Are parents justified in spending on low-cost private schools?

In this paper, I explore the preference for private schools among parents in developing countries, foregoing the option of free schooling, in many cases. I delve into the rationality of this decision in the face of mixed evidence suggesting the efficacy of private schools in improving learning outcomes.

The case of Indian parents is illustrative, as evidenced by the decreasing trend in public school enrollment and booming private schools enrollment. A recent paper by Karthik Muralidharan and Venkatesh Sundararaman finds that learning outcomes are approximately the same across groups of kids randomly given scholarships to attend private schools and those that attended a public school. Given this, why would parents choose to allocate a significant portion of their household budget on private schools when they can obtain the same learning outcomes for their children from free public schools?

I’ll be looking at this through the lens of “Judgement”, specifically how hard is it to be Bayesian about the quality of schools (what the appropriate signal is and how much noise is mixed in?) and also does “availability” play a role in parent choices.

Do private schools provide better education than public schools?

An interesting trend in India over the past two decades has been the shift from public schools to private schools — many of them low cost. Data gleaned from ASER reports show that 29% of children between the ages of 6 and 14 in rural India attended fee-charging private schools in 2013 compared to 18.7% in 2006, indicating that private schools were growing very rapidly (ASER 2013). The majority of these schools are low-cost private schools which charge a modest fee that is still a significant portion of the household budget (Tooley 2009). The alternative is to send children to public schools which provide amenities like free lunch and free school supplies in addition to being completely free.

Muralidharan et al present experimental evidence on “the impact of a school choice program in the Indian state of Andhra Pradesh that provided students with a voucher to finance attending a private school of their choice.” AP is the fifth most populous state in India, with a population of 85 million (70% rural). Recent estimates suggest that over 35% of students in rural AP are enrolled in private schools, compared with an all-India average of 28% (ASER 2013).

After two and four years of the program, the authors find no difference between test scores of lottery winners and losers on Telugu (native language), Math, English, and Science/Social Studies, suggesting that the large cross-sectional differences in test scores across public and private schools mostly reflect omitted variables. Private schools, however, are much more cost-efficient, delivering the same outcomes for children at one-third the cost but this has no bearing on the decision matrix for parents.

Standard economic theory would suggest that, in light of negligible differences in learning outcomes for public schools and private schools, parents should choose to keep their money and send their children to public schools. But the increasing enrollment in low-cost private schools belies this, indicating a possible breakdown in rational decision-making.

How easy is judgement?

Let’s say judgement is required in making a decision about a new practice or a new situation that you have observational data for from your peers. A useful framework to model this is the Target Input model (Jovanovic & Nyarko 1996, Foster & Rosenzweig 2001) where optimal input use in any given year varies and this directly affects the profits. Over the course of several such periods, mental models of how much inputs to use are updated — by observing your own situation as well as of those around you.

One of the challenges with judgement, and learning is a Bayesian manner, is poor diagnostic information — how does one determine what distinguishes the noise from the signal to understand what’s a success and what’s a failure, especially in situations where networks of communication are weak. As Bjorkman-Nyqvist et al find with regards to the prevalence of fake anti-malarials in Uganda, achieving desirable equilibria in the presence of “bad apples” in the market is not so straightforward. Generalizing from the anti-malarial case, learning is hampered by: a) the wrong (subjective) model to arrive at a decision b) absence of accurate diagnostic information c) difficulty of distinguishing success from failure especially when it isn’t clear what success and failure is. These observations suggest that there exists a scope for actors to facilitate interventions that make it easy for individuals or communities to learn “correctly” and make judgements that are reflect decisions they’d have made if they possessed the correct models.

Judgement in the Real World

Several studies document the difficulty of learning. Udry et al find that pineapple farmers respond to both signal and noise — especially when farmers are novices. This stems from a model of social learning where farmers adjust the level of their inputs based on observations they make of outputs around them. Hanna et al find that seaweed farmers ignore particularly important dimensions that could maximize yield and fail to optimize for them. Cases of parents foregoing providing medicine to sick children and instead approaching the local healer abound — partially because learning about the right approach is difficult. Rockoff et al, in an experiment to understand how employers learn about worker productivity through an experiment involving teachers, find that a) the correlation between performance estimates and prior beliefs rises with more precise objective estimates and more precise subjective priors b) New information exerts greater influence on posterior beliefs when it is more precise and when priors are less precise. c) Employer learning affects job separation and productivity in schools, increasing turnover for teachers with low performance estimates and producing small test score improvements.

Bayesian Learning in a Schools Context

Choosing a school for its kids is one of the most important decisions parents have to make, assuming they care about the welfare of their children. In countries with reasonably bad public schools and a proliferation of low-cost schools, this can be a cognitively taxing process for households in the lower quintiles of income (perhaps even higher quintiles but that’s beside the point). The presence of private schools — while adding, perhaps, a desirable option of school choice — also implies the necessity of ‘judgement’ to maximize utility for the child. A good model to think about this issue is to think about this in terms of Bayesian Learning.

Let’s define belief about the quality of schools in the neighborhood as a continuously distributed variable, Y, where: Y = x + e where x is the prior and e is the noise associated with the signal and both are normally distributed. Both the variables on the right are of interest, in the sense that, in the context of school choice, how do parents formulate their beliefs about the quality of schools and what signals the quality of a school including how much noise is mixed in?

