EDUCATION WORKING PAPER No. 3 | APRIL 2024 Evaluation of Achievement Change in Egypt using PIRLS Data Harry Patrinos, Maciej Jakubowski and Tomasz Gajderowicz © 2024 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy, completeness, or currency of the data included in this work and does not assume responsibility for any errors, omissions, or discrepancies in the information, or liability with respect to the use of or failure to use the information, methods, processes, or conclusions set forth. The boundaries, colors, denominations, links/footnotes and other information shown in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. The citation of works authored by others does not mean the World Bank endorses the views expressed by those authors or the content of their works. Nothing herein shall constitute or be construed or considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. Cover design: Marianne Siblini Evaluation of Achievement Change in Egypt using PIRLS Data 1 Harry Anthony Patrinos, World Bank Maciej Jakubowski, University of Warsaw Tomasz Gajderowicz, University of Warsaw April 2024 Abstract Amidst global COVID-19 school closures, student outcomes varied worldwide, but usually declined. Egypt notably defied this trend, showing significant improvement in reading scores. Utilizing data from the Progress in International Reading Literacy Study (PIRLS), one observes a remarkable 48-point increase from 2016 to 2021, akin to two years' academic growth. Gender disparities narrowed, with boys enhancing scores notably. The increase is confirmed using a model that analyzes the change in student achievement in 2021 by predicting the deviation of the most recent 2021 results from an estimated linear trend in reading achievement using data from all PIRLS rounds. Employing the Oaxaca-Blinder technique, reading gaps are decomposed. Notably, one-third of the progress stemmed from unexplained variables, hinting at possible policy reforms. Since most of the difference is unobservable, then it is probably due to policy or system changes. JEL codes: J15, I20 Keywords: education, test scores, Egypt, PIRLS 1 Address correspondence to Harry Anthony Patrinos (hpatrinos@worldbank.org). The comments of Shwetlena Sabarwal and Andreas Bloom are gratefully appreciated. The views expressed here are those of the authors and should not be addressed to the World Bank Group. 1 I. Introduction Students in most countries were negatively affected by the COVID-19 pandemic. Using global student achievement data, it was shown that student achievement declined by an equivalent of 7 months of education in countries with average length of school closures (Jakubowski et al 2023). Countries that closed schools for longer experienced even larger learning losses. However, achievement levels were similar or even improved in some countries. Here we use evidence from the Progress in International Reading Literacy Study (PIRLS), the world’s largest international test of 4th-grade students’ reading achievement. The 2021 edition was implemented when most countries re-opened schools after closures related to the COVID-19 pandemic. We estimate achievement change in Egypt, controlling for the overall time trend in PIRLS- participating countries and changes in student characteristics in PIRLS samples. We also report results for different social background groups, age groups, boys and girls, and student at different achievement levels. Reading scores in Egypt between 2016 and 2021 increased by 48 points, almost two years’ worth of learning. We model the change in student achievement in 2021 by predicting the deviation of the most recent 2021 results from an estimated linear trend in reading achievement using data from all rounds. Part of the improvement is due to boys improving their reading scores and reducing the gap with girls, which still outperform boys. The performance of the lowest-achieving boys improved by around 80 points, which is around 2/3 of a standard deviation (SD). We apply the Oaxaca-Blinder technique to decompose reading gaps to assess how much of the gap is due to differences in characteristics (explained variation) and how much is due to policy or system changes (unexplained variation). The decomposition method, the technique used for analyzing earnings differentials, was popularized in the economics literature