Alcohol Consumption Decisions in Korea Deanna L. Sharpe University of Missouri Mohamed Abdel-Ghany University of Alabama Hye-Yeon Kim Cheju National University Gong-Soog Hong Utah State University

ABSTRACT: Data from the 1996 Korean Household Panel Study and a double-hurdle model were used to analyze alcohol consumption decisions in Korea. The Korean most likely to consume alcoholic beverages is a young male, and a college graduate. Among those who drink, being older, male, self-employed, or a farm or fishery worker were positively associated, and higher education levels were negatively associated with consumption level. The findings for gender and education were generally similar to those of studies using U.S. or Canadian data on alcohol consumption. The role of household level economic factors, family composition, and residence in alcohol consumption decisions appears to differ across cultures, however. KEY WORDS: alcohol consumption; Korea; double-hurdle model. Deanna L. Sharpe is Assistant Professor in the Department of Consumer and Family Economics, University of Missouri, Columbia, MO 65211; e-mail: SharpeD噝 missouri.edu. Her research interests include consumption economics, influence of lifetime labor market participation and financial decisions on later life economic wellbeing, health care, and economics of aging. Mohamed Abdel-Ghany is Professor in the Consumer Sciences Department and Director of International Affairs, College of Human Environmental Sciences, University of Alabama, Tuscaloosa, AL 35487-0158; e-mail: mabdel-g噝ches.ua.edu. His research interests include consumption economics, income inequality, and quality of life. Hye-Yeon Kim is Associate Professor in the Department of Home Management, Cheju National University, Cheju, Korea; e-mail: morning噝cheju.cheju.ac.kr. Her research interests include consumption economics and health care. Gong-Soog Hong is Professor and Head in the Department of Human Environments, Utah State University, Logan, UT 84322. Her research interests include economic wellbeing of older individuals, health care, and consumption and saving behavior. Journal of Family and Economic Issues, Vol. 22(1), Spring 2001 䉷 2001 Human Sciences Press, Inc.

7

8

Journal of Family and Economic Issues

Introduction Economic analysis of alcohol demand in Korea is important for at least two major reasons. First, while a large body of empirical literature related to alcohol demand exists, virtually all of this work has focused on Western developed nations, specifically, the United States, the United Kingdom, Canada, and Australia. (For reviews of such studies, see Ornstein, 1980; Ornstein & Levy, 1983; Leung & Phelps, 1993.) Factors associated with alcohol demand in Korea have not been analyzed. As a Far Eastern country experiencing recent and rapid economic growth, Korea stands in contrast to the long established Western, developed economies that have been the focus of previous studies of alcohol demand. Second, between 1985 and 1995, real spending on alcoholic beverages by Korean households increased 156% (Bank of Korea, 2000; National Statistical Office of Korea, 1985; National Statistical Office of Korea, 1995). Evidence exists that changing social norms in Korea may be contributing to higher levels of alcohol expenditures and consumption (“Statistics On,” 1996; “The Expenditures For,” 1996; Inje University Center for Alcohol Studies, 1999; “The Rapid Increase,” 1997). It is a well-established fact that excessive alcohol consumption can have negative health and social implications (Sindelar, 1998). Knowledge of factors associated with alcohol demand can be crucial in developing programs to address alcohol abuse and alcohol related problems (Hilton, 1993). Knowledge of such factors can also have public policy implications for influencing the type and amount of alcohol consumption or for generating tax revenue (Ornstein, 1980; Ornstein & Hanssens, 1985). This study uses data from the 1996 Korean Household Panel Study to examine the socioeconomic and demographic factors associated with demand for alcohol in Korea. A double-hurdle model is used to permit separate analysis of the factors associated with the decision to consume alcohol or not and with the decision of how much alcohol to consume. Findings using Korean data are compared and contrasted with findings for similar studies using data from Western developed nations. Policy implications are discussed.

Background Traditional Korean culture encourages drinking among men. When a young Korean man’s father offers him his first drink, it is a rite of passage into adult life and adult drinking customs. After school or

