Multinomial logistic regression is conducted to define the determinants of the provision of parental support, (its) lack of provision due to the lack of means, or non-provision due to the lack of need on both the 2003 and the 2015 datasets. Further, the provision of certain types of support by the group of support-providing parents is analysed through four logistic regressions performed separately on both the 2003 and 2015 datasets to cover four major types of support (provided either alone or in combination with some other type of support): labour support, temporary accommodation, financial support and the provision of a dwelling. The reason why only the above four types of support categories are examined is that other categories were chosen by a very small amount of respondents and could therefore not be examined through logistic regression. The provision of a dwelling is considered a type of financial support in other analyses, however, as this kind of support is provided by a relatively large share of households in the sample, financial support (excluding the provision of dwellings) and the provision of a dwelling are analysed separately. Altogether two multinomial and eight binomial logistic regressions are conducted.

Independent variables cover factors most likely influencing parental support such as class (EGP occupational classification), type of settlement of residence, tenure, and the number of inhabitants in the dwelling. There are three types of settlements of residence: Budapest (including suburbia), large provincial cities (including cities with county rights, their suburbs and the Balaton conurbation) and the rest of settlements classified as provincial towns and villages. Age of respondents is not recorded in the database, however, the year of establishing the first independent household of the head of household can be used as a proxy of age. The year of the provision of support is included in the model only when examining the determinants of support types. Year of support is applied as a categorical variable to account for the period in which support was provided.

Of the above variables, only the number of inhabitants in the dwelling and the year of establishing independent household are continuous variables, the rest are categorical ones. Values of the two continuous variables were normalised to fall in the range between 0 and 1 in each regression. Full regression output tables are available in the Appendix.