou strengthened the enforce­
Peking University (Peking, China; IRB00001052–14010). ment mechanism in June, 2018, and Xi’an passed the
All CFPS participants provided written informed consent comprehensive smoke-free law in November, 2018. We
before participation, and all personal information has excluded observations in these two cities to test the
been de-identified. Ethics approval was not required for robustness of our results. Furthermore, we excluded
this present study because only secondary datasets were respondents from Beijing, since a similar smoke-free
used. policy was implemented in 2015.37 Respondents in rural
The CFPS dataset is a nationally representative areas were also excluded because only 10% of residents
longitudinal survey that uses multistage probability in Shanghai live in rural areas. Only respondents aged
proportional to size sampling with implicit stratification. 18 years and older were included in the difference-in-
The survey has been done biannually between 2010 and differences analysis because the smoking ban primarily
2020 by the Institute of Social Science Survey at Peking targeted adults in public venues. Further exclusions were
University (Peking, China). The CFPS covers 25 of made for missing data on key variables such as smoking
34 provincial-level administrative regions in China and behaviours, ethnicity, and city of residence. The primary

www.thelancet.com/public-health Published online November 19, 2024 https://doi.org/10.1016/S2468-2667(24)00262-7 3

 Articles

variable of interest in the difference-in-differences education database curated by Barro and Lee.48 We used
analysis was self-reported smoking status, which was these educational data and parameters previously
assessed by the following question, “Did you smoke described by Zhang,49 to estimate the duration of work
cigarettes in the past month?”. This question was experience for each age group, and applied the Mincer
consistent with that in the Global Adult Tobacco Survey Equation to calculate individual human capital. For
adopted by WHO,38,39 and it remained unchanged across modelling physical investment, we derived economic
CFPS waves. Based on this variable, we estimated the growth and capital projections for China from the
smoking prevalence (measured by the share of Organisation for Economic Co-operation and
respondents who reported that they had smoked during Development’s GDP long-term forecast,50 and we
the previous month before the survey). Additionally, we obtained data on China’s saving rate from the World
assessed the intensity of smoking using the question, Bank World Development Indicators database.51 Initial
“How many cigarettes do you smoke on average data on physical capital and its output elasticity were
per day?”. Non-smokers did not need to answer this extracted from the Penn World Table.52 Medical
question. To control for the risk factors of smoking expenditure estimates were based on total health
behaviours, we also include a set of covariates in the expenditure projections published by Zhai and
difference-in-differences analysis (appendix p 25). colleagues,53 with the smoking-attributable fraction, as
Considering the change in smoking prevalence calculated by Chen and colleagues.42 To ensure
estimated by the difference-in-differences approach, we consistency in our analysis, all economic and expenditure
used the health-augmented macroeconomic model data were standardised to 2017 ¥.
developed by Bloom and colleagues40 to calculate the
potential macroeconomic gains from implementing Statistical analysis
such a comprehensive public smoking ban nationwide. We adopted the difference-in-differences model to
This model integrates health capital as a key determinant quantify the effect of the comprehensive public smoking
of physical investment and overall economic productivity, ban on smoking behaviours in Shanghai. Specifically, we
enabling a thorough assessment of the economic benefits used the following linear regression without using survey
of the smoking ban. We simulated and compared the weights, fitted with the least-squares approach:
aggregate gross domestic product (GDP) from 2017 to
2035 under two scenarios: one with a nationwide Yict=α + β × Policyct + γXict + μc + δt + ϵit
smoking ban and one without. This model adopts a
production function-based framework that considers where i indicates the individual, c indicates the city, and t
two critical pathways through which smoking-attributable indicates the year. Yict is the dependent variable indicating
diseases affect economic growth (appendix p 7). The first self-reported smoking status or daily cigarette
pathway pertains to human capital. Smoking increases consumption. The key explanatory variable is Policyct, a
morbidity and mortality, hence reducing the size of the dummy variable representing the implementation of the
working-age population and labour participation. The smoking ban in Shanghai (one for Shanghai respondents
economic value of each worker’s human capital is after 2017 and zero otherwise). The β coefficient captures
assessed by the Mincer Equation,41 which links the average treatment effect of the smoking ban in
educational attainment and work experience to workers’ Shanghai. Xict refers to a series of individuals’
income by estimating the average monetary returns of characteristics and household-level factors that could
one additional year of education or employment. The potentially affect smoking behaviours. Specifically, we
second pathway involves physical capital accumulation, included sex, age, ethnicity, education level, marital
through which smoking-attributed diseases lead to status, employment status, household income,
higher medical expenditure that could reduce investment household size, and living arrangements as control
in physical capital, jeopardising long-term economic variables in the baseline analysis, which have been
growth. Previous studies have applied this model to associated with smoking status.34,54,55 City fixed effects µc
estimate the macroeconomic burden of diseases and the and year fixed effects δt were controls for the unobserved
potential savings from policy interventions.40.42–44 time-invariant heterogeneity across cities and flexible
We extracted data from multiple resources for our time effects. ε refers to the error term. SEs were clustered
modelling analysis on potential macroeconomic gains at the city-year level. Additionally, we tested the
(appendix pp 6–20). To model the labour force in China, robustness of our results to various model specifications,
we used population projections from the UN Population including the addition of other control variables
Division,45 and labour participation rates from the World (individual health, the logarithm of GDP per capita,
Bank dataset.46 Smoking-attributable mortality, years of cigarette prices, and the logarithm of the city population;
life lost, and years lived with disability in 2017 were scenario 2), individual fixed effects (scenario 3),
sourced from the 2021 Global Burden of Disease, provincial linear trends (scenario 4), the 2020 wave
Injuries, and Risk Factors Study.47 Age-specific data on (scenario 5), exclusion of respondents from controlled
mean duration of education were obtained from the cities that strengthened local smoking bans after 2017