D. L. Sharpe, M. Abdel-Ghany, H.-Y. Kim, and G.-S. Hong

9

work, men get together and drink at special drinking establishments and may move from establishment to establishment, drinking through the night. In these establishments, alcohol consumption is part of social custom and exchange, contributing to a shared group identity, fostering frank conversations, and easing conflict resolution. In this environment supportive of male drinking, it is not surprising that over 80% of Korean men report consuming alcohol on a regular basis (“The Expenditures For,” 1996). In the past, women have been excluded from drinking establishments. Alcohol consumption by women was not considered socially acceptable behavior. However, there is evidence that attitudes and behavior are changing. A public opinion poll conducted in 1998 asked 500 randomly selected adult Korean women about women’s drinking. Nearly half (48.6%) asserted “it is up to the individual” whereas only 14.4% thought “they should never do it” (People and Society Research, 1998). This permissive attitude is clearly reflected in women’s behavior. The Korean Gallup Institute found that 61.2% of Korean women reported drinking alcohol in 1997, up from 26.1% just five years earlier (“The Rapid Increase,” 1997). Drinking is more prevalent among young women. The same source notes that 83.3% of women in their early 20s reported drinking alcohol as compared with 32.9% of women in their 50s. These figures suggest that gender differences in alcohol consumption may be decreasing. A recent survey of 1,685 adult male and female Koreans found that 22.6% drank less than once per month, 49.5% drank once or twice a week, and 27.9% drank three or more times per week. A closer examination of the characteristics of the group that drank three or more times per week revealed most were over age 55 (39.6%), had only an elementary school education (50%), and lived in a rural area (35.1%) (“Drinking Behavior,” 1997). Another study by the Korean Statistics Institute conducted in 1995 found that nearly two-thirds of Koreans over age 20 drank alcohol once a month or more. More men than women reported drinking alcohol, 83% compared to 44.6%, respectively. These figures represented a slight decrease for men but an 11.6% increase for women compared to 1992. Men were more likely to be frequent drinkers; 35.1% reported drinking two to four times per month. Most women who reported drinking said they drank less than once a month (60.3%) (“Statistics On,” 1996; “The Expenditures For,” 1996). To the extent that changing social attitudes and norms lead to increased alcohol consumption, it is likely that alcohol-related problems will also increase. The Center for Alcohol Studies at Inje University estimates there are currently 3.5 million alcoholics in Korea but this

10

Journal of Family and Economic Issues

number could be a serious understatement given cultural reticence to seek help for alcohol-related problems (Inje University Center for Alcohol Studies, 1997). Recently, Korean media has focused attention on the social costs of alcohol consumption including family problems, criminal behavior, and death, injury or destruction of property due to drunk driving (SBS Broadcast, 1997). This attention has been unusual in a culture that has tended to discount and conceal alcoholrelated problems. Korean public records indicate that social costs associated with alcohol consumption are high. Between 1990 and 1997, the percentage of injuries in vehicular accidents due to alcohol rose from 3.3 to 10.5 while the percentage of alcohol-related deaths rose from 3.1 to 8.7 (Korean Road Traffic Safety Authority, 1996; Korean Road Traffic Safety Authority, 1997). A survey of 340 individuals charged with murder revealed that nearly 3 out of 4 were drunk at the time of the murder (Korean Institute of Criminology, 1991). Korea leads other developed nations in number of deaths per 100,000 population due to alcohol-related liver disease. The statistics for Korea are almost five times higher than those for the United Kingdom, almost three times higher than those for the United States, and slightly over two times higher than those for Japan (Korean Institute for Health and Social Affairs, 1993). The Korean Institute for Health and Social Affairs estimates that the direct and indirect economic costs of drinking in the form of such things as decreased labor productivity, injury, property damage, and death amounted to almost 4% of gross domestic production in 1995 (Korean Institute for Health and Social Affairs, 1997). Taken together, these social trends point to several issues. First, as social sanctions against women drinking alcohol relax, overall spending on and consumption of alcohol are likely to increase. Second, increased consumption of alcohol especially by young women is a concern. The health risks of alcohol consumption appear to be greater for women than for men because of the adverse effects alcohol can have during pregnancy (Floyd, Ebrahim, & Boyle, 1999). Third, increased alcohol consumption may exacerbate alcohol-related problems already recognized in Korea. These issues highlight the importance of conducting an economic analysis of alcohol demand in Korea.

Review of Literature Many early economic studies of alcohol demand used aggregate time series data from the United States, Canada, United Kingdom, or

D. L. Sharpe, M. Abdel-Ghany, H.-Y. Kim, and G.-S. Hong

11

Australia to obtain single equation demand estimates. Some of these studies focused on aggregate demand (see, for example, Duffy, 1983; Johnson & Oksanen, 1974; McGuinness, 1980) while others have examined demand for various types of alcoholic beverages such as beer (Hogarty & Elzinga, 1972), distilled spirits (Ornstein & Hanssens, 1985), and imported and domestic alcoholic beverages (Adrian & Ferguson, 1987). Other researchers have used aggregate time series data to estimate demand systems for alcoholic beverages. Some of this work focused on demand for alcohol in general (Thom, 1984; Clements & Selvanathan, 1987; 1991) while some focused on specific types of alcohol, including beer, wine, and spirits (Clements & Johnson, 1983). These studies give insight into the relationship between aggregate expenditure levels for alcoholic beverages and macro-level factors such as the aggregate price of alcohol and aggregate advertising expenditures for alcohol. Aggregate data, however, cannot answer questions about the relationship between household level social, economic, and demographic characteristics and alcohol demand. Consequently, several researchers have utilized household expenditure data from the United States, Canada, United Kingdom, and Australia (Abdel-Ghany & Silver, 1998; Atkinson, Gomulka, & Stern, 1990; Johnson & Oksanen, 1974; Labys, 1976; Yen & Jensen, 1996) and household food consumption data from the United States (Blaylock & Blisard, 1993a, 1993b; Heien & Pompelli, 1989; Yen, 1994, 1995). The advantage of household survey data is that detailed individual and household characteristics are reported. This information is not available in aggregate time series on alcohol demand. A major limitation of household survey data for research on alcohol demand, however, is that the decision to use alcohol is not only influenced by economic factors, but by social and health considerations as well. Many individuals would not purchase or consume alcohol even if their income increased substantially or the price dropped dramatically. Consequently, measures of alcohol expenditures or consumption at the individual or household level may contain a large proportion of zeros. In such cases, ordinary least squares regression is inappropriate as it will lead to biased and inconsistent estimates (Maddala, 1983). Thus, another empirical model is needed. The double-hurdle model has generally proven superior to Tobit models (Tobin, 1958) for modeling alcohol demand given the large proportion of zeros in measures of the dependent variable (Blaylock & Blisard, 1993a, 1993b; Yen & Jensen, 1996). The double-hurdle model features two separate stochastic processes. These two processes allow the decision to participate (i.e., to drink alcohol) to be analyzed sep-

