Temperature and Teacher Absence : Evidence from 4,085 Schools in India

Abstract

Theory of human capital suggests that human capital is an asset composed of both education (schooling years) and talent, but also health and many other characteristics. The amount of talent that one is endowed is fixed, but education must be acquired through learning. Teachers should thus be one of the key roles in education, which in turn should determine one’s chances of succeeding in life. The economic production is commonly composed of capital and labour. This second component is also known as human capital and tends to be more efficient when the workers have had higher education. The level of high-skilled workers – people who had the opportunity of getting good education – is essential in understanding how human capital is related to the level of economic development, especially since human capital contributes to its own growth through innovation and technical change. As a result, several authors have attempted to identify different determinants that could affect human capital through teacher absence. For instance, Kremer et al. (2005) study teacher absence in India. They found that 25% of teachers were absent during unannounced visits to schools. In some states, absenteeism rates can even exceed 40%. Kremer et al. (2005) find no evidence that teacher absence is correlated with higher salary. However, they find that some incentives actually work: schools with better infrastructure contribute positively to teacher attendance. In addition, using a randomized experiment in India, Duflo et al. (2012) find that teachers react positively to financial incentives. More specifically, in their experiment, the group exposed to financial incentives was 21% less absent relative to the control group. The two above-mentioned studies show a certain importance of different determinants in the absenteeism of teachers. It is therefore interesting to look at other potential determinants. An interesting and still very under-studied candidate is temperature. We already know that high temperatures are associated with lower levels of output, especially in poor countries (Dell et al., 2012), and India is not escaping the current global warming trend (IMD, 2015). Therefore, the aim of this paper is to study the impact of temperature on teacher presence in India.We address this issue by conducting an analysis that estimates the effects of temperature on teacher absence in India. We find an inverted U-shaped relationship between temperature and teacher attendance, with a peak around 25°C. Above this threshold, our results show that, if on average 85% of teachers are present at school, one additional degree (°C) will raise absences by 8.1%. In other words, the higher the temperature, the bigger the effect of one additional degree. Our result is robust to different regression methods (logit, probit, and OLS) and to several specification checks. This result is crucial as it presents one key factor that affects future economic outcomes through the future workforce’s education. With higher teacher absenteeism due to higher temperatures, children are less educated. This leads to a larger proportion of unskilled workers without human capital, which generates lower GDP. Slow growth will in turn limit the capacity to invest in health and education, creating a vicious cycle. In other words, it is likely that high temperatures play a key role in this issue where human capital, economic development, and welfare are closely interconnected. Our analysis also demonstrates that the presence of fans in classrooms and other investments in better infrastructure allow to tackle this problem, by reducing the strong negative impact of heat on attendance. The rest of the paper is organised as follows. Section 2 summarizes the main results of the previous literature. Section 3 describes the data used for teacher information and weather. Having matched each interview to all the weather information of the location of the interview of the day, Section 4 details the chosen specification. Section 5 presents the benchmark results followed by several robustness checks. Section 6 extends our main results by including different variables that could logically affect the estimated effect. Section 7 concludes the paper.