Air pollution, weather factors, and realized volatility forecasts of agricultural commodity futures

• Published date: 01 February 2024
• Author: Jiawen Luo,Qun Zhang
• Date: 01 February 2024
• DOI: http://doi.org/10.1002/fut.22467

Received: 6 June 2023

|

Accepted: 25 September 2023

DOI: 10.1002/fut.22467

RESEARCH ARTICLE

Air pollution, weather factors, and realized volatility

forecasts of agricultural commodity futures

Jiawen Luo

1

|Qun Zhang

2,3,4

1

School of Business Administration, South China University of Technology, Guangzhou, Guangdong, China

2

School of Finance, Guangdong University of Foreign Studies, Guangzhou, Guangdong, China

3

Southern China Institute of Fortune Management Research, Guangzhou, Guangdong, China

4

Institute of Financial Openness and Asset Management, Guangzhou, Guangdong, China

Correspondence

Qun Zhang, School of Finance,

Guangdong University of Foreign Studies,

University Town, Panyu District,

Guangzhou, Guangdong 510006, China.

Email: qunzhang@gdufs.edu.cn

Funding information

National Natural Science Foundation of

China, Grant/Award Numbers: 72171088,

71801057, 71803049; Natural Science

Foundation of Guangdong Province,

Grant/Award Numbers:

2021A1515011337, 2022A1515012093,

2023A1515012527; Guangdong Philosophy

and Social Sciences Foundation,

Grant/Award Number: GD23CGL01;

Guangzhou Municipal Science and

Technology Bureau,

Grant/Award Number: 2023A04J1298,

SL2024A04J01647

Abstract

This study investigates the potential effects of environmental factors on

fluctuations in agricultural commodity futures markets, by constructing a new

category of daily exogenous predictors related to air pollution, weather,

climate change, and investor attention. The empirical results from out‐of‐

sample analyses suggest that the heterogeneous autoregressive (HAR) model

incorporating all these exogenous predictors is more likely to outperform other

HAR‐type models. Additionally, economic evaluations demonstrate the

superior performance of models incorporating investors' attention to climate

change or extreme weather as predictors. While not all exogenous predictors

are equally important for volatility forecasts, adopting appropriate variable

selection methods to handle different sets of exogenous predictors can lead to

better performance than the HAR benchmark. With the inclusion of air

pollution or weather factors in the HAR model, a portfolio with an annualized

average excess return of 16.2068% or a Sharpe ratio of 10.0431 can be achieved

for the wheat futures, respectively.

KEYWORDS

agricultural commodity futures, air pollution, Bayesian model averaging, LASSO, volatility

forecasting

JEL CLASSIFICATION

C53, G11, O13, Q53, Q54

1|INTRODUCTION

Accurate forecasts of financial time series volatility are crucial for asset pricing, portfolio optimization, risk

management, and policy making. The advent of readily available high‐frequency data has spurred a revived interest in

alternative ways to model and forecast realized volatility (RV) more accurately for different forecast horizons

(Andersen & Bollerslev, 1998; Sévi, 2014). Despite the fact that a large number of models have been designed, a

considerable gap still exists in the literature for models aimed at RV forecasting (Andersen et al., 2006). The

heterogeneous autoregressive model (henceforth HAR model) proposed by Corsi (2009) elicits a considerable

J Futures Markets. 2024;44:151–217. wileyonlinelibrary.com/journal/fut © 2023 Wiley Periodicals LLC.

|

151

advancement in this field, as it allows for approximating the long memory features of the data while also responding to

short‐term shocks. Due to its simple estimation procedure, interpretability, and expansibility, the HAR model and its

extensions have been widely employed in empirical research and have demonstrated superior performance when

compared with generalized autoregressive conditional heteroskedasticity (GARCH)‐type models (Sévi, 2014)or

extended by the GARCH‐jump process (Dutta & Das, 2022).