12

Journal of Family and Economic Issues

arately and apart from the decision of how much to consume, thus providing more meaningful insights on consumer behavior than the Tobit model (Cragg, 1971). Research using a double-hurdle model with household level data has relied on U.S. and Canadian data sources. This research includes analysis of wine consumption by American men (Blaylock & Blisard, 1993a), alcohol consumption by American women (Blaylock & Blisard, 1993b; Yen, 1995), and assessments of level and determinants of household alcohol demand in the United States (Yen, 1994; Yen & Jensen, 1996) and in Canada (AbdelGhany & Silver, 1998). Studies that have applied a double-hurdle model to data from the USDA Nationwide Food Consumption Survey (Blaylock & Blisard, 1993a, 1993b; Yen, 1994) have generally found that likelihood of participation is higher for males and females of prime drinking age (25 to 35 and 36 to 49 for males, 18 to 24 for females), those who are white, a homeowner, reside in urban areas in the Western portion of the United States, have relatively high income, smoke, and do not receive food stamps. In addition, research that focused on women found that those who were pregnant or lactating or believed they were in poor health were less likely to consume alcohol (Yen, 1995). Among those who consume alcohol, relatively heavy drinkers were likely to be a male over age 17 or a female between the ages of 25 to 49, and white. A price increase in alcohol, food stamp receipt, rural residence, residence in regions other than the West, and presence of children under age 18 in the home were associated with lower levels of alcohol consumption among those who drank. Analyses using expenditure data have generally yielded results similar to analyses using consumption data. Yen and Jensen (1996) used U. S. Consumer Expenditure Diary Surveys for 1989 and 1990 to examine alcohol demand, asserting that they expected “the expenditure decision to coincide with the consumption decision” (p. 50). They found that the decision to consume was positively associated with having a household member aged 19 to 64, higher income levels, and being male. Being older and residing in the urban northeast as compared with rural residence made alcohol consumption less likely. Among alcohol consumers, relatively heavy drinkers were apt to be between age 19 and 64, although older ages were negatively associated with alcohol consumption. Higher levels of income and education, being white, and urban residence were also characteristic of relatively heavy drinkers. Having children under age 18 in the home and being married were negatively associated with level of alcohol consumption. Inter-

D. L. Sharpe, M. Abdel-Ghany, H.-Y. Kim, and G.-S. Hong

13

estingly, homeownership, as a proxy for asset holdings, was significant and positive in the Tobit model that Yen and Jensen (1996) examined as an alternative but it was not significant in the double-hurdle model. Abdel-Ghany and Silver (1998) examined the economic and demographic factors associated with use of and spending on alcohol in Canadian households. Using the 1990 Survey of Family Expenditures published by Statistics Canada, they found that the probability of spending on alcohol was higher if the reference person was male, reported having higher levels of education and, interestingly, if the reference person’s occupation was a teacher. Among households reporting alcohol expenditures, age was a significant factor in decreasing expenditures while higher income and education levels and being male were associated with increasing expenditure levels. Some regional differences in alcohol expenditures were also noted. Residents in the Atlantic Provinces spent significantly less on alcohol than British Columbia residents did. Prior research suggests that household composition and individual and household characteristics have significant effects on alcohol consumption. In general, the participation decision is negatively associated with age, but positively associated with education, income, and health. Race, ethnicity, and region also have effects, perhaps influencing social norms relative to alcohol consumption among various groups and in various residential locations. Decisions regarding the level of alcohol to consume appear to be positively associated with presence of individuals of prime age for drinking and relatively higher levels of income and education. For the United States and Canada, there is evidence that regional differences exist. Higher prices for alcohol or having children under age 18 in the home were negatively associated with alcohol consumption among those who drank.

Method Data Data for this study are from the 1996 Korean Household Panel Study, the fourth national data collection by the Daewoo Economic Research Institute in Seoul, Korea (Daewoo Economic Research Institute, 1996). The survey, conducted yearly since 1993, was designed to obtain information about the social and economic activities of individuals and households in Korea. The survey sample was selected by multi-stage stratified random sampling techniques. Because these data report both individual and household characteristics, they provide a unique opportunity to examine Korean alcohol drinking behavior.