by Oaxaca (1973) and Blinder (1973), although it was used earlier in demography (Kitagawa 1955). Although in the economics literature it was first used to analyze the determinants of male/female earnings differentials, the decomposition technique has been used since to analyze ethnic earnings differentials, public/private sector earnings differentials, earnings 2 differentials by socioeconomic background, to test the screening hypothesis, and to test the effectiveness of a job training program, among other uses. It has been applied since in other social issues, including education, where it can be used to assess how much of a gap is due to differences in characteristics (explained variation) and how much is due to policy or system changes (unexplained variation). Only one-third of the improvement in reading achievement is associated with changes in the explanatory variables. Of these, the improved results are mainly explained by increased socioeconomic status. Improved school characteristics and improved access to digital textbooks are associated with small positive changes. This means that most of the test score increase was unexplained, or, in other words, was due to changes in the returns to the characteristics rather than due to changes in the characteristics themselves. To put it another way, Egypt was able to better educate its students in 2021 regardless of the characteristics of those students. Since most of the difference is unobservable, then it is probably due to policy or system changes. Information provided by school principals can help identify sources of positive change and provide avenues for further research. Principals claim that collaboration to plan instruction and parental involvement increased. While in 2016 more than 30 percent of students where in schools with medium or low collaboration, in 2021 less than 10 percent were in such schools. Reading performance in schools with high collaboration increased by 40 points between 2016 and 2021. II. Data The PIRLS 2021 data include results for 65 countries and benchmarking participants. This is the first international picture of student achievement after the COVID-19 pandemic. Our analysis uses data from 56 education systems. We compare PIRLS results from assessment rounds in 2001, 2006, 2011, 2016 and 2021. These data allow us to estimate the overall time trend in reading across PIRLS-participating countries. This is contrasted with changes in results of Egypt between 2016 and 2021. 3 PIRLS data include the plausible values measuring reading achievement and information on student socio-demographic characteristics (e.g., age, grade, and gender). In addition, microdata from PIRLS allow to estimate effects for students at different achievement levels, testing whether the achievement changes across time are similar for low- and high-achieving students. III. Empirical Strategy We model the change in student achievement in 2021 by predicting the deviation of the most recent 2021 results from an estimated linear trend in reading achievement using data from all PIRLS rounds (Jakubowski et al 2023). We estimate a linear regression model on pooled data from all assessments and all countries, assuming a common time trend for all countries. The model is given by equation below: = � + ∗ + � ∗ 2021 + + =1 =1 where i denotes students, j schools, and k countries, with n equal to the number of countries. In this case, D equals 1 for 2021 data collected after the pandemic and zero for previous years. Thus, captures the change in student achievement between 2021 and PIRLS rounds before 2021. Country fixed effects are denoted by and time trend is captured by , while is a vector of demographic variables (student age, grade, and gender). This model allows obtaining country-specific estimates for countries, which has only data from two rounds of PIRLS. By including country-level fixed effects, we control for unobserved time-invariant country characteristics. By controlling for time trend, we account for the overall change in student achievement over time, but with a possibility to estimate country- specific changes in achievement in the last round. In general, PIRLS surveys 4th-grade students who are around ten years old at the assessment time. However, students might take the test earlier or later. Some are repeating, and some are advancing a