Our research examines these questions using data from Pakistan featured in the Learning and Educational Achievements in Punjab Schools report between 2003 and 2007. These data cover all 820 schools—public and private—in 112 villages in the province of Punjab, Pakistan, and contain over 71,000 test scores of children tested in English, mathematics, and Urdu (the vernacular) every year between the third and sixth grades. Using these data, we constructed a measure of school quality—value-added—defined as the predicted average test-score gains in English, mathematics, and Urdu that a randomly selected child will experience if enrolled in that school. Multiple analyses confirm that this measure reflects the quality of the school and not the preexisting characteristics of children enrolled in the school.
Our research has produced several findings. First, school quality varied substantially, even within villages. The students we studied increased their test scores by 0.39 standard deviations (SD) on average every year. But if those children had attended the best school in their village compared with their current school, they would have increased their test scores by 0.24 SD more—equivalent to about half a school year of additional learning for the average student. That effect, accumulated over five years of primary schooling, produced a test-score gap similar to the gap between students in high- and low-income countries and the gap between black and white students in the United States. Furthermore, those children would have increased their test scores by 0.43 SD more if they were in the best school compared with the worst school in their village.
Differences between public and private schools drove some of this variation, but there was also considerable variation within these sectors. An increase of 1.00 SD in value-added improved average test scores by 0.32 SD in public schools and 0.21 SD in private schools. The gains from moving to a public school that was 1.00 SD higher in the value-added distribution were therefore similar to three-quarters of a school year, while the gains from moving to a private school that was 1.00 SD higher were equivalent to roughly half a school year.
Second, attending an average-quality private school increased average yearly test score gains by 0.15 SD relative to attending an average-quality public school. However, the substantial differences in value-added between schools in the same sector indicate multiple possible estimates of the so-called private school premium. For instance, moving all children in public schools to the worst private school in the same village would have increased their average test scores by 0.08 SD. However, the precise effect varies between children. Children at the 10th percentile of the distribution of test score effects would have experienced a substantial loss of 0.22 SD. On the other hand, moving all children in public schools to the best local private school would have increased their test scores by more than 0.25 SD, and the losses at the 10th percentile of the distribution would have been only 0.06 SD—this loss applies to children who would have moved from better-performing public schools. More-targeted policies could produce larger gains; for instance, moving children from the worst public school to the best private school in their village would have increased their test scores by 0.38 SD on average. This range of effects demonstrates that estimating a single private premium can be misleading and will only be valid for a specific reallocation of children.
Third, our findings suggest that parents respond to school quality. This is especially pronounced in private schools, where prices—which reflect parents’ valuations of schools—are strongly correlated with value-added. Furthermore, a private school with value-added that was 1.00 SD higher in student test scores gained 6 percentage points in market share and was 65 percentage points less likely to shut down by the 2011–2012 school year. Additionally, the correlation between private schools’ value-added (calculated using data from 2003 to 2007) and test scores in 2011 is high despite significant turnover among teachers in private schools between 2003 and 2011. This suggests that school owners work to preserve their schools’ quality and market position. By 2011, public schools with higher value-added gained a greater market share. However, though schools’ value-added persisted over time, the correlation between school closures and value-added was weak—the 12 percent of public schools that closed between 2003 and 2011 had low enrollments but not low value-added.
Fourth, households with a high socioeconomic status—measured by parental education and household assets—were more likely to enroll their children in schools with high value-added. Although this finding may seem driven by the fact that households with high socioeconomic status are more likely to live near schools with high value-added, the relationship holds even within villages and sectors. This variation in parents’ responsiveness to quality based on socioeconomic status partly explains why the average value-added of schools differs between villages. Indeed, private schools had higher value-added on average in villages with higher socioeconomic status. Relative to a village at the 25th percentile in socioeconomic status, a village at the 75th percentile had an average private-sector value-added that was 0.05 SD higher. Markets also evolved in different ways in villages with low socioeconomic status than in villages with high socioeconomic status. Private schools in villages at the 75th percentile of the socioeconomic status distribution experienced gains in market share over time that were almost double those of private schools in villages at the 25th percentile. Additionally, the negative correlation between value-added and the likelihood of school closure was 50 percent higher.
Unlike the private sector, there is no evidence that the public sector responds to variation in parents’ demand for quality. There was no significant relationship between a village’s average socioeconomic status and public schools’ value-added or public schools’ likelihood of closure. This finding was not driven by all parents who responded to quality differences withdrawing their children from public school; there were families of high socioeconomic status attending public schools with higher value-added. Thus, although parents demand quality from the public sector, there is little evidence that the government responds to this demand. This lack of response implies that the quality of public schools remains relatively the same no matter the socioeconomic status of the village, while the quality of private schools is systematically higher in villages with high socioeconomic status.
Finally, even if the government does not respond to the demand for quality, parents can still send their children to high-value-added public schools within villages because all public schools in Pakistan are free and because parents can enroll their children in any (sex-segregated) public school. Strikingly, there was no relationship between enrollment and value-added in the public sector. Why do some low-value-added public schools remain large and some higher-value-added public schools remain small? Our findings suggest that parents have less agency in school choice than the number of public schools in each village would indicate. Specifically, low-value-added public schools can maintain a higher market share when they are in smaller markets that are unable to sustain a large private sector, when they are the only boys’ or girls’ school in the village, or when they are close to the village center—reflecting parents’ strong desire for proximate schools. Conversely, a higher-value-added public school can remain small if it is not the only boys’ or girls’ school in the village or if it is far from the village center. Therefore, parental agency is effectively constrained despite the appearance of a substantial number of options.
Overall, our findings demonstrate that the educational systems in emerging economies are increasingly characterized by variation in choice and quality, that parents recognize and demand quality, and that the private sector is more responsive to this demand than the public sector. These findings have important policy implications for various educational initiatives, including educational vouchers and direct interventions in public and private schools.
