Going Beyond Test Scores: The Gender Gap in Italian Children’s Mathematical Capability

Abstract

This paper investigates the relationship between gender, attitudes, and test scores in mathematics. The study argues that measures of children’s capability in mathematics must include some indicators of attitudes toward the subject. These are particularly important when analyzing gender gaps because attitudes toward mathematics differ by gender. To this end, the study first analyzes the gender gap in attitudes and test scores separately using school fixed effects models. Second, it estimates a structural equation model, which takes into account that mathematical capability is a latent construct for which some indicators (test scores and attitudes) are observed. Using data from the Italian National Institute for the Evaluation of Education Systems (INVALSI) for school years 5 and 10 in 2014 and 2015, results confirm that when mathematics capability, including both attitudes and test scores, is measured, the gap between boys and girls changes, and it is therefore relevant to consider both concepts.

HIGHLIGHTS

• Italy has one of the highest gender gaps in mathematics in the OECD.

• Gender gaps are substantial both in children's attitudes and their test scores.

• Tackling gender stereotypes may improve women's self-confidence in mathematics and the gender gap in scores.

• This may also help close the gender gap in STEM occupations.

KEYWORDS:

• Math gender gap
• attitudes
• structural equation models
• school achievement

JEL Codes:

• J16
• C31
• I21

Supplemental data

Supplemental data for this article can be accessed at https://doi.org/10.1080/13545701.2021.1908574.

Notes

1 An extensive review of the literature can be found in Judy Larsen (2013) and in Pietro Di Martino and Rosetta Zan (2011).

2 The 2013 survey also includes some information about attitudes, but it uses different questions with respect to 2014 and 2015 and therefore it has not been considered in this analysis.

3 We have also performed a preliminary descriptive analysis of girls’ attitudes toward Italian, and we found that girls report higher preferences than boys for this subject.

4 In Principal Component Analysis, we reverse values for questions “Mathematics is boring” and “Mathematics is harder for me than for most of my classmates,” so that higher values indicate higher preferences for mathematics.

5 In order to test the stability of our main findings, we construct the index of attitudes in four different ways: Polychoric Principal Component Analysis; Factor Analysis; a standardized indicator ranging 1–6 based on dichotomised attitudes variables; and a factor derived from the estimation of a measurement model using STATA package gsem, which allows specifying that the attitudes indicators should be treated as ordinal variables. We present the results using all these indicators in the Online Appendix A (Table A4).

6 The cumulative density function of the indicator of attitudes toward math by gender is presented in Online Appendix A (Figures A1 and A2).

7 On the operationalization of the capability approach see, for example, Jaya Krishnakumar and Florian Chávez-Juárez (2016).

8 Unfortunately, no information is available in the data on whether the child lives with both parents.

9 We have also estimated an ordered probit given that the items for attitudes vary on a Likert scale 1–4. Results are very similar to the OLS and are available from the authors upon request.

10 It should be noted that the software STATA gives only unstandardized coefficients and completely standardized coefficients. Completely standardized coefficients can be read as the standard deviation change in the dependent variable that follows one standard deviation change in the explanatory variable. The relationship between standardized and completely standardized coefficient is the following: $b_{STD\mathrm{\_}YX}=b_{UNSTD}\ast \left(\frac{SD\left(x\right)}{SD\left(y\right)}\right)$. Nevertheless, as suggested by Linda K. Muthén and Bengt O. Muthén (1998–2010), in the case of binary covariates, standard deviation changes in x are meaningless and coefficients should be standardized only with respect to y: $b_{STD\mathrm{\_}Y}=b_{UNSTD}\ast \left(\frac{1}{SD\left(y\right)}\right)$. It is straightforward to see that $b_{STD\mathrm{\_}Y}=b_{UNSTD}\ast \left(\frac{1}{SD\left(y\right)}\right)=b_{ST{D}_{YX}}\ast \left(\frac{1}{SD\left(x\right)}\right)$. We have therefore computed $b_{STD\mathrm{\_}Y}$ efficients and presented these results in Table 4 and Table B4. It should also be noted that the SEM standardized coefficients are numerically comparable with the results of the OLS.

11 This could be achieved by using a fixed effects model. However, adding school fixed effects in our MIMIC model has not been possible, as the STATA gsem package does not allow for correlations between exogenous variables and the residuals of the dependent variable.

12 Dipartimento delle pari opportunità della Presidenza del Consiglio dei Ministri. http://www.pariopportunita.gov.it/cultura-scientifica-e-stereotipi-di-genere/.

Additional information

Notes on contributors

Maria Laura Di Tommaso

Maria Laura Di Tommaso is Full Professor of Economics at the University of Turin in Italy and Affiliate to Collegio Carlo Alberto. Previous positions include College Lecturer in Economics and Fellow at Robinson College, University of Cambridge, and Research Associate in the Department of Applied Economics, University of Cambridge (UK). She has published extensively in the fields of gender economics, labor economics, and feminist economics.

Anna Maccagnan

Anna Maccagnan has a PhD in labor economics from the University of Modena and Reggio Emilia, Italy. She has carried out this work while working as a Research Fellow at the University of Exeter Medical School in the UK. Anna has previously held research posts at the University of Modena and Reggio Emilia and the University of Turin, working in the fields of happiness and well-being, measurement of capabilities, economics of education, household economics and labor economics.

Silvia Mendolia

Silvia Mendolia is Senior Lecturer in Economics at the University of Wollongong in Australia. She holds a PhD in Economics from the University of New South Wales (UNSW) and a Master of Science (Distinction) in Economics from the University College of London (UCL). Before joining the University of Wollongong in 2012, Silvia worked as a Lecturer in Economics at the University of Aberdeen and as a Research Associate at the Social Policy Research Centre (UNSW). She joined IZA – Institute for the Study of Labor as Research Fellow in August 2014. Silvia has published extensively in the fields of health economics, economics of education, and applied microeconometrics.