grade. Also, some countries decided to test students in 3rd, 5th, or even 6th 4 grade. Moreover, while most countries sampled the same population over the years, some changed the target population. Finally, due to the pandemic, some countries postponed the assessment to the beginning of the fifth grade. To consider these differences, we additionally control for student age (10 dummies, one for each age decile, to allow for a non-linear age effect), student grade, and for the average age effect calculated as a mean age for each round and country. PIRLS data are collected through a complex stratified survey with schools sampled as primary sampling units and classrooms with students sampled at the second stage. We used the jackknife replicate weights provided by the PIRLS database to estimate the sampling error. We use plausible values for reading achievement to adjust for the measurement error. The Oaxaca-Blinder technique was originally used in labor economics to decompose earnings gaps and to estimate the level of discrimination. It has been applied since in other social issues, including education, where it can be used to assess how much of a gap is due to differences in characteristics (explained variation) and how much is due to policy or system changes (unexplained variation) (Ammermueller 2007; Barrera-Osorio et al 2011). We apply the decomposition technique to analyze the increase in Egypt’s reading score. The gap over time is decomposed into its constituent components based on the estimation of cognitive achievement production functions. IV. Results While most countries suffered significant learning losses during COVID-19 as a result of the school closures (Jakubowski et al. 2023), a few improved their scores. The biggest improvements are in Malta and several Middle East and North African countries. Figure 1 presents the change in achievement between 2016 and 2021 in Egypt and other countries participating in these two rounds of PIRLS. Egypt is among the countries with improved achievement. 5 Figure 1: PIRLS 2016 to 2021 Achievement Change in Egypt and Other Participating Countries 80 60 40 20 0 -20 -40 Spain Kazakhstan Latvia Russian Federation Slovenia Israel Germany Sweden Austria Denmark Oman Australia Belgium (French) New Zealand Northern Ireland France Singapore Bahrain Poland South Africa Hungary Italy England Egypt Slovak Republic Lithuania UAE (Dubai) Qatar Malta Azerbaijan Iran Bulgaria Georgia Netherlands Macao SAR Chinese Taipei Czech Republic Canada (Quebec) Hong Kong, SAR Morocco Saudi Arabia UAE Norway Finland Belgium (Flemish) Moscow (Russia) Portugal United States UAE (Abu Dhabi) Ireland The model-based estimates are obtained by assuming a common time trend in PIRLS achievement but with country-specific fixed effects (Table 1). The estimates for Egypt are displayed in Figure 2. This compares the unadjusted achievement change with model-based estimates of the achievement change in 2021, comparing Egypt to all other participants. The first bar shows the raw unadjusted change in PIRLS results for Egypt, as shown in Figure 1. The next three bars show model-based estimates. The model-based estimates for Egypt are larger than raw estimates as other PIRLS-participating countries experienced a large decline in reading achievement in 2021 (see Jakubowski et al. 2023). Table 1: Model-based Estimates of Achievement Change (1) (2) (3) Variable b/se b/se b/se 1.14*** 1.18*** 1.16*** Time (0.12) (0.13) (0.13) -13.47*** -18.82*** -18.4*** Departure from the linear trend in 2021 (2.17) (1.9) 1.93) Change in Egypt in 2021 (relative to 55.52*** 58.54*** 60.72*** change in other countries) (7.31) (7.24) (7.93) PIRLS literacy countries in 2016 (compared to 2.48 countries in the main study) (2.97) Country fixed effects Yes Yes Yes Age and grade effects No Yes Yes N 1,112,257 1,110,821 1,110,821 Legend: * p<0.05; ** p<0.01; *** p<0.001 6 Figure 2: Model-based Estimates of Achievement Change in Egypt 60.7 58.5 55.5 47.8 raw change common time trend and country common time trend, country common time trend, country fixed effects fixed effects, age and grade fixed effects, age and grade effects effects, and control for PIRLS literacy countries Differences between student groups The differences in reading achievement within Egypt are larger than on average in PIRLS. Table 2 compares standard deviation