Ulf-G. Gerdtham, Magnus Johannesson, and Bengt Jönsson
The demand-for-health model by Grossman (1972a, b) has become a cornerstone in the field of health economics. 1 The model has been tested in a number of empirical applications (Grossman, 1972a; Cropper, 1981; Wagstaff, 1986, 1993; van Doorslaer, 1987; Leu and Gerfin, 1992; Erbsland et al., 1995; Nocera and Zweifel, 1998). An important problem in the empirical analyses has been the unobservability of health capital (health status). Measures of health status have been constructed based on various health indicators, e.g. various health problems and symptoms. This has led to problems in interpreting the resulting health measure and the size of the regression coefficients in the estimated demand for health equations.
To overcome these problems Gerdtham et al. (1999) recently carried out an analysis with some direct measures of overall health status from a dataset from Uppsala County in Sweden. Two continuous measures of health status based on the rating scale method and the time trade-off method were used. 2 A categorical measure of overall health status was also used that divided health status into five categories: poor health, fair health, good health, very good health and excellent health. 3 The demand for health was estimated as a function of age, gender, income, education, being single, the distance to the health care provider, unemployment, overweight, smoking, alcohol consumption, and sporting activities. Especially the categorical health measure yielded results consistent with theoretical predictions. In the regression equation with the categorical health measure the effects of all variables were in the expected direction, with the exception of alcohol consumption that had a positive effect on health. The effect of gender and unemployment did not reach statistical significance.
Gerdtham and Johannesson (1999) carried out a further study based on a categorical health measure that divided health status into three categories. That study was based on data from a random sample of the Swedish population, the Level of Living Survey (LNU) from 1991. The demand for health was estimated as a function of age, gender, income, being single, unemployment, overweight, living in big cities, and the initial inherited stock of health. According to the results the demand for health increased with income, education and the initial inherited