Noise pollution is a significant source of psychological stress that can adversely affect mental and physical health1–6. Prior studies have examined the harmful effects of noise on the auditory systems of both animals and humans, showing that noise pollution can alter auditory perception and behavior7–11. The World Health Organization (WHO) highlights noise pollution as a serious public health issue and qualifies it as the third most harmful factor to mental and physical health in large cities12–14. Because noise permeates the physical environment, it is not surprising that a great deal of research has explored the effect of noise on health outcomes15–18. Beyond health outcomes, studies have also examined the impact on economic and policy outcomes such as productivity, creativity, and welfare19–22.
However, most prior studies have either investigated the impact on adults or analyzed the context of developed countries23–25. Many cities across the developing world are facing rising levels of noise pollution owing to infrastructure development, such as the construction of buildings, bridges, roads, and railways. Governments in developing countries like India are installing noise-monitoring stations across major cities. Yet, there remains an acute absence of empirical evidence on the effects of noise pollution that rely on an analysis of longitudinal data from developing countries. This is an unfortunate oversight because many cognitive tasks are often performed in the presence of task-irrelevant environmental noise11,26. Although some studies have begun to investigate the impact of noise in developing countries, this limited research has relied on either small-sample data or on analyzing data on adults12,27. Empirical analysis examining noise’s effect on children’s academic performance in developing countries has been sparse.
Research on this topic, primarily investigating the context of developed countries, has produced mixed results. On the one hand, some studies suggest that noise can severely impair children’s cognitive abilities as it can cause an attentional burden and interfere with short-term memory22,28−30. On the other hand, some studies conducted in laboratory settings suggest that a moderate level of ambient noise is likely to enhance creative performance as it can activate abstract cognition19,31,32. Thus, there is no clear understanding of how noise may affect children’s academic performance in developed countries. Mixed evidence from developed countries, when combined with the fact that little research has investigated the context of developing countries facing rapid increases in noise pollution, makes the relationship between noise and learning in developing countries, a topic worthy of a scientific investigation using large-scale data.
In this short paper, we examine the association between noise pollution and academic performance by analytically leveraging the spatial-and-temporal variation in the noise pollution recorded by the noise monitoring stations located across major cities in India and the spatial-and-temporal variation across schools in the academic performance of students in Class 9–12 examinations. We obtained academic performance data from the District Information System for Education (DISE), maintained by the National Institute of Educational Planning and Administration. Our dataset contains information on the failure percentage of Class 9–12 examinations across a representative panel of schools for four years from 2014 to 2017. We construct the dependent variables Boys Class 12 Fail % and Girls Class 12 Fail % to measure the percentage of boys and girls who fail the Class 12 examination in a school s in the year t (see summary statistics in Appendix 1 in the online supplementary material; we also construct similar variables for Class 11, 10, and 9).
We obtained noise pollution data from India’s Central Pollution Control Board (CPCB). Our dataset contains noise levels data for every month from January 2014 to December 2017 from CPCB’s National Ambient Noise Monitoring Network (NANMN), which covers noise monitoring stations across major cities: Bengaluru, Chennai, Delhi, Hyderabad, Lucknow, and Mumbai (Appendix 2 shows the list of the noise monitoring stations in our sample). By definition, any unwanted sound is called noise. Sound travels as an oscillation of pressure. The vibrations of pressures within a frequency range stimulate sensation in the ears19. Hearing is thus sensitive to the noise level, measured in decibels (dB). As per the WHO guidelines and the EU definition, many people are seriously annoyed by noise levels above 55 dB30. At an average daily exposure of above 55 dB, noise can lead to severe health effects such as increased stress hormones and the risk of cardiovascular and metabolic diseases2,33. Recent research, conducted in Spain, shows that children attending schools exposed to noise levels above 55 dB during daytime revealed greater inattentiveness than those exposed to lower noise levels30. Therefore, we construct the independent variable Noisy Days % (Jan-Mar Quarter in Calendar Year t) to measure the percentage of days with high noise (i.e., above 55 dB) in the January-March quarter of calendar year t as recorded by the nearest noise monitoring station of a school. School-level examinations in India are typically conducted during the January-March quarter. We also construct variables to measure Noisy Days % for the other three quarters and each of the 12 months.
To examine the association between noise pollution and the academic performance of Class 12 boys, we use a Poisson Pseudo Maximum Likelihood (PPML) regression model because the distribution of the dependent variable Class 12 Boys Fail % has several zeros and the PPML estimator does not make any assumptions about the distribution of the non-negative dependent variable34,35. The distributions of the key variables are shown in Appendix 3–7. For ease of interpretation, we also conduct additional analyses using the Ordinary Least Squares (OLS) regression model. We cluster standard errors at the school level. In our estimation (Eq. (1)), the unit of analysis is school-year. We control for time-invariant school characteristics and economy-wide common shocks. We also control for time-varying characteristics such as class size and school size. Our analysis suggests that an increase in Noisy Days % (in Jan-Mar Quarter in Calendar Year t) is associated with an increase in Class 12 Boys Fail % (β = 0.0059; P = 0.050; 95% CI (0.0001, 0.0117)), see Column 4 of Table 1. Figure 1a reveals that a ten percentage point increase in Noisy Days % (in Jan-Mar Quarter in Calendar Year t) is associated with an increase of 6.1 percent in Class 12 Boys Fail %. The results remain qualitatively similar when we use OLS instead of the PPML model. We do not find any association between Noisy Days % (in Jan-Mar Quarter in Calendar Year t) and Class 12 Girls Fail % (β = -0.0023; P = 0.621; 95% CI (-0.0113, 0.0068)); see Column 4 (and also Column 8) of Table 2 and Fig. 1b.
[Insert Fig. 1 and Tables 1–2 here]
As a falsification test, we investigate the relationship between noise pollution and change in enrollment at both class and school levels (see Appendix 8 for the distributions of change in enrollment in Class 12 and in school). We do not find any association between Noisy Days % (in Jan-Mar Quarter in Calendar Year t) and change in enrollment in either Class 12 (see Columns 1–2 of Appendix 9) or school (see Columns 3–4 of Appendix 9). As another falsification test, we investigate the effect of noise pollution recorded in quarters away from examination days on the academic performance of Class 12 boys and girls. We do not find any association between Noisy Days % (in Oct-Dec quarter in Calendar Year t – 1) and Class 12 Boys Fail % (see Column 1 of <link rid="Sec12">Appendix 10</link>) or Class 12 Girls Fail % (see Column 4 of <link rid="Sec12">Appendix 10</link>). Similarly, we do not find any effect of noisy days % of the other two quarters, i.e., Oct-Dec quarter (in calendar year t – 1) and Jul-Sep quarter (in calendar year t – 1).
Figure 2 and Table 3 show the results of the analysis examining the relationship between monthly noise and the academic performance of Class 12 boys. We find that increases in Noisy Days % in March, February, and January in calendar year t are associated with an increase in Class 12 Boys Fail % (see Columns 1–3 of Table 3 and Fig. 2a), whereas no such association is observed with Class 12 Girls % (see Columns 1–3 of Table 4 and Fig. 2b). We find no effect of Noisy Days % in the other nine months in calendar year t – 1 on Class 12 Boys Fail % (see Columns 4–12 of Table 3 and Fig. 2a) or on Class 12 Girls Fail % (see Columns 4–12 of Table 4 and Fig. 2b).