of student achievement in 2016 and 2021. As the standard deviation on the PIRLS scale is equal to 100, these data show variation in student achievement is larger in Egypt. Table 2 also provides results by gender, showing a larger variation among boys. Student reading achievement variation decreased from 2016 to 2021. Moreover, the decrease in reading achievement variation between 2016 and 2021 is larger among boys. On the other hand, achievement still varies more among boys than among girls. Table 2: Standard Deviation of Reading Achievement in Egypt in 2016 and 2021 Year All Girls Boys 2016 123.6 115.3 128.6 2021 109.9 106.3 112.8 In Egypt, as in most countries, girls perform better in reading. In 2016, the gender gap was around 37 points. However, between 2016 and 2021, the gender gap diminished to 16 points. Figure 3 shows that the overall improvement of reading achievement in Egypt between 2016 and 2021 was driven by large improvements among the lowest-achieving students, especially 7 boys. Figure 3 also compares reading achievement at quantiles, from the 1st decile (the lowest-achieving students) to the 9th decile (the highest-achieving students). Performance of the lowest-achieving boys (1st and 2nd deciles) improved by around 80 points, which is around 2/3 of standard deviation. Performance of the lowest-achieving girls improved by around 40- 50 points, which is less than ½ of standard deviation. At the top of the reading achievement distribution, performance of girls improved by around 30 points and of boys around 40 points. Figure 3: Achievement Changes between 2016 and 2021 by Quantiles of Reading Achievement and Sex 90 80 70 60 50 40 girls 30 boys 20 10 0 1 2 3 4 5 6 7 8 9 Achievement quantile The changes in performance at achievement deciles presented in Figure 3 explain why the overall differences in achievement have changed between 2016 and 2021. Larger improvements among the lowest-achieving students resulted in overall declining reading achievement variation. Moreover, larger improvements among boys, especially the lowest- performing boys, result in a larger decline in boys' reading achievement variation compared to girls. V. Decomposing the Changes in Reading Scores Over Time We compare PIRLS 2016 and PIRLS 2021 data to explain achievement change using selected variables. To facilitate regression and decomposition analysis, individual items measuring reading activities, instruction, and post-reading questioning were used to construct three 8 indices. These indices were estimated using IRT 2PL and IRT GPRM models on the 2016 and 2021 data pooled. Similarly, the SES index was constructed using information on the number of books, parents’ education, and home possessions. Finally, an index of school characteristics based on principal opinions was constructed. Annex Table 1 provides descriptive statistics for all variables, including single questionnaire items for each IRT-derived index. All explanatory variables were standardized as z-scores for the pooled 2016 and 2021 samples using survey weights to allow the interpretation of regression and decomposition coefficients. Moreover, all variables were recoded, so higher values are expected to correlate with reading achievement positively. Reading assessment results were kept on the original international PIRLS scale. Thus, a coefficient of 13.5, for example, can be interpreted as showing a difference equal to 13.5 points on the PIRLS international scale related to a change of 1 standard deviation in the explanatory variable for the 4th-grade population in Egypt. Considering that the standard deviation of reading achievement in Egypt was 108.4 points in 2021, the coefficient of 13.5 points is equivalent to the effect of 12.5 percent of a SD. 9 The means of the variables used in regression and decomposition for 2016 and 2021 are presented in Table 3. Comparing the values of the variables across 2016 and 2021, one can note a considerable improvement in reading achievement. The values of the three indices capturing characteristics of instruction in the national language decreased between 2016 and 2021. However, only the index of reading activities and frequency of homework showed statistically significantly lower values in 2021. This can indicate a more balanced approach to reading instruction, with less focus on single activities. Principal