New Partner, New Order? Multipartnered Fertility and Birth Order Effects on Educational Achievement

A substantial amount of research shows that younger siblings perform worse than their older sisters and brothers in several socioeconomic outcomes, including educational achievement. Most of these studies examined stable families and excluded half-siblings. However, the increasing prevalence of multipartnered fertility implies that many children grow up in nonnuclear families. We examine whether there is evidence for birth order effects in this context, which offers an opportunity to test and potentially expand the explanatory scope of the two main theories on birth order effects. We use comprehensive Norwegian registry data to study siblings in the 1985–1998 cohorts born to mothers or fathers who parented children with at least two partners. We provide evidence for negative effects of birth order on lower secondary school grades in both cases. Children born to fathers displaying multipartnered fertility tend to have lower grades than older full siblings but perform more similarly or better compared with older half-siblings. For siblings born to mothers with the multipartnered fertility pattern, later-born siblings do worse in school compared with all older siblings. This indicates that negative birth order effects tend to operate either within or across sets of full siblings, depending on the sex of the parent displaying multipartnered fertility. We argue that these findings can be explained by a combination of resource dilution/confluence theory and sex differences in residential arrangements following union dissolutions. We also suggest an alternative interpretation: maternal resources could be more important for generating negative birth order effects. Electronic supplementary material The online version of this article (10.1007/s13524-020-00905-4) contains supplementary material, which is available to authorized users.


Table A1
Main results for different combinations of number of children and co-parents from mother-based samples: Estimated difference in lower-secondary school grade average for each higher overall birth order 1-2 1-3 2-1 2-2 (1) (2) Number of observations 9,276 1,173 6,339 1,176 Notes: Results in panel A are based on Model 1, which includes dummy variables for overall birth order; results in panel B are based on Model 2, which includes dummy variables for full biological birth order. All models include controls for fathers' and mothers' age at birth, own cohort, and gender. Standard errors are shown in parentheses. Source: The authors' own calculations based on registry data. *p < .05; **p < .01; ***p < .001 Table A2 Main results for different combinations of number of children and co-parents from father-based samples: Estimated difference in lower-secondary school grade average for each higher overall birth order (panel A) and each higher full biological birth order (panel B) 1-2 1-3 2-1 2-2 3-1 (1) (2) Because there is no distinction between overall and full biological birth order in this sample, we refer to the coefficients as "birth order." The model was estimated with fixed effects for combinations of the two parents' IDs. All models include controls for fathers' and mothers' age at birth, own cohort, and gender. Standard errors are shown in parentheses Source: The authors' own calculations based on registry data. *p < .05; **p < .01; ***p < .001 Section 2: Additional results Table A6 Associations between birth order and school point average, including controls for the educational level of the co-parents. Mother-based sample Notes: The results of both columns are estimated based on model 3.The results presented in the first column includes controls for the educational level of all co-parents, measured in 2014. All models include controls for the father`s and mother`s age at birth, own cohort and gender. Source: The authors` own calculations based on registry data. Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0   Notes: The oldest of second set dummy has the value 0 for the youngest of the first set and 1 for the oldest of the second set. This dummy is interacted with each combination of children and partners to see whether the difference in school points between the youngest of the first and the oldest of the second varies for children between the different combinations. The combination variable includes four combinations ( 1-2, 1-3, 2-1,2-2, 3-1) with over 800 observations, with the rest of the combinations as the reference group. All models include controls for fathers` and mothers age at birth, own cohort and gender. Source: The authors` own calculations based on registry data. Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001 Notes: The oldest of second set dummy has the value 0 for the youngest of the first set and 1 for the oldest of the second set. This dummy is interacted with each combination of children and partners to see whether the difference in school points between the youngest of the first and the oldest of the second varies for children between the different combinations. The combination variable includes four combinations ( 1-2, 1-3, 2-1,2-2, 3-1) with over 800 observations, with the rest of the combinations as the reference group. All models include controls for fathers` and mothers age at birth, own cohort and gender. Source: The authors` own calculations based on registry data. Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001  19508 19508 19508 19508 19508 19508 Notes: The spacing variable measures the years since the birth of the last-born older sibling, and is included as a set of dummy variables with zero (the value for the first born) as the reference category. Controlling for spacing of births generally does not seem to matter much in connection to birth order effects (see e.g. Black et al. 2005;Härkönen 2014). We still chose to include these variables here as conditioning on multipartnered fertility leaves us with a different sample compared to that of previous studies. In this context spacing could be related to differential investment in different children based on their birth order for example or matter in other ways. All models include controls for fathers` and mothers age at birth, own cohort and gender. Source: The authors` own calculations based on registry data. Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001 Notes: The spacing variable measures the years since the birth of the last-born older sibling, and is included as a set of dummy variables with zero (the value for the first born) as the reference category. Controlling for spacing of births generally does not seem to matter much in connection to birth order effects (see e.g. Black et al. 2005;Härkönen 2014). We still chose to include these variables here, as conditioning on multipartnered fertility leaves us with a different sample compared to that of previous studies. In this context spacing could be related to differential investment in different children based on their birth order for example or matter in other ways. All models include controls for fathers` and mothers age at birth, own cohort and gender. Source: The authors` own calculations based on registry data. Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001 12959 Notes: Based on reduced samples that include siblings in the 1987-1998 cohorts. Models 1-3 were estimated. The registries provides us with information on cohabitation for all individuals with common children, and we use this to construct a measure of the number of years (before age 16) that a child has lived with both parents or only the father, included as a continuous variable. All models include controls for fathers` and mothers age at birth, own cohort and gender. Source: The authors` own calculations based on registry data Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001 Notes: Based on reduced samples that include siblings in the 1987-1998 cohorts. The registries provide us with information on cohabitation for all individuals with common children, and we use this to construct a measure of the number of years (before age 16) that a child has lived with both parents or only the father. The variable we use to define the sub-sample in each column is based on the average of this measure, calculated for all children with the same father. All coefficients are reported for each subsample based on model 3, which includes both types of birth order simultaneously. All models include controls for fathers` and mothers age at birth, own cohort and gender. Source: The authors` own calculations based on registry data. Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001 Notes: All estimates based on model specification 3. All models include controls for fathers` and mothers age at birth, own cohort and gender. Source: The authors` own calculations based on registry data. Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001 Notes: All estimates based on model specification 3. All models include controls for fathers` and mothers age at birth, own cohort and gender. Source: The authors` own calculations based on registry data. Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001 Notes: These findings are based on a reduced version of the nuclear family sample that only includes the 1987-1998 cohorts, because of the need for information on cohabitation. In this sample, there is no distinction between the two types of birth order, so we simply refer to our measure as "Birth Order". The models were estimated as fixed effects analyses based on the mother`s and father`s identities. The second model only includes siblings from families where a union dissolution has taken place before the oldest child turned 15 (several thresholds were tested, and the findings were highly similar). In the third model controls for age at union dissolution are included for each separate sibling, with no union dissolution as the reference category. All models include controls for fathers` and mothers age at birth, own cohort and gender. Source: The authors` own calculations based on registry data. Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001