[Insert Fig. 2 and Tables 3–4 here]
We also investigate the relationship between noise pollution and academic performance of Class 12 boys and girls by varying proximity of schools from their respective nearest noise monitoring stations. Figure 3 and Table 5 show the results. Figure 3a shows that the strongest effect of Noisy Days % (in Jan-Mar Quarter in Calendar Year t) on Class 12 Boys Fail % is observed for schools located within 1 km of the nearest noise monitoring station. Column 3 of Table 5 suggests that a ten percentage point increase in Noisy Days % (in Jan-Mar Quarter in Calendar Year t) is associated with an increase of 11.3 percent in Class 12 Boys Fail % for schools located within 1 km of the nearest noise monitoring station. Column 6 of Table 5 suggests that for schools located within 1 km of the nearest noise monitoring station, an increase in Noisy Days % (in Jan-Mar Quarter in Calendar Year t) is also associated with an increase in Class 12 Girls Fail %, but Fig. 3b importantly reveals that this estimate does not have high precision.
[Insert Fig. 3 and Table 5 here]
In an additional analysis, we also investigate the relationship between noise pollution and the academic performance of Class 11, Class 10, and Class 9 boys and girls for schools lying within 1 km distance from their nearest noise monitoring stations (see <link rid="Sec13">Appendix 11</link>). <link rid="Sec13">Appendix 11</link> reveals no association between noise and academic performance of Class 11 and Class 9 boys and girls, respectively (see Columns 1, 3, 4, and 6). However, Column 2 of <link rid="Sec13">Appendix 11</link> suggests that an increase in Noisy Days % (in Jan-Mar Quarter in Calendar Year t) is associated with an increase in Class 10 Boys Fail % (β = 0.0423; P = 0.025; 95% CI (0.0053, 0.0794)) for schools located within 1 km of the nearest noise monitoring station. Column 5 of <link rid="Sec13">Appendix 11</link> suggests that for schools located within 1 km of the nearest noise monitoring station, an increase in Noisy Days % (in Jan-Mar Quarter in Calendar Year t) is also associated with an increase in Class 10 Girls Fail % (β = 0.0616; P = 0.080; 95% CI (-0.0072, 0.1306)), but these estimates for girls do not have high precision.
We now turn to the mechanisms that could generate an adverse effect of noise pollution on academic performance. In other words, why are noise pollution and academic performance negatively related? We explore three potential mechanisms and discuss their empirical validity; note that our regression analyses include school-fixed effects, class size, and school size, which rule out many alternative explanations that rely on unobserved school-specific characteristics or unobserved school-and-time varying characteristics that are likely correlated with school size.
Mechanism 1 (Cognitive burden explanation). Because attention matters more when the cognitive burden is high, noise pollution hurts academic performance.
The logic of this mechanism is that when the cognitive burden is high, attention plays a much more crucial role in academic performance. A natural implication of this mechanism is that the negative effect of noise pollution should be stronger when the cognitive burden is higher. Several studies conducted in laboratory settings have shown that cognition can be impaired by noise22–24. Cognitive function includes the ability to direct one’s attention and manipulate information in memory. Therefore, a plausible channel through which noise adversely affects learning outcomes is via the effect of noise on the impairment of cognitive function.
Taken together, our results (see Table 1 and <link rid="Sec13">Appendix 11</link>) reveal that the negative effect of noise pollution is observed for Class 12 and 10 exams but not for Class 11 and 9 exams. The performance in Class 12 and 10 exams places a high cognitive burden on students in India because these exams are cumulative36. Unlike Class 11 and 9 exams, Class 12 and 10 exams have substantial implications on subsequent entry into colleges and competitive entrance examinations. The performance in Class 12 and 10 exams matters substantially because these exams have enormous cumulative implications with a potential to affect life-long outcomes.
Thus, the cognitive burden on students due to Class 12 and 10 exams is high. That the adverse effect of noise pollution on academic performance is observed only in months closer to examination days (Table 3) further reinforces the idea that noise pollution hurts academic performance through its effect on cognitive function. Thus, the overall evidence we find in our data is consistent with Mechanism 1 (cognitive burden explanation).