4 www.thelancet.com/public-health Published online November 19, 2024 https://doi.org/10.1016/S2468-2667(24)00262-7

 Articles

(scenario 6), and using a logit model (scenario 7). We

Total sample Shanghai Control
conducted a series of sensitivity analyses to test the (n=14 688) (n=5766) (n=8922)
validity of the results derived from the difference-in-
Smoking status
differences analysis, which included survey weights,
Smoker 3844 (26·2%) 1515 (26·3%) 2329 (26·1%)
sample attrition, variation in tobacco prices, the 2015
Non-smoker 10 844 (73·8%) 4251 (73·7%) 6593 (73·9%)
nationwide tobacco tax policy, and other regulations and
Daily cigarette consumption 4·1 (8·6) 4·4 (9·1) 3·9 (8·3)
programmes related to smoke control (appendix p 21).
Sex
We then applied a health-augmented macroeconomic
Female 7679 (52·3%) 2930 (50·8%) 4749 (53·2%)
model to estimate national macroeconomic gains if a
Male 7009 (47·7%) 2836 (49·2%) 4173 (46·8%)
comprehensive public smoking ban was implemented
Age, years 50·5 (16·8) 52·8 (16·6) 49·0 (16·7)
nationwide in China. Full details about the application of
this model are in the appendix (pp 6–20). We compared Education

annual aggregate output in the form of GDP across Senior high school and above 5647 (38·4%) 2324 (40·3%) 3323 (37·2%)

two scenarios over the period from 2017 to 2035: (1) the Junior high school and below 9041 (61·6%) 3442 (59·7%) 5599 (62·8%)
status quo with no further interventions to reduce Marital status
smoking prevalence relative to current and projected Married 11 992 (81·6%) 4748 (82·3%) 7244 (81·2%)
rates, and (2) a counterfactual scenario where China Unmarried or divorced 2696 (18·4%) 1018 (17·7%) 1678 (18·8%)
implemented a nationally comprehensive public Ethnicity
smoking ban as effective as the 2017 ban in Shanghai. Han 14 359 (97·8%) 5739 (99·5%) 8620 (96·6%)
The national macroeconomic gains from implementing Other 329 (2·2%) 27 (0·5%) 302 (3·4%)
such a public smoking ban were calculated as the Employment status
cumulative difference in projected annual GDP between Employed 8132 (55·4%) 3081 (53·4%) 5051 (56·6%)
these two scenarios over the time span of 2017–35. Unemployed 6556 (44·6%) 2685 (46·6%) 3871 (43·4%)
Additionally, we used parameters specific to Shanghai to Log of income, ¥ 9·8 (1·2) 10·2 (1·0) 9·5 (1·0)
directly calculate the macroeconomic gains for Shanghai Household size 3·6 (1·5) 3·4 (1·5) 3·7 (1·6)
after the 2017 smoking ban (appendix p 18). We conducted Living arrangements
sensitivity analyses to test the robustness of our results Living with children younger 2318 (15·8%) 738 (12·8%) 1580 (17·7%)
derived from the macroeconomic model, which than 16 years
considered changes in second-hand smoke exposure, Other arrangement 12 370 (84·2%) 5028 (87·2%) 7342 (82·3%)
alternative calculations of the policy effect, uncertainty in Data are n (%) or mean (SD). An individual was classified as a smoker if they reported smoking at any time during the
mortality rates, alternative depreciation rates of physical month preceding the survey. Household size was defined as the total number of individuals in the respondent’s family,
capital, forecasting of urbanisation, and parameter including the respondent. Living arrangements indicates whether the respondent resides with a child younger than
16 years, coded as 1 for “yes” and 0 for “no”. CFPS=China Family Panel Studies.
variation (appendix pp 16–18).
Table 1: Baseline characteristics of CFPS respondents
Role of the funding source
The funder of this study had no role in study design, data
collection, data analysis, data interpretation, or writing of The difference-in-differences analysis showed that the
the manuscript. smoking prevalence in Shanghai decreased compared
with the control group after the comprehensive public
Results smoking ban in 2017. Smoking prevalence decreased in
14 688 respondents were included in the analysis: Shanghai by 2·2 percentage points (95% CI 2·1–2·3;
5766 from Shanghai and 8922 from the control group p=0·001), equating to an 8·4% reduction (2·2% divided
(table 1). 1515 (26·3%) of 5766 respondents in Shanghai by 26·3% of CFPS respondents in Shanghai who were
and 2329 (26·1%) of 8922 respondents in the control smokers [table 1]) in the number of current smokers in
group were smokers. The daily cigarette consumption Shanghai (table 2). This finding was robust to different
was higher in Shanghai than in the control group (mean model specifications. The β coefficient for smoking
4·42 cigarettes [SD 9·08] vs 3·91 cigarettes [8·30]). The status remained largely unchanged across various model
mean age of respondents in Shanghai was slightly higher specifications, including the addition of other control
than that in the control group (52·8 years [SD 16·6] vs variables (scenario 2), individual fixed effects (scenario 3),
49·0 years [16·7]). Respondents in Shanghai had a higher provincial linear trends (scenario 4), the 2020 wave
level of education (2324 [40·3%] of 5766 respondents vs (scenario 5), and exclusion of respondents from cities
3323 [37·2%] of 8922 respondents with high school that implemented local smoking bans after 2017
education or above) and had a higher mean income than (scenario 6; table 2). This robustness also persisted when
respondents in the control group. The mean household using survey weights, considering the potential bias of
size in Shanghai was smaller than the control group, and sample attrition, and controlling for other smoking-
respondents in Shanghai were less likely to live with related regulations and programmes (appendix p 26).
children younger than 16 years. The daily cigarette consumption among respondents in