Table A17
Results for models 1-3 with a union dissolution taking place in both sets. Reduced father-based sample.
( Notes: These findings are based on a reduced version of the father-based sample that only includes the 1987-1998 cohorts, because of the need for information on cohabitation. The estimates are based on model specifications 1-3. The analyses include siblings born to mothers who experienced two union dissolutions taking place before the oldest child/children of both sets turned 15 (several thresholds were tested, and the findings were highly similar). All models include controls for fathers` and mothers age at birth, own cohort and gender. Source: The authors` own calculations based on registry data. Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001

Table A18
Results for models 1-3 with a union dissolution taking place in both sets. Reduced mother-based sample.
( Notes: These findings are based on a reduced version of the mother-based sample that only includes the 1987-1998 cohorts, because of the need for information on cohabitation. The estimates are based on model specifications 1-3. The analyses include siblings born to mothers who experienced two union dissolutions taking place before the oldest child/children of both sets turned 15 (several thresholds were tested, and the findings were highly similar). All models include controls for fathers` and mothers age at birth, own cohort and gender. Source: The authors` own calculations based on registry data. Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001 Notes: Each column display coefficients estimated based on model 3, with both types of birth order included. The father`s level of education is measured in 2014. "Lower" refers to grades 1-10, i.e. elementary school and lower secondary school. "Secondary" refers to grades 11-13, i.e. upper secondary school, and "Higher" to a bachelor degree or higher. All models include controls for the father`s and mother`s age at birth, own cohort and gender. Source: The authors` calculations based on registry data. Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001 Notes: Each column display coefficients estimated based on model 3, with both types of birth order included. The mothers` level of education is measured in 2014. Lower refers to grades 1-10. Secondary to grades 11-13 and higher to a bachelor degree or higher. All models include controls for the father`s and mother`s age at birth, own cohort and gender. Source: The authors` calculations based on registry data. Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001