Modeling the household milk consumption data by endogenous Bayesian Tobit Quantile (BTQ) regression in sidoarjo

摘要

In surveys conducted by Badan Pusat Statistik (BPS), such as SUSENAS, many households do not allocate expenditures for certain types of consumer goods. This causes a lot of censored data. An example of such expenditure is spending on milk consumption. Previous studies that analyzed household expenditure for milk consumption were conducted using the Bayesian Tobit Quantile (BTQ) model. However, the research has not been able to include household income variable that is not linear. This raises the assumption that the variable is endogenous. If the endogenous variable is used in BTQ model, the result of parameter estimation will be biased. Thus, an alternative model that can accommodate the endogenous variables is required. Based on the descriptions, this research used Endogenous Bayesian Tobit Quantile (BTQ) model. The variables used are household expenditures for milk consumption as the censored response variable, household income as the endogenous variable, household head education, percentage of household expenditure for food, numbers of household member, percentage of household member aged ≤ 12 years, average of per capita expenditure, and percentage of working household member as the exogenous variables, and working hours of household head as the instrumantal variable. Furthermore, Endogenous BTQ models for household expenditure data for milk consumption are compared with both Tobit Quantile (TQ) and BTQ models using RMSE and the result is Endogenous BTQ models perform better when endogenous problems arise. In addition, it was found that from the data, significant endogeneity was found on the left side of 0.55-th quantile and on the right side of 0.80-th quantile. It's also found that in the lower quantiles, the percentage of household expenditure on food does not significantly affect household expenditure for milk consumption. Besides, in the upper quantile the percentage of household member aged 收起