Data Policy Analyst

This post reviews the paper “Is the Smoking Reduction Effect of the Tobacco Consumption Tax Temporary?” (https://www.dbpia.co.kr/Journal/articleDetail?nodeId=NODE11044324). The study looks at the short-term and long-term effects of South Korea’s tobacco tax increase on January 1, 2015, which raised cigarette prices by 80%. While research shows that tobacco taxes reduce smoking rates in the short term, few studies have examined the lasting effects.

Data and Variables

The analysis uses data from the Korea Welfare Panel Study conducted by the Korea Institute for Health and Social Affairs (KIHASA), covering 7,072 households in 446 survey districts nationwide from 2008 to 2019. Key variables include:

Table 1. Variables and Measurements

Type	Variables	Measurement
Dependent	Smoking Amount	average smoking amount for a day
Independent	Policy	introduction of tax: before 2015 =0, 2015 and after=1
	Year	slope if there was no tax: coded as 2008=1, 2009=2 ~ 2019=12
	Policy*Year	Slope change after the tax
Control	Sex	male=1, female=2
	Education	elementary=0, middle=1, high=2, college(2y)=3, 4y college(4y)=4
	Region	Seoul=1, mega city=2, city=3, gun=4, mixed-gun=5
	Income	disposable income/CPI(2015=100)
	Health	good=1, bad=2
	Drinking	Less than once in a month =1, 2~4 times in a month=2, 2~3 times in a week=3, more than 3 times in a week=4, none=5
	Spouse	no =0, yes=1
	Satisfactory in Life	1=very dissatisfied, 2= dissatisfied, 3=neutral, 4=satisfied, 5=very satisfied

Methodology

To estimate the causal effects, the researchers applied a one-way fixed-effects panel model with an interrupted time-series design. This approach allows comparison between two distinct time periods which are before and after the policy intervention while controlling for unobserved individual characteristics. The fixed-effects model was selected over OLS and random-effects models based on both the F-test and Hausman test, which confirmed the presence of significant individual-specific effects.

To estimate the short-term and long-term effects of the tobacco consumption tax, we used one-way fixed effect panel analysis with interrupted time-series design. The nationwide policy such as tobacco consumption tax in South Korea is a classic subject for interrupted time-series design. The purpose of the interrupted time-series design is to evaluate possible differential performance under the two conditions of the time series. In our case, before and after the tax are the two conditions.

The fixed-effect model assumes that Cov(X, α_i) ≠ 0. When this is analyzed with traditional OLS, the result cannot obtain consistent estimates because of endogenous problem. The fixed-effect model controls the fixed effect of individual observation by removing α_i. To estimate the long-term effect of the tax, we input the Year_t and Policy_t*Year_t. The β_y refers to the slope of smoking amount if there was no tobacco consumption tax while the β_py is a measure of change in slope after the tax. Thus, β_y+β_py is the slope after the tax. The equation is as follows:

Results

 The F-test and the Hausman test of the panel analysis were statistically significant at the 1% confidence level and rejected the hypothesis ‘fixed-effects are zero’ and Cov(X, α_i) = 0. When this is analyzed with traditional OLS, the result cannot obtain consistent estimates because of end. Table 2 presents the estimates by income divisions. In all income divisions, β_p which is the short-term effect was statistically significant at 1% confidence level and on average it reduced smoking by 3.39 cigarettes. Also, both the slope without tax (β_y), and slope change after the tax (β_py) were statistically significant, that is, the long-term effect exists. The slope after the tax (β_y + β_py) is -0.12 on average. This means that the amount of smoking decreases by 0.12 annually.

Table 2. Estimates by Income

	All	Low Income	Middle Income	High Income
Yeart (βy)	-0.2992***	-0.3940***	-0.2133***	-0.3005***
Policyt(βp)	-3.3883***	-3.5580***	-3.3570***	-3.2240***
Policyt*Yeart(βpy)	0.1837***	0.2099**	0.1477*	0.1880**
*** p<0.01, **p<0.05, *p<0.1
Slope after tax (βy + βpy)	-0.1154	-0.1841	-0.0657	-0.1124

With the previous results, we can evaluate that the tobacco consumption tax has both short-term and long-term effects; thus, it was an effective policy tool that reduced smoking rates. Meanwhile, if the result is compared with the counterfactual, the evaluation may yield different conclusions. Figure 1 illustrates the slope of the average smoking trend without the tax (red) and with the tax (blue). The blue line indicates a short-term effect in 2015, where the amount decreased by 3.39, followed by a long-term effect of an annual reduction of 0.12. However, the counterfactual (red) shows that the slope of the reduction trend was -0.3 and would have reached a comparable smoking amount by 2032, even without the tax.

Counterfactual Comparison

However, when compared with the counterfactual scenario (the trend projected without the tax increase), the long-term effectiveness becomes less clear. The simulated “no-tax” trend suggests that smoking rates would have declined at a faster rate (–0.30 per year) even without the policy, converging with the actual post-tax trend around 2032. Hence, while the tax successfully triggered an immediate reduction, its relative long-term impact may diminish over time.

Policy Implications and Limitations

The results imply that tobacco taxation is effective in reducing smoking in both the short and medium term, but its impact could weaken in the long run unless complemented by continuous policy interventions. Given that tobacco consumption is highly price-sensitive, indexing the tax rate to inflation could help sustain its deterrent effect. Furthermore, analysis across income groups reveals evidence of tax regressivity. In other words, lower-income smokers are more price-responsive yet spend a larger share of income on tobacco, highlighting the need for non-price measures to mitigate disproportionate burdens on low-income households.