How Free is Free Education? The Impact of Tuition on Exam Marks

Photo courtesy Rose Sri Lanka

Sri Lanka’s Free Education Scheme aims to provide each child an equal opportunity at educational success, but there is debate over to what extent it succeeds in this endeavor. Notably, some critics are concerned that private spending on education (for private and international schools, universities, and tuition) may be diluting the Free Education Scheme’s ability to provide equal opportunities to all. My doctoral dissertation is composed of three studies exploring these issues. In my first study, summarized here, I found evidence that educational stratification grew stronger in Sri Lanka from 1985 to 2010. In my second study, summarized here, I found it is unlikely that private and international schools or private universities are causing educational stratification, but it is possible that tuition is an important cause. This article summarizes the final study, estimating the impact of tuition on exam marks.


Private tuition has been documented worldwide extensively, but it is of special concern in the Sri Lankan context. Local researchers have documented the overwhelming prevalence of tuition, while a steady stream of newspaper editorials bemoan its potential to cause educational stratification. Until we know how effective tuition is at raising exam marks it is not clear to what extent private tuition causes stratification or whether other dynamics, such as uneven distribution of educational resources and teacher quality, are more important causal factors.

Sri Lankan Context

Sri Lanka achieved substantial successes in providing mass education, closing the gender gap, and accomplishing near-universal adult literacy earlier than many other developing nations. Early in Sri Lanka’s independence the state invested heavily in schools spread broadly across the island and mandated that state schools would not charge fees for primary, secondary, or tertiary education. As in many developing nations, however, Sri Lanka’s welfare state changed substantially when economic policies opened the economy in a series of reforms starting in 1977. Critics of the education system argue that because of these reforms, state investments in the education system have decayed and new regulations have allowed private educational providers to attract socially advantaged students away from the state system. In the logic of such critics, because private and ‘international’ schools, private universities, or private tuition meet advantaged families’ educational needs, these politically powerful families do not pressure the state to improve the public system.

The extended conflict between the government of Sri Lanka and the Liberation Tigers of Tamil Eelam (1983 to 2009) may also have implications for learning in schools and in tuition classes. Diversion of government resources from education to military spending, the emigration of educated professionals, particularly ethnic Tamils, and the conflict’s violence and social disruptions in the Northern and Eastern provinces all likely had a negative impact on educational processes, in and out of school, especially for ethnic minority children. As the armed conflict ended in 2009, it is an ideal time to investigate the possibility of ethnic differences in learning.

Methodological challenges in estimating the impact of tuition on performance

As with many social phenomena, it is challenging to estimate the impact of tuition on exam performance because children who attend tuition are different from those who do not. In general children who attend tuition are more advantaged than children who didn’t attend tuition: their parents are better educated; their households have higher expenditures and help slightly more with homework. Their schools are more frequently located in more urban settings and they attend higher grade schools. Notably, their exam performance prior to attending tuition is dramatically higher. Because of these initial differences, we cannot simply compare post-tuition exam marks to estimate the impact of tuition on performance.

In my study I use statistical analysis to “control” for these initial differences.[i] In effect, for each student who attended tuition I find a similar student who did not attend tuition: one whose exam marks were similar before tuition, whose parents have a similar level of education, whose home environment supports learning to a similar degree, whose school is of a similar quality, etc. Once I have identified the similar student, I can compare the students’ marks for an exam they took after one attended tuition and the other did not. I conduct this matching and comparison process for all the children in my data. Depending on how tuition impacts exam marks, across all these comparisons I may find that the students who attended tuition got higher or lower marks than those that did not attend tuition.

I also explore the possibility that different kinds of tuition may help students’ exam marks more or less, and the possibility that certain groups of students benefit more from tuition than others. I conduct this analysis because there is great variation in the kinds of tuition offered in Sri Lanka. This is an exploratory analysis, providing suggestive evidence only to inform further research.


Once I statistically control for the initial differences between students who attend tutoring and those who do not, I find no difference in their exam marks of the two groups after tuition. This means that on average, tuition has no impact on exam marks, at least for Year 5 students. This result is striking, given the large amounts of time and money students and their families are dedicating to tuition.

