The literatures on health care expenditure have demonstrated that most of the studies
used parametric techniques to estimate the elasticities of income and prices (Matteo,
2003). The parametric technique assumes a functional form, normal distribution and
linear relationships; however, true shape of the functional form is unknown and it
is highly sensitive to the choice of the functional form. The OLS method is highly
sensitive to outlier values of expenditures. Sometimes the researchers have applied
weighted least squares or generalized linear model, to get rid of these problems but
this does not able to solve all problems that arises from nature of health expenditure
data. The log linear specifications are used to minimize these problems in estimations
of demand functions, as well. Again, the researchers use log for dependent variable
(log linear) to avoid the normality problems however, this technique cannot ruled out
but reduces the problems (Wooldridge, 1992).The log linear form, applying natural log
of both dependent and independent, facilitates the estimation of elasticities. However,
from theoretical prospective, both linear and log linear specifications are inconsistent
with budget constraint (Hunt-McCool, et al 1994; Deaton and Muellbauer, 1980). In
this case, sum of all price and income elasticities are equal to one. We can’t get the
information from the estimated elasticities on luxurious goods or necessity goods or
inferior goods.