opinions about the school's characteristics are more positive in 2021, but the difference is not statistically significant. Teacher and principal experience increased, as did their formal education levels, which might be related to teacher age. Finally, access to digital books and the number of computers available for the 4th graders' instruction increased between 2016 and 2021. Table 3: Means of Variables used in Regression and Oaxaca Analyses 2016 2021 2021 minus 2016 Coef SE Coef SE Diff SE Reading achievement 331.54 6.86 377.73 5.91 46.19*** 8.80 Index of reading activities 0.28 0.10 -0.19 0.08 -0.48*** 0.13 Index of reading instruction 0.14 0.10 -0.10 0.10 -0.24 0.14 Index of post-reading questioning 0.12 0.09 -0.08 0.09 -0.20 0.13 Reading Homework 0.26 0.09 -0.18 0.09 -0.44** 0.14 Teacher’s years of experience -0.16 0.10 0.11 0.09 0.27* 0.14 Teacher’s formal education -0.21 0.09 0.14 0.08 0.36** 0.11 SES index (books, parent education, possessions) -0.17 0.05 0.11 0.05 0.28*** 0.08 Gender (1=Boy) -0.01 0.03 0.01 0.03 0.02 0.04 Years in preprimary/ECE -0.06 0.05 0.04 0.04 0.11 0.06 Index of school characteristics (principal opinion) -0.13 -0.10 0.09 0.08 0.22 0.12 Principal experience (in years) -0.13 0.06 0.09 0.12 0.22 0.13 Principal formal education -0.04 0.09 0.03 0.09 0.08 0.12 The school has access to digital books -0.30 0.10 0.20 0.07 0.50*** 0.13 Number of computers for 4th graders' instruction -0.12 0.05 0.08 0.11 0.20 0.13 Sample size 4683 5723 10406 Note: Bolded differences show changes between 2016 and 2021 that are statistically significant at a level denoted by: * p<0.05; ** p<0.01; *** p<0.001 The results using the same variables but estimated separately for 2016 and 2021 data are presented in Table 4. Note that the estimation sample is much smaller than the available 10 sample due to missing values (mainly SES and preschool data). Only the SES index and gender are strongly associated with reading achievement. Table 4: Regression Results Variable 2016 2021 Index of reading activities 1.1 -4.7 Index of reading instruction 2.7 -1.5 Index of post-reading questioning -10.2 2.4 Reading homework 5.6 -10.9* Teacher experience -4.4 -13.7** Teacher formal education 1.3 -8.7 SES index (books, parent education, possessions) 40.9*** 21.8*** Boy (baseline=girls) -15.1*** -11.5*** Years in preprimary/ECE 5.7 3.5 Index of school characteristics (principal opinion) 11.4* 13.8* Principal experience (in years) 11.5 1.7 Principal formal education 8.5 -10.3 School has access to digital books 9.1 7.2 Number of computers for 4th graders instruction 8.4 -5.9 Constant 344.3*** 373.1*** N of students 4683 5723 Legend: * p<0.05; ** p<0.01; *** p<0.001 Oaxaca Decomposition The Oaxaca decomposition results are presented in Tables 5 and 6, assuming the same regression coefficients for the pooled sample from 2016 and 2021. Table 5 is based on a model with all explanatory variables. Table 6 reports a more parsimonious model, only with variables significantly associated with changes in reading achievement over time. In Table 4, the first column shows effects for all variables, while the second column reports combined effects for groups of variables. One-third of the improvement in reading achievement (15.2 points out of 46.2 points) is associated with changes in the explanatory variables. Thus, the decomposition model associates the increase in achievement of around 14 percent of a SD with the explanatory variables. The improved results are mainly explained by increased SES (socioeconomic status). Around 9 points (one-fifth of the improvement, or the effect of 8 percent of SD) can be explained by the higher SES of student families in 2021. Improved school characteristics, as reported by principals, are associated with a change of 3.4 points (7 percent of the 11 improvement or 3 percent of SD). Finally, improved access to digital textbooks is associated with 4.1 points (9 percent of the improvement or 4 percent SD). This effect might be related to relatively more effective learning during the pandemic or improved instruction using digital resources afterward. On the other hand, increases in teacher experience and formal education level are not associated with improvements. One possible explanation is that experienced teachers lack further motivation to excel in the classroom, which results in lower achievement. Moreover, reporting less frequent reading activities or homework is positively associated with improving reading achievement, although the effects are minor and not statistically significant. Based on a correlational study like PIRLS, further understanding of the reasons behind these effects is limited. Below, we provide a detailed analysis of indicators associated with improved reading achievement, which might help interpret some effects. 