Preschool and school performance of children from immigrant families

Excellent contributions include Chetty et al. (2011), Dahl and Lockner (2012), Duncan et al. (2011), Heckman et al. (2010a, (2010b), Deming (2009), Heckman (2006), Garces et al. (2002) and Duncan and Brooks-Gunn (1997). Almond and Currie (2011) provide an overview of the literature on children below age five.

See Almond and Currie (2011), Ruhm and Waldfogel (2012) or Heckman and Kautz (2014) for excellent reviews of this rapidly expanding literature.

There are a few studies exploring the related question of how the achievement gap between persons with and without immigrant background is affected by preschool/kindergarten, see next section.

See Havnes and Mogstad (2011b) for a recent and notable exception. Using Norwegian data, they find that expansions of publicly certified child care only marginally increased maternal labor supply. They argue, however, that the expansion crowded out low-quality and informal day care. Another recent Norwegian study (Drange and Telle 2015) considers short-term effects of offering free child care and finds no or negligible effects on maternal employment despite significant effects on child care uptake for their children.

Cohorts born 1990 and before started school at age 7, and graduated at the same age (16). They thus have completed 9 years of schooling upon graduation from compulsory school, while our later cohorts complete 10 years.

All languages except Norwegian, Sami, English, Danish and Swedish.

The decision to providing free child care in Gamle Oslo was announced in October 1997 when the National Budget for 1998 was published (Prop 1, 1997/1998). In Grünerløkka, the final decision came later, with a decision in the city district council early autumn 1998 (see Appendix 2 for details).

The figures in Fig. 1 might understate the effect of the intervention on the increase in preschool attendance; for example, as child care slots are measured 15 December in these data, children who started later (like in January or February) will not be included. Note that the figures do not present shares in short-term care for children of immigrants only. These statistics are unfortunately not available.

Some records contain missing observations for one or more marks. As there was no legal opportunity to fail students in this age group, the missing observations are most likely due to problems with registration. If there are more than five missing observations on an individual, we exclude the record. This similarly applies to the other measures defined below, though below we only require that they have at least one mark to be included. To make sure that excluding missing observations are not endogenous, we have performed a robustness test where we run our main regression (see Sect. 5) with a missing observation of GPA as the outcome. We find no significant differences between treatment and control groups before and after the reform.

The measure of behavior is only available for children born 1992–1995, i.e., we only have one pre-cohort.

A considerable share with missing on educational achievement is not surprising for these parents, since they are born outside Norway (typically in developing countries like Pakistan and Turkey; cf. Sect. 4.1). Moreover, survey data indicate that education of immigrants is not well captured and typically underreported in register data (Blom and Henriksen 2008).

Decade of immigration could also be a proxy for language proficiency. Whether the child is an immigrant should also to some extent be expected to pick up language proficiency among parents (everything else equal, parents with a 5-year-old child born in Norway would presumably have a better proficiency in the new language than parents with a 5-year-old child born outside Norway as they have resided in the country for longer).

To explore this further, we have run regressions (like Eq. (1)) using the dummy for mother finished high school as the dependent variable, see Appendix 3. The results show that without control variables the “effect” estimate is positive and significant. However, the estimate becomes much lower when including the controls for country of origin and decade of immigration, and it becomes statistically insignificant at the 10 % level when we also include the other control variables (except maternal education).We also see from Table 3 that controlling for parental education reduces the effect estimates somewhat.

While misclassification is likely to result in a downward bias, the direction of the self-selection bias is not obvious. The bias could be upwards if higher-skilled families are more able (and willing) to relocate into the treatment area to get the free preschool. The bias could be downwards if the relative value of the free preschool is higher for lower-skilled families, for example since the value of the free preschool constitutes a larger share of their family income and this makes them move more.

