Ordered and tobit models for different expenditure groups comparison between regions

In most of the applied studies, data of continuos dependent variable and explanatory variables cannot be observed. However sometimes, it is possible to observe the continuos data as sub-set of population. Social scientists often meet censored, truncated or limited dependent variables. Censored or truncated regression models are known as Tobit model. Tobit models refer to regression model with dependent variables which have lower and upper limits. In addition to this, in the cases where dependent variable is categorical and ordinal, ordered logit probability estimates can be used. The ordered logit models, which are generally labeled as ordered regression models, are seen mostly in economic and financial research areas. In this study, firstly logarithmic models for seven geographic regions were obtained by using Engel functions and the results were interpreted. Then, elasticities were tried to obtain by applying OLS to seven regions. The aim of this study is to attract the attention to the OLS usage problems after continuous or ordinal dependent variables with consumption expenditures and food-sub items for regions using the income, sex, age, education level, marital status and information about the jobs of household heads benefiting from the consumption data of the Turkish households in 1994, and to estimate the parameters with Tobit model. For Tobit model, limit zero "0" was given to households who do not spend. Thus, estimation was made by censoring the models. Then, the household consumptions were divided into 20 % groups and continuous variables were converted to categorical variable and ordered logit model was used. The aim of using ordered logit model is to estimate the probability of expenses of individuals who have certain characteristics according to several scenarios. VI