14

Journal of Family and Economic Issues

Sample The sample for this study consisted of 6,048 individuals. Of these, 3,102 had responded “yes” to the question “Do you drink?” while 2,946 had responded “no.”

Model Consumer demand theory postulates that a household chooses among consumer goods with a goal of maximizing utility subject to a budget constraint. Thus, the household acts to solve the constrained utility maximization problem: Max [u (y, h) ⱍ py ⳱ b] y

[1]

where y is a vector of consumer goods, p is a vector of corresponding prices, h is a vector of household characteristics, and b is the household budget. If the utility function u (y, h) is assumed to be continuous, increasing, and quasiconcave, then the demand for a given consumer good such as alcohol can be expressed as a demand function of the form f (p, h, b). Since price information is usually not available in cross-sectional data and the Korean Household Panel Study is no exception, of necessity, it was assumed that all households face the same relative prices. The consumption equation then becomes g (h, b). Analysis of alcohol demand raises an important consideration. A large proportion of those surveyed may report zero consumption of alcohol during the survey period. Tobit analysis would permit parameter estimation in such instances (Madalla, 1983). Tobit has been rejected, however, as an appropriate model for alcohol consumption because it forces zero observations to represent corner solutions and presumes that the same set of variables and parameter estimates determine both the discrete probability of a nonzero outcome and the level of positive expenditures (Blaylock & Blisard, 1993b; Yen, 1994; Yen & Jensen, 1996). The double-hurdle model specifies a participation equation, X␣ Ⳮ ␮, and a consumption equation, Y␤ Ⳮ ε, such that consumption, C, is modeled as: C ⳱ Y␤ Ⳮ ε if X␣ Ⳮ ␮ ⬎ 0 and Y␤ Ⳮ ε ⬎ 0 ⳱0 otherwise

[2]

where X and Y are vectors of explanatory variables, ␣ and ␤ are vectors of parameters, and ␮ and ε are the error terms (Cragg, 1971).

Variables Respondents in the 1996 Korean Household Panel Study reported whether they did or did not drink alcohol. Thus, the dependent variable of the partici-

D. L. Sharpe, M. Abdel-Ghany, H.-Y. Kim, and G.-S. Hong

15

pation equation is binary: 1 if drink, 0 otherwise. Following standard practice in the literature, this equation was estimated as a Probit model (Blaylock & Blisard, 1993a, 1993b; Cragg, 1971; Yen & Jensen, 1996). While it would be preferred to measure volume of consumption, data limitations necessitated use of the number of drinks consumed per week by the survey respondent as the dependent variable. The consumption equation is estimated using truncated regression analysis. The explanatory variables were selected on the basis of demand theory and precedent in the literature. Continuous variables included yearly household income in won, household assets in won, number of children under age 18 in the home, and respondent’s age. Dummy variables were used to control for marital status (1 if married, 0 otherwise), respondent gender (1 if male, 0 otherwise), and residence (1 if urban, 0 otherwise). Education level of respondent has three categories: middle school graduate (reference category), high school graduate, and college graduate. Occupation is categorized as salaried worker, self-employed worker, farm & fishery worker, housewife (reference category), and student, part-time or temporary worker.

Hypotheses In general, it was hypothesized that the sign and significance of parameter estimates using Korean data will parallel previous research using U.S. and Canadian data and the double-hurdle model. Thus, higher income and assets are expected to increase both probability of participation and level of consumption. Younger respondents, it was hypothesized, are more likely to participate in alcohol consumption. But, because alcohol consumption can be habit-forming, it was expected that the association between age and level of consumption is positive. Having young children in the home or being married is expected to discourage both participation and consumption. Because of cultural norms in Korea, it is hypothesized that males will both be more likely to drink and will drink more than females. The effect of education in the Korean culture is difficult to predict. A priori, a negative relationship between educational level and drinking of alcohol was expected. More educated individuals should be more cognizant of the damaging effects of drinking on health. However, evidence from studies based on U.S. and Canadian data is that higher levels of education are associated with participation (Abdel-Ghany & Silver, 1998; Blaylock & Blisard, 1993b; Yen, 1995) and higher levels of consumption (Yen & Jensen, 1996). Therefore, it was hypothesized that a significant relationship exists between educational level and alcohol consumption decisions in Korea, but the direction of that relationship was not specified. Little precedent exists to make hypotheses regarding the relationship between occupations and alcohol consumption. Previous research using U.S. data only distinguished between white collar and other occupations. In some cases it was not a significant factor (Yen, 1995) while in others, it was a positive factor in the participation decision (Blaylock & Blisard, 1993a). AbdelGhany and Silver (1998) used six occupational categories of Canadian consumers: manager/professional, sales/services, teacher, blue collar, retired or

