ENH: Remove `eigenbedarf_gedeckt` in Kindergeldübertrag

[MImmesberger]

Kindergeldübertrag uses eigenbedarf_gedeckt to make sure that the child is not in the same Bedarfsgemeinschaft as parents (because then, Kindergeld may be counted as income twice). Conditioning on this exogenous variable is not necessary, we can either

• Make sure that the child is not in the same BG as parents via an endogenous variable
• Calculate income in a way that the double Kindergeld bug doesn't occur (preferred)

[hmgaudecker]

Isn't this "just" a matter of:

• in bg_id_numpy, use a condition derived from something like _diff_kindergeld_kindbedarf_m (I guess, using kindergeld_anspruch and _mean_kindergeld_per_child_m_fg) instead of eigenbedarf_gedeckt

  Btw, the implementation kindergeld_anz_ansprüche does not seem to make much sense at the individual level, right? Will always return the value for one child instead of the average in whatever grouping?

• Then continue with _mean_kindergeld_per_child_m_bg when doing any stuff in SGB-II realm

?

[MImmesberger]

in bg_id_numpy, use a condition derived from something like _diff_kindergeld_kindbedarf_m (I guess, using kindergeld_anspruch and _mean_kindergeld_per_child_m_fg) instead of eigenbedarf_gedeckt

Sounds good! In _diff_kindergeld_kindbedarf_m we would need to use an individual version of arbeitsl_geld_2_regelbedarf_m_bg. Then we could throw out the exogenous eigenbedarf_gedeckt entirely, or am I missing something?

Btw, the implementation kindergeld_anz_ansprüche does not seem to make much sense at the individual level, right? Will always return the value for one child instead of the average in whatever grouping?

Maybe I don't understand your comment correctly, but kindergeld_anz_ansprüche also returns values larger than 1. It is created via

aggregate_by_p_id_kindergeld = {
    "kindergeld_anz_ansprüche": {
        "p_id_to_aggregate_by": "p_id_kindergeld_empf",
        "source_col": "kindergeld_anspruch",
        "aggr": "sum",
    },
}

Then continue with _mean_kindergeld_per_child_m_bg when doing any stuff in SGB-II realm

I don't see why we would want to use the mean Kindergeld on the parental level? I think we're fine with using kindergeld_zur_bedarfsdeckung_m. If Kindergeldübertrag is 0, all is fine. If it is positive, we have to be careful when aggregating to the household level (to not count Kindergeldübertrag and kindergeld_zur_bedarfsdeckung_m twice. But I already added a comment regarding this to the relevant docstring in #753

[hmgaudecker]

in bg_id_numpy, use a condition derived from something like _diff_kindergeld_kindbedarf_m (I guess, using kindergeld_anspruch and _mean_kindergeld_per_child_m_fg) instead of eigenbedarf_gedeckt

Sounds good! In _diff_kindergeld_kindbedarf_m we would need to use an individual version of arbeitsl_geld_2_regelbedarf_m_bg. Then we could throw out the exogenous eigenbedarf_gedeckt entirely, or am I missing something?

Exactly. eigenbedarf_gedeckt is just a Boolean based on _diff_kindergeld_kindbedarf_m < 0 (I might have a sign error here)

[hmgaudecker]

Btw, the implementation kindergeld_anz_ansprüche does not seem to make much sense at the individual level, right? Will always return the value for one child instead of the average in whatever grouping?

Maybe I don't understand your comment correctly, but kindergeld_anz_ansprüche also returns values larger than 1. It is created via

aggregate_by_p_id_kindergeld = {
    "kindergeld_anz_ansprüche": {
        "p_id_to_aggregate_by": "p_id_kindergeld_empf",
        "source_col": "kindergeld_anspruch",
        "aggr": "sum",
    },
}

My bad. I meant _mean_kindergeld_per_child_gestaffelt_m, which uses it as an input.

[hmgaudecker]

Then continue with _mean_kindergeld_per_child_m_bg when doing any stuff in SGB-II realm

I don't see why we would want to use the mean Kindergeld on the parental level? I think we're fine with using kindergeld_zur_bedarfsdeckung_m. If Kindergeldübertrag is 0, all is fine. If it is positive, we have to be careful when aggregating to the household level (to not count Kindergeldübertrag and kindergeld_zur_bedarfsdeckung_m twice. But I already added a comment regarding this to the relevant docstring in #753

Apologies, got confused what is at the child and what is at the parent level. What I meant was "aggregate a child-level variable in its mean version to the Bedarfsgemeinschaft-level"

[MImmesberger]

My bad. I meant _mean_kindergeld_per_child_gestaffelt_m, which uses it as an input.

Ah I see. The purpose of that function is getting the Kindergeld that is available to the children to cover their own needs. While before 2023, Kindergeld was not the same for each child, still each child got the same share of overall Kindergeld to cover its own needs. In other words, we don't want to use a grouped version of _mean_kindergeld_per_child_gestaffelt_m in this context ever.

The correct target for what you're having in mind is kindergeld_zur_bedarfsdeckung_m

[cpz-research]

During the policy project, I used this function to check whether my manual setting of eigenbedarf_gedeckt was correct. Maybe helpful..

def _arbeitsl_geld_2_bedarf_über_einkommen_kind_m(
    arbeitsl_geld_2_regelbedarf_m_bg: float,
    _arbeitsl_geld_2_nettoeink_ohne_transfers_m: float,
    kindergeld_zur_bedarfsdeckung_m: float,
    kind_unterh_erhalt_m: float,
    unterhaltsvors_m: float,
) -> float:
    """ALG 2 needs of a child that exceeds their own income.

    If this value is negative, the child's income is sufficient to cover their needs
    and the child constitutes a Bedarfsgemeinschaft of their own.

    # TODO (@MImmesberger): Use this function to endogenize `eigenbedarf_gedeckt` and,
    hence, the creation of Bedarfsgemeinschaft.
    https://github.com/iza-institute-of-labor-economics/gettsim/issues/758

    Note: arbeitsl_geld_2_regelbedarf_m_bg has to be replaced by
    arbeitsl_geld_2_regelbedarf_m. Until then, the function is only correct if
    `eigenbedarf_gedeckt == True` for the child.
    
    Parameters
    ----------
    arbeitsl_geld_2_regelbedarf_m_bg
        See :func:`arbeitsl_geld_2_regelbedarf_m_bg`.
    _arbeitsl_geld_2_nettoeink_ohne_transfers_m
        See :func:`_arbeitsl_geld_2
    kindergeld_zur_bedarfsdeckung_m
        See :func:`kindergeld_zur_bedarfsdeckung_m`.
    kind_unterh_erhalt_m
        See :func:`kind_unterh_erhalt_m`.
    unterhaltsvors_m
        See :func:`unterhaltsvors_m`.

    Returns
    -------

    """
    out = (
        arbeitsl_geld_2_regelbedarf_m_bg
        - _arbeitsl_geld_2_nettoeink_ohne_transfers_m
        - kindergeld_zur_bedarfsdeckung_m
        - kind_unterh_erhalt_m
        - unterhaltsvors_m
    )

    return out

[hmgaudecker]

Just linking related issue, which also proposes calculating the regelbedarf at the individual level: #668