Measuring school income segregation in Spain

University of the Basque Country UPV/EHU Master in Economics: Empirical Applications and Policies 2019/2020 Master Thesis Measuring School Income Segregation in Spain Author: LANDER SAN VICENTE LARRONDO Supervisor: CASILDA LASSO DE LA VEGA MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 2 Index Abstract .............................................................................................................. 3 1. Introduction ..................................................................................................... 4 2. Variance Separation Index and Income Segregation Axioms......................... 8 3. Database ...................................................................................................... 12 4. Results ......................................................................................................... 15 5. Conclusions .................................................................................................. 28 References ....................................................................................................... 29 MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 3 Abstract In this work, our goal is to answer if Spanish students are differently located among schools depending on their standard of living. In other words, we analyse the segregation by income in schools in Spain. To do so, we use data from Programme for International Student Assessment (PISA) databases. We consider all autonomous communities and types of school of Spain for year 2018. Although PISA does not collect data about the income level of students, it creates an index which measures the socioeconomic status of pupils. We use that index as a proxy of the income level for the analysis. This index is standardized by PISA, therefore we use the Variance Separation Index (VSI) to measure the segregation, that is an invariant and absolute index. The results show that there is low income segregation level among the schools of Spain for 2018. We find the largest segregation level by income in schools in Madrid and the lowest in Cantabria. We also analyse the contribution of the type of school -public school, private-government school and private-government schooland we find that the school type may significate a big part of the school income segregation for some regions. Finally, we examine the segregation due to the language type of school in the Basque Country and we find that it represents around 28% of the total segregation. Keywords: Segregation; PISA database; ESCS; Variance Separation Index. Acknowledgement: We want to thank Agurtzane Lekuona for her help and support. Also, we consider convenient to thank the professor Petr Mariel for the suggestion about measuring the income segregation by the language model of schools. MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 4 1. Introduction The Cambridge Dictionary defines segregation as “the act of keeping one person or thing separate from another person or thing”. The goal of this work is to analyse the segregation by income in Spanish schools. This topic is a recurring in papers (Rubia, 2013) and press (Hernández, 2019 for La Vanguardia) in Spain and the reason is obvious; it causes concern in the population for the negative effects it may cause, as the achievement gaps it may cause between advantaged and disadvantaged students, Owens (2018). However, this disruption has existed in the whole world for decades, see Douglas and Denton (1988) for instance. Before delving into the analysis, there are two notions we have to understand about segregation. The first one is that there are different types of segregation, according to the criterion used to classify the groups, Reskin (1993). For instance, ethnic segregation means that demographic groups are classified according to ethnicity. Likewise, when groups are classified by gender, it refers to gender segregation. These examples have something in common, and it is that there is no natural order of groups. In other cases, individuals could be classified according to an ordered criterion, such as the educational level of the parents, depending if they have completed the primary, secondary or higher education. In these cases, it would not be correct to treat groups symmetrically. In the same vein, we could also distinguish segregation by income. Even if this is a very important topic, there is wide disagreement about how to measure the income segregation, because the variable used to classify people is not discrete as the ones aforementioned above. There had been proposed many indices, some of them are based on ethnic segregation indices, like Jahn, Schmid, & Schrag (1947) with the Dissimilarity Index. But some other indices treated