The different data sets — the current data set is much larger — lead to a first significant difference. Column 3 uses their categories but the ordered probit model for estimation. This leads to two further consumption categories that becomesignificant: health care and dining out. Column 4 uses my consumption categories but their OLS regression.25 In comparison to column 2, household products are additionally significant. Column 5 shows my baseline results. Clearly, there are stark differences, particularly, there are many more significant consumption categories in column 5. Moving from column 1 to 5 shows that the difference in the results comes from a combination of varying data, categorization ofconsumption goods, and the estimation model.