Preschool for all? Enrollment and maternal labour supply implications of a bilingual preschool policy

ABSTRACT

Previous research supports the effectiveness of preschool in various contexts, yet there is limited evidence whether universal-type preschool policies induce changes in enrollment. While certain states have enacted universal preschool policies, some have also considered bilingual preschool mandates, either as a supplementary or stand-alone policy, requiring schools to open up bilingual classrooms for children from non-English speaking families. The question of whether bilingual preschool policies can induce enrollment and close achievement gaps between English learners and English speakers is particularly important today for urban cities and states with large immigrant populations. In this study, I exploit exogenous variation from the first bilingual prekindergarten mandate in Illinois to estimate the causal effects on preschool enrollment and maternal labour supply of recently immigrated and Hispanic families. Utilizing a difference-in-differences strategy, estimates suggest significant effects on preschool enrollment between 18% and 20% and no effects of increasing maternal labour supply in Illinois. Estimates are robust to various specifications, control groups, and timeframes. I use the analysis to further discuss whether universal preschool policies are designed sufficiently for access and inclusion of various student types, and contribute to our understanding on the effectiveness of using childcare subsidies to increase the welfare of low-income families.

KEYWORDS:

• Maternal labor supply
• universal preschool
• early childhood education
• education policy

JEL CLASSIFICATION:

• I20
• I28
• J22
• J24

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ The authors described rigorous studies as those that included atreatment and control group, and where participants were either randomly assigned or pre-treatment differences were at a minimum controlled for when not randomly assigned..

² The language achievement gap is vast in Illinois. Ninety-seven per cent of fourth-graders with limited English proficiency (LEP) in 200,809 scored in the two lowest levels of the statewide standardized assessment, compared to 58% of White students. For the senior class of 2008–09, 63% of students with limited English proficiency and 28% of migrant students graduated from high school, compared to 93% White/Asians, 77% of Blacks/Hispanics, and even 78% of special education students (Illinois Board of Education 2011).

³ ACS datasets are publically available and can be obtained from IPUMS for research purposes. Here, I used the 1% nationally representative samples from 2008–2013 (See Ruggles et al. 2015).

⁴ In the sample, a large majority of non-English speaking families are from non-White Hispanic families. Because of that, around 95% of the samples are composed of non-English speaking Hispanic households..

⁵ This can be tested by examining how non-White and White Hispanics compare when it comes to the effects of the initial 2007 universal preschool policy. It will be a part of future drafts.

⁶ While officially, 5-year olds are also eligible to be in a preschool program, most are eligible for kindergarten based on birthdate cutoffs. For the most part, 5-year olds in preschool are those who started when they were four. Therefore, the effects for 5-year olds may not be seen in the estimates and may attenuate estimates so I leave them out of the analyses.

⁷ The counterfactual states used in this specification are Indiana, Iowa, Kentucky, Missouri, and Wisconsin.

⁸ The results for the propensity-matched states are not shown in this paper but are available upon request..

⁹ As a test of robustness, I provide estimates using (a) a counterfactual consisting of targeted populations in states that already have universal preschool, and (b) a synthetic control method that produces and weighs a group of states that are similar to Illinois on demographic and economic observables. These estimates are available in the Table A1 and show that estimates are robustly significant regardless of the control group used.

¹⁰ The estimates using ACS data are limited to the 2010–2011 timeframe, but the 2010 data includes a portion of surveys completed during both the prior and after mandate implementation in September 2010, which may cause attenuation of estimates. As a secondary test, I also ran DD estimates by restricting data to just 2009 and 2010 ACS data. Once again, the estimate is similarly in the 8% range.