income as a cardinal variable, like the Neighbourhood Sorting Index of Jargowsky (1996). Another way to measure income segregation is to stablish a threshold or a MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 5 poverty line to create two income status groups, such as in Fong & Shibuya (2000), or to divide the groups by income percentiles, Watson (2009), but due to income is a dynamic variable, it is hard to control the distribution of the groups. And more importantly, the resulting segregation orders do not satisfy continuity, which is a basic property for any income segregation order. In this work, we use the properties proposed by Lasso de la Vega & Volij (2019), leveraging that they “adapt the properties of standard ethnic segregation measures to the new context and investigate their implications”, instead of adapting an existing ethnic segregation to the context of income segregation. Through that method, we can see how the proposed properties characterize an absolute index of income segregation, which we introduce afterwards. The second notion is the perception about the income segregation and its relation to income inequality. As we work about schools, lets explain this approach using schools as example: Suppose two districts X and Y. Each of the districts have two schools A and B as follows. In both districts poor people attend school A and rich pupils attend school B, so it could be thought that in both situations the segregation by income is equal and maximum, according to the Scale Interpretability proposed by Reardon (2011). Nevertheless, the idea about income segregation we are working at is different. We not only measure the scale interpretability but also the income inequality between students. Therefore, for us in district Y the income segregation is higher as the income difference in schools is much higher. We will consolidate this idea with the axioms we will explain later. Let’s see another example about the difference of the pure segregation and the income segregation we propose along this work: District X 10€ 100€ A 100 0 B 0 100 District Y 10€ 1.000.000€ A 100 0 B 0 100 MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 6 Once again, some people could consider that there is pure and equal segregation in the schools of the two districts, as the rich and poor people are separated, Reardon (2011). However, we consider that segregation in district Y is larger, since the income is more distributed, in the same way as Denton & Massey (1988). These two examples show the main difference between the segregation by income and the ethnic segregation, and detail how our index works. For our analysis, we use data from PISA. PISA is a worldwide study by the Organisation for Economic Co-operation and Development (OECD), which includes the evaluation of educational systems by measuring 15-year-old school pupils’ performance on science, mathematics and reading of both member and non-member nations, Schleicher (2019). It was first performed in 2000 and then repeated every three years. Although there is no data about the income level of pupils’ families, this programme creates the indicator of Economic, Social and Cultural Status (ESCS) index 1 , that provides a comprehensive measure of student socioeconomic background according to the OECD, Rutkowski & Rutkowski (2013). In that way, they allocate a value to each student, that we use as a proxy of the income level. This value is standardized, in order to get a mean of 0 and standard deviation of 1 for the OECD countries, for that reason we use an absolute and invariant index 2 . The goal of our work is to analyse school income segregation in Spain and its regions as measured by the VSI and using the ESCS index as a proxy of income. We further examine to what extent the type of school, Public, Private-government and Private-independent schools is a source of segregation. In addition, we analyse if in the Basque Country the attendance of students to schools classified according to the language model contributes to the school segregation. 1 Check Section 3 for more details. 