Hunger can lead to lower cognition not only in the short term but sustained spells of hunger may affect the brain in the long run as well. Hunger in the classroom, in particular, can adversely affect the learning outcomes of children by lowering their effort and cognition during school hours.
The premise of school feeding programmes, which were first implemented in developed countries in the early 1960s, such as the School Breakfast Program in the US, was that students tend to come to school hungry. The programmes, therefore, sought to increase students’ concentration by easing hunger and improving cognitive functions during school hours. To the best of our knowledge, ours is the first study to assess the impact of a large-scale school meal programme on students’ performance in the classroom.
We examine the impact of school meals on students’ short-term performance in the classroom. We measure students’ cognition by their ability to solve maze puzzles administered in a classroom setting. We find that school meals enhanced students’ effort and cognition and those in treatment schools improved their performance in solving maze puzzles. Performance, assessed by the number of mazes solved correctly, increased by 13–16 per cent for treated students.
We design a novel school-based experiment in India to study whether the provision of meals to malnourished students during school hours can improve their performance in cognitive tasks in the classroom. Our data come from tests of effort and cognition we designed and administered to students in randomly selected public schools of Delhi. We conducted our tests before and after the extension of an ongoing free school meal programme to upper primary grades (6 to 8) in public schools in Delhi which were providing free cooked meals to primary grades (1 to 5), as part of a federal programme popularly known as Mid-Day Meal Scheme, since 2003.
The exact date of extension of the programme, 29th September 2009, was unanticipated. Since our study was initiated before the extension became effective in August 2009, schools whose randomly selected date of the first round of data collection (called baseline) fell before 29 September 2009 had not yet started the school meal scheme for upper primary grades, while those visited after 29th September were serving meals to upper primary grades. Sampled schools were re-visited for a second round of data collection (called endline) between February and April 2010 when all public schools were providing the meals in upper primary grades.
We define schools that changed their meal implementation status in the upper grades between the two visits as treatment schools and those that did not as control schools. Thus, children in Grade 7 in treatment schools were offered school meals at endline while those in control schools were receiving meals in both rounds. This allows us to use a double difference, intention-to-treat estimation strategy while accounting for unobserved heterogeneity in students’ ability. We use Grade 5 students who were receiving school meals throughout the study period (i.e. both rounds) as the placebo group.
We used maze puzzles as tests of students’ effort and cognition in the classroom. Thus our main outcome of interest is the number of maze puzzles solved correctly during regular school hours by a student in two sessions we administered during a school visit – a session before school meals are offered to students and a session after the distribution of school meals. These cognitive tests were administered by us again after a span of 4 to 6 months to students in all sampled schools after the extension of the programme at endline.
Besides being able to compare the performance of students between the two visits, we are also able to compare their performance before and after the distribution of school meals during recess on the same visit. Since the maze puzzles did not require either reading or writing skills but rather skills such as attention, perseverance and patience we are able to focus on the impact of the meal programme on classroom effort as opposed to standard tests of learning such as reading or math.
Our results show that the consumption of school meals improved students’ performance on non-curriculum related cognitive tasks that required attention and concentration in the classroom. We interpret these results as indicative of improved cognition in the short-run due to the supplementary meals attenuating fluctuations in food intake during school hours. Could school meals have led to some other impacts and thereby affected students’ performance on our assigned task? We explore possible changes in factors measured at the school or student level that may have impacted students’ performance, besides our proposed mechanism of classroom attention, in Table 1.
Table 1: Impact of school meals on student and school-level characteristics at endline
First, the availability of school meals may have improved the students’ average attendance rate, which in turn may have had an impact on classroom performance. Using data on the school level attendance rate in the difference-in-differences (DID) specification in column 1, we find an insignificant effect of treatment on average Grade 7 attendance rate. This result is in line with the existing literature which suggests that school meals improved attendance rates at lower grades (Grades 1-2) but not in higher grades in primary schools (Afridi, 2011). Moreover, our maze puzzles were not curriculum-based and therefore our measure of performance is unlikely to be affected by regular school attendance.
We may, however, be worried that the number of participants in our sessions in the section to which we administered the mazes may have fallen at endline, affecting performance positively in the treatment group due to fewer classroom disruptions. However, the number of participants in our classroom sessions were not significantly different between baseline and endline in the treatment schools vis-à-vis control (column 2, Table 1).