www.thelancet.com/public-health Published online November 19, 2024 https://doi.org/10.1016/S2468-2667(24)00262-7 5

 Articles

Scenario 1 Scenario 2 Scenario 3 Scenario 4 Scenario 5 Scenario 6 Scenario 7

Observations, n 14 688 14 688 14 688 14 688 16 710 12 503 14 688
Smoking prevalence, –0·022† (0·006) –0·033† (0·008) –0·018† (0·006) –0·027† (0·006) –0·019† (0·005) –0·019‡ (0·008) –0·317†§ (0·066)
β coefficient (SE*)
p value 0·001 <0·0001 0·002 <0·001 0·001 0·014 <0·0001

Each column represents a sample of urban residents from Shanghai and the control group for the period 2012–18. The outcome variable (smoking status) indicates whether a respondent smoked in the previous
month. Scenario 1 was the baseline model, which included city and year fixed effects, and control variables (table 1). Scenario 2 included additional controls for individual health, per-package cigarette price, the
logarithm of the gross domestic product per capita, and the logarithm of the population at the city level. Scenario 3 included individual fixed effects: due to multicollinearity, sex, age, ethnicity, and city fixed
effects were absorbed. Scenario 4 included provincial linear trends as controls. Scenario 5 extended the sample of analysis to the 2020 wave of the China Family Panel Studies surveys, but the model specification
was the same as scenario 1 (baseline). Scenario 6 excluded respondents from cities that implemented comprehensive public smoking bans or strengthened enforcement after the implementation of the smoking
ban in Shanghai in 2017 (Xi’an and Lanzhou). Scenario 7 reports the log-odds estimation using the Logit regression. *Robust SEs, clustered at the city-year level. †p<0·01. ‡p<0·05. §Log-odds effect.

Table 2: Effect of the comprehensive public smoking ban on the prevalence of smoking in Shanghai

β coefficient 95% CI high school education or higher was 3·1 percentage

points (95% CI 0·9–5·3; p=0·005), whereas no sig­
Male nificant difference was identified among the respondents
Female
with a junior high school education or lower before and
after the 2017 smoking ban. Third, larger decreases in
Higher level of education smoking prevalence were observed among unmarried
respondents than married respondents after the
Lower level of education
comprehensive public smoking ban. Fourth, the
Unmarried reduction in smoking prevalence was significant among
respondents younger than 40 years. In contrast, no
Married
significant difference was identified among respondents
Younger (age <40 years) aged 40 years and older before and after the 2017 smoking
ban. Fifth, we found that the effects of the smoking ban
Older (age ≥40 years)
varied by smoking status before its implementation. The
Previous smoker ban was effective in preventing non-smokers from
starting to smoke tobacco; however, the reduction in
Previous non-smoker overall smoking prevalence was more likely to be driven
–0·08 –0·06 –0·04 –0·02 0 0·02 by previous smokers who quit smoking after the smoking
Effect on smoking prevalence ban. The estimated coefficients for interactions between
the dummy indicating the smoking ban implementation
Figure 1: Effects of the comprehensive public smoking ban on smoking prevalence, by subgroups in Shanghai and variables indicating subgroups suggested that larger
The β coefficients for smoking prevalence, segmented by demographic and socioeconomic groups (sex,
educational level, marital status, age, and previous smoking status). 95% CIs were computed using SEs clustered at
reductions in smoking prevalence were observed among
the city-year level. The regression model incorporated city and year fixed effects and other covariates included in male respondents, people with a higher level of
table 1. The estimated coefficients for interactions indicating the smoking ban implementation and variables education, unmarried respondents, younger people, and
indicating subgroups are shown in the appendix (p 29). smokers before 2017 than among their respective
counterparts (appendix p 29).
Shanghai reduced by 4·4% (95% CI 1·7–7·5; p=0·008) Three additional tests were done to assess the validity
when compared with the control group after 2017 of difference-in-differences results. First, we used an
(appendix p 27). The finding was also robust to alternative event study framework to show the dynamic effects
model specifications (appendix p 27). relative to the year before the ban. Patterns in smoking
Smoking prevalence varied among subgroups prevalence between the Shanghai group and the control
(appendix p 28); therefore, we assessed the heterogeneous group before 2017 were parallel (figure 2). The estimates
effects of this comprehensive public smoking ban across in the placebo tests assuming that the public smoking
different groups. Figure 1 shows the estimates of policy ban was enacted in 2013 or 2015 were also not statistically
effects for each group. First, the reduction in smoking significant (appendix p 30). These results suggest that
prevalence was mainly driven by male respondents. the key identifying assumption of the difference-in-
The smoking prevalence among men decreased by differences method (ie, that in the absence of treatment,
4·0 percentage points (95% CI 1·3–6·7; p=0·004), the difference between the treatment and control group
whereas no significant difference was identified among is constant over time) holds in this study. Second, we
female respondents before and after the 2017 smoking conducted a permutation test, in which we reassigned
ban. Second, the comprehensive public smoking ban had the treatment status to one city in the control group
a greater effect on respondents with higher education iteratively, to address the concern that our results were
attainment. The estimate for respondents with senior driven by characteristics idiosyncratic to specific cities.