2 Check Section 2 for more details. District X 200€ 300€ A 100 0 B 0 100 District Y 100€ 200€ 300€ 400€ A 50 50 0 0 B 0 0 50 50 MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 7 The work is organized as follows. The next Section introduces the index we use for the analysis and the list of axioms fulfilled by the index. After that, in Section 3 we describe the data and its characteristics. Section 4 shows the results and finally we highlight some conclusions about the analysis. MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 8 2. Variance Separation Index and Income Segregation Axioms As we have explained in the previous section, the ESCS value is standardized by PISA, for that reason we use an absolute and invariant segregation index. This index is the Variance Separation Index (VSI) characterized by Lasso de la Vega & Volij, (2019). This index measures income segregation through the variance between schools. A segregation index S defines a segregation order ≽ as follows. For any two districts X,Y, 𝑋 ≽ 𝑌 if and only if 𝑆(𝑋) ≥ 𝑆(𝑌). For any district X= {c1, …, ck} where {c1, …, ck} is the list of schools in the district, the Variance Segregation Index is defined as follows 𝑉(𝑋)=1 𝑛𝑋∑𝑛𝑐(𝜇𝑐− 𝜇𝑋)2 𝑐∈𝑋 , where 𝑛𝑐 is the total enrolment of school 𝑐𝑐 and 𝜇𝑐 is the mean income of the school. In the same way, 𝑛𝑋 represents the total attendance of the district X and 𝜇𝑋 its mean income. Now we present some properties that are desirable for an income segregation index. We begin with three fundamental axioms that transmit the idea of what a district means to be segregated. Particularly, these axioms express the idea that there cannot exist segregation unless there are at least two schools with different income distributions. For any district X, R(X) is the district obtained from reallocating students so that the schools keep their initial enrolment while sharing the same relative income distribution. In other words, R(X) district does not have any segregation as all the schools have the same income distribution. MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 9 The first axiom Equal Allocation Property (EAP) demands that if students are reallocated with the purpose of all schools to have the same income distributions, the segregation does not increase. That is, EAP requires that for any district 𝑋, 𝑋 ≽ 𝑅(𝑋). It does not talk about different allocations of students across income groups nor about different number of schools. The second axiom recognizes a class of districts all of whose members display the same level of segregation. The Equivalence of Single-School Districts (SSD) stands that if X and Y are single-school districts, then X ∼ Y. The next axiom deals with simple districts and egalitarian districts. Simple districts are those with no income variation within schools. That is, all pupils that attend the same school belong to the same income group, while egalitarian districts are known to have an egalitarian income distribution if all pupils have the same income. Egalitarian districts are equally segregated and less segregated than any other simple district, unless it is an egalitarian simple district. The Equivalence of Uniform Distribution Districts (UDD) assumes two simple districts X and Y with the same income and number of pupils. Assume also that X has an egalitarian income distribution. Then Y ∼ X if and only if Y also has an egalitarian income distribution. In consequence of these two axioms, the egalitarian districts and the singleschool districts are equally segregated. The next two axioms require invariance to certain changes in units of measurement. The first one expresses that changes in population that leave the relative attendances of the schools unchanged do not affect segregation. Population Homogeneity (PH) states that for any district X and scalar λ>0, X∼ λX. The λX district is equally segregated as X district. School 100€ 200€ 300€ 400€ A 120λ 50λ 30λ 20λ B 30λ 30λ 30λ 30λ C 10λ 20λ 50λ 100λ MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 16 Remember that PISA standardizes the ESCS indicator so that the mean for the OECD countries takes a value equal to 0 and a standard deviation equal to 1. So, from that point on, negative values mean that the region or country is socioeconomically under the OECD mean level. The ESCS values are between -1 and 1. Hence, we can see that Spain is socioeconomically under the OECD mean level. 13 out of 19 regions have negative values (68.42%), but Ceuta and Melilla for example have a low weight, that is why Spain is still close to the zero mean. Spain also has a large standard deviation. That means that the values are not concentrated near the mean. If we go over the regions in a general view, we can observe that all have large standard deviations. That means that all the regions have large differences between the most socioeconomic advantaged student and the less ones. The most iconic regions, Madrid and Catalonia, have 0.18 and 0.10 positive values respectively. However, there are more regions with positive and large values, as Graph 1 shows. Graph 1: ESCS mean by Regions in Spain in 2018. Source: Own elaboration with data from PISA (2018). -0,6 -0,5 -0,4 -0,3 -0,2 -0,1 0 0,1 0,2 0,3 ESCS mean by Regions in Spain MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 17 So many regions have the ESCS mean over Spain mean, but only 6 of them have positive mean, all in the north half of the country. This could be for many reasons, such as the