16

Journal of Family and Economic Issues

not working, other (reference group). They found teachers were more likely than other occupations to drink alcohol. Thus, it seems reasonable to speculate that, in Korea, certain occupations may be associated with a greater likelihood of drinking alcohol, although it would be difficult to identify those occupations, a priori. There does not seem to be a basis for speculating that higher levels of consumption are associated with certain occupations, therefore, no significant relationship between occupation and the level of alcohol consumed was expected. In general, previous studies have found that rural residents were less likely to drink alcohol than urban residents were. It was not always a significant factor in level of consumption, however. Measures of residence have varied. Blaylock and Blisard (1993a, 1993b) compared city to non-city residence. Yen (1994, 1995) and Yen and Jensen (1996) compared rural residence to residence in various urban regions. It was expected that rural residents in Korea would be less likely to drink alcohol than urban residents would, but it was not clear that rural residence would affect level of alcohol consumed.

Findings and Discussion Sample Characteristics Table 1 shows the independent variables used to specify the participation and expenditure equations. For each variable, the sample average or percentage over the full sample, over the subset of 3,102 observations on those who chose to drink, and over the 2,946 observations on those who did not choose to drink is given. In general, average yearly income for households in the drinkers’ subset was more than the average income in the non-drinkers’ households. Conversely, the total assets of nondrinkers were slightly higher than the total assets of drinkers. (In the table, figures for income and assets are also presented in terms of U.S. dollars.) The respondents in the nondrinkers’ subset were older, less educated, and more apt to be married, females, and housewives compared with their counterparts in the drinkers’ subset. Parameter Estimates for Participation and Consumption Equations The double hurdle model was estimated using LIMDEP (Greene, 1995). The parameter estimates of the model are shown in Table 2. In general, the expected relationships were confirmed. Findings of this research suggest that, in Korea, individual and household characteristics are significantly related to both the decision to drink alcohol and the decision about how much alcohol to consume as they have

D. L. Sharpe, M. Abdel-Ghany, H.-Y. Kim, and G.-S. Hong

17

TABLE 1 Means and Proportions of the Independent Variables for Non-drinkers, Drinkers, and Total Sample

Means Household incomea Household assets

a

Respondent age Number of children aged 18 and under in the household Proportions Respondent married Respondent male Education level of respondent middle school graduate high school graduate college graduate Occupation of respondent salaried worker self-employed worker farm & fishery worker housewife student, part-time, or temporary worker Urban residence

Non-drinkers (N ⳱ 2,946)

Drinkers (N ⳱ 3,102)

Total sample

2470.91 ($3,082.95) 10647.70 ($13,285.12) 42.55

2753.68 ($3,435.76) 10543.07 ($13,154.57) 38.43

2593.20 ($3,235.53) 10538.00 ($13,148.25) 40.44

0.50

0.55

0.53

82 24

77 72

79 49

46 38 16

26 45 29

35 42 23

18 11 5 48

38 20 7 15

28 16 6 32

18 48

19 53

18 50

a Korean Won reported in units of 1,000. U.S. dollar equivalent, calculated using the 336 day average exchange rate for 1996, is given in parenthesis (Oanda, 2000).

been in Western developed nations. Some interesting differences exist, however, all of which suggest that cultural norms regarding alcohol consumption may dominate the decision-making process in Korea. Income was a significant and positive factor in both the participation and consumption decision when U.S. expenditure data were used (Yen & Jensen, 1996) and in the participation decision when U.S. consumption data were used (Blaylock & Blisard, 1993a; Yen, 1995). Its lack of significance in the analysis of Korean data suggests that alcohol consumption is not sensitive to a change in income. Thus, if reduction of alcohol consumption becomes a national goal in Korea, all else being equal, imposing a higher tax on alcohol is not likely to have much effect. Asset holdings (proxied as homeownership) were not a significant factor for U.S. alcohol expenditures in Yen and Jensen (1996). However, it was a positive factor in the decision to consume

18

Journal of Family and Economic Issues

TABLE 2 Results of Binomial Probit and Truncated Regression for Alcohol Consumption

Household income Household assets Respondent age Respondent married Respondent male Number of children aged 18 and under in the household Education level of respondent middle school graduate (reference category) high school graduate college graduate Occupation of respondent housewife (reference category) salaried worker self-employed worker Occupation of respondent farm & fishery worker student, part-time, or temporary worker Urban residence Constant ␴a log likelihood N

Probit (Participation decision)

Truncated regression (Consumption decision)

0.90E-05 (0.664E-05) ⳮ0.12E-05 (0.131E-05) ⳮ0.02*** (0.002) 0.02 (0.056) 1.18*** (0.047) 0.01 (0.024)

0.74E-05 (0.782E-05) 0.56E-06 (0.144E-05) 0.01*** (0.002) 0.02 (0.071) 1.23*** (0.067) ⳮ0.03 (0.029)

0.10 (0.050) 0.12* (0.600) 0.30*** (0.057) 0.23*** (0.064) 0.26** (0.089) 0.12* (0.061) 0.05 (0.037) ⳮ0.15 (0.103) — ⳮ3328.51*** 6048

ⳮ0.17** (0.062) ⳮ0.34*** (0.071) 0.16 (0.086) 0.40*** (0.091) 0.50*** (0.114) 0.13 (0.093) ⳮ0.06 (0.044) 0.86*** (0.131) 1.08*** (0.017) ⳮ4314.579*** 3102

a For a technical explanation of the role of ␴ in ML estimation, see Greene (1995). ***p ⬍ .001. **p ⬍ .01. *p ⬍ .05.