6 www.thelancet.com/public-health Published online November 19, 2024 https://doi.org/10.1016/S2468-2667(24)00262-7

 Articles

18 of 1000 tests had greater estimated effects than the β coefficient 95% CI
baseline analysis, reinforcing our finding that the 0·04
smoking ban in Shanghai generated true effects on
smoking behaviours (appendix p 22). Third, we

Change in smoking prevalence

conducted the difference-in-differences analysis with 0·02
propensity score matching to address the observable
difference between the Shanghai group and the control
group before the ban, which showed that the matching 0

method could largely reduce the difference between the

treatment and the control group (appendix p 23). The
–0·02
reductions in smoking prevalence and cigarette
consumption remained statistically significant in the
propensity score matching difference-in-differences
–0·04
setting, albeit with a slightly smaller magnitude 2012 2014 2016 2018 2020
(appendix p 31). These analyses also suggest that the Year
cigarette tax introduced in China in 2015 was not a
Figure 2: Dynamic effects of the comprehensive public smoking ban on
confounding factor, because our findings remained smoking prevalence in Shanghai
consistent when we excluded data from 2015 and earlier. The dynamic effects of the smoking ban implemented in Shanghai in 2017 on
We then estimated the macroeconomic effects of such smoking prevalence compared with the control group. 2016 was defined as the
reference year, with the x-axis indicating the survey year relative to the reference
bans. First, we estimated the total macroeconomic
year. 95% CIs were computed using SEs clustered at the city-year level. Estimates
burden of smoking-attributed diseases in China over the were controlled for city and year fixed effects, and control variables detailed in
period of 2017–35, if no interventions were implemented. the baseline model.
The total burden of smoking-attributed diseases was
estimated to be ¥25·6 trillion at constant 2017 prices from imple­menting the comprehensive public smoking
(approximately $3·8 trillion), which was equivalent to ban across China would equate to 0·04–0·07% in the
0·885% of the GDP, with total economic loss per capita national GDP between 2017 and 2035. Estimates for the
of ¥18 319 (approximately $2714) during the period from direct economic gains from the 2017 smoking ban in
2017 to 2035. The estimate of a counterfactual scenario Shanghai suggest that economic savings would equal
showed that if a nationwide public smoking ban was 0·05–0·07% of the total GDP in Shanghai between 2017
implemented in China between 2017 and 2035, economic and 2035 (appendix p 32).
gains of ¥1675·98 billion (approximately $248·3 billion) We then did a preliminary calculation of the
would be observed (table 3). These macroeconomic gains enforcement costs in Shanghai, using the approach from
were equivalent to 0·058% of the national GDP during Da Mata and Drugowick56 and used the following
this period. information: (1) the mean number of inspections per year
Panels B–G in table 3 report the results of sensitivity (n=300 466), (2) the mean salary for a health inspector,
analysis for economic gains. In Panel B, we assumed (3) the size of the inspection team (six inspectors at a
that the second-hand smoke exposure rate would time), and (4) the mean number of establishments
proportionally decrease with the smoking prevalence, inspected per inspection team per day (three on
and it showed that reduced second-hand smoke exposure average).57 With these parameters, the cost of
due to the ban would result in additional macroeconomic 300 466 inspections in Shanghai in 2019 did not exceed
gain equivalent to 0·006% (gains for scenario 2 ¥600 million, accounting for less than 0·015% of the
[0·064%]–gains for scenario 1 [0·058%]) of the national total GDP in Shanghai for 2019. Considering the higher
GDP. The education-adjusted and age-sex-adjusted labour costs in Shanghai, the enforcement costs as a
results in Panel C suggest that economic gains would be share of total GDP at the national level should be even
equivalent to 0·049% and 0·066% of the GDP, lower.
respectively. When considering the lagged health effects However, fiscal revenue would likely decrease due to
of reducing smoking prevalence, the economic gains the ban, although it would result in economic gains for
were smaller (0·053% of total GDP). Panel D reports the the whole society. In 2019 alone, cigarette sales generated
updated estimates for economic gains using the lower approximately ¥1·2 trillion ($0·178 trillion) in
and upper bounds of 95% CIs for smoking-attributable government revenue, accounting for 6·3% of the total
mortality rates. Panels E and F show the corresponding fiscal revenues (¥19·04 trillion in 2019). Assuming that
results with alternative assumptions of depreciation total cigarette sales decrease proportionally with smoking
rates and the urbanisation process. Panel G reports prevalence, total fiscal revenues would decrease by
estimates from the permutation test using random 0·53% (6·3% × 8·4%). A more conservative estimate
values ranging from 50% to 150% of the initial value. based on the change in per capita cigarette consumption
These results suggest that when accounting for the indicated a decline of 0·28% (6·3% × 4·4%).58 In contrast,
uncertainty in parameters, the macroeconomic gains even if all macroeconomic gains translated to government

www.thelancet.com/public-health Published online November 19, 2024 https://doi.org/10.1016/S2468-2667(24)00262-7 7