differences in GDP, income and employment in absolute and per capita terms, Caballero (2019). Likewise, the regions in the worst situation are in the South, or even farther like Canary Islands or Ceuta and Melilla. Now we check whether those ESCS differences have an effect on school segregation. We use the Variance Separation Index to measure the school segregation. Table 2 below shows the results. TABLE 2: Variance Separation Index in Spain and its regions in 2018. Region Variance Separation Index ESCS indicator; Variance ESCS (total inequality) Andalusia 0.2316 (21%) 1.0947 Aragon 0.1757 (18%) 0.9757 Asturias 0.2541 (25%) 1.0369 Balearic Islands 0.1888 (20%) 0.9673 Canary Islands 0.2242 (22%) 1.0282 Cantabria 0.1383 (17%) 0.8285 Castile and Leon 0.1629 (17%) 0.9820 Castile La Mancha 0.2207 (20%) 1.1289 Catalonia 0.2383 (25%) 0.9393 Extremadura 0.1885 (18%) 1.0637 Galicia 0.1749 (18%) 0.9962 La Rioja 0.1607 (16%) 0.9990 Madrid 0.3267 (32%) 1.0268 Murcia 0.2375 (20%) 1.1596 Navarre 0.2024 (20%) 1.0186 Basque Country 0.1627 (21%) 0.7823 Valencian Community 0.1984 (19%) 1.0394 Ceuta 0.2748 (22%) 1.2328 Melilla 0.4660 (30%) 1.5629 Spain 0.2706 (25%) 1.0627 Source: Own elaboration with data from PISA (2018). The second column in Table 2 shows income segregation disaggregated by regions jointly with the percentage of total inequality captured by the segregation MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 18 in parenthesis. As can be seen in the table, the mean of Spain is 0.27, which does not show too many school income segregation. Madrid and Melilla have a value larger than the mean, but Melilla has a low weight to be considered in an analysis. Catalonia has a 0.24 income segregation value, that is quite low to be Catalonia. We find the lowest income segregation in Cantabria. According to the percentages, the 25% of the total school inequality is due to income segregation. In general, all the regions are between 0.15 and 0.30 points more or less, what means the income segregation to be low. We find Madrid with the largest percentage (32%), followed by Catalonia and Asturias, with the 25% each. The difference between the income segregation and the total inequality corresponds to the inequality within schools by region. With these results, we can conclude whether there is income inequality within schools. In this case, the variance is much higher than between schools. This means that the main source of the variance in the ESCS indicator is the within-schools. We can then conclude that students are not separated by schools by the socioeconomic status, but they are mixed among them. The variance of Spain is almost three times larger in terms of within schools than between them. Only Madrid, Andalusia and Murcia have larger values, but two of those have a high weight in the country’s population. In the last column, we find the ESCS indicator variance by region. Through these results, we conclude the inequality by income in the regions. We use this to conclude the segregation by socioeconomic status in schools in general terms for the region. While the ESCS variance is 1.0627, there are some regions whose variance is over that number. In first places we find the two autonomous cities Ceuta and Melilla, whose ESCS level is -0.55 in both cases. Those cities besides having the worst ESCS situations they also have the largest variation in the index. This means that there are some students whose socioeconomical situation is even worse than the mean of the region. MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 19 We can extrapolate this idea to other regions like the Basque Country, which is the region with the less ESCS variation. In this case, the total variation is 0.782, that is composed by 0.16 variation between schools and 0.62 within schools. In the case of this region, the ESCS mean is 0.14, with the lowest standard deviation. In general, in the region the socioeconomic level is better than in other regions and there is low segregation level by income between schools. Once we have analysed the school segregation by regions, we go deeper in the analysis and check the segregation by type of schools. In Spain we distinguish three type of schools: Public schools, Private-government schools and Private-independent schools. Our goal is to observe the distribution to conclude whether this distribution affects the segregation by income, that is, whether the type of school separates richer pupils from poorest ones. It would be understandable to think that the less income the student has, it will study in a public school. Likewise, richer pupils are expected to go to private independent schools. The following table 3 shows the student distribution in Spanish schools. TABLE 3: Student distribution in Spanish schools in 2018: Region Public school Private Governmentdependence school Private independent school TOTAL Andalusia 2,282,533 (82.13%) 496,697 (17.87%) - 2,779,230 (100%) Aragon 250,946 (66.42%) 96,352 (25.50%) 30,493 (8.07%) 377,791 (100%) Asturias 182,116 (72%) 69,329 (27.41%) 1,506 (0.60%) 252,951 (100%) Balearic Islands 226361 (69.86%) 72,071 (22.24%) 25,603 (7.90%) 324,035 (100%) Canary Islands 502,752 (77.47%) 72,949 (11.24%) 73,272 (11.29%) 648,973 (100%) Cantabria 117,143 (74.72%) 36,187 (23.08%) 3,441 (2.19%) 156,771 (100%) MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 20 Castile and Leon 434,023 (67.95%) 179,595 (28.12%) 25,096 (3.93%) 638,714 (100%) Castile La Mancha 536,000 (83.32%) 81,256 (12.63%) 26,046 (4.05%) 643,302 (100%) Catalonia 1,535,817 (65.24%) 680,533 (28.91%) 137,683 (5.85%) 2,354,033 (100%) Extremadura 267,917 (81.53%) 60,691 (18.47%) - 328,608 (100%) Galicia 499,684 (76.56%) 125,002 (19.15%) 27,995 (4.29%) 652,681 (100%) La Rioja 61,426 (61.67%) 38,181 (38.33%) - 99,607 (100%) Madrid 1,077,402 (57.08%) 462,561 (24.51%) 347,456 (18.41%) 1,887,419 (100%) Murcia 371,228 (72.74%) 139,087 (27.26%) - 510,315 (100%) Navarre 127,681 (60.67%) 82,770 (39.33%) - 210,451 (100%) Basque Country 278,718 (50.54%) 272,777 (49.46%) - 551,495 (100%) Valencian Community 1,025,774 (68.72%) 358,249 (24%) 108,720 (7.28%) 1,492,743 (100%) Ceuta 26,647 (77.02%) 7,950 (22.98%) - 34,597 (100%) Melilla 31,637 (88.92%) 3,941 (11.08%) - 35,578 (100%) Spain 9,835,805 (70.36%) 3,336,178 (23.87%) 807,311 (5.78%) 13,979,294 (100%) Source: Own elaboration with data from PISA (2018). This table may insinuate that in some regions there are no private-independent school, but by a long shot, that is because this database is made by sample weights. In those regions, PISA has not coincided in private-independent schools, that is why we have no data. MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 21 Looking the results obtained for the whole country, we can observe that more than the 70% of the students attend public schools, while the 23.87% go to private government-dependence schools and 5,78% to private-independent schools. We also had “no-response” data because this PISA questionnaire was answered by 15 years old students. That is why we must understand that some of them might not know exactly what type of schools they belong to. Nevertheless, we have dropped all missing data for the analysis. The Basque Country is the region where less students attend public schools, only the 50%. In the Basque Country also, we find the larger private-government dependence school attendance, with the 49.46% of the students of the region. Whereas in Madrid, we find the highest private-independent school attending rate, 18.41%, far above the rest of regions. Now that we know the distribution of pupils, let’s examine the ESCS by types of school and regions in Spain. It will be interesting to conclude whether the socioeconomic status is related with the school type attendance. Table 4: ESCS by type of school in Spain in 2018. Region ESCS mean in Pu. Sch. Std. Dev. ESCS in Priv.- Gov. School Std. Dev. ESCS mean in Pr. Ind. Sch. Std. Dev. Andalusia -0.42 1.07 0.01 0.87 - - Aragon -0.15 0.99 0.26 0.90 0.51 0.86 Asturias -0.22 1.00 0.39 0.92 1.00 0.53 Balearic Islands -0.22 0.99 0.20 0.91 0.31 0.85 Canary Islands -0.57 0.97 0.13 0.89 0.34 0.88 Cantabria -0.09 0.89 0.36 0.90 -0.05 0.70 Castile and Leon -0.11 1.00 0.19 0.94 0.45 0.82 Castile La Mancha -0.39 1.04 0.43 0.90 0.41 0.94 Catalonia -0.09 0.98 0.37 0.81 0.98 0.68 Extremadura -0.48 1.02 0.02 0.99 - - MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 22 Galicia -0.18 1.01 0.11 0.84 0.81 0.88 La Rioja -0.30 0.99 0.13 0.96 - - Madrid -0.13 1.02 0.25 0.91 0.92 0.65 Murcia -0.49 1.08 0.06 0.96 - - Navarre -0.22 1.04 0.44 0.82 - - Basque Country -0.04 0.90 0.29 0.84 - - Valencian Community -0.42 1.00 0.19 0.86 0.56 0.89 Ceuta -0.78 1.06 0.19 0.92 - - Melilla -0.70 1.22 0.65 0.73 - - Spain -0.29 1.03 0.22 0.86 0.76 0.79 Source: Own elaboration with data from PISA (2018). Public schools have negative values for all regions. That means that the students of public schools are in mean socioeconomically disadvantaged in comparison with the OECD mean. For private-government schools, the ESCS results are all positive, some of them far from the zero mean. The mean for Spain is 0.22, which means that students of private-government schools tend to be socioeconomically advantaged with respect to the OECD mean. At last, we observe that pupils of private-independent schools are much more socioeconomically advantaged, as the ESCS mean of Spain is 0.76, being 1 the maximum. Therefore, we could conclude that in mean there are socioeconomical differences between students from different type of schools. On the other hand, we