Alcohol consumption, smoking and wages

Abstract

The good health of an individual is a combination of uncontrollable factors that includes genetics and random events and controllable factors through the regulation of activities such as smoking and drinking. Since the work of Grossman in the 1970s, a significant relationship between health and earnings has been predicted. In this present paper the 1995 Australian National Health Survey is used to examine simultaneously the effects of drinking and smoking on wages. To model the interaction of smoking with alcohol consumption separate models are fitted for smokers and nonsmokers. These models account for potential selectivity bias resulting from the decision to smoke, and endogeneity arising from a potential causal relationship between earnings and alcohol consumption.

Notes

A standard drink consists of one 12 ounce can of beer, one 5 ounce glass of wine, or a mixed drink containing 1.5 ounces of 80% proof spirits.

Note that the pattern of drinking is also very important. The pattern of drinking that includes episodes of heavy intake (i.e. weekend splurges and weekday abstinence) even if the individual's total weekly alcohol intake is within the limits of moderation can be very harmful (ACSH, 1999).

Among middle-aged and elderly people moderate drinkers have lower mortality rates than do abstainers which has been shown to be largely due to the protective effect of alcohol against coronary heart disease. Among younger people, that is, men who have not reached their 40s and premenopausal women no beneficial effect of moderate drinking on mortality has been demonstrated (ACSH, 1999).

Alcoholism is used to summarize both the ‘alcohol dependence’ and ‘alcohol abuse’ disorders as defined by the American Psychiatric Association (1980, 1987).

Problem drinking is generally defined as heavy drinking and also the diagnosis of an alcohol-related disorder, alcohol abuse and/or dependence.

While it is possible that smoking just a few cigarettes a day might not present a significant health risk to most people there is evidence to suggest that there are relatively few smokers who limit their smoking that much. It is estimated that the vast majority of current smokers smoke more than 15 cigarettes a day (London et al., 1996).

Employees get distracted by thinking about having a cigarette, three or four minutes go on closing up the current work task, ten minutes are spent smoking the cigarette, and seven or eight minutes are spent regaining the momentum.

Drinkers are defined as those who drink liquor, beer or wine once or twice a week or more, and non-drinkers as those who drink less. Smoking is separated into three categories these being non-smokers who have never smoked cigarettes during their lifetime, smokers who consume at least one cigarette on average a day and ex-smokers.

A private dwelling was defined as a house, flat, home unit, caravan, garage, tent and any other structure being used as a private place of residence at the time of the survey. Non-private dwellings included hotels and motels, hostels, boarding houses and caravan parks. Hospitals, nursing homes and convalescent homes were excluded from the scope and/or coverage of the survey, as were prisons, reformatories and single quarters of military establishments. A supplementary sample of the indigenous population was included to enhance the reliability of data.

This excluded diplomatic personnel of overseas governments and non-Australian members of their household; non-Australian service personnel stationed in Australia and their dependents and overseas visitors whose usual place of residence is outside Australia.

Information on family characteristics includes the number and age of children in the household and the number of hours the partner works.

Body mass index (BMI) is calculated as weight divided by the square of height. The National Health and Medical Research Council define an individual whose bmi is <20 as being underweight, bmi 25–30 as being overweight, bmi >30 as being obese and bmi between 20 and 25 as being of acceptable weight.

A frequency count of the days of the week on which the survey was asked revealed that Saturday and Sunday interviews were quite rare.

To determine the influence of the change in the question for 1995 a series of results for the 1989/90 data were simulated based on the equivalent total alcohol consumption that may have been generated if the 1995 method was used in 1989/90. Thus each response for 1989/90 was randomly given a day of the interview based on the distribution of the days of the interview as observed in the 1995 survey (mostly weekdays with a few weekend days). Then we repeated this simulated value 100 times for each household. The ratio of the simulated weekly total and the actual reported total was computed for each interview day and by gender over all the replications of the 19,170 adults in the 1989/90 survey. It was found that the method used in 1995 will under report the consumption by 0.80 for men and 0.73 for women when using the distribution of interview days of the week from the 1995 survey. This average ratio varied from 1.03 for men surveyed on a Monday to 0.45 for women interviewed on a Friday.

Results based on the estimate provided in the survey give similar results and are available on request.

Ex-smokers were included with non-smokers. Since the dummy variable for this group was not significant they were not distinguished from non-smokers.

The variables are similar to those to be defined for the wage equation except for the exclusion of the smoking variables and instead of the inclusion of the married/defacto and divorced/widowed/separated dummy variables a more detailed set of family relationship variables were included.

Note though as in conventional simultaneous equations it is necessary to use the original data, not the predicted values when computing the estimate of the disturbance variance. The estimates of these equations are not of primary interest here and are available on request.