Exploring Public Montessori Education: Equity and Achievement in South Carolina

ABSTRACT

This article examines the expansion of public Montessori education and its implications for student participation and outcomes. The study focuses on the state of South Carolina, which has the largest number of public Montessori programs in the United States. Through a comprehensive analysis of demographic characteristics and standardized test scores, we investigate the participation of different student groups in public Montessori programs and compare the academic achievement of public Montessori students to their peers in traditional public schools. The findings indicate that public Montessori attracts a diverse range of students, but there is an underrepresentation of less-resourced students and students of color in public Montessori programs. Using matching procedures, we find that Montessori students demonstrated higher achievement growth in ELA and math compared to similar traditional public school students. Subgroup analyses find that higher achievement growth for Montessori students is consistent across many student groups. This study provides a comprehensive analysis of public Montessori and highlights the importance of considering curriculum and educational philosophy when evaluating the impact of education policies and programs.

KEYWORDS:

• Education segregation
• equity
• Montessori
• school choice
• student achievement

Acknowledgments

The authors thank Mattie MacDonald and Shekinah Lightner for their research assistance on this paper. The authors acknowledge Gina Riga for her advice on this project. This article was produced as part of a larger evaluation of Montessori: Culclasure, B., Fleming, D. J., Riga, G., & Sprogis, A. (2018). An evaluation of Montessori education in South Carolina’s public schools. The Riley Institute at Furman University. https://riley.furman.edu/sites/default/files/docs/MontessoriOverallResultsFINAL.pdf.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. Fleming’s (2023) analysis suggests that the underenrollment of students of color and students from less-resourced families in Montessori programs could be partly explained by differential knowledge levels about Montessori. For example, Black parents were the most supportive racial group of Montessori education after becoming knowledgeable about it.

2. While the CREDO analyses use a method similar to coarsened exact matching, they do not cite the literature on this approach. See Gaertner and Kirshner (2017).

3. The CREDO evaluations and other studies (Ackerman & Egalite, 2017; Fortson et al., 2012) have gone even further and have limited the comparison pool to students who attended the same feeder schools that treatment students attended before enrolling in their treatment schools. However, Hoxby and Murarka (2008) contend that using students from baseline feeder schools may actually enhance selection bias as those who attended the same feeder schools but choose different schools at the next level are doing so for nonrandom reasons and are fundamentally different types of students. In their within-study comparison, Fortson et al. (2012) find that using the school district, rather than feeder schools, as a limiting characteristic when creating the possible matched pools does not substantively change the estimated causal effect. Therefore, we chose a middle ground and matched within district. We reexamine this decision as a robustness check later in this study.

4. Our analysis uses the baseline test score (t-1) for matching purposes. Unlike RCT designs in which students are assigned to conditions before experiencing Montessori education, our design does not allow for such comparisons. For example, Montessori students may have already experienced many years in Montessori before they are matched to non-Montessori students, which may attenuate any effect of Montessori. We thank a reviewer for highlighting this fact.

5. Critiquing the CREDO method, Hoxby (2009) cautions that this 0.10 bandwidth is too large, especially for students with extreme test scores. Fortson et al. (2012) examine this issue. They compare coarsened exact matching methods with 0.10 and 0.05 bandwidths and find no substantive difference in estimated charter school impacts.

6. While critics argue that this creates an unequal distribution of measurement error (Hoxby, 2009), an independent analysis found that using the CREDO procedure to match one-to-one or up to one-to-seven created estimated effects within 0.003 standard deviations and standard errors within 0.001 (Ackerman & Egalite, 2017).

7. The number of Montessori programs within TPS that are included in these analyses ranges from 22 programs in 2013–14 to 30 in 2015–16.

8. In an effort to relax the matching criteria, additional within-school matches were conducted that increased the test bandwidth to 0.20 standard deviations. Using this sample, the results for ELA, Math, and Writing are substantively similar to the results for the matching with the 0.10 bandwidth. Using the 0.20 standard deviation bandwidth matching sample, we find that the Montessori coefficient is 0.07 for ELA (s.e.: 0.04, p-value: 0.06), 0.08 for Math (s.e.: 0.04, p-value: 0.08), and 0.07 for Writing (s.e.: 0.05, p-value: 0.12). This more liberal approach produced samples in which between 42% − 47% of Montessori students were matched. Full results are available from the authors upon request.

Additional information

Funding

This project was funded by a grant from the Self Family Foundation.