3.4. Economic analysis of tobacco demand and price and income elasticity of cigarettes
(by International Centre For Policy Studies)
3.4.1. Demand equation
We made an attempt to estimate the demand function based on micro-level data. The data comes from the survey of individuals (both smokers and non-smokers) conducted in all regions of Ukraine. Over 2700 individuals took part in the survey.
In performing the research we aimed at estimating individual price elasticity of cigarette demand. We tried to control for many factors that may potentially affect the amount of cigarettes smoked. These include age, sex, price, household income, strength of addiction, etc. In particular, we divided smokers into different income groups and estimated the effect of price changes on them.
The number of cigarettes smoked per day was used as a dependent variable in our estimates. The survey questionnaire included the question regarding the number of cigarettes smoked per day. However, the data seems not to be precise. The distribution of answers to this question envisioned that smokers tend to round the number of cigarettes consumed. This behavior led to about a half of smokers reporting that they smoke one cigarette pack per day, or half of a pack per day. Consequently, due to such distribution of answers we faced problems in estimating the demand equation based on smokers’ answers about cigarette consumption.
Hence, we derived another method of estimating the volume of smoking. The survey questionnaire included questions regarding monthly cigarette expenditures by smokers and cigarette prices. Assuming that smokers are more accurate in keeping track of their expenditures, we divided cigarette expenditures by cigarette prices to obtain the number of cigarettes consumed. Such calculations gave us the dependent variable, i.e. individual cigarette demand.
In the demand equation, we included the following factors that potentially affect cigarette demand:
 | Price of cigarettes |
| Sex (standing for the hypothesis that on average women smoke less) |
| Age (standing for the hypothesis that price elasticity may differ among age groups) |
| Income groups (standing for the hypothesis that price elasticity may differ across income groups) |
| Strength of addiction (standing for the hypothesis that on average heavily addicted smokers smoke more) |
| Household income (to estimate the income elasticity of cigarette demand) |
| Region dummies (standing for the hypothesis that smoking patterns may differ across regions of Ukraine) |
| A dummy reflecting whether a smoker has under-age children (standing for the hypothesis that parents may smoke less if they have under-age children) |
The estimation was performed using the OLS method. We now turn to analyzing the results of our estimation.
3.4.2. Estimation results
Price elasticity of demand
We estimated the price elasticity for different income and age groups. We separated three income groups (high, middle, and low income) and three age groups (14-17, 18-28, and 29+ years). For the base group we chose individuals aged 18-28 years having high levels of income. The group was chosen arbitrarily.
From our estimation, we can conclude that for the base group the price elasticity of demand equals –0.24, meaning that a ten-percent increase in cigarette price leads to approximately 2.4-percent decrease in cigarette consumption.
Our estimates showed that price elasticity of demand indeed differs across income and age groups. We summarize the results in the table 3.4.1. The numbers in the table should be interpreted in the same manner: that is by what percentage the demand will fall given that (1) cigarette prices rise by one percent and (2) an individual belongs to particular income and age groups.
Table 3.4.1. Price elasticity of demand
Price elasticity |
low income |
middle income |
High income |
14-17 years |
-0.65 |
-0.7 |
-0.52 |
18-28 years |
-0.37 |
-0.42 |
-0.24 |
29+ years |
-0.28 |
-0.33 |
-0.15 |
From the table 3.4.1. we can make two general conclusions:
1. As smokers become older, they are less sensitive to price changes.
2. For high-income individuals belonging to a particular age group, price elasticity is the lowest (i.e. they are least price sensitive). At the same time, low-income individuals have a little bit smaller elasticity as compared to middle-income individuals .
|
Box. Price elasticity for Kiev teenagers (by ADIC-Ukraine) According to the analysis of the Global Youth Tobacco Survey (GYTS) conducted in Kiev in spring 1999 among teenagers of 13-16 years old in one hundred schools (4,156 participants) price elasticity was –0.5 (Hana Ross, personal communication). The analysis brings some evidence that both cigarette prices and tobacco control policies can reduce cigarette demand among teenage students in Ukraine. Price simulation based on the estimates of participation price elasticity indicates that 100 percent increase in cigarette excise tax in Ukraine (from current 0.2 UAH per pack to 0.4 UAH per pack) would lead to 8.8% increase in cigarette prices and consequently to an average 4.4 percentage point decrease in smoking participation among Ukraine teenagers. This would reduce the current smoking prevalence (defined as smoking in the last 30 days before the survey) among youth from 41.1 percent to 36.6 percent. However, the estimate is based on a short-term reaction to a price change. It can be expected that the long-term effects of this price increase would be even larger thanks to the addictive nature of cigarette consumption. |
Income elasticity
As was discussed earlier, smokers’ reaction to price changes is different depending on the level of household income. However, we did not find any evidence of the difference in income elasticities of cigarette demand for different income groups. This fact may be explained by a relatively low level of household incomes in Ukraine. For instance, 50% of individuals from our sample reported household income of less or equal to approximately 350 UAH. Hence, we do not observe differences in income elasticities across income groups. Alternatively, consumer reaction to income changes is similar across our sample.