12 Table 5: Oaxaca Decomposition Individual Variable variables Group effects Overall scores Group_1 (2021) 377.7*** 377.7*** Group_2 (2016) 331.5*** 331.5*** Difference 46.2*** 46.2*** Explained 15.2** 15.2** Unexplained 31.0*** 31.0*** Explained Reading 3.7 Reading activities index 0.4 Reading instruction index 0.1 Index of post-reading questioning 0.6 Reading homework 2.5 Teacher -4.9* Experience -3.3 Formal education -1.6 Family and individual 8.8*** SES 9.0*** gender -0.3 Preschool time 0.6 0.6 School characteristics (index of principal opinions) 3.4 3.4 Principal characteristics 0.1 Principal experience 0.5 Principal education -0.4 Digital 3.5 Access to digital textbooks 4.1 Number of computers -0.6 N 10,406 10,406 Legend: * p<0.05; ** p<0.01; *** p<0.001 A simpler model is also estimated (Table 6). It includes only the variables with statistically significant effects or coefficients above 3 points. Qualitatively, the results show similar effects as for the model presented in Table 5. 13 Table 6: Oaxaca Decomposition: Parsimonious Model Variable Individual variables Overall scores Group_1 (2021) 377.7*** Group_2 (2016) 331.5*** Difference 46.2*** Explained 15.5** Unexplained 30.7*** Explained Reading (activities, instruction, post-reading questioning, homework)) 3.4 Teacher characteristics (experience and formal education) -4.5* Socioeconomic Status (SES) 9.2*** School characteristics (index of principal opinions) 3.4 Access to digital textbooks 4.0 N 10,406 Legend: * p<0.05; ** p<0.01; *** p<0.001 The effects estimated in the decomposition analysis are summarized in Figure 4. In what follows, we focus on the principal-provided characteristics of schools and how they are associated with reading improvements. Figure 4: Summary of Decomposition Analysis 10 9.2 8 6 4 4 3.4 3.4 2 0 -2 -4 -4.5 -6 Teacher characteristics Reading (activities, School characteristics Access to digital Socioeconomic Status (experience and instruction, post- (index of principal textbooks (SES) formal education) reading questioning, opinions) homework)) 14 Regarding school characteristics as evaluated by school principals, only two significantly improved between 2016 and 2021. Principals claim that (1) collaboration to plan instruction and (2) parental involvement increased. Figure 5 compares percentages of students in schools with different responses regarding collaboration to plan instruction in 2016 and 2021. While in 2016 more than 30 percent of students where in schools with medium or low collaboration, in 2021 less than 10 percent were in such schools. Despite the increase in the share of schools claiming high teacher collaboration, and despite that this group of schools scored higher already in 2016, the reading performance in schools with high collaboration increased further by around 40 points between 2016 and 2021 (about 37 percent of a SD). The reading achievement also improved among students in schools reporting medium or low collaboration, but the increase was lower (by 16 points or 15 percent of a SD). Figure 5: Principal’s Opinion about Teacher Collaboration to Plan Instruction (percentage of low/medium vs. high responses (left panel) and reading achievement (right panel) in 2016 and 2021) 400 382.5 342.6 2021 91.8 350 324.6 308.5 300 2016 67.6 250 200 0% 20% 40% 60% 80% 100% medium or low high medium or low high 2016 2021 A similar analysis of the principal’s opinion about parental involvement is shown in Figure 6, which is less favorable than the opinion about teacher collaboration. Anyhow, in 2021, around 27 percent of students in schools where principals evaluated parental involvement positively, compared to only 14 percent in 2016. The reading achievement improved in this group despite that it doubled in size, but the improvement was relatively small, by around 14 points (13 percent of a SD). On the other hand, the reading achievement improved significantly in the group with medium or low involvement, from 326 to 374 points. Consequently, in 2021, the difference between schools with low, medium, or high parental involvement was much smaller than in 2016. 