While the inconsistency of RV estimators remains under intense debate (see Liu et al., 2015), one direction to

improve the prediction performance of the HAR model is to explore certain variables or factors that are related to

equity returns or volatility. A growing body of research documents that some local environmental factors have

biophysical and emotional effects, which can influence investors' mood, trading decisions, and equity returns. For

instance, Saunders (1993) utilizes the cloud cover over New York City as a proxy for investors' mood and observes that

such a factor has a strong, negative association with the returns of three major US stock market indices. Analogously,

Kamstra et al. (2003) note that stock returns exhibit a seasonal cycle that is linked to seasonal affective disorder (SAD),

and confirm that the investors suffering from SAD effects avoid risky portfolios in the fall and resume risky assets in

winter, finally giving rise to a decline in returns in the fall and an increase in returns following the longest night of the

year. Along similar lines, alternative environmental factors have been used as proxies in an attempt to measure

collective mood‐swing patterns, such as temperature (Cao & Wei, 2005), humidity (Pardo & Valor, 2003), rain, snow

(Hirshleifer & Shumway, 2003), and the seasonal light cycle (Kamstra et al., 2003). For a review of this literature, see

Lucey and Dowling (2005). The rationale behind all these studies is that environmental changes may trigger mood

swings that, ultimately, may impact investment decisions. Dong and Tremblay (2022) confirm that multiple premarket

weather conditions (sunshine, wind, rain, snow, and temperature) exert economically important impacts on stock

returns worldwide during 1993–2012, and weather effects are related predominantly to investor emotion rather than

economic fundamentals.

A more recent—and more closely related—body of work links air pollution to investors' mood, cognition, mental

well‐being, trading behavior, and stock prices, with investors' mood posited as the mediating channel (e.g., Levy &

Yagil, 2011; Lepori, 2016). Here the focus is on ambient air pollution, which is one of the most critical environmental

stressors to which individuals are exposed and has been found to be responsible for a broad spectrum of physical and

psychological effects on human beings. Heyes et al. (2016) provide detailed empirical evidence of a direct effect of air

pollution on the efficient operation of the New York Stock Exchange, linking short‐term variations in fine particulate

matter (PM

2.5

) in Manhattan to movements in the S&P 500. They find that PM

2.5

was significantly negatively correlated

with stock index returns, and provide evidence of decreased risk tolerance operating through pollution‐induced

changes in mood or cognitive function. Dong et al. (2021) document a negative relation between air pollution during

corporate site visits by investment analysts in China and subsequent earnings forecasts, but pollution only affects

forecasts that are announced in the weeks immediately following a visit, indicating that mood likely plays an essential

role in treatment response to mood stabilizers. Particularly, Li et al. (2021) find that air pollution significantly increases

investors' disposition effects or investor biases. These findings further reinforce the possibility that pollution may

impact investors' attention and behaviors.

There are several motivations to link air pollution, climate, or weather related factors to fluctuations in the

agricultural commodity futures markets. First, atmospheric pollution might have adverse effects on the supply side of

agricultural production. On the one hand, toxic air pollutants, such as sulfates, nitrates, dust, and heavy metals can

accumulate in the soil through precipitation and consequently harm the production of agricultural plants

(Chatzopoulos et al., 2020; Nagajyoti et al., 2010). On the other hand, air pollution would harm the health of workers

and reduce food productivity indirectly (Chang et al., 2016; Graff Zivin & Neidell, 2012; Sun et al., 2017). Second,

sudden climate events (e.g., floods, tornadoes, hurricanes, and high temperatures) or long‐term climate changes (e.g.,

global warming, sea level rise, precipitation change, and ocean acidification) can change the distributions and

behaviors of creatures and animals, which would influence agricultural productivity in the long term (Chen et al., 2011;

Lobell & Gourdji, 2012). Extreme climate change also has a significant tail‐risk spillover effect on commodity futures

markets (Jia et al., 2023). By analyzing granular daily data on temperatures across the continental United States from

1990 to 2015, Addoum et al. (2020) find no significant evidence to suggest that temperature exposures have a direct

effect on sales or productivity at the establishment level. Considering the relatively backward production level in

developing countries and the higher cost of coping with these events or changes, the negative impact on agricultural

productivity is more likely to occur in developing countries. Chen et al. (2023) evaluate the regional and sectoral

sensitivities and adaptabilities of Chinese agriculture to rising temperatures and find that temperature effects differ

between northern and southern China, and between sectors (cropping, livestock, forