Background The objective of this study is to estimate the sensitivity of cigarette quantity demanded to price and income changes in Montenegro.
Data and methods The sensitivity of cigarette quantity demanded to price and income changes was estimated using micro and macro analysis. Micro analysis implied the use of Deaton’s model on Household Budget Survey data (2006–2017). In macro analysis, conventional static demand model is applied using error correction and autoregressive distributed lag time series methodology on annual time series aggregated data (2001–2017).
Results The same results were obtained using micro and macro analysis which contributes to the objectivity of the conducted research. Results derived from the Deaton’s model indicate a negative price elasticity of cigarettes in the range between −0.62 and −0.80 (conditional and unconditional), while in macro model estimated price elasticity is in that range and equals −0.68. Simulation results confirm the efficiency of excise tax policy changes, having an evident decrease in consumption and increase of public revenues.
Conclusion Analysis of the tobacco market and regulatory environment suggests that the increase of excise and other taxes on tobacco have an important direct impact on the reduction of cigarettes and other tobacco products consumption. Our estimates of long and short-run price elasticity show that direct impact is strong and very much in accordance with the results obtained so far for other low-income and middle-income countries. This paper gives a contribution to the analysis of price elasticity of demand for cigarettes, which was for the first time conducted in Montenegro.
- low/middle income country
- public policy
- global health
Statistics from Altmetric.com
The prevalence of smoking in Montenegro is high, with an increasing trend in the recent period. Results from Montenegrin Institute of Public Health,1 given in table 1, show a decreasing trend of smoking prevalence in the adult population from years 2000 to 2012, while increasing afterward. The increase is mostly due to the female smoking rates increase in 2017 compared to 2012, having a stable trend of male prevalence in the same period.1 Results from various sources and research1–6 showed an increasing trend of smokers in youth population in Montenegro, also particularly among females.
Total legal consumption (table 1), which is calculated based on used excise stamps, shows decreasing trend from 2000 to 2017 (78% decrease; data obtained from Montenegrin Ministry of Finance). The possible reasons for higher smoking rates in 2017 compared to 2012, while having a decreasing trend of legal consumption at the same time, could be potentially in cross-border transactions and illicit trade. Additional reason could be that consumers may smoke fewer cigarettes, even if number of smokers is increasing.
The Government of Montenegro, as a signatory to the WHO Framework Convention on Tobacco Control (WHO FCTC),7 8 encourages the reduction of tobacco use through a set of smoking control measures. All relevant legislation,8–15 as well as strategies, was adopted. The tobacco tax structure includes a specific tax that currently amounts €30 on 1000 cigarettes and an ad valorem tax of 32% of the weighted average price of cigarettes (WAPC). Montenegro is aiming to fulfil the European Union (EU) Directives16 17 and excise tax requirements are partially in line with the Council Directive 2011/64/EU. WAPC in 2018 amounted €2.10 compared to the European Union average of €4.80 in the same year, while minimal excise per pack of cigarettes was €1.27 comparing to European Union directive of €1.80. Even though the European Union levels are not still being fulfilled, the excise tax increase was followed by decreasing trend of consumption, and increasing trend of public revenues share in Gross Domestic Product (GDP) collected from tobacco, from 0.70% to 1.70%.18
The objective of this study is to give the first estimates of the sensitivity of cigarette demand to price and income changes, by applying micro and macro analysis. Moreover, the goal is to verify the suitability of the methods applied, Deaton’s19 20 and time series, in case of limited availability of data, especially market price. The data related to cigarettes consumption is derived from the Household Budget Survey (HBS), covering a longer period, which allows us to apply Deaton’s methodology. Even though there are no available longer time series aggregated data on consumption and prices of cigarettes, times series methodology was applied, as it represents the most widely applied methodology in low-income and middle-income countries (LMICs). This methodology is also applied with the aim to check robustness of micro analysis. Additionally, the study will provide simulation effects of an excise tax increase on consumption and public revenues, using estimated price elasticities.
Generally, the results of studies in LMICs, found price elasticity estimates ranging between 0 and −1, depending on methodology applied.21 22 Deaton’s model in LMICs is mostly used on HBS data: Asia,23–28 Africa29 30 Latin America.31 32 Most of these researches were done from 2000 onwards. Some studies in LMICs use sample of households with positive cigarettes purchases, or estimate conditional price elasticities in range between –0.20 and –0.60.23 29 On the other hand, others use the whole sample, including households with zero purchases, or estimate unconditional elasticities in range between –0.50 and –1.30.24 25 27 28 30 32 To the best of our knowledge, there is a lack of research using Deaton’s method in LMICs in Western Balkan, as well as in Eastern Europe.
If we summarise results of a number of studies in LMICs from 2000 and onwards, which use time series methodology,21 31 33–38 we conclude that they are in line with estimated elasticities using Deaton’s method, being in the range between –0.20 and –0.80.
This paper contributes to the existing literature of the analysis of price effects on cigarettes consumption and public revenues, providing conditional and unconditional estimates of price and income elasticities for the first time in Montenegro. Being the first scientific research of this kind, the simulations on consumption and public revenues can be used as a crucial input for evidence-based policy-making, such as legislative changes and compliance with European Union directives. This fact is even more important, taking into consideration that Montenegro, in its region, has the highest adult’s prevalence of tobacco smoking, according to WHO estimates.6 Due to previously mentioned, our research represents the inception phase for further research related to tobacco prevalence and diverse health effects.
Data and methods
In micro analysis, we use HBS data from 2006 to 2017 (excluding 2016 when Survey was not conducted), collected by the Statistical Office of Montenegro.39 HBS is the national annual survey which provides data for average households’ expenditure and consumption size and structure. Households are surveyed during 1 month per year which means that each household is surveyed only once a year. This survey covers 21 municipalities in three regions in Montenegro—North, Central and South. The database for each year provides detailed characteristics on households, comprising of socio-economic and demographic features (on average 1202 households surveyed annually). There are 12 503 households in the whole sample and 5775 households with positive consumption or 46.33% of the whole sample, after removing outliers. Even though the price effect may trigger the transition from cigarettes to other tobacco products, we have excluded households who report spending on other tobacco products, having a negligible share—only 3.59%.
Deaton’s model,19 20 applied in micro analysis, is a model of consumer behaviour in which households within a specific cluster choose products and spend the budget on a product depending on its price, quantity and quality. Using this model, a system of demand equations for deriving own and cross-price elasticities can be estimated, based on expenditure and quantity data from household surveys. The model defines a proxy for cigarette prices, or unit value, as a ratio of cigarette expenditures and quantity. However, it is important to point out that this ratio is only unit value, not price. The reason is found in the fact that unit value, besides information about the price, also reflects the choice of quality (quality shading) and possible measurement error. Quality shading means that consumers could retain the same level of consumption, switching to cheaper brands, due to price increase. Considering the relationship of harmfulness, price and quality of cigarettes, all cigarettes are considered to be harmful and while there may be very minor differences in how harmful, this is captured by the Deaton model as part of how it deals with differences in prices and quality. On the other hand, measurement error refers to misreport expenditure of goods purchased. This is why Deaton’s model proposes a methodology on how to correct the final estimation of price elasticity for quality shading and measurement error. Additionally, the model assumes that unit values vary spatially between clusters, but not within the cluster. In order to capture the effect of the spatial price variation, the households in the sample must be geographically clustered.
The unit values are expressed in euro per cigarette pack. Moreover, the model includes a variable budget share which is calculated as a ratio of monthly household expenditure on cigarettes and the total monthly household expenditure.
The model relies on two equations19 20:
The dependent variables in (1) and (2) are budget share and unit value, respectively, both household i in the cluster c. We define clusters c based on the information on municipalities and years, that is, the cluster is defined as a municipality x in the year t. The independent variables are: is the logarithm of total household expenditure; is the vector of household socio-economic characteristics (household size, adult ratio, male ratio, maximum and mean of years of education and household type by activity); is the logarithm of prices; is the idiosyncratic error.
Real values for budget share, unit value and all expenditure variables were constructed using annual Consumer Price Index (CPI) for the period 2006–2017.18 Regression beta coefficient of variable in the unit value equation represents expenditure elasticity of quality which measures choice of quality due to changes in the household’s wealth. The degree of unit value change due to 1 unit change in prices is represented by the coefficient ψ. Using the information on the quality effects magnitude (), it is possible to correct final price elasticity for quality shading. In case of no variations due to the product’s quality, the change of unit value would reflect only the changes in price ( → 0, ψ = 1). The model consists of three stages. In the first stage, the effects of household socio-economic characteristics are removed from both dependent variables, so that estimation of price elasticity in the second stage is not influenced by these characteristics (ie, purging quality effects). In the second stage, cluster averages of budget shares and unit values are generated. Finally, in the third stage the weak separability assumption is introduced, as to separate the price and quality effects within the unit value.
According to Deaton,19 20 demand model formulation depends on whether we want to model demand conditional on making purchases, or whether unconditional with consumers and non-consumers covered. Some previous empirical research in LMICs23 40 confirmed that spending decisions of households with zero consumption do not depend on cigarette prices. Still, considering that there is no evidence for Montenegro that cigarettes do not exist in non-consumers utility function, no matter what the price is, we decided to estimate both conditional and unconditional elasticity. In that manner, we can estimate excise tax impact on consumption of consumers, as well as on the whole sample, because some non-consumers may begin to consume tobacco products when the price decreases.
In the macro analysis, we used annual time series aggregate data on cigarette consumption, cigarette prices, income and public policies from 2001 to 2017. We estimated cigarette demand function using a conventional static demand model, presented in linear functional form. According to the time-series properties of data, the Engle-Granger two-step method was applied.41 In the first step we estimated long-run price elasticity by employing the following linear functional form of cigarette demand:
where the dependent variable is legal cigarette consumption per adult population. was calculated by multiplying the number of used excise stamps by number of cigarettes in one pack (20 cigarettes) and divided by the total number of adults (15+).18 The obtained data of used excise stamps was the only available data on legal cigarettes consumption. However, taking into account that industry may take advantage of frontloading, there is a possibility that this quantity does not completely correspond to the real consumption. Having regard to still low price and high availability of cigarettes in Montenegro, we may assume that the deviations of legal consumption from the real values are not high.
Independent variables comprised:
Cigarette prices () approximated by real tobacco CPI18—tobacco CPI divided with general CPI.
Income () approximated by GDP per capita in constant prices.42
A single regulatory variable () comprising of Law on Limiting Use of Tobacco Products amendments14 introduced at once in 2011 (stricter bans of tobacco advertising, sponsorship, marketing, selling and promotion, as well as on tobacco use in public spaces).
Tested impact on consumption of other regulatory changes, covering all tobacco control laws and strategies was not statistically significant. Due to estimation problems (multicollinearity, endogeneity), other control variables (enrolment in tertiary education, employment and unemployment percentage, male to female ratio, survival to age 65 for male and female) were not included. In the second step of the Engle-Granger method—error correction model (ECM), we used the first-differenced variables and lagged residuals from the first step to estimate the short-run price elasticity of cigarette quantity demanded. We checked the sensitivity of the results using an alternative estimating methodology in cointegration framework - autoregressive distributed lag (ARDL).
Estimation of the price elasticity using micro data
Table 2 summarises the results of regression analysis related to unit values and budget shares regional and time variations. These variations are significant in case of both variables. The exception is the consumers’ budget share in the period 2011–2015. In both samples, the budget share has a decreasing trend, while we can notice an increasing trend of the average price of cigarettes per pack during the period observed. The reason for the budget share decrease is the higher growth rate of total expenditure comparing to the growth rate of expenses on cigarettes. Decrease of budget share could also be explained by the decrease of the quantity purchased or substitute to cheaper quality due to price increase.
The results from the regression are expected, showing the lowest prices per pack paid by household in the North region, being the least developed region in Montenegro. The prices in South and Centre are approximately the same. On the other hand, even though the North region is the least developed, the households spend a larger share of the available budget on cigarettes compared to other two regions, which was also confirmed in previous research.27 43 44
Descriptive statistics of variables used in the first stage regression is given in table 3. According to t statistics, there is a statistically significant difference between variables means of consumers and the whole sample. Results of both samples indicate almost equal representation of man and woman in households, while adults represent 88% of the household members. The majority of households’ members have a secondary level of education, while on average 3% of households are ‘unemployed’, 19% to 25% are ‘pensioners’, and more than 70% ‘employed’. The lower budget share in the whole sample is expected, 2.10% compared to 4.60% in the sample of consumers.
First stage: household level regression
Summary of the first stage regression results is presented in table 4.
The expenditure elasticity of quality amounts 0.22, indicating that quality effects are present in unit value data. With 10% higher total expenditure, the households tend to buy cigarettes that are 2.22% more expensive, controlling for other sociodemographic variables. Considering that these results confirm quality shading in Montenegro, the use of Deaton’s model is appropriate, as it enables control for quality effects. When it comes to influencing of total expenditure on cigarettes budget share, the 10% higher total expenditure will lead to a decrease of 0.25% (consumers) or 0.04% (whole sample) share of budget spent on cigarettes.
Based on the sign and the value of other socio-demographic variables, we can conclude that larger households, households with more men and elderly members pay lower prices. Pensioners and unemployed households pay less for cigarettes, compared to employed. Households with more men and adults spend a larger share of the budget on cigarettes. Cluster fixed effects confirm the existence of time and spatial variation.
Based on the results from the first stage regression, we generate the results of the total expenditure elasticity of demand which is positive and amounts 0.23 (conditional) and 0.63 (unconditional). More precisely, 10% higher total expenditure of household will lead to 2.30% (conditional) or 6.30% (unconditional) higher demand for cigarettes.
Second stage: cluster level estimates
In the second stage regression, the analysis additionally purged the regional effects, due to differences between the regions (North, Centre and South) in the level of development, which could influence consumer preferences. Results of the estimation indicate a negative price elasticity of cigarettes, which amounts −0.62 (conditional) and −0.80 (unconditional). SE of the elasticity calculated via bootstrapping procedure (1.000 replications) indicates that the elasticity is significantly lower than zero (ϵc= -0.62; SEϵ=0.11, t=−5.53; ϵuc= -0.80; SEϵ=0.13, t=−6.27). These results are in line with previous relevant scientific research done applying Deaton’s model on HBS data23–32 which showed that the price elasticity of the demand in LMICs varies from −0.2 to −0.6 (conditional) and −0.5 to – 1.3 (unconditional).
Estimation of the price elasticity using macro data
Examination of time-series properties of data showed that all variables presented in equation 3 are non-stationary at the level, stationary at the first difference, cointegrated and that price exogeneity cannot be rejected which allow us to use the Engle-Granger two-step methodology.41 From regulatory framework point of view, variable that represents changes in tobacco control regulation from 2011,14 was negative, suggesting that these changes had a negative impact on cigarette consumption. Price coefficient was also negative, as it was expected. We derived long-run and short-run price elasticities using estimated price coefficients in the first and the second step of the Engle-Granger method. Our model passed all post-diagnostic and specification test implemented, and the results along with estimated elasticities in table 5.
Significant coefficient of lagged residuals, given in table 5, indicates that after a short-run response to a change, consumption monotonically converges back towards its long-run equilibrium. This adjustment parameter amounts to −0.52, that is, 52% of deviation from the equilibrium value is eliminated in the following year. If we consider the bootstrapped values, we can conclude that a 10% increase in price will lead to a 6.84% decline in cigarette quantity demanded in the long-run. After checking the stability of price elasticity using an alternative methodology ARDL, we can conclude that the ECM estimated the price elasticity of −0.68 is reliable, as it is close to ARDL estimate of −0.64. Estimated price elasticity in the short-run is –0.46. These results are confirmed in previous relevant scientific research, pointing out that the price elasticity of the demand in LMICs varies from −0.20 to −0.8021 31 33–38
Tobacco tax simulation model: the impact of tax changes on public revenues and consumption
Simulation is based on a realistic assumption that supply function is perfectly elastic and that consequently, the whole tax burden is on the consumers. Even if this assumption is not totally realistic, its conclusions regarding the impact of excise on public revenue and cigarette consumption are the same. In simulations, we use estimated conditional and unconditional price and income elasticities from micro model, due to the more representative sample compared to macro. The simulation of consumption and public revenues change due to the excise taxes increase comprises of two scenarios, made according to Montenegro’s excise taxes timetable. Consumption is calculated multiplying baseline consumption (1 000 000 cigarettes) with the sum of price and income effect. Price effect is derived from the product of the retail price percentage change and price elasticity (conditional −0.62 and unconditional −0.80) while income effect is defined as a product of GDP growth (3.5%) and income elasticity (0.23 and 0.63). Total excise revenue was estimated as a product of total excise tax and changed consumption. Total tax revenue from cigarettes was obtained by summing up Value Added Tax (VAT) and total excise revenue. The scenarios results are given in table 6.
Simulation results of both scenarios confirm the positive effects of excise tax policy changes, having inelastic demand: consumption is decreasing, while public revenues increase. These results compare favourably with estimates in other LMICs.22 26–28 33 44 45
Conclusions and discussion
The main objective of this paper was to provide the first estimates of conditional and unconditional price and income elasticity of cigarettes quantity demanded in Montenegro. Additionally, the goal was to verify the suitability of the methods applied, Deaton and time series, in case of limited data, especially price. Results of the estimation indicate a price and income elasticity of cigarettes ranging from −0.62 to –0.80 and 0.23 and 0.63, respectively, depending on the methodology and sample used. These results are comparable with the international evidence in LMICs. According to the simulation results, an increase in tobacco excise taxes would lead to a decrease in tobacco products consumption. Depending on the applied policy scenario and dynamics of the excise tax increase, the increase of prices would positively impact not only population health but also the fiscal situation. Additionally, the impact could be even greater in the case of faster and more dynamic increases in tobacco excise taxes. In that manner, the revenues from excise taxes increases could be invested in health programme. Moreover, the change in price could reduce the overall adults and youth prevalence and decrease direct healthcare costs and number of patients.
As the increase in the standard of living could potentially neutralise the effects of tobacco tax increases, the importance of other instruments for tobacco consumption reduction has to be more emphasised. Even tough non-price measures such as anti-tobacco campaigns and other defined in WHO FCTC7 are recognised by Montenegrin regulatory framework, they need to be more efficiently implemented. This process, along with stricter tobacco market control, would be feasible through synergy and widened dialogue between various stakeholders. Definitely, the increasing prevalence sets the alarm for immediate changes.
The policy-makers in Montenegro could benefit from the very first scientific results of conditional and unconditional elasticity and simulations on consumption and public revenues as a crucial input for evidence-based policy.
Potential limitations of the study could be found in the unavailability of a longer time series of price and consumption in macro model, which does not contain the data of illicit trade. Considering that industry may take advantage of frontloading, there is a possibility that consumption variable from macro model does not completely correspond to the real legal consumption. However, the macro model limitations are mostly neutralised through micro model, due to larger sample and data which covers illicit market (the surveyed households reported each purchase regardless of the point of sale—legal or illegal). Due to unavailability of data, the effect of switching to cheaper brands when price increases was not taken into account. This aspect is very important, as the brand substitution could potentially, in some extent, decrease the price elasticity coefficient.46 47 The future research needs to address youth cigarettes consumption and the issues of these products availability within this population.
What this paper adds
While much is known about the smoking’s negative impact on health and increasing prevalence in the Western Balkans region, there is no evidence about the impact of excise taxes increase on consumption in Montenegro. This is the first paper which estimates the sensitivity of cigarette quantity demanded to price and income changes in Montenegro. Based on macro and micro (Household Budget Survey) analysis, the research provides relevant and objective results considering the impact of tobacco taxation and control policies on cigarettes consumption.
Policies implying faster and more dynamic increases in tobacco excise taxes could positively impact the overall prevalence of adults and youth, by reducing prevalence rate and decreasing direct healthcare costs and number of patients. Additionally, there would be an evident and strong positive impact on public revenues.
The authors would like to thank Violeta Vulović, Frank J Chaloupka, Hana Ross and Rijo M John to their excellent research assistance. We would also like to thank and express our appreciation to the State institutions and WHO, who were actively participating and supporting the research process.
Contributors AM and MČ collected data, conducted the data analysis and wrote the first draft. TL and MP contributed in developing ideas, data interpretation, revision and approval of the different paper drafts.
Funding This research was funded by the University of Illinois at Chicago (UIC) Institute for Health Policy and Research. UIC is a core partner of the Bloomberg Initiative to Reduce Tobacco Use.
Disclaimer The statements and conclusions expressed in this research are those of the authors and do not necessarily reflect those of the sponsors.
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement Publicly available data were obtained from the Statistical Office of Montenegro,
open access repository, linkhttps://www.monstat.org/cg/publikacije.php?id=100
Some data are obtained from a third party and are not publicly available. Household Budget Survey Data is not publicly available (data obtained from the Statistical Office of Montenegro).
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.