Effects of Heat on Mathematics Test Performance in Vietnam

• Published date: 01 March 2022
• Author: Tien Manh Vu
• Date: 01 March 2022
• DOI: http://doi.org/10.1111/asej.12259

Effects of Heat on Mathematics Test

Performance in Vietnam

Tien Manh Vu

Received 7 February 2021; Revised 31 July 2021; Accepted 19 December 2021

We examine the effects of average test-day temperature on the mathematics test

performance of all Vietnamese students who took the national university and col-

lege entrance examinations in 2009. Using individual fixed effects, we find that an

increase of 1F results in an approximate 0.006 standard deviation loss of in stan-

dardized test scores by age and test problem. The negative effects are persistent

regardless of whether students were from the hottest or coolest climate regions in

the country. We also find that female students and students from rural areas and

townships are most vulnerable to the effects of heat.

Keywords: heat, test performance, test score, Vietnam.

JEL classification codes: I29, J24, J16, O15.

doi: 10.1111/asej.12259

I. Introduction

Climate and cognitive skills are important life factors that humans have not been

able to alter immediately. Both are objects of great concern. Various studies

regarding their interactions have examined how climate change, particularly heat

waves, might influence cognitive skills and corresponding task performance

(Garg et al., 2017; Graff Zivin et al., 2018; Park et al., 2018; Park, 2020; Park

et al., 2020). Specifically, Horowitz (2009) suggested that a 1C (1.8F) increase

in global temperature would reduce global income by approximately 3.8 percent.

Examining approximately 10 million test-takers in the USA who repeated the

PSAT, Park et al. (2020) showed that a 1F (0.56C) hotter school year leading

up to the tests decreased learning outcomes by approximately 1 percent. Mean-

while, Park et al. (2020) demonstrated that hotter temperatures might have

lowered the test performance of New York City high school students during

1999–2011 by up to 11 percent of the standard deviation (SD).

This line of research is crucial for developing countries. Poor countries are

more likely to be in hot regions (Horowitz, 2009; Acemoglu and Dell, 2010). In

addition, poor households are more likely to be involved in agriculture, to be

*Vu (corresponding author): Miyazaki International College, Miyazaki, Japan. Email: mvu@sky.

miyazaki-mic.ac.jp.

© 2022 East Asian Economic Association and John Wiley & Sons Australia, Ltd.

Asian Economic Journal 2022, Vol.36 No.1, 72–94 72

exposed more to heat (Park et al., 2018) and to suffer more from extreme

weather because of residential ‘sorting problems’(Graff Zivin and

Neidell, 2013). Poor households have fewer means of prevention and protection

(e.g. housing with cooling systems and favorable working conditions; i.e., with

air-conditioning), fewer safety buffers (private savings) and fewer means of loss

reduction (insurance) to combat the effects of climate change.

The link between cognitive skills and climate in developing countries is still

being investigated. The main obstacles are inappropriate priorities and a lack of

reliable data on cognitive skills that can be geocoded. Among the few studies on

this topic, Park et al. (2018) suggested a sorting problem; that is, a negative rela-

tionship between hotter locations and lower household income within some

developing countries with hot climates. Using repeated cross-sectional data from

India for 2006–2014, Garg et al. (2017) showed a similar negative relationship

between temperature in the calendar year prior to a standardized test and the test

scores. Specifically, an additional 10 days with an average temperature above

29C (85F) relative to 15C–17C (59F–63F) decreased test scores by

approximately 0.02–0.03 SD. However, without individual fixed effects, the esti-

mations suffer from bias because past weather can be correlated with weather

information used in the estimations. Horowitz (2009) showed a negative rela-

tionship between long-run average temperature and logarithm per capita GDP.

In Horowitz’s analyses, Vietnam was listed in the group with a relatively higher

temperature but a relatively lower income. We chose Vietnam for our analyses

for that reason and because approximately 21.26 out of Vietnam’s 85.84 million

people work in the agricultural sector (GSO, 2006).

We investigated the effects of test day average temperature on test-taking per-

formance using the 2009 Vietnamese national university and college

1

test score

data. The data contained 1 125 953 math test scores (779 123 test-takers). Using

individual fixed effects, we examined 294 623 test-takers who took two or three

separate mathematics examinations and created an unbalanced panel dataset

(641 453 math test scores). We also investigated subsamples, including the hot-

test regions and coolest regions, heterogenous numbers of mathematics tests

taken, gender, regional economic disparities and policy groups. We repor ted the

results using standardized scores (by test problem and year of birth) but also

checked the results against the zero-value-adjusted logarithm of the raw scores.

We found that a 1F increase in the average temperature on the test day

resulted in an approximately 0.006 SD decrease in the standardized test score.

Specifically, the decrease was approximately 1.64 percent of the raw mathemat-

ics score of a typical student in the largest group of test-takers (born in 1991) if

the temperature changed from 77.9F (25.5C)—the average temperature

between 1950 and 2009—to 82.8F (28.2C). This temperature (82.8F) was

75 percent of the highest recorded average daily temperature in Vietnam in the

1

The exams were centrally controlled and had uniform test problems. Test-takers had up to three

times to sit in different tests and needed to earn top scores each time in order to earn admittance.

EFFECTS OF HEAT ON TEST PERFORMANCE 73