over time. The first wave focuses on teens and young adults aged 12 to 24. Later on
the upper age limit was successively raised up to 39 in 1990. Since 1995 the target
population solely consists of adults aged 18 to 59. As a consequence, consumers’
family background increasingly became a minor issue and therefore smoking as well
as drinking habits at parental home are not reported in waves more recent than 1992.
The recent waves therefore lack those instrumental variables that are decisive for our
econometric model and, consequently, our analysis has to rely on data collected in
1980, 1986, 1990, and 1992. We do not consider individuals younger than 16 years
for estimating the model. Though numerous people from this age group do report
having consumed alcohol or tobacco this often may reflect experimenting rather than
already settled consumption patterns. After excluding observations with missing data
the sample consists of 26516 individuals.16
4.2 Variables
In our analysis, the quantity of tobacco consumed is measured by the average number
of smoked cigarettes per day. The variable takes the value zero, if the individual
answers to be an ex- or never smoker. Alcohol consumption is defined as grams of
alcohol drunken per day which is calculated from the reported glasses of beer, wine
and spirits per week.17
Numerous consumers do report to be a drinker or smoker but do not report the
amount of alcohol or nicotine consumed. This, for instance, applies to all individuals
that smoke cigars, small cigars or pipes, since only cigarette consumers were asked
about frequency and amount of tobacco consumption. In our sample, quantitative
information regarding consumption is missing for 20 percent of all drinkers and for 17
percent of all smokers. Nevertheless, we do not exclude these observation from our
analysis but let the probability to either drink or smoke enter the overall likelihood
function being maximized.18
16 25695 observations are used for estimating the equations explaining alcohol consumption and
26353 are used for estimating the equations explaining tobacco consumption because of missing
information about either dependent variable.
17We use standard values for beverages’ alcohol content: one glass of beer (0.3l) contains 12 grams
of alcohol, one glass of wine (0.25l) 20 grams, and one glass of spirits (0.02l) 5.6 grams.
18 For the univariate Tobit model this can quite easily be implemented by recoding consumers with
no information about quantitative consumption as non-consumers and multiplying the explanatory
variables by minus one.
12
More intriguing information
1. Mergers and the changing landscape of commercial banking (Part II)2. The Prohibition of the Proposed Springer-ProSiebenSat.1-Merger: How much Economics in German Merger Control?
3. MULTIMODAL SEMIOTICS OF SPIRITUAL EXPERIENCES: REPRESENTING BELIEFS, METAPHORS, AND ACTIONS
4. Nach der Einführung von Arbeitslosengeld II: deutlich mehr Verlierer als Gewinner unter den Hilfeempfängern
5. The name is absent
6. Opciones de política económica en el Perú 2011-2015
7. Consumer Networks and Firm Reputation: A First Experimental Investigation
8. The name is absent
9. Demand Potential for Goat Meat in Southern States: Empirical Evidence from a Multi-State Goat Meat Consumer Survey
10. Commuting in multinodal urban systems: An empirical comparison of three alternative models