Application estimator robust in unbalanced panel data of Indonesian household consumption expenditure

Outliers are often found in data, including panel data. However, research on the robust method of outliers for panel data is still limited, especially for unbalanced panel data. This study discusses robust estimator for the one-way unbalanced panel data regression model with a fixed-effects approach. The estimation method used is the within transformation method, and the least trimmed squares method, both previously went through median centering of the data. This study uses data on monthly average household consumption expenditure per capita by the province in Indonesia from 2013 to 2020. Based on the analysis results, we found that the best method to determine the robust estimator is the within transformation method with median centering because it has the smallest mean squares error of 5.362 × 109. This best model has a determination coefficient of 80%. We can interpret that an increase of one thousand rupiahs from the monthly average wage/net salary for workers/employees (X1) and average health expenditure per capita a month (X3) will increase monthly average household consumption expenditure per capita by 256 and 8,705 rupiahs, respectively. The addition of one person in the population (X4) will increase household consumption expenditure per capita in a month by 0.077 rupiahs.