 Articles

respondents who were male, more educated, unmarried,

Total economic Economic gains Proportion of total
gains (¥billion) per capita (¥) GDP, 2017–35 (%) and younger were more likely to change their smoking
behaviours after the implementation of the smoking ban.
Panel A: Baseline calculation*
The heterogeneous effects observed might be attributable
Gains from reduction in smoking 1675·98 1201·42 0·058%
prevalence to disparate levels of exposure to the smoking ban and
Panel B: Incorporating second-hand smoke exposure varying baseline smoking prevalence, which were also
Gains from smoking reduction 1854·12 1329·12 0·064%
documented in other studies.8,56 Our results also suggest
Panel C: Alternative specification of policy effects
that comprehensive smoking bans would generate
Education-adjusted effect 1416·98 1015·75 0·049%
differential impacts across Chinese cities, since there are
large differences in social and economic development
Age-sex-specific calculation 1905·08 1365·65 0·066%
within China, and different cities might be at varying
Lagged health effects over 10 years 1534·23 1099·80 0·053%
stages of the tobacco epidemic.36
Panel D: Mortality variation
Importantly, our study shows that implementation of a
Lower bound of 95% CIs 1499·19 1074·69 0·052%
nationwide comprehensive public smoking ban as
Upper bound of 95% CIs 1877·64 1345·98 0·065%
effective as the 2017 ban in Shanghai could generate
Panel E: Alternative depreciation rates of physical capital
economic gains equating to 0·04–0·07% of the total GDP
Depreciation rate of 0 1496·41 1072·70 0·052%
in China between 2017 and 2035, which contrast with the
Depreciation rate of 3% 1600·70 1147·46 0·055%
estimated 0·8–0·9% reduction in total GDP for the same
Depreciation rate of 10% 1878·21 1346·39 0·065%
period without such interventions. This calculation has
Panel F: Alternative urbanisation rate
internalised enforcement costs of the smoke-free policy,
No growth after 2017 1561·61 1119·44 0·054%
which are borne by the government and passed on to the
Panel G: Permutation tests of parameters private sector. However, the tobacco-related tax revenues
Random bootstrap† (95% CI) 1701·04 1219·39 0·059% would decrease accordingly. Since the tobacco-related tax
(1470·59–2080·76) (1054·18–1491·58) (0·051–0·073)
revenues are shared between central and local
Total economic gains and gains per capita between 2017 and 2035 were adjusted to the price level of general goods in governments under which local governments have
2017 using the GDP deflator.50 The specific specifications used in Panels B–G are described in the appendix (pp 16–17). strong incentives to overproduce tobacco products,62 such
In Panel G, a permutation test was performed 1000 times. GDP=gross domestic product. *The difference between the
economic output in the counterfactual scenario and that in the status quo scenario. †Using random values ranging
potential reductions in government revenues might
from 50% to 150% of the initial value. partly explain why comprehensive public smoking bans
are selectively implemented in wealthy cities, which are
Table 3: Estimated macroeconomic gains from a national comprehensive public smoking ban, 2017–35
less dependent on tobacco tax revenues.
Our study has some limitations. First, we did not assess
revenue, they account for merely 0·30% of the total fiscal the direct impact of the smoking ban on health due to data
revenues: limitations. Health outcomes would have taken longer to
GDP in 2019 manifest and the COVID-19 pandemic prevented longer-
macroeconomic gains × = term impact assessments. Second, we did not directly
Fiscal revenues in 2019 examine changes in second-hand smoking exposure.
¥ 98·65 trillion Instead, we imposed a more conservative assumption on
0·058% × reductions in second-hand smoke exposure by assuming
¥ 19·04 trillion
that the second-hand smoke exposure rate would
When macroeconomic gains are multiplied by the ratio proportionally decrease with the smoking prevalence.
of government revenues to GDP (19·3% in 2019), the Therefore, macroeconomic gains in the baseline analysis
comprehensive public smoking ban would lead to an should be treated as a lower bound. Third, the accuracy of
increase of 0·06% in government revenue. parameters has a large effect on the estimated
macroeconomic benefits of reduced smoking prevalence.
Discussion We obtained values of parameters from historical data or
Our analysis showed that the 2017 smoking ban reduced previous studies. Although we conducted a series of
the smoking prevalence in Shanghai by 2·2 percentage sensitivity analyses, the estimated macroeconomic benefits
points, equivalent to a reduction of 8·4% in current of reduced smoking prevalence should be interpreted with
smokers in Shanghai. Daily cigarette consumption caution, and results should be updated accordingly when
among all respondents was reduced by 4·4% after the new data are available. Fourth, although we investigated
implementation of the smoking ban. Previous studies heterogeneity across different groups, the exact impacts of
from China, South Korea, Argentina, Ireland, implementing comprehensive smoking bans in small
the Netherlands, and England all showed that the cities were unknown. We also did not have robust evidence
effectiveness of smoking bans was associated with the on the impacts among rural residents. Fifth, the
comprehensive­ ness and strictness of the bans.17,59–61 production-based macroeconomic model only focused on
Partial smoking bans are less effective at changing supply-side impacts. Incorporating demand-side responses
smoking behaviours.59 The study findings show that and market equilibrium dynamics would substantially