For the first post-intervention cohort born 1993, this alleviates concerns of self-selection bias (since the reform was not known 2 years in advance), and the scope for attenuation bias is limited (since few change residency from age four to five). For the later cohorts, especially those born 1996, self-selection remains an issue (since the intervention was known to them before they turned five). However, to eliminate the possibility of self-selection for the 1996-cohort, we would have had to use residency at age one (since this is the most recent time when the intervention was not known to them), which could introduce serious attenuation bias as changing of residency from age one to five is more frequent. While in our main results we have chosen to use residency at age four for all cohorts, we will also show robustness results when using age one for all cohorts. In addition, we have estimated our main results by instrumenting residency in the treatment area at age four (as well as at age five) with the district of residency at age two (and instrumenting the interaction in Eq. (1) with the residency at age two interacted with the post variable). This yielded qualitatively similar estimates as those presented, but precision was lower.

In the 1992 cohort, among all children with immigrant background residing in the whole of Norway at age four (like in our main sample) only 39 children graduated 1 year too late. Twenty-two of these were born in the last quarter of the year.

Also, the intervention occurred around the same time as a program called Plan of Action Oslo Centre/East (PAOCE). Maybe of relevance to our children, PAOCE included subsidies to after-school clubs and an expansion of outdoor activities. But PAOCE started in 1997, so children born in for example 1992 and 1993 should be similarly affected by the program.

Note, however, that we do not include school fixed effects in our main regressions since what school a student attends may be endogenous to the intervention: If the intervention improves school performance of the child or the integration of the parents, it is possible that the family might be better at identifying and getting the child into a higher-quality school. Public schools in Norway are generally of high quality, and there are very few private schools. The public schools have geographically defined catchment areas and getting into a school outside the catchment area is rarely possible. Segregation by socioeconomic status results in the abilities of the students to vary substantially across schools, but this is largely compensated by more public funding directed toward schools with low-performing students (Hægeland et al. 2005).

Throughout the paper, we report results from one regression model with interactions for gender treatment. Our main result becomes almost identical if we run regressions on the sample of boys and girls separately.

When adding all cohorts (last row), there seems to be a rather large positive reform effect for native girls. When looking closer into this, it turns out that this is due to a very good result for native girls in treatment city districts born 1996, i.e., the last cohort to be included in our analysis. We find it unlikely that this is related to the intervention, as the estimates during the first years post-reform are nonsignificant and modest.

In 2003, only 8 % of teachers and assistants working in Norwegian preschools were men (St. meld. 2008–2009).

Some studies indicate that females from immigrant families are kept under stricter control by their parents than males from immigrant families and non-immigrants (Yuval-Davis 1997; Prieur 2004). However, this excess control of females is found to be connected to controlling female sexuality (Prieur 2004), which makes it hard to assess the relevance of these studies for the 5-year-old girls in our sample.

Drange and Telle (2015) do not estimate causal effects on child development (only on uptake), since these recent data are only available for cohorts that were all affected by this recent intervention.

Lov av 5. mai 1995 nr. 19 om barnehager (Act of May 5, 1995, no. 19 about child care institutions).

Rammeplan for barnehagen, 1995 (Curriculum for childcare institutions, 1995).

Informasjon om prosjektet gratis kortidsbarnehage i bydel Gamle Oslo (Plan for the project free part-time preschool in Gamle Oslo city district.) Gamle Oslo city district.

Unless anything else is specified, the content of this subsection is based on information from the qualitative evaluation report “Gratis barnehage for alle femåringer i bydel Gamle Oslo” by Trude Brita Nergård (NOVA report 3/02).

From “Rapport februar 1999”: Prosjekt gratis korttidsplass for alle femåringer i bydel Gamle Oslo.

City district council case 136/98 and 137/98.

Handlingsprogram for Oslo indre øst, årsrapport 1999 (Plan of Action Oslo Centre/East, annual report 1999)

Satsinger på oppvekst i bydel Grünerløkka under Handlingsprogram Oslo indre øst 1997-2005. En intern evalueringsrapport. Internal evaluation report.

City district council meeting November 26, 1998.

Suggestion from the children and youth committee to the city district meeting December 16, 1999.

This subsection is based on an interview with the former unit director for child care centers in Stovner city district, Mary Kristensen.