observe that the standard deviation of the variance index is larger in public schools and lower in private-independent schools. This means that there is more within-school segregation in public schools than in private-independent schools. At last, and before analysing these results through a graphic, we have to point out that Cantabria has a -0.05 ESCS mean value for private-independent schools. MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 23 There can be several reasons for this, but as we have no additional data about this issue, we cannot state anything. In the following Graph 2, we do not only see the ESCS mean evolution but also the relation between ESCS region mean and ESCS region mean by school type. Graph 2: ESCS mean in Spain by regions and type of school in 2018. Source: Own elaboration with data from PISA (2018). Blue columns are the ESCS mean by regions. That shows what we have seen in the previous Graph 1. Next to those, we find the orange columns, which represent the ESCS mean of public schools. We can observe the negative tendency they have for the less ESCS region mean. There are some exceptions like the Basque Country, Catalonia, Cantabria, Castile and Leon and Galicia. Those regions’ public school ESCS mean is larger than the expected value considering the negative tendency. Nevertheless, they all still are negative values. The next grey columns show the private-government schools’ ESCS mean, which has no continuity though the graph. That is, it is not related to the region’s ESCS mean. Some regions have larger values and others lowers. MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 24 At last, we find the yellow columns, that symbolizes the ESCS mean in privateindependent schools. It is more complicated to conclude a tendency due to not all regions have private-independent school data. These values are expected to be the highest ones, but we can clearly see that not always is like that. In Cantabria and Castile la Mancha, the private-government school ESCS mean value is larger. In the following Table 5 we find results of the income segregation according to the type of schools and its implication in the total Income Segregation. In other words, in the following table we find whether students are segregated by income among type of schools and how much of this segregation explains the total school segregation. Table 5: Income Segregation between type of schools in Spain in 2018. Region Income Segregation according to type of schools % of the total Income Segregation Andalusia 0.0344 14.85 Aragon 0.0474 26.99 Asturias 0.1292 50.83 Balearic Islands 0.0421 22.30 Canary Islands 0.1179 52.58 Cantabria 0.0366 26.47 Castile and Leon 0.0284 17.47 Castile La Mancha 0.1375 62.32 Catalonia 0.0889 37.29 Extremadura 0.0463 24.56 Galicia 0.0622 35.58 La Rioja 0.0358 22.29 Madrid 0.0753 23.06 Murcia 0.0582 24.51 Navarre 0.0813 40.15 Basque Country 0.0206 12.67 Valencian Community 0.1044 52.63 Ceuta 0.1789 65.09 Melilla 0.3450 74.03 Spain 0.0931 34.40 Source: Own elaboration with data from PISA (2018). MASTER THESIS MEASURING SCHOOL INCOME SEGREGATION 25 The second column exhibits the segregation between type of schools, which are the public school, the private government-dependence school and the private independence school. The results show that in some cases the type of school is not relevant for the segregation between type of schools, while it is for other cases. Let’s see that in regions as the Basque Country, Andalusia or Castile and Leon the segregation by type of school does not even suppose the 20% of the income school segregation. However, in Valencian Community, Castile La Mancha or Asturias it signifies more than the 50% of the total income school segregation. To finish with the analysis, we test whether the language and the type of school are relevant in the income segregation for the Basque Country. We do this analysis for the Basque Country because the language may be significant for students and their parents when deciding where they study. In the Basque Country the culture is very important and there are two types of schools in respect of the language: AB model that mixes Spanish and Basque languages and the D model, that teaches only in Basque except the Spanish language subject. In the next Table 6 we find results about the type of schools mentioned above in the work and the type of language of the teaching. Table 6: Type of schools and language in the Basque Country in 2018. Number of students ESCS mean Income Segregation according to language and type of schools % of the total Income Segregation Public, AB 24,922 -0.5846 0.0460063 0.2828 Public, D 246,422 -0.0034 Privategovernment, AB 152,488 0.2429 Privategovernment, D 127,663 0.3531 Source: Own elaboration with data from PISA (2018).