This contribution extends the analysis of Pollet et al. (2009) to ten European countries (Austria, Belgium, Denmark, France, Germany, Greece, Italy, Netherlands, Spain and Sweden). It questions whether lineage can explain the variation in the intensity of grandparental time investments regarding their grandchildren.

The data shown in Table 1 (Model 1) presents a replication of the findings of Pollet et al. (Reference Pollet, Nelissen and Nettle2009) on SHARE data (first wave 2004). It shows a selection of results of a multilevel analysis of the number of hours grandparents spent weekly with grandchildren belonging to one particular child (a grandchild set, see Ghysels, Reference Ghysels2011). It should be noted that the perspective of the analysis is quite different from the analytic perspective of Pollet et al. SHARE reports on responses by grandparents (the serving side), while Pollet et al. reflect the viewpoint of parents with young children (the receiving side). Theoretically, however, these are two sides of the same coin and the lineage hypothesis remains that maternal grandchildren receive more help than paternal grandchildren.

Table 1.

Multilevel estimates of the number of hours grandparents spent weekly with grandchildren belonging to one particular child^a

N=6950.

Statistical significance levels: **>99% and *>95%.

Neg. and Pos. summarize statistically significant effects of control variables (>95%).

a Method: random effects estimates controlling for clustering at the level of grandparents (level 2: 4042 clusters) within countries (level 3: 10 clusters).

b International Standard Classification of Education.

Indeed, in Model 1, a marked difference can be observed between child care assistance given to daughters and sons, even when controlling for a variety of characteristics of the grandparent and his/her child. Daughters (and their children) get almost two and a half hours a week more of grandparental help with child care (55% of the average weekly child care help of grandparents). Similar results using the same data set were reported earlier by Hank & Buber (Reference Hank and Buber2009) and Danielsbacka et al. (Reference Danielsbacka, Tanskanen, Jokela and Rotkirch2011).

Model 2, however, tests an alternative specification. The latter looks into a potential interaction between lineage and the need for child care stemming from paid work by the receiving child. The estimates of Model 2 strongly qualify the previous lineage effect. In interaction with employment, there is no homogeneous lineage effect anymore. In fact, daughters get more assistance only if they are gainfully employed and self-employed sons get more assistance than unemployed or home-making daughters. Perceived need for assistance rather than lineage seems to be the driving factor here. If relative certainty about gene transmission were the explanation, there is no reason why grandparents would not invest more in the grandchildren of a non-working daughter than those of a non-working son.

Rather, the perception of need seems to follow a gendered hierarchy: mothers come first and maternal grandparents fill the gaps. In Western societies, the care of children is largely seen as a female responsibility (Glass, Reference Glass1998; Wheelock & Jones, Reference Wheelock and Jones2002) and the latter is reflected in the data. If mothers' time for care is limited by paid employment, then grandparents step in and provide help. For paternal grandchildren, this only happens if the need is exceptionally high (i.e. in the case of self-employment, an atypically time-consuming activity). Wald coefficient tests indicate that for employed children the parameter estimates of daughters are higher than those of sons. This is not the case for unemployed daughters and sons. Moreover, Model 2 fits the data significantly better than Model 1 (Likelihood Ratio test).

In short, a ‘gendered needs' explanation fits the data better than a lineage explanation.