D. L. Sharpe, M. Abdel-Ghany, H.-Y. Kim, and G.-S. Hong

19

alcohol in Yen (1995). Household asset holdings were not a significant factor in either the participation or consumption decision in Korea. In contrast to studies using U.S. data, the number of children aged 18 and under, marital status and urban residence were not significant factors in alcohol use and consumption levels in Korea. This finding may imply that alcohol consumption is an activity rather evenly distributed throughout the population, regardless of family composition or residence. Consequently, this research provides no support for segmenting the Korean population on the basis of family composition or residence for either alcohol marketing or providing treatment programs for alcohol-related problems. In contrast to studies using U.S. and Canadian data (Abdel-Ghany & Silver, 1998; Yen & Jensen, 1996; Yen, 1995) the sign and significance of the coefficient on respondent’s age and respondent’s education differs across the participation and consumption equations. Ceteris paribus, the probability of drinking was negatively related to age, however, the level of drinking increased with age for those who chose to drink. The probability of drinking increased for respondents who were college graduates compared with those respondents who were middle school graduates. However, the level of consumption of alcohol decreased as educational level of respondent increased, supporting our initial hypothesis regarding the negative relationship between the amount of drinking and the level of education. These findings suggest that education can play a role in Korean public policy designed to reduce alcohol consumption. Given the cultural norms in Korea, it is not surprising that gender differences in drinking behavior were found. Male respondents were more likely to drink than female respondents were. Among those who chose to drink, male respondents were likely to drink more than female respondents were. These results support our stated hypothesis regarding Korean culture and its restrictive norms on women drinking. Having gender be a significant factor in both the participation and consumption decisions was unique to Korea. Using U.S. data, Yen and Jensen (1996) found that being male was a significant factor in the participation decision, but not the consumption decision. Conversely, using Canadian data, Abdel-Ghany and Silver (1998) found that being male was significant in the consumption decision, but not the participation decision. Interacting gender with various age categories, Yen (1994) noted both participation and consumption in the U.S. were significantly higher for males aged 25–34 and 35–49 and significantly lower for females aged 64 and older. The use of interac-

20

Journal of Family and Economic Issues

tion terms, however, precludes a direct comparison with the results of this study. The probability of the respondent drinking alcohol was significantly higher if their occupation was different from housewife. Since only Korean women would state “housewife” as an occupation, this finding for occupation further underscored gender differences in the choice to drink alcohol or not. It is interesting that, among those who drink, the self-employed and the farm and fishery worker consumed significantly more than the housewife while the salaried worker and student or part-time or temporary worker did not. Further research on the effect of occupation on alcohol consumption decisions might shed some light on reasons for these differences. Comparison of results related to occupation between this study and studies using Western data is difficult to make. Occupational categories may differ between countries. Occupation is both influenced by and, in turn, influences a variety of cultural factors that may be unique to a given location. Also, relatively few researchers using data from Western countries considered the possible influence of occupation on decisions to drink or how much to drink. Of those who did, Yen (1995) did not find it to be a significant factor in analysis of the alcohol consumption decisions of American women. Abdel-Ghany and Silver (1998) noted that teachers in Canada were more likely than other occupations to drink alcohol, however, they did not find occupation had a significant influence on how much one chose to drink. Considered as a whole, the findings of this research suggest that in Korea, as in Western developed nations, it is the young men with higher levels of education who generally choose to drink. Unlike findings for Western nations, however, higher education appears to discourage relatively heavy consumption in Korea among those who choose to drink. Whether the findings for gender reflect a cohort or a lifecycle effect cannot be ascertained from cross-sectional data. It is interesting, however, that despite long-standing liberalized attitudes regarding women drinking in Western cultures, gender difference in consumption and participation persists. Since attitudes toward women drinking in Korea are just now beginning to be challenged, it is likely that gender differences in alcohol consumption decisions will persist for quite some time. Significant cultural change does not usually happen quickly. In Korea, the decision to drink and the decision of how much to consume did not appear to be affected by household economic resources, family composition, or residence. These findings stood in stark contrast to those using data from Western countries where relatively