Carneiro et al find that a one-standard deviation increase in the quality of basic facilities is valued at about 210 rupees, which corresponds to an increase of the average school fee by around 45 per cent. In terms of distance, a one standard deviation decrease (on average an 800 meter decrease) is valued at about 180 rupees. It could be that the quality of basic facilities is a given preference for parents, but it could certainly be the case that basic facilities are viewed as a proxy for a school that does better on learning, which may or may not be true. According to Yaacob et al, parents place emphasis on the importance of private schools’ syllabus, schools’ environment and facilities when selecting to enroll their children in private schools. The academic performance of the school was placed third in preference, with fourth factor considered being the quality teachers that the school possessed.

Admission into private schools could be a case of paying a premium for entry into a club- Larkin (2009) found that salespeople in a U.S. company were willing to trade off approximately $30,000 in income to achieve membership in the firm’s “club” for top performers — the benefits of which were a gold star on their name card, companywide recognition, and an e-mail from the chief executive officer. It is plausible that parents are paying for the social capital that could be accrued by sending their kids to private schools.

Going back to the target-input model, remember that it depends on agents updating their beliefs based on observations of self and others and how much “off” of the profit target the consumer is. With schools, because of weak feedback loops, it might be difficult for parents to confirm if “profits” (say, learning outcomes) have been attained, even in the presence of school administered tests. This exacerbates the judgement problem in choosing to allocate money for private schools. The implications of this model are that likelihood of new ideas adoption increases upon observations and the usage patterns move in the direction of “successful” peers. In the case of school choice, these implications could mean that parents choose private schools upon observing other children attending private schools doing well even though it isn’t clear if the success of children attending private schools is attributable to the schools themselves. This could have the adverse effect of parents allocating a portion of their income based on incomplete information or information which has a lot of noise in it.

Assessing need for intervention and potential intervention

Assuming the state or other actors are interested in maximizing the quality of education for each child, there is a case to be made for making sure that parents are equipped to make optimal choices for their children. A quick diagnostic check to evaluate if an intervention is actually required could be done using routine administrative data that is collected in India. Data on inputs in schools is collected by DISE (District Information System for Education). A spatial analysis that overlays school enrollment on school facilities for both private and public schools and analyses if parents are sending their kids to private schools when there exist public schools in the vicinity can help establish if low-income parents are choosing to send their kids to private schools even when they have cheaper alternatives. If learning assessments could be conducted across schools (as they are likely to happen in the near future), looking at learning outcomes across schools controlling for household income, since consumption data already exists, could give a good sense of parents are making the right decisions.

A low-cost policy intervention that could be piloted is the dissemination of school report cards consisting of information on school test scores and measures related to school inputs. Andrabi et al find that providing information about school and child test scores increased subsequent test scores for children whose families were provided this information. This should be a relatively cost-effective intervention that requires the regulatory body to conduct low-cost assessments and subsequently disseminate that information, perhaps through mobile phones.

Conclusion

It is unclear if allocation from already tight household budgets towards private schools maximizes welfare for large sections of poor households. It is even possible that there is a breakdown in Bayesian learning which is leading to these sub-optimal outcomes. Interventions targeted towards improving “learning” for parents in order to make better decisions for their children would be desirable in such a situation.

References:

Andrabi, Tahir, Jishnu Das, and Asim Ijaz Khwaja. “Report cards: The impact of providing school and child test scores on educational markets.” (2014).

Banerji, Rukmini, Suman Bhattacharjea, and Wilima Wadhwa. “The annual status of education report (ASER).” Research in Comparative and International Education 8.3 (2013): 387–396.

Bast, Joseph L., and Herbert J. Walberg. “Can parents choose the best schools for their children?.” Economics of Education Review 23.4 (2004): 431–440.

Bjorkman Nyqvist, Martina, Jakob Svensson, and David Yanagizawa-Drott. “Can Good Products Drive Out Bad? Evidence from Local Markets for (Fake?) Antimalarial Medicine in Uganda.” (2012).

Carneiro, Pedro, Jishnu Das, and Hugo Reis. “Estimating the demand for school attributes in Pakistan.” Progress. 2010.

Foster, Andrew D., and Mark R. Rosenzweig. “Learning by doing and learning from others: Human capital and technical change in agriculture.”Journal of political Economy (1995): 1176–1209.

Jovanovic, Boyan, and Yaw Nyarko. Learning By Doing and the Choice of Technology. No. w4739. National Bureau of Economic Research, 1994.

Muralidharan, Karthik, and Venkatesh Sundararaman. The Aggregate Effect of School Choice: Evidence from a two-stage experiment in India. No. w19441. National Bureau of Economic Research, 2013.

Rockoff, Jonah E., et al. “Information and employee evaluation: Evidence from a randomized intervention in public schools.” The American Economic Review 102.7 (2012): 3184–3213.

Tooley, James. The beautiful tree. Washington, DC: Cato Institute, 2009.

Yaacob, Noor Alyani, Mariana Mohamed Osman, and Syahirah Bachok. “Factors influencing parents’ decision in choosing private schools.”Procedia-Social and Behavioral Sciences 153 (2014): 242–253.