• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
1163 demand analysis on tobacco consumption.full

1163 demand analysis on tobacco consumption.full






Total Views
Views on SlideShare
Embed Views



0 Embeds 0

No embeds



Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
Post Comment
Edit your comment

    1163 demand analysis on tobacco consumption.full 1163 demand analysis on tobacco consumption.full Document Transcript

    • Nicotine & Tobacco Research Volume 9, Number 11 (November 2007) 1163–1169Demand analysis of tobacco consumption inMalaysiaHana Ross, Nabilla A. M. Al-SadatReceived 18 May 2006; accepted 12 February 2007 Downloaded from http://ntr.oxfordjournals.org/ at Perpustakaan Universiti Sains Malaysia on October 24, 2011We estimated the price and income elasticity of cigarette demand and the impact of cigarette taxes on cigarettedemand and cigarette tax revenue in Malaysia. The data on cigarette consumption, cigarette prices, and publicpolicies between 1990 and 2004 were subjected to a time-series regression analysis applying the error-correctionmodel. The preferred cigarette demand model specification resulted in long-run and short-run price elasticitiesestimates of 20.57 and 20.08, respectively. Income was positively related to cigarette consumption: A 1% increasein real income increased cigarette consumption by 1.46%. The model predicted that an increase in cigarette excisetax from Malaysian ringgit (RM) 1.60 to RM2.00 per pack would reduce cigarette consumption in Malaysia by3.37%, or by 806,468,873 cigarettes. This reduction would translate to almost 165 fewer tobacco-related lungcancer deaths per year and a 20.8% increase in the government excise tax revenue. We conclude that taxation is aneffective method of reducing cigarette consumption and tobacco-related deaths while increasing revenue for thegovernment of Malaysia.Introduction Tobacco use is currently one of the leading causes of death in Malaysia, accounting for 19% and 11.5%Tobacco use has reached epidemic proportions of deaths among men and women, respectivelyworldwide (Jha, 1999). Although the prevalence of (World Health Organization, 2003). The economicsmoking has decreased in countries with higher per- costs of tobacco use are equally high and consistcapita income over the past two decades, cigarette primarily of the healthcare costs of treating tobacco-use has increased in countries with low- and mid- related diseases (often covered by public funds) andlevel per-capita income (Gajalakshmi, Jha, Ranson, lower labor productivity.& Nguyen, 2000). Malaysia is no exception to this Some government interventions have been showntrend. Smoking prevalence there has increased from to reduce tobacco use (Ranson, Jha, Chaloupka, &21.5% in 1986 to 24.8% in 1996 (Institute of Public Nguyen, 2000), and the Malaysian government hasHealth, 1987, 1997). Smoking is much more pre- taken steps to leverage that fact. In 2004, thevalent among males than females (49.2% vs. 3.5%; government introduced a total ban on all forms ofInstitute of Public Health, 1997). Youth smoking is a tobacco advertising and launched a 5-year multi-particularly acute problem in Malaysia, where as million-dollar smoking prevention media campaign.many as 60% of young males from lower socio- Malaysia also bans smoking in many public areas.economic backgrounds smoke (Ahmad, Jaafar, & However, Malaysia does not yet have a clear tobaccoMusa, 1997). tax policy, which is one of the most effective methods to combat smoking behavior (Chaloupka, Hu, Warner, Jacobs, & Yurekli, 2000). The motivationHana Ross, Ph.D., International Tobacco Surveillance, American for several cigarette tax increases in the past decadeCancer Society, Atlanta, GA; Nabilla A. M. Al-Sadat, M.P.H.,Department of Social and Preventive Medicine, Faculty of Medicine, was primarily to raise government revenue (Table 1).University of Malaya, Malaysia. The 2005 excise tax on locally produced cigarettes, Correspondence: Hana Ross, Ph.D., Epidemiology and Surveillance which constitute over 95% of the market, representsResearch, National Home Office, American Cancer Society, 250Williams St. NW, Atlanta, GA 30303-1002, USA. Tel: +1 (404) 329- only about 25% of the retail price. This rate is far7990; Fax: +1 (404) 327-6450; E-mail: hana.ross@cancer.org below the tax level in some of Malaysia’s neighboringISSN 1462-2203 print/ISSN 1469-994X online # 2007 Society for Research on Nicotine and TobaccoDOI: 10.1080/14622200701648433