D. L. Sharpe, M. Abdel-Ghany, H.-Y. Kim, and G.-S. Hong

21

lower household income levels, presence of young children or a spouse, and rural residence generally discouraged either participation or consumption levels (Abdel-Ghany & Silver, 1998; Blaylock & Blisard, 1993a, 1993b; Yen, 1994, 1995; Yen & Jensen, 1996). It may be that Korean culture is more homogeneous than that of the U.S. or Canada. Or, the cultural norms regarding alcohol consumption may dominate the decision making process. Whatever the reason for the differences between Korea and Western developed nations relative to alcohol consumption decisions, it is clear that targeted methods to market alcohol or to reach those with alcohol-related problems cannot simply be exported from the West to the Far East. At the same time, the Korean experience with direct and indirect costs of excessive alcohol consumption does seem to shadow that of the West. It is fair to say that the health and social costs of excessive alcohol consumption transcend cultural differences. Therefore, Korea may need to develop methods of reaching its own at-risk population that differ from methods used in Western cultures. A recent issue regarding alcohol and Korea points to a direction for future research. In 1999, the nation of Korea came under severe criticism from the World Trade Organization (WTO) for the wide gap between tax rates imposed on imported and domestic alcohol. Imported whisky and brandy faced tax rates of 100%, while imported vodka, rum, and gin were taxed at 80%. The liquor tax on domestically produced soju was only 35%. In addition to the liquor tax, Korea levies an “education tax” of 30% on imported and 10% on domestic spirits (European Business Community, 1999). These facts suggest that the Korean government has used alcohol-related taxes both to raise revenues and to encourage purchase of domestic alcohol. It would be of interest to know what proportion of Korean alcohol consumption is related to imported versus domestic spirits. If, as WTO sanctions propose, the tax on imported spirits is reduced, will consumption of alcohol in Korea increase overall or will the proportion of existing consumption simply change? Price data would be necessary to estimate consumer response to a change in tax rate. However, the fact that income was not a significant factor in alcohol consumption decisions in this research provides at least indirect evidence that Korean alcohol consumers are not price sensitive.

Summary Findings of this research suggest that the Korean most likely to purchase alcoholic beverages is a young male with a college degree. A relatively heavy drinker in Korea can be typified as a male who has a

22

Journal of Family and Economic Issues

middle school education and who is self-employed or works in farming or fishery. The results of this research underscore the importance of using a two-step approach to alcohol consumption modeling. Parameter estimates of the double-hurdle model show that the effect of age, education level and occupation of the respondent was different in each decision step. Thus, this approach provided more information than the Tobit model would have provided regarding the unique role of each variable in the participation and consumption decisions. This additional information may be valuable for understanding consumer behavior in the alcoholic beverages market and for developing public awareness programs. Given differences found in this analysis and analyses using data from Western nations, it is clear that Korea will need to develop its own approach to segmenting the alcohol market for marketing or policy.

References Abdel-Ghany, M., & Silver, J. L. (1998). Economic and demographic determinants of Canadian households’ use of and spending on alcohol. Family and Consumer Sciences Research Journal, 27(1), 62–90. Adrian, M., & Ferguson, B. S. (1987). Demand for domestic and imported alcohol in Canada. Applied Economics, 19, 531–540. Atkinson, A. B., Gomulka, J., & Stern, N. H. (1990). Spending on alcohol: Evidence from the family expenditure survey 1970–1983. The Economic Journal, 100, 808– 827. Bank of Korea. (2000). Principal economic indicators: Prices. [Online]. Available: http:// www.bok.or.kr/index e.html [2000, August 8]. Blaylock, J. R., & Blisard, W. N. (1993a). Wine consumption by US men. Applied Economics, 25, 645–651. Blaylock, J. R., & Blisard, W. N. (1993b). Women and the demand for alcohol: Estimating participation and consumption. The Journal of Consumer Affairs, 27, 319–334. Clements, K. W., & Johnson, L. W. (1983). The demand for beer, wine and spirits: A system-wide analysis. Journal of Business, 56, 273–304. Clements, K. W., & Selvanathan, E. A. (1987). Alcohol consumption. In H. Theil & K.W. Clemets (Eds.), Applied demand analysis: Results from system-wide approaches (pp. 185–264). Cambridge, MA: Ballinger. Clements, K. W., & Selvanathan, S. (1991). The economic determinants of alcohol consumption. Australian Journal of Agricultural Economics, 35, 209–231. Cragg, J. G. (1971). Some statistical models for limited dependent variables with applications to the demand for durable goods. Econometrica, 39, 829–844. Daewoo Economic Research Institute. (1996). Korean household panel study. Seoul, Korea: Daewoo Economic Research Institute. Drinking behavior among people over 18 years. Author? (1997, October 14). Drinking behavior among people over 18 years. Hankook Ilbo. [Online]. Available: http:// www.hk.co.kr [1998, August 21].