Next, I explored whether the impact of tuition varies for different kinds of tuition or for different groups of students. One notable pattern I found is that impact estimates for socially advantaged groups tend to be higher than for disadvantaged groups. This trend obtains for categories of parental education, household expenditure, average school achievement, and school facilities. Similarly, impact estimates for more expensive tuition are higher than estimates for cheaper tuition, and impact estimates of private classes are higher than for classes given in school. In both cases, more advantaged groups tend to take classes whose impact estimates are higher. This analysis provides strong suggestive evidence that advantaged children get a more positive impact out of tuition than disadvantaged children.

The differences in impact estimates by ethnicity are also interesting. For both math and language, the impact of tuition for Tamil[ii] and Muslim children is worse than for Sinhala children. Because the data I used were collected in 2003 I thought this result might be driven by the impact of the ethnic conflict on educational processes. However, even when I removed data from the Northern and Eastern provinces from my analysis the finding still held. This finding is more remarkable because the incidence of tuition is higher for Tamil children (77%) and Muslim children (71%) than for Sinhala children (61%).

Other results of my exploratory analysis suggest what types of tuition are most helpful. Mass classes appear to be more helpful than group classes, and classes given in private tutories appear to be more helpful than those given in schools. Finally, for both subjects, the impact estimates by the number of weekly hours of tuition follow the same pattern: the lowest impact comes from attending more than six hours of tuition per week, a middling impact comes from attending less than three hours per week, and the highest impact comes from attending three to six hours per week.

Discussion and Conclusions

While my study cannot definitively say that tuition doesn’t help students’ exam marks, it raises many questions. On the strength of this analysis and psychologists’ opinions that tuition may be harmful for young children, Sri Lankan parents of primary school children may question whether tuition is worth the time, expense, and stress for children. The findings of my exploratory study may also provide guidance for such decisions, insofar as they confirm parents’ existing doubts about the effectiveness of their child’s tuition: group classes, in-school classes, and more than six hours per week of tuition may be less effective than other choices. Similarly, government or NGO programs targeting disadvantaged communities may want to encourage or subsidize, other after-school activities, more developmentally appropriate for young children instead of tuition. If they do support tuition, they should carefully consider what kind of tuition is most likely to positively impact children’s performance, perhaps first investigating this question through research.

In considering these results and suggestions, the reader should note that this study considers only the impact of tuition for Year Five students. Given the different kinds of skills that are examined at higher levels and the developmental differences between young children and teenagers, it is quite possible that tuition has a different impact on Ordinary and Advanced Level performance. However, further research should study this question.

Although there is no difference in the exam marks of students with and without tuition, there are differences in impact estimates between advantaged tuition and disadvantaged tuition which could potentially be a cause of educational stratification in Sri Lanka. This finding suggests tuition is one mechanism by which advantaged children are learning more than their disadvantaged counterparts. However, the differences in impact estimates associated with measures of socioeconomic advantage are small to moderate in size compared to most educational interventions. The overall zero impact and the moderate magnitude of differences in impact estimates by socioeconomic status suggest it is unlikely that tuition plays a strong role in causing educational stratification in Sri Lanka. Researchers interested in the causes of stratification should explore other potential mechanisms, such as the unequal distribution of school resources and teacher quality.

Likewise, my finding that Sinhala children benefit more from tuition than Tamil and Muslim children could have important equity implications. The estimated differences associated with ethnicity are somewhat larger and several differences are statistically significant. Along with ethnic differences in educational attainment and in household expenditure on education, these findings raise questions about the quality of instruction available to ethnic minorities in Sri Lanka. If ethnic minority families find it difficult to find high quality instruction both in schools and in tuition, future conflict could arise.

My investigation of the possibility of differences in impact—by both socioeconomic advantage and ethnicity—must be seen as purely exploratory. I cannot recommend government policy based on these tentative findings, but further research to more rigorously test these dynamics is a good idea. Especially in the aftermath of a prolonged ethnic conflict, sparked in part by differences in educational opportunities, these tentative findings are worthy of further exploration.


[i] I conducted a propensity score analysis, using a rich array of control variables collected for the National Education Commission’s Inter-Sectoral Study on Health and Education (2005) and academic performance data from the National Education Research and Evaluation Center’s assessment of Year 4 skills. I used these controls to model performance on the Year Five Scholarship Exam, taken 5 months after the NEREC test. Full details of the analysis are available on request.

[ii] Unfortunately the NEC data did not distinguish between Tamil and Up Country Tamil ethnic groups.