    • 1164 DEMAND ANALYSIS OF TOBACCO CONSUMPTION IN MALAYSIATable 1. Import, excise, and sales taxes, 1990–2005. Import tax (non-ASEAN Import tax (ASEAN countries) Excise tax (local cigarettes)Year countries) RM/KG or RM/stick RM/kg or RM/stick RM/kg or RM/stick Sales tax (%)1990 85 85 13 151991 135 135 14 151992–1997 162 162 29 151998–2000 180 180 40 152001–2002 216 216 48 252003 259 108 58 252004 200 100 58 252005 0.20 0.10 0.08 25Note. ASEAN, Association of Southeast Asian Nations; kg, kilogram; RM, Malaysian Ringgit. Tax in 2005 is in RM/stick; tax for all otheryeas is in RM/kg.countries. In Thailand, for example, the cigarette smoking prevalence and smoking intensity while Downloaded from http://ntr.oxfordjournals.org/ at Perpustakaan Universiti Sains Malaysia on October 24, 2011excise tax represents 78% of the retail price. controlling for the population growth, served as the International research has shown that a 10% dependent variable in our demand model. The realincrease in cigarette prices can reduce cigarette tobacco consumer price index (CPI), which representsconsumption by 4%–8% (Jha, 1999). Most countries the cost of all tobacco products in Malaysia adjustedfall into this range, but some countries or regions for inflation, was provided by the Department ofmay exhibit different price sensitivity because of Statistics. It is based on the price of one of the mostcultural or social factors. Nevertheless, only a few popular cigarette brands in Malaysia, Benson &low- and middle-income countries have calculated Hedges (ACNielsen, 2002), which was collectedtheir country-specific estimates of the price respon- monthly by the Department of Statistics in randomlysiveness of the cigarette market. Lack of data or selected shops across the country. We adjusted theresearch capacity is often the reason why this tobacco CPI for inflation using the general CPI. Ourinformation is not available. Having a country- model of cigarette demand controlled for the impactspecific estimate of responsiveness to cigarette tax of income and tobacco control policies on cigarettechanges is useful for planning purposes because the consumption. We measured income by real grossimpact of a tax increase on tax collection can be domestic product (GDP) per capita.predicted with a higher degree of precision. Tobacco control policies other than cigarette taxes This study is the first to estimate the responsive- can be important determinants of cigarette consump-ness of Malaysians to a change in cigarette prices. It tion. We created a set of policy or event variablesdemonstrates how cigarette excise tax policy can be that capture the tobacco control environment inused to curb the tobacco epidemic in Malaysia, Malaysia between 1990 and 2004. Variable ‘‘tlaw1’’predicts the impact of higher cigarette taxes on future takes the value of 1 for 1994–1996 and the value of 2tobacco-related mortality, and estimates the impact for 1997–2004 to reflect the adoption of the Controlof cigarette tax policy on budget revenue. of Tobacco Products Regulation law and its amend- ment in 1997 that expanded smoke-free areas and banned minors’ smoking. Variable ‘‘relig’’ is assignedMethod the value of 1 for 1995–2002 to mark the NationalThe secondary aggregate time-series data for 1990 to Fatwa Council announcement that ‘‘Smoking Is2004 used in this study are summarized in Table 2. Haram (Forbidden),’’ and the value of 2 for 2003–The per-capita consumption of domestic and 2004 to capture the additional impact of the Newimported cigarettes was calculated using the excise Breath Beginning Ramadan Campaign calling fortax and import duties collected by the Malaysian smoking cessation during Ramadan. Variable ‘‘ban-government and the size of the adult population derol’’ takes the value of 1 for 2003, when the(aged 15 years or older). Since the excise tax and government introduced special stickers to curb illegalimport duties were levied per kilogram until 2004, we tobacco products, and the value of 2 for 2004, whendetermined the consumption of both domestic and security marks were placed on cigarette packs toimported cigarettes in kg per year. To convert the improve the control of cigarette smuggling. Variableweight amount to the number of cigarettes, we ‘‘taknak’’ assumes the value of 1 for 2004, when theassumed, as did the Malaysian Department of national media anti-tobacco campaign Tak Nak wasCustoms, that each kilogram of cigarettes is equal launched. Variable ‘‘tcmeas’’ is a dichotomousto 1,100 sticks. Per-capita consumption is obtained indicator for every year in which a new tobaccoby dividing the total consumption (in sticks) by the control policy was adopted or a new tobacco controlsize of the adult population (defined as population event occurred. The rationale for this variable is thataged 15 years or older). This variable, which reflects the impact of a new policy or event lasts only one