D. L. Sharpe, M. Abdel-Ghany, H.-Y. Kim, and G.-S. Hong

23

Duffy, M. (1983). The demand for alcoholic drink in the United Kingdom, 1963–77. Applied Economics, 15, 125–140. European Business Community. (1999). The EU wins WTO case against Korea’s alcohol law. [Online]. Available: http://www.jpn.cec.eu.int/english/info-resources/6-6-6-08. html [2000, February 9]. Floyd, R. L., Ebrahim, S. H., & Boyle C. A. (1999). Observations from the CDC. Preventing alcohol-exposed pregnancies among women of childbearing age: The necessity of a preconceptional approach. Journal of Women’s Health & Gender-Based Medicine, 8(6), 733–736. Greene, W. H. (1995). LIMDEP version 7 user’s manual. Bellport, NY: Econometric Software, Inc. Heien, D., & Pompelli, G. (1989). The demand for alcoholic beverages: Economic and demographic effects. Southern Economic Journal, 55, 759–770. Hilton, M. E. (Ed.). (1993). Economics and the prevention of alcohol-related problems. (Research Monograph No. 25). Research Monograph no. 25. Rockville, MD: National Institute on Alcohol Abuse and Alcoholism. Hogarty, T. F., & Elzinga, K. G. (1972). The demand for beer. Review of Economics and Statistics, 54, 195–198. Inje University Center for Alcohol Studies. (1997). No title. [Online]. Available: http:// www.iucas.com [1999, September 8]. Johnson, J. A., & Oksanen, E. H. (1974). Socio-economic determinants of the consumption of alcoholic beverages. Applied Economics, 6, 293–301. Korean Institute for Health and Social Affairs. (1993). A survey on public health and health-related behaviors. Seoul, Korea: Korean Institute for Health and Social Affairs. Korean Institute for Health and Social Affairs. (1997). The social and economic costs of drinking and related policies. Seoul, Korea: Korean Institute for Health and Social Affairs. Korean Institute of Criminology. (1991). The culture and violent crime in Korea. (4th Criminal Policy Seminar). Seoul, Korea: Korean Institute of Criminology. Korean Road Traffic Safety Authority. (1996). The statistics on car accidents. Seoul, Korea: Korean Road Traffic Safety. Korean Road Traffic Safety Authority. (1997). Analysis of large car accidents. Seoul, Korea: Korean Road Traffic Safety. Labys, W. C. (1976). An international comparison of price and income elasticities for wine consumption. Australian Journal of Agricultural Economics, 20, 33–36. Leung, S. F., & Phelps, C. E. (1993). ‘My Kingdom for a Drink . . .?’: A review of the estimates of the price sensitivity of demand for alcoholic beverages. In M. E. Hilton (Ed.), Economics and the prevention of alcohol-related problems. (Research Monograph No. 25.) Rockville, MD: National Institute on Alcohol Abuse and Alcoholism. Maddala, G. S. (1983). Limited dependent and qualitative variables in econometrics. Cambridge: Cambridge University Press. McGuinness, A. J. (1980). An econometric analysis of total demand for alcoholic beverages in the UK, 1956–75. Journal of Industrial Economics, 29, 85–109. National Statistical Office of Korea. (1985). Social statistical survey, 1985. Seoul, Korea: National Statistical Office of Korea. National Statistical Office of Korea. (1995). Social statistical survey, 1995. Seoul, Korea: National Statistical Office of Korea. Oanda, Inc. (2000). Historical currency table. [Online]. Available: http://www.oanda. com/converter/cc table [2000, February 21]. Ornstein, S. I. (1980). Control of alcohol consumption through price increases. Journal of Studies on Alcohol, 41(9), 807–818. Ornstein, S. I., & Hanssens, D. M. (1985). Alcohol control laws and consumption of distilled spirits and beer. Journal of Consumer Research, 12, 200–213.

24

Journal of Family and Economic Issues

Ornstein, S. I., & Levy, D. (1983). Price and income elasticities and the demand for alcohol beverages in M. Galanter (Ed.), Recent developments in alcoholism, vol. 1 (pp. ?). New York: Plenum. People and Society Research. (1998, July 18). The bad effect of smoking and drinking among women. Seoul, Korea: People and Society Research. SBS Broadcast. (1997). The 48 reports on Koreans and their drinking. Seoul, Korea: SeoJiWon Press. Sindelar, J. (1998). Social costs of alcohol. Journal of Drug Issues, 28, 763–780. Statistics on smoking and drinking. (1996, May 23). Chosun Ilbo. [Online]. Available: http://chosun.co.kr [1998, August 21]. The expenditures for cigarettes are decreasing and the expenditures for alcoholic beverages are increasing. (1996, December 19). Donga Ilbo. [Online]. Available: http:// www.dongailbo.co.kr [1998, August 21]. The rapid increase in women’s drinking may be caused by their extended social activities. (1997, February 7). Joongang Ilbo. [Online]. Available: http://www.joongang. co.kr [1999, September 8]. Thom, D. R. (1984). The demand for alcohol in Ireland. The Economic and Social Review, 15, 325–336. Tobin, J. (1958). Estimation of relationships for limited dependent variables. Econometrica, 26, 24–36. Yen, S. T. (1994). Cross-section estimation of U.S. demand for alcoholic beverage. Applied Economics, 26, 381–392. Yen, S. T. (1995). Alternative transformations in a class of limited dependent variables models: Alcohol consumption by U.S. women. Applied Economics Letters, 2, 258– 262. Yen, S. T., & Jensen, H. H. (1996). Determinants of household expenditures on alcohol. The Journal of Consumer Affairs, 30, 48–67.