15 Figure 6: Principal’s Opinion about Parental Involvement (percentage of low/medium vs. high responses (left panel) and reading achievement (right panel) in 2016 and 2021) 400 388 373.9 367.3 380 2021 27.1 360 340 325.9 320 2016 13.7 300 280 0% 20% 40% 60% 80% 100% medium or low high medium or low high 2016 2021 VI. Conclusion Students in most countries were adversely affected by the COVID-19 pandemic-induced school closures. However, not all countries experienced losses. In fact, some improved outcomes. Egypt is one such country. We use evidence from the Progress in International Reading Literacy Study (PIRLS) to estimate achievement change controlling for the overall time trend in participating countries and changes in student characteristics. Reading scores in Egypt between 2016 and 2021 increased by 48 points, almost two years’ worth of learning. We model the change in student achievement in 2021 by predicting the deviation of the most recent 2021 results from an estimated linear trend in reading achievement using data from all rounds. Part of the improvement is due to boys improving their reading scores and reducing the gap with girls, which still outperform boys. The performance of the lowest- achieving boys improved by around 80 points, which is around 2/3 of a standard deviation. We apply the Oaxaca-Blinder technique to decompose reading gaps to assess how much of the gap is due to differences in characteristics (explained variation) and how much is due to policy or system changes (unexplained variation). Only one-third of the improvement in reading achievement is associated with changes in the explanatory variables. Of these, the improved results are mainly explained by increased socioeconomic status. Improved school characteristics and improved access to digital textbooks are associated with small positive changes. 16 This means that most of the test score increase was unexplained, or, in other words, was due to changes in the returns to the characteristics rather than due to changes in the characteristics themselves. To put it another way, Egypt was able to better educate its students in 2021 regardless of the characteristics of those students. Since most of the difference is unobservable, then it is probably due to policy or system changes. Information provided by school principals can help identify sources of positive change and provide avenues for further research. Principals claim that collaboration to plan instruction and parental involvement increased. While in 2016 more than 30 percent of students where in schools with medium or low collaboration, in 2021 less than 10 percent were in such schools. Reading performance in schools with high collaboration increased by 40 points between 2016 and 2021. 17 Annex Table 1: Descriptive Statistics for Decomposition Analysis 2016 2021 Variable Mean SD N Mean SD N Index of reading activities Read\activities\read aloud to students 1.16 0.39 6867 1.31 0.64 7900 Read\activities\ask stud to read aloud 1.24 0.43 6823 1.33 0.60 7945 Read\activities\read silently 1.65 0.66 6884 1.69 0.72 7891 Read\activities\decoding words 1.98 0.89 6727 2.20 0.93 7735 Read\activities\new vocabulary 1.40 0.57 6884 1.40 0.62 7945 Read\activities\skimming or scanning activities 1.88 0.80 6884 2.39 1.03 7820 Index of reading instruction Read\how often\provide reading materials 2.05 0.94 6830 1.82 0.93 7872 Read\how often\provide materials for reading levels 2.44 0.97 6919 2.24 0.96 7872 Read\how often\link new content 1.35 0.67 6903 1.37 0.63 7773 Read\how often\encourage to develop understanding 1.20 0.49 6957 1.34 0.69 7872 Read\how often\encourage discussions 1.36 0.67 6907 1.43 0.70 7872 Read\how often\encourage to challenge opinion 1.89 0.94 6907 1.94 0.90 7669 Read\how often\multiple perspective 1.65 0.79 6957 2.27 0.98 7872 Read\how often\time to read books 2.56 0.98 6910 2.45 0.95 7872 Read\how often\individual feedback 2.08 1.00 6938 2.11 1.00 7872 Index of post-reading questioning Read\after reading\write response 1.83 0.78 6924 1.87 0.76 7877 Read\after reading\summarize 1.38 0.62 6957 1.66 0.72 7877 Read\after reading\talk with each other 1.96 0.88 