    • NICOTINE & TOBACCO RESEARCH 1165Table 2. Cigarette consumption, cigarette prices, and real income in Malaysia, 1990–2004. Consumption (cigarettes/Year person) Real tobacco CPI Real GDP per capita (RM) Tobacco policy index1990 1,476 77.6 8,292 01991 1,679 78.3 8,504 01992 1,034 81.2 8,610 01993 1,554 91.8 8,887 01994 1,456 94.0 9,110 11995 1,549 93.1 9,398 21996 1,579 92.5 9,762 21997 1,607 92.0 9,977 31998 1,179 91.6 8,576 31999 1,393 98.8 8,642 32000 1,360 100.0 9,000 32001 1,175 105.6 9,027 32002 1,278 111.1 9,397 3 Downloaded from http://ntr.oxfordjournals.org/ at Perpustakaan Universiti Sains Malaysia on October 24, 20112003 1,335 112.7 9,895 52004 1,402 124.6 10,588 7Mean (SD) 1,404 (181.6) 96.3 (13.1) 9,178 (651.0) 2.13 (1.60)Note. CPI, consumer price index; RM, Malaysian Ringgit; GDP, gross domestic product.period because of its weak enforcement, and that the selecting the two model versions can be found inimpact is related mostly to publicity and public the Results section.health advocates’ lobbying efforts surrounding pol- We began by evaluating stationarity of our time-icy enactment or a tobacco control event. All events series data. A nonstationary time series can lead toand policies are summarized by a tobacco policy spurious regression, which confuses long-term rela-index (variable ‘‘tcindex’’) defined as the sum of tionships, such as correlation over time, with causaldichotomous indicators ‘‘tlaw1,’’ ‘‘relig,’’ ‘‘ban- relationships. We applied the Dickey–Fuller test forderol,’’ and ‘‘taknak.’’ unit root and found that our measure of consump- To estimate the demand for cigarettes, we used the tion was integrated at zero order I (0), that is, it wasfollowing conventional model in linear functional stationary since the 10% critical value for theform: reported Z(t) test statistic was 22.630. The price and income variables were integrated at first order I Yst ~azb0 Xpt zb1 Xgt zb2 Xtct zeð1Þ (1); they were stationary in their first differences. Since the variables were not integrated at the same WhereYst5aggregate consumption of cigarettes order, we proceeded with the Engle–Granger test forper capita; Xpt5real tobacco CPI; Xgt5real GDP cointegration. This test is based on the stationarity ofper capita; Xtct5tobacco control policy/event. the model’s residuals and detects a possibility of We estimated several versions of this model in a spurious regression. We found that the model’ssearch for our preferred specification. We were residuals were stationary and that cointegrationlimited by the degrees of freedom and thus could existed, given that the 10% critical value for thenot estimate a model controlling for all individual reported Z(t) test statistics was 21.60. This allowedpolicies and events. Therefore, we adopted three us to proceed with the ordinary least squares (OLS)different strategies: First, we estimated a model that model.included only price and income variables to assess Given that our OLS model describes tobacco usethe impact of price on cigarette consumption without in the entire country (macro level), the marketa possible distortion related to the high degree of clearance price could be determined by the interac-correlation between price and other tobacco control tion of both demand and supply sides of the market.policies or events. Second, we augmented the model In that case, price would be determined endogen-by controlling for one policy or event variable at a ously and OLS estimates would be biased. We testedtime. Third, we estimated the model with the tobacco this possibility using Hausman’s test. The m testpolicy index representing the summary measure for statistic for Model II was 23.447, which is below theall tobacco control policies and events. critical value of 6.63. Therefore, we could not reject We subjected our key variables and two selected the null hypothesis of exogenous price. This result ismodel versions to a battery of tests to verify the consistent with the theory of open economy andaccuracy of our specifications (Table 3 and Table 4). perfect competition, whereby cigarette price isModel I included only price and income variables. determined exogenously by costs of production atModel II was similar to Model I but controlled for the world market and by cigarette taxes. Hausman’sthe impact of tobacco control policies or events by test could not be performed on Model I becausethe tobacco policy index. The justification for of the small number of independent variables.
    • 1166 DEMAND ANALYSIS OF TOBACCO CONSUMPTION IN MALAYSIATable 3. Test for nonstationarity. Engle–Granger method. ECM is based on the notion that deviations from long-run equilibrium tend to Consumption Yst Price Xpt Income Xgt partially revert to the equilibrium position in the following period. ECM uses stationary data (in thisAutocorrelation No Test No case, first differences of price and income measures) inconclusiveDickey–Fuller 23.939 0.753 20.918 and includes the lagged residuals (of the long-runtest: Z(t) relationship) as an explanatory variable. CoefficientsDickey–Fuller 22.717 22.836 from ECM represent the relationship in the shorttest firstdifference: Z(t) run, and the coefficient on the lagged residualResults Variable Variable Variable measures the speed of convergence to the long-run integrated at integrated at integrated at equilibrium (as a percentage). zero order I (0) first order I (1) first order I (1) Long-run price elasticity is derived by multiplying the relevant price coefficient estimated in the firstHowever, the exogeneity of price has been confirmed step of the Engle–Granger method by the fitted Downloaded from http://ntr.oxfordjournals.org/ at Perpustakaan Universiti Sains Malaysia on October 24, 2011in Model II and its variations with different policy values of price, and then dividing this expression by the fitted values for quantity. The use of fitted valuesvariables. instead of actual average values is required to obtain Further, we applied the Ramsey regression speci- results based on the long-run equilibrium. Incomefication error test and the omitted variable test. The elasticity is calculated similarly but using the incomeomitted variable test is conducted by regressing coefficient and the income fitted values instead.trend-stationary variables on time and using (sta- Short-run elasticities are calculated using coefficientstionary) residuals from this regression in the model from the short-run ECM equation and the means ofwith a time trend. Both tests indicated that we did variables representing consumption, price, andnot exclude any important variables from our model. income.Such exclusion would result in biased estimates. The Durbin–Watson test assessed the autocorrela-tion of OLS model residuals. If residuals are Resultscorrelated, OLS estimates are unbiased and consis-tent, but they are inefficient. We found the value of Results for different versions of our model arethe reported d -statistic to be closer to the value 2 (no summarized in Table 5. Each model includes price,serial correlation) than to the value 0 (positive serial income and one of the policy or event variables,correlation) or 4 (negative serial correlation). except for Model I. The results in Table 5 show that The Breusch–Pagan/Cook–Weisberg test deter- price has a negative and statistically significantmined that residuals of the OLS model have constant impact on cigarette use in four out of seven modelvariance. Therefore, no heteroscedasticity exists that specifications. The impact of income is quitewould reduce the reliability of our hypothesis testing consistent across different model specifications. Itsand cause OLS estimators to be inefficient. coefficient is statistically significant in six out of Because our model passed the specification tests, seven model specifications. The lack of the signifi-we proceeded with estimating both long-run and cance of the price variable can be explained by theshort-run relationships in the tobacco market using high degree of correlation between the measure ofthe Engle–Granger two-step method (Engle & price and policies or events, given that in most casesGranger, 1987). The first step estimates a long-run the adoption of a new policy also has beenequation without time trend. Given that a cointe- accompanied by a price increase. For example, thegrating relationship exists, we proceeded with an correlation coefficient between price and the tobaccoerror-correction model (ECM), the second step of the Table 5. Linear demand model: Impact of tobacco control policies and events.Table 4. Test results. Policy/ Model I Model II Model: Price Income event Yst5a+b0Xpt+b1Xgt+b3Xtct+e coefficient coefficient coefficientEngle–Granger test for 26.248 25.965cointegration: Z(t) Xtct not included (Model I) 211.05** 0.21** —Hausman test: m — 23.447 Xtct5tlaw1 29.18 0.21** 233.64Ramsey specification error 0.05 0.23 Xtct5relig 29.18 0.23** 258.33test: F Xtct5banderol 210.80* 0.22** 213.29Omitted variable test: time 213.85 (21.45) 240.36 (21.05 ) Xtct5taknak 211.15** 0.21** 14.91coefficient (t value) Xtct5tcmeas 210.75** 0.19 32.86Durbin–Watson test: d 2.98 1.85 Xtct5tcindex (Model II) 28.31 0.22** 228.96Breusch–Pagan/Cook– 0.21 0.01 2 Note. *Statistically significant at 10% level; **statistically sig-Weisberg test: x nificant at 5% level.