8 www.thelancet.com/public-health Published online November 19, 2024 https://doi.org/10.1016/S2468-2667(24)00262-7

 Articles

increase model complexity. Moreover, due to data and gains from national scale up. Such a step-by-step
limitations, we did not calculate the reduced loss of unpaid policy implementation approach could improve public
household production from comprehensive smoke-free awareness and reduce resistance.
legislation.63 If this part is considered, estimated Contributors
macroeconomic gains would become even larger. HF and DX contributed to the study design and conceptualisation.
Our study had the following strengths. First, this study HF and DX contributed to the methodology. DX and SZ implemented
the data analysis. HF and DX wrote the original draft. ST and WY
provided robust evidence on smoke-free legislation from revised and edited the draft. HF, DX and SZ have directly accessed and
a large middle-income country. Existing evidence on the verified the underlying data reported in the manuscript. HF and DX
impact of public smoking bans mainly focused on high- were responsible for the decision to submit the manuscript. All authors
income countries. A better understanding of the impacts discussed the results, commented on the manuscript, and approved the
final version.
of public smoking bans in LMICs is needed, since
tobacco-associated deaths and diseases disproportionately Declaration of interests
We declare no competing interests.
affect people from these countries. Second, we
investigated the causal effects of the smoking ban on Data sharing
Requests for CFPS data can be made at https://www.isss.pku.edu.cn/
smoking behaviours. The unique setting in China cfps/gycfps/cfpsjj/index.htm. Codes for empirical analysis and other
allowed us to apply a difference-in-differences design by data used in this study are available on reasonable request to the
comparing changes in smoking behaviours in Shanghai corresponding author.
with comparison groups. Previous findings from Acknowledgments
interrupted time-series analyses have often been We acknowledge funding by the National Social Science Fund of China
(grant number 23&ZD182). We thank Yuan Xin and Yishan Zhu for
questioned, because changes in smoking rates could also
their research assistance. We also thank the China Family Panel Studies
be attributed to factors other than the smoking ban. project team for granting permission to use restricted city codes held on
Additionally, the individual data used in this study their database.
allowed us to do heterogeneity analyses to assess the References
distributional impacts. Third, to the best of our 1 WHO. WHO report on the global tobacco epidemic, 2023: protect
knowledge, this study is the first to estimate the economic people from tobacco smoke. World Health Organization, 2023.
https://www.who.int/publications/i/item/9789240077164 (accessed
gains from implementing public smoking bans in Oct 18, 2023).
LMICs, and provide much needed support to scale up a 2 Reitsma MB, Kendrick PJ, Ababneh E, et al. Spatial, temporal, and
comprehensive public smoking ban in China and will demographic patterns in prevalence of smoking tobacco use and
attributable disease burden in 204 countries and territories,
further raise awareness among policy makers about the 1990–2019: a systematic analysis from the Global Burden of Disease
necessity of smoke-free legislation. Study 2019. Lancet 2021; 397: 2337–60.
Our findings have important policy implications. First, 3 Mehrotra R, Yadav A, Sinha DN, et al. Smokeless tobacco control in
180 countries across the globe: call to action for full implementation
LMICs, including China, should accelerate the of WHO FCTC measures. Lancet Oncol 2019; 20: e208–17.
nationwide implementation of comprehensive public 4 Flor LS, Reitsma MB, Gupta V, Ng M, Gakidou E. The effects of
smoking bans. Smoking bans effectively curb smoking tobacco control policies on global smoking prevalence. Nat Med
2021; 27: 239–43.
behaviours and can generate substantial economic 5 WHO. WHO report on the global tobacco epidemic, 2009:
benefits. Second, the comprehensive nature of the ban implementing smoke-free environments. Geneva: World Health
might be an important feature, since our findings Organization, 2009. https://www.who.int/publications/i/
item/9789241563918 (accessed Oct 18, 2023).
contrast with the limited impacts of partial or voluntary
6 DeCicca P, Kenkel D, Lovenheim MF. The economics of tobacco
smoking bans from other studies in China and other regulation: a comprehensive review. J Econ Lit 2022; 60: 883–970.
countries. Therefore, strong political leadership, 7 Joossens L, Raw M. The Tobacco Control Scale: a new scale to
adequate resources for enforcement, and a monitoring measure country activity. Tob Control 2006; 15: 247–53.
8 Anger S, Kvasnicka M, Siedler T. One last puff? Public smoking
and evaluation system need to be in place to ensure bans and smoking behavior. J Health Econ 2011; 30: 591–601.
effective implementation. Third, the heterogeneous 9 Been JV, Nurmatov UB, Cox B, Nawrot TS, van Schayck CP,
effect of the comprehensive public smoking ban implies Sheikh A. Effect of smoke-free legislation on perinatal and child
health: a systematic review and meta-analysis. Lancet 2014;
that the effectiveness and corresponding macroeconomic 383: 1549–60.
gains might be country-specific, due to large differences 10 Bharadwaj P, Johnsen JV, Løken KV. Smoking bans, maternal
in social and economic development across countries. smoking and birth outcomes. J Public Econ 2014; 115: 72–93.
Smoke-free policies had a small effect on people with 11 Boes S, Marti J, Maclean JC. The impact of smoking bans on
smoking and consumer behavior: quasi-experimental evidence
less education and older age groups, suggesting that from Switzerland. Health Econ 2015; 24: 1502–16.
other tobacco control approaches, such as raising tobacco 12 Carpenter C, Postolek S, Warman C. Public-place smoking laws and
taxes, increasing public health literacy, and banning exposure to environmental tobacco smoke (ETS).
Am Econ J Econ Policy 2011; 3: 35–61.
tobacco advertisements should be considered. Fourth,
13 Jones AM, Laporte A, Rice N, Zucchelli E. Do public smoking bans
decreasing government revenues from tobacco sales have an impact on active smoking? Evidence from the UK.
might be a common dilemma encountered by many Health Econ 2015; 24: 175–92.
LMICs when they plan to scale-up comprehensive 14 Mackay D, Haw S, Ayres JG, Fischbacher C, Pell JP. Smoke-free
legislation and hospitalizations for childhood asthma. N Engl J Med
smoke-free policies. LMICs can pilot smoke-free 2010; 363: 1139–45.
legislation in some areas, and identify the potential costs