The overall income elasticity was estimated to be 0.06. This means that a ten-percent increase in the household income causes the cigarette demand to rise by approximately 0.6 percent.
Other factors affecting cigarette demand
Sex. Our estimates support the hypothesis that on average women smoke less than men. Women smoke approximately 15.9% less per day.
Children. One interesting phenomena is that presence of under-age children in a family affects smokers’ behavior. Those smokers who do not have children consume approximately 6.2% more cigarettes.
Region of residence. The highest average cigarette consumption was estimated to be in the Southern region of Ukraine (which is the base region in our estimations). In the West of Ukraine, average smoker consumes approximately 14.5% less cigarettes per day compared to a smoker from the South; in the Center – approximately 9.9% less cigarettes; in the North – 14.8% less cigarettes; in the East – 12.9% less.
Strength of addiction. The survey questionnaire included the following question: How would you respond to a significant increase in cigarette prices? There were four options to choose from: do nothing, switch to a cheaper brand, reduce smoking, and quit smoking. In our opinion, the first two answers may indicate a high degree of addiction. Hence, we tested the hypothesis that heavily addicted smokers smoke more. According to our estimates, those who will do nothing on average smoke 17.8% more compared to individuals who will reduce smoking or quit. At the same time, those who will switch to cheaper brands on average smoke 11.5% more. Hence, we may conclude that heavily addicted smokers are indeed heavy smokers.
3.4.3. Participation Model
Using the previous model, we have described the impact of various factors on the number of cigarettes consumed by a smoker. The purpose of the following model, so called participation equation, is to analyze the impact of various factors on odds of being a smoker, given the survey sample. We try to analyze the characteristics that determine the smoking status (smoke or not smoke) in the survey sample. We assume that odds of being a smoker are correctly represented by the participation rate, i.e. the share of smokers in the sample.
In the participation equation we control for age, sex, education, marital status, observed price, etc. The objective is to analyze the marginal effect of changes in explanatory variables on odds of smoking. Also, we try to determine the broad-sense price elasticity of smoking odds. Moreover, the price marginal effect is analyzed across different income groups.
Binary probit regression has been estimated using Maximum Likelihood method. We have estimated two regressions, with and without differentiation by income group. Specifically, we regress binary indicator of smoking on age, sex, high education dummy, marriage status (either registered or not) dummy, urban/rural dummy, smoking status of a close relative or a friend dummy, income group, and price. For the latter, we have reconstructed the price variable to include non-smokers into consideration. This variable is equal to actual price reported for smokers, price reported by smoking friend or relative for non-smoker respondent. The missing price observations were imputed with an average price. While using such transformations we assume that the decision of non-smokers to participate is based on price he/she observes from smokers-peers, or from the market. In the second version of participation equation the income groups were used to differentiate among the price marginal effects.
The results of regression imply the following key findings:
| Women have some 38% less chances to smoke than men. |
| With each year of age the odds of smoking reduce by 0.3%. Say, “50 and above” age cohort has 2.8% less chances to smoke than other cohorts. |
| High education decreases chances to smoke by 9.6%. |
| Rural residence of an individual decreases odds by 8.8%. |
| The existence of friend-smoker increases chances by 30%, while relative’s smoking adds 17.5% more chances. |
| Being a low-income individual decreases smoking odds by about 8%. |
| Price elasticity of smoking odds is –4.7%, i.e. 1% increase in price decreases smoking chances by 4.7%. |
Finally, considering the effects of smoking by income groups, we infer that the high-income group has the highest price elasticity of smoking odds of –11%. The low-income group has meanwhile the lowest price elasticity, equal to –2.7%. It may be explained by the fact that low-income group has more smoking participation, so their income considerations are justified by entering the low price cigarette group. Indeed, the low-income smokers participation rate is 72%, rather than 31% average participation rate, and among them about two thirds prefer cigarettes of first two low price quartiles (priced below 1.5 UAH).
1 Note that the average price elasticity across all income and age groups equals –0.4. This figure is higher (in absolute value) compared to the elasticity reported in our macro-level research. However, such results are supported by economic theory. Aggregate (macro-level) elasticity of demand is usually found to be smaller than the individual (micro-level) elasticity since in macro-estimates we determine the elasticity of demand for cigarettes as a single good. Given that in our micro-estimates we have cross-section data, the elasticity obtained is likely to refer to the demand for a particular cigarette brand preferred by a consumer. Hence, it is not surprising that the latter elasticity is found to be higher.
2 This may be explained by the fact that middle-income individuals are more “rational” in planning their expenditures, while low-income smokers may be more addicted, and their poor standards of living promote smoking.
| Who we are? |
 |
 |
 |
 |
 |
 |
 |
 |
|
|