    • NICOTINE & TOBACCO RESEARCH 1167policy index is 0.87. Both price and income are impact of price was statistically significant at a 5%statistically significant in the model that does not level in all models except for the long-term relation-include a tobacco control policy. None of the policy ship based on Model II (because of the high degree ofvariables were statistically significant. This finding correlation, as explained earlier). The impact ofcan be explained by the lack of enforcement of the income was statistically significant at a 5% level in allpolicies and the short-lived impact of health promo- models. As expected, the long-run elasticities weretion campaigns. greater than the short-run elasticities, which is typical We selected two model specifications to calculate for an addictive product such as cigarettes. Priceour price and income elasticity estimates. Model I elasticity was larger in Model I because thisincluded only price and income variables, thus specification does not control for the impact ofavoiding the problem of the high degree of correlation policies or events, and we considered this value to bebetween price and a policy or event. Both price and the upper bound of our price elasticity estimate.income coefficients were significant in Model I. The Model II price elasticity was considered the lowerresults based on Model I can be considered an upper bound of our estimate because of the high level of Downloaded from http://ntr.oxfordjournals.org/ at Perpustakaan Universiti Sains Malaysia on October 24, 2011bound of our price elasticity estimate since the impact correlation between price and the tobacco policyof a tobacco control policy or event was not taken into index.account. Model II was similar to Model I, but it The coefficients on the lagged residual of the short-controlled for the impact of tobacco control policies run equations were 20.89 and 20.86 for Model I andor events by including the tobacco policy index, the Model II, respectively. This indicates that, onmost comprehensive measure of these policies and average, about 86% to 89% of the deviation fromevents. Model II may underestimate the impact of long-run equilibrium will be corrected in the follow-price because of the high degree of correlation ing year. This is a large speed of adjustment,between the index and the price variable. Therefore, reflecting the addictive nature of tobacco use.the results based on Model II are considered the lower We used our price elasticity estimate to calculatebound of our price elasticity estimate. This lower the impact of a 25% cigarette tax increase (raising thebound is used for predicting the impact of a tax tax to RM0.1 per stick from its current level ofincrease on budget revenue and on reduced mortality. RM0.08) on cigarette consumption, revenue fromThe impact of income on the demand for tobacco tobacco taxes, and long-term health outcomes. First,seemed to be quite stable across models. To be we estimated the impact of this tax increase on theconsistent, we also used the income elasticity based average cigarette price using the 2005 tax incidence,on Model II in our simulation of future growth in average cigarette price of Benson & Hedges brandtobacco consumption because of GDP growth. (the base for our tobacco CPI), and market share of Table 6 summarizes results of the long-run and domestic and imported cigarettes. If the tobaccoshort-run elasticities based on Model I and Model II. industry passes all of the tax increase on toThe results for long-run elasticities also have been consumers, cigarette prices can be expected tobootstrapped to calculate the confidence interval for increase by about 5.9%. We applied the lower boundthe estimates. The bootstrap method failed in of our price elasticity estimate, 20.57, to predict thecalculating the results for short-run price elasticities impact of a tax increase to compensate for a possiblebecause of an insufficient number of observations upward bias in our estimates. This bias could have(one data point is lost in the short-run equation since occurred