Mechanism 2 (School selection explanation). Because poorer quality schools may select into high-noise locations, noise pollution may show a negative effect on academic performance.
This mechanism contends that noise pollution and academic performance show a negative relationship because of schools’ sorting into different kinds of locations. In particular, poorer quality schools may self-select into higher noise locations for various reasons (e.g., because such locations may be cheaper). Our research design, however, rules out this possibility. The inclusion of school fixed-effects in our regression analyses rules out the confounding effect of any unobserved school-specific time-invariant characteristic (e.g., school quality). However, one may argue that the time-varying component of school quality is still unobserved and not ruled out by the inclusion of school-fixed effects. Thus, an alternative explanation exists that the changes in school quality over time are related to both the changes in noise pollution in the location of the school and the changes in academic performance over time. However, the results documented earlier (Appendix 9) do not show any evidence indicating that an increase in noise pollution is associated with a decrease (or an increase) in enrollment at the school level. This suggests that noise pollution and school enrollment follow independent dynamics over time. The evidence we find in our data is thus inconsistent with Mechanism 2 (school selection explanation).
Mechanism 3 (Economic growth explanation). Higher economic growth may lead to higher noise pollution, and higher economic growth may also lead to attention depletion on its own. Thus, economic growth is an omitted variable driving the observed negative effect of noise pollution on academic performance.
Changes in noise pollution and academic performance could both be simultaneously driven by economic growth. Faster-growing cities may witness a greater increase in noise pollution. Faster-growing cities may also offer students more channels for distraction, which could result in poorer academic performance. Thus, economic growth could be an omitted variable that drives the negative effect of noise pollution on academic performance. We test this prediction empirically. We classify the cities in our dataset into higher versus lower economic growth. Higher economic growth cities in our dataset include Bengaluru, Hyderabad, and Chennai, as they are among the top 10 fastest-growing economies globally as reported by Oxford Economics research institute37. The remaining three cities Delhi, Mumbai, and Lucknow, are not among the top 10 fastest-growing economies; thus, these are classified as lower economic growth cities. As <link rid="Sec14">Appendix 12</link> shows, the coefficient estimate of the interaction between noise pollution and higher economic growth (relative to lower economic growth) is materially small and statistically indistinguishable from zero in both PPML and OLS regression models. That is, the negative effect of noise pollution on academic performance remains similar for schools located in higher versus lower economic growth locations. Thus, the evidence is inconsistent with Mechanism 3 (economic growth explanation).
In conclusion, we examined the relationship between noise and learning using unique data on noise pollution and student performance from a large developing country. Our analysis, examining the spatial-and-temporal variation in noise pollution and student performance, shows that noise pollution matters: An increase in noise pollution is associated with a deterioration in academic performance. We ask why noise pollution harms student performance. Our analyses reveal results consistent with the cognitive burden explanation. When the cognitive burden is high, noise pollution can hurt academic performance, given its adverse effect on attention. The finding that the association between noise pollution and academic performance is not observed in months away from examination days rules out the possibility of school selection or economic growth mechanisms. Our findings also suggest that the adverse effect of noise pollution may not be gender-neutral. While prior studies conducted in laboratory settings have shown that noise can easily impair cognitive function24,38, it is important for future research to investigate whether the adverse effect of noise on cognitive functions may vary across gender in school and work settings.
Our findings have important implications. To improve academic performance, policymakers may design rigorous policies to check noise pollution. Such efforts are likely to be particularly valuable in locations with high noise levels; for example, most cities across developing countries are going through major infrastructure development with the construction of buildings, roads, and railways, which create noise pollution. Our findings also have implications for contributing to cumulative knowledge building on the question of who might suffer more from noise pollution. Our study underscores the idea that policymakers need to consider the heterogeneity in the impact of noise on the cognitive function and development of different kinds of populations. We hope our study inspires further research on noise pollution, which is widely prevalent globally but remains underexamined in large-sample empirical research.