6915 1.87 0.86 7851 Read\after reading\take quiz 1.63 0.75 6957 1.70 0.79 7877 Homework: recode of atbr15 (read\reading assigned as homework) 4.08 0.93 6840 3.46 1.16 7789 Teacher characteristics recode of atbg01 (gen\years been teaching) 17.19 9.58 6544 19.68 9.74 7475 recode of atbg04 (gen\level of formal education completed) 4.56 0.85 6873 4.84 0.71 7779 SES and gender recode of asbg04 (gen\amount of books in your home) 1.43 0.80 6716 1.77 1.06 6712 Gen\home possess\shared computer or tablet 0.58 0.49 6811 0.69 0.46 7368 Gen\home possess\study desk 0.67 0.47 6745 0.64 0.48 7262 Gen\home possess\own room 0.38 0.49 6778 0.52 0.50 7259 Gen\home possess\internet connection 0.38 0.49 6767 0.64 0.48 7256 recode of asbh15a (gen\lvl of education\guardian a) 4.17 2.16 6058 4.45 2.04 5312 recode of asbh15b (gen\lvl of education\guardian b) 3.77 2.24 5413 4.24 2.21 4111 Itsex (1=boy 0=girl) 0.51 0.50 6957 0.51 0.50 7979 Preschool recode of asbh05aa (gen\attend\early childhood educational program) 0.45 0.50 5716 0.58 0.49 5942 recode of asbh05ab (gen\attend\pre-primary educational program) 0.73 0.45 6009 0.82 0.38 6447 recode of asbh05b (gen\how long attended programs) 1.74 1.39 6527 1.81 1.31 7979 School characteristics (principal opinion) gen\sch character\tch understanding 1.81 0.66 6957 1.76 0.71 7876 gen\sch character\tch success 1.89 0.62 6957 1.94 0.70 7855 gen\sch character\tch expectations 2.27 0.70 6957 2.33 0.71 7771 gen\sch character\tch ability to inspire students 1.83 0.71 6957 1.80 0.62 7876 gen\sch character\collaboration to plan instruction 2.21 0.75 6957 1.66 0.66 7876 gen\sch character\parental involvement 3.42 0.99 6957 2.93 0.98 7876 gen\sch character\parental commitment 2.83 1.07 6910 2.55 0.96 7760 gen\sch character\parental expectations 2.73 0.85 6957 2.53 0.82 7876 gen\sch character\parental support 2.92 0.97 6957 2.61 0.90 7876 gen\sch character\std desire to do well 2.34 0.74 6957 2.32 0.79 7876 gen\sch character\ability to reach goals 2.71 0.79 6957 2.65 0.71 7876 gen\sch character\respect for classmates 2.28 0.81 6897 1.98 0.69 7833 Principal characteristics recode of acbg15 (gen\years principal altogether) 4.92 4.67 6797 7.07 7.55 7622 recode of acbg17 (gen\highest level of formal education) 1.64 0.56 6766 1.80 0.52 7306 Digital recode of acbg08 (gen\access to digital books) 0.37 0.48 6262 0.63 0.48 7459 recode of acbg09 (gen\total number computers) 7.59 11.04 6355 11.00 17.81 7687 18 VII. References Ammermueller, A. 2007. PISA: What makes the difference? Empirical Economics 33(2): 263. Barrera-Osorio, F., Garcia-Moreno, V., Patrinos, H.A. and Porta, E. 2011. Using the Oaxaca- Blinder decomposition technique to analyze learning outcomes changes over time: an application to Indonesia's results in PISA mathematics. Regional and Sectoral Studies 11(3): 65-84. Blinder, A. 1973. Wage discrimination: Reduced form and structural estimates. Journal of Human Resources 8(4): 436–455. Jakubowski, M., T. Gajderowicz and H.A. Patrinos. (2023). Global Learning Loss in Student Achievement: First Estimates using Comparable Reading Scores. Economics Letters 232. https://doi.org/10.1016/j.econlet.2023.111313. Kitagawa, E. M. 1955. Components of a difference between two rates. Journal of the American Statistical Association 50: 1168–1194. Oaxaca, R. 1973. Male-female wages differentials in urban labor markets. International Economic Review 14(3): 693–709. 19 ABSTRACT Amidst global COVID-19 school closures, student outcomes varied worldwide, but usually declined. Egypt notably de ed this trend, showing signi cant improvement in reading scores. Utilizing data from the Progress in International Reading Literacy Study (PIRLS), one observes a remarkable 48-point increase from 2016 to 2021, akin to two years' academic growth. Gender disparities narrowed, with boys enhancing scores notably. The increase is con rmed using a model that analyzes the change in student achievement in 2021 by predicting the deviation of the most recent 2021 results from an estimated linear trend in reading achievement using data from all PIRLS rounds. Employing the Oaxaca-Blinder technique, reading gaps are decomposed. Notably, one-third of the progress stemmed from unexplained variables, hinting at possible policy reforms. Since most of the di erence is unobservable, then it is probably due to policy or system changes.