because we were unable to control forfirst differences of income and price are used). The cigarette smuggling in the model. We predict that the proposed tax increase will result in a 3.37% reduction in cigarette consumption in the long run. This changeTable 6. Price elasticity estimates. translates to a reduction of about 47 cigarettes per Model I Model II person per year, or 806,468,873 fewer cigarettes consumed in Malaysia per year.Price elasticity Long-run 20.758* 20.571 The reduced consumption of cigarettes would have Long-run bootstrapped 20.745 20.537 many health benefits for the Malaysian population. (¡ 0.059)* (¡0.079) Research shows that for every cigarette per person Short-run 20.083* 20.077*Income elasticity not smoked, lung cancer mortality decreases by Long-run 1.403* 1.464* 0.0248 per 100,000 adults aged 35–69 years within 20 Long-run bootstrapped 1.413 1.495 years (Gajalakshmi et al., 2000). This estimate is (¡ 0.089)* (¡0.124)* Short-run 0.028* 0.025* based on regressing 1990 tobacco-attributable lung Coefficient on lagged 20.891 20.862 cancer mortality per 100,000 adults aged 35–69 years residual (¡0.772)* (¡0.809) * on 1970 cigarette consumption in industrializedNote. *Two-tailed test used to determine 5% level of statistical countries with a history of prolonged smoking.significance. Assuming that the current population growth of
    • 1168 DEMAND ANALYSIS OF TOBACCO CONSUMPTION IN MALAYSIA2.8% continues for the next 20 years, there will be lower bound of our elasticity estimate for predicting14.17 million people in Malaysia in the 35–69 age the impact of a tax increase on cigarette consumptioncategory by 2026. Therefore, a 25% cigarette tax and government revenue.increase in 2006 would prevent about 165 premature Simulation of the impact of a 25% cigarette exciselung cancer deaths per year among that age group by tax increase predicted a 5.9% increase in the average2026. Additional premature deaths would be pre- price of cigarettes and a 3.37% reduction in cigarettevented thanks to reduced mortality from other consumption. This reduced cigarette consumptiontobacco-related diseases. could prevent about 165 premature tobacco-related In addition to reducing the number of premature deaths related to lung cancer per year by 2026 anddeaths, the cigarette tax increase would raise increase government tax revenue by RM434 million,government revenue. With the current cigarette tax or 20.8%.level, population, and income growth, Malaysia can The estimate of the tax revenue increase is close toexpect to collect about RM2,088 million in cigarette the World Bank’s prediction of 17.5% (Jha, 1999),excise tax in 2006. A 25% tax increase would generate based on its global experience, and in accordance Downloaded from http://ntr.oxfordjournals.org/ at Perpustakaan Universiti Sains Malaysia on October 24, 2011RM2,522 million in cigarette tax revenue in 2006, an with a mathematical model of tax revenue and priceincrease of Malaysian ringgit (RM)434 million elasticity (Merriman, 2002) predicting a 20.5%(US$115 million using the exchange rate increase in tax revenue. We conclude that a cigaretteUS$15RM3.77), or 20.8%. tax increase in Malaysia will result in improved public health and increased tax revenue. Ideally these newly obtained resources would be used to helpDiscussion smokers quit, strengthen the enforcement of the current tobacco control laws, and to public health inOur preferred lower bound estimate of price elasticity general. They also could be used to support tobaccoof 20.57 based on macro-level data is comparable farmers in switching to alternative crops.with results from neighboring countries based onmicro-level data, such as Thailand (price elasti-city520.39; Sarntisart, 2003) or Vietnam (price Acknowledgmentselasticity520.53; Eozenou, 2001). According to ourresults, a 1% increase in income in Malaysia will lead The authors gratefully acknowledge funding support from the Rockefeller Foundation and from the ThaiHealth Foundation.to a 1.46% increase in cigarette demand. Again, thisestimate is comparable with those from other middle-income countries (Sarntisart, 2003; Van Walbeek, References2000). It suggests that the income effect in Malaysiais quite strong. Given the real GDP and population ACNielsen. (2002). 2001 Market research. The Star, 7–19. Ahmad, Z., Jaafar, R., & Musa, R. (1997, December 6–7). Cigarettegrowth rates of 5.3% and 1.78%, respectively smoking among Malaysian youth: Problems and prospects. Paper(Central Intelligence Agency, 2006), per-capita cigar- presented at the Proceedings of the Malaysian Society of Health 21stette consumption will increase by 5.12% every year. Scientific Symposium, Kuala Lumpur. Central Intelligence Agency. (2006). The world factbook, Malaysia.This increase will translate to both higher smoking Retrieved September 21, 2006, from www.cia.gov/cia/publications/prevalence and higher smoking intensity. The overall factbook/geos/my.html#Econconsumption of cigarettes in Malaysia will increase Chaloupka, F. J., Hu, T., Warner, K. E., Jacobs, R., & Yurekli, A. (2000). The taxation of tobacco products. In: P. Jha, & F. J.by 7.0% per year. This is good news for the tobacco Chaloupka (Eds.), Tobacco control in developing countries. Oxford:industry but not for public health. There is a danger Oxford University Press.that the tobacco epidemic will spread quite rapidly if Engle, R. F., & Granger, C. W. J. (1987). Co-integration and error correction representation. Econometrica, 55, 251–276.no tobacco control measures are taken. Eozenou, P. (2001). Price elasticity estimations for cigarette demand The present study has some limitations. For in Vietnam. Retrieved from http://pathcanada.org/vietnam/tobacco/example, we could not test whether the price and research/docs/PriceElasticityEstimatesForCigaretteDemandInVietnam- EN.pdfincome elasticities of cigarette demand changed over Gajalakshmi, C., Jha, P., Ranson, K., & Nguyen, S. (2000). Globaltime, as suggested by previous research (Van patterns of smoking and smoking attributable mortality. In: P. Jha,Walbeek, 2000). Our time-series data cover only a & F. Chaloupka (Eds.), Tobacco control in developing countries. Oxford: Oxford University Press.short time period, which is not suitable for that type Institute of Public Health. (1987). National Health and Morbidityof analysis. In addition, our model estimated the Survey 1986. Kuala Lumpur: Author, Ministry of Health Malaysia.impact of price on taxable cigarette sales and thus did Institute of Public Health. (1997). National Health and Morbidity Survey 1996. Kuala Lumpur: Author, Ministry of Health Malaysia.not control for illegal cigarette sales. Therefore, we Jha, P. (1999). Curbing the epidemic: Governments and the economics ofmay have overestimated the impact of price on tobacco control. Washington, DC: World Bank.cigarette demand because some of the measured Merriman, D. (2002). Methods for studying tobacco smuggling withreduction in consumption may be attributed to applications to Southeast Asia. Presented at the Southeast Asia Tobacco Control Workshop, Kanchanaburi, Thailand.substitution with smuggled cigarettes. This possible Ranson, K., Jha, P., Chaloupka, F., & Nguyen, S. (2000). Theupward bias in our estimates led us to apply the effectiveness and cost-effectiveness of price increases and other
    • NICOTINE & TOBACCO RESEARCH 1169 tobacco control policies. In: P. Jha, & F. Chaloupka (Eds.), Tobacco Van Walbeek, C. (2000). The economics of tobacco control in South control in developing countries. Oxford: Oxford University Press. Africa. (Research Release No.1). Cape Town, South Africa:Sarntisart, I. (2003). An economic analysis of tobacco control in Economics of Tobacco Control Project, University of Cape Thailand. (Economics of Tobacco Control Paper No.15; Health, Town. Nutrition and Population Discussion Paper). Washington, DC: World Health Organization. (2003). WHO mortality database. In, World Bank Human Development Network. Tobacco control country profiles. (2nd ed.). Geneva: Author. Downloaded from http://ntr.oxfordjournals.org/ at Perpustakaan Universiti Sains Malaysia on October 24, 2011