Furthermore, any classroom behavioural changes, such as reductions in disruptions because students may be more attentive after receiving supplementary nutrition, are unlikely to explain our findings given that the tests were administered in a relatively ‘controlled’ environment. Hence spillover effects through changes in programme uptake, do not appear to have impacted performance on our cognitive task. We also rule out gains in performance due to any systematic changes in administrative efficiency (column 3) between the two groups of schools*.
Second, did the school meals improve students’ health and thereby affected their classroom performance directly? In columns 4-5 the coefficient on the interaction term is insignificant, suggesting that there were no improvements in students’ height-for-age Z score or BMI-for-age Z score as a result of school meals. Note that the gaps between the baseline and endline varied between 4 to 6 months (including 1-2 months of school vacations). Neither was this a long enough period to result in improvements in long-term health or learning nor was our study intended to capture these outcomes.
Overall, our findings indicate that the smoothening of temporary variation in nutrient availability, rather than fundamental improvements health indicators or regular attendance, was a channel through which classroom effort improved. Not surprisingly, therefore, students did not substitute school meals with meals from home by reducing consumption of lunch from home during recess (column 6). Hence school meals did appear to have provided supplementary nutrition to the students. These results reinforce our interpretation that the students’ attention and concentration improved due to the alleviation of hunger in the classroom.
Table 2: Impact of school meals on math and language test scores at endline
We find suggestive evidence that even in the absence of improvements in school participation or health indicators, learning can improve due to the alleviation of fluctuations in food intake. Table 2 shows that the proportion of math questions correctly answered by Grade 7 students improved by 19.6 percentage points on the two questions that were common between rounds and 18.3 percentage points for all three questions asked at endline. We do not find any significant improvements in the language score although the point estimate on the DID term is positive. We interpret these results as suggestive since the math score of students in the treatment group was significantly lower at baseline. The estimated improvement in math score could, therefore, be biased upwards.
There are two possible mechanisms through which school meals can improve cognition. One is the health-nutrition channel- school meals lead to an improvement in health status and, thereby, cognition. The second is the hunger alleviation channel – school meals reduce hunger in classroom. Our study suggests that the increments we observed were due to the hunger alleviation mechanism. First, we see no improvement in students’ health outcomes in treatment schools compared to the control group. Second, we only see improvements in scores in the puzzle-solving sessions held after the meals were distributed on a school day. We see no significant gains in maze scores in sessions held before meal distribution, something we would have expected to see if this was driven by improved health status.
Students in our sample were receiving school meals for 6 to 8 months. This may be insufficient to cause improvements in health outcomes. Additionally, we do not have any information on the actual quantities of meals consumed by students to be able to assess if these are adequate for students of this age group. Perhaps it is not surprising then that we are unable to find any impact on stunted students’ performance even though these students are more likely to consume meals.
However, we are able to establish that consumption of school meals can have short term impacts on students’ effort through higher attention in the classroom which may be one of the mechanisms through which learning outcomes could improve (Chakraborty and Jayaraman, 2019) even without fundamental improvements in child health. Hence the longer-term effects of school meals on learning may be even larger.
*Note: We analysed the impact of programme extension to Grade 7 on the time spent on meal distribution during recess in column 3 of Table 1. We do not find a significant coefficient on the DID term, suggesting that schools’ administrative efficiency was not impacted due to the extension of the school meal programme. Instead, school meals were distributed to lower and upper primary grades simultaneously. If anything, treatment schools would have more administrative learning to do and so meal distribution efficiency or time may be longer in these schools which could adversely affect instruction time and thereby possibly classroom performance of students.
Farzana Afridi is Associate Professor, Economics and Planning Unit, Indian Statistical Institute. She is a PhD in Economics from the University of Michigan, US.
Bidisha Barooah is a researcher at the Delhi School of Economics, and Senior Evaluation Specialist, International Initiative for Impact Evaluation.
Rohini Somanathan is Professor at the Delhi School of Economics. She received her PhD from Boston University, US.
This is an edited extract from the authors’ paper Hunger and Performance in the Classroom published by IZA – Institute of Labor Economics. Read the full paper here.