www.thelancet.com/public-health Published online November 19, 2024 https://doi.org/10.1016/S2468-2667(24)00262-7 9

 Articles

15 Pell JP, Haw S, Cobbe S, et al. Smoke-free legislation and 40 Bloom DE, Chen S, Kuhn M, McGovern ME, Oxley L, Prettner K.
hospitalizations for acute coronary syndrome. N Engl J Med 2008; The economic burden of chronic diseases: estimates and projections
359: 482–91. for China, Japan, and South Korea. J Econ Ageing 2020; 17: 100163.
16 Radó MK, van Lenthe FJ, Laverty AA, et al. Effect of comprehensive 41 Deming DJ. Four facts about human capital. J Econ Perspect 2022;
smoke-free legislation on neonatal mortality and infant mortality 36: 75–102.
across 106 middle-income countries: a synthetic control study. 42 Chen S, Kuhn M, Prettner K, Bloom DE. Noncommunicable diseases
Lancet Public Health 2022; 7: e616–25. attributable to tobacco use in China: macroeconomic burden and
17 Nagelhout GE, de Vries H, Boudreau C, et al. Comparative impact tobacco control policies. Health Aff (Millwood) 2019; 38: 1832–39.
of smoke-free legislation on smoking cessation in three European 43 Chen S, Cao Z, Prettner K, et al. Estimates and projections of the
countries. Eur J Public Health 2012; 22 (suppl 1): 4–9. global economic cost of 29 cancers in 204 countries and territories
18 Chan KH, Xiao D, Zhou M, Peto R, Chen Z. Tobacco control in from 2020 to 2050. JAMA Oncol 2023; 9: 465–72.
China. Lancet Public Health 2023; 8: e1006–15. 44 Chen S, Kuhn M, Prettner K, et al. The global economic burden of
19 Zhang M, Yang L, Wang L, et al. Trends in smoking prevalence in chronic obstructive pulmonary disease for 204 countries and
urban and rural China, 2007 to 2018: findings from 5 consecutive territories in 2020–50: a health-augmented macroeconomic
nationally representative cross-sectional surveys. PLoS Med 2022; modelling study. Lancet Glob Health 2023; 11: e1183–93.
19: e1004064. 45 UN Population Division. World Population Prospects 2022. https://
20 Zhou M, Wang H, Zeng X, et al. Mortality, morbidity, and risk factors population.un.org/wpp/Download/Archive/Standard/(accessed
in China and its provinces, 1990–2017: a systematic analysis for the Oct 18, 2023).
Global Burden of Disease Study 2017. Lancet 2019; 394: 1145–58. 46 The World Bank. Labor force participation rate, total (% of total
21 Chen Z, Peto R, Zhou M, et al. Contrasting male and female trends population ages 15–64) (modeled ILO estimate). https://data.
in tobacco-attributed mortality in China: evidence from successive worldbank.org/indicator/SL.TLF.ACTI.ZS?locations=CN (accessed
nationwide prospective cohort studies. Lancet 2015; 386: 1447–56. Oct 18, 2023).
22 Yang G, Wang Y, Wu Y, Yang J, Wan X. The road to effective tobacco 47 Institute for Health Metrics and Evaluation. Global Burden of
control in China. Lancet 2015; 385: 1019–28. Disease Study 2021 (GBD 2021) results. 2021. http://ghdx.
23 Wan X. Protecting people from tobacco smoke in China: current healthdata.org/gbd-results-tool (accessed Oct 5, 2024).
status and challenges. China CDC Wkly 2022; 4: 460–64. 48 Barro RJ, Lee JW. A new data set of educational attainment in the
24 Lin H, Chang C, Liu Z, Zheng Y. Subnational smoke-free laws in world, 1950–2010. J Dev Econ 2013; 104: 184–98.
China. Tob Induc Dis 2019; 17: 78. 49 Zhang X. The rate of returns to schooling: a case study of urban
25 Yang J. China Report on Smoke-Free Law Enforcement. Beijing, China. Int J Humanit Soc Sci 2011; 1: 173–80.
China: China Democratic and Legal Publishing House, 2015 50 Organisation for Economic Co-operation and Development. Long-
(in Chinese). term baseline projections, no. 109 (edition 2021). https://www.oecd-
26 Rose TC, Daras K, Manley J, et al. The mental health and wellbeing ilibrary.org/content/data/cbdb49e6-en (accessed Oct 5, 2024).
impact of a Community Wealth Building programme in England: 51 The World Bank. Gross savings (% of GDP) - China. https://data.
a difference-in-differences study. Lancet Public Health 2023; 8: e403–10. worldbank.org/indicator/NY.GNS.ICTR.ZS?view=chart&­
27 Epure AM, Courtin E, Wanner P, Chiolero A, Cullati S, Carmeli C. locations=­CN (accessed Oct 18, 2023).
Effect of covering perinatal health-care costs on neonatal outcomes 52 Feenstra R, Inklaar R, Timmer MP. The next generation of the
in Switzerland: a quasi-experimental population-based study. Penn World Table. Am Econ Rev 2015; 105: 3150–82.
Lancet Public Health 2023; 8: e194–202. 53 Zhai T, Goss J, Dmytraczenko T, Zhang Y, Li J, Chai P. China’s
28 Yao Y, Lv X, Qiu C, et al. The effect of China’s Clean Air Act on health expenditure projections to 2035: future trajectory and the
cognitive function in older adults: a population-based, quasi- estimated impact of reforms. Health Aff 2019; 38: 835–43.
experimental study. Lancet Healthy Longev 2022; 3: e98–108. 54 Wang M, Luo X, Xu S, et al. Trends in smoking prevalence and
29 Lai Y, Fu H, Li L, Yip W. Hospital response to a case-based payment implication for chronic diseases in China: serial national cross-
scheme under regional global budget: the case of Guangzhou in sectional surveys from 2003 to 2013. Lancet Respir Med 2019;
China. Soc Sci Med 2022; 292: 114601. 7: 35–45.
30 Angrist JD, Krueger AB. Empirical strategies in labor economics. 55 Liu Z, Li YH, Cui ZY, et al. Prevalence of tobacco dependence and
In: Ashenfelter O, Card D, eds. Handbook of labor economics. associated factors in China: findings from nationwide China Health
Amsterdam: Elsevier, 1999: 1277–366. Literacy Survey during 2018–19. Lancet Reg Health West Pac 2022;
31 Roth J, Sant’Anna PH, Bilinski A, Poe J. What’s trending in 24: 100464.
difference-in-differences? A synthesis of the recent econometrics 56 Da Mata D, Drugowick P. The consequences of health mandates on
literature. J Econom 2023; 235: 2218–44. infant health: evidence from a smoking-ban regulation. J Dev Econ
32 Xie Y, Hu J. An introduction to the China Family Panel Studies 2023; 167: 103171.
(CFPS). Chin Sociol Rev 2014; 47: 3–29. 57 Huang J. The penalty amount for tobacco control in Shanghai in
33 Xie Y, Lu P. The sampling design of the China Family Panel Studies 2019 was more than 2·68 million yuan. Jiemian News, 2019.
(CFPS). Chin J Sociol 2015; 1: 471–84. https://www.jiemian.com/article/4448472.html (accessed
34 Cai W, Zhou Y. Men smoke less under the COVID-19 closure Oct 18, 2023; in Chinese).
policies: the role of altruism. Soc Sci Med 2022; 306: 115159. 58 Nargis N, Stoklosa M, Shang C, Drope J. Price, income, and
35 Lopez AD, Collishaw NE, Piha T. A descriptive model of the cigarette affordability as the determinants of tobacco consumption:
epidemic in developed countries. Tob Control 1994; 3: 242–47. a practitioner’s guide to tobacco taxation. Nicotine Tob Res 2021;
23: 40–47.
36 Thun M, Peto R, Boreham J, Lopez AD. Stages of the cigarette
epidemic on entering its second century. Tob Control 2012; 59 Catalano MA, Gilleskie DB. Impacts of local public smoking bans
21: 96–101. on smoking behaviors and tobacco smoke exposure. Health Econ
2021; 30: 1719–44.
37 Zheng Y, Wu Y, Wang M, et al. Impact of a comprehensive tobacco
control policy package on acute myocardial infarction and stroke 60 Ko H. The effect of outdoor smoking ban: evidence from Korea.
hospital admissions in Beijing, China: interrupted time series Health Econ 2020; 29: 278–93.
study. Tob Control 2020; 30: 434–42. 61 Fong GT, Sansone G, Yan M, Craig L, Quah ACK, Jiang Y.
38 WHO, Centers for Disease Control. Tobacco questions for surveys: Evaluation of smoke-free policies in seven cities in China,
a subset of key questions from the Global Adult Tobacco Survey 2007–2012. Tob Control 2015; 24 (suppl 4): iv14–20.
(GATS): global tobacco surveillance system. Geneva: World Health 62 Hu TW, Mao Z, Shi J, Chen W. The role of taxation in tobacco
Organization, 2011. https://iris.who.int/bitstream/handle/10665/ control and its potential economic impact in China. Tob Control
87331/9789241500951_eng.pdf (accessed Oct 5, 2024). 2010; 19: 58–64.
39 Palipudi KM, Morton J, Hsia J, et al. Methodology of the Global 63 Nargis N, Hussain AKMG, Asare S, et al. Economic loss attributable
Adult Tobacco Survey–2008–2010. Glob Health Promot 2016; to cigarette smoking in the USA: an economic modelling study.
23 (suppl): 3–23. Lancet Public Health 2022; 7: e834–43.

10 www.thelancet.com/public-health Published online November 19, 2024 https://doi.org/10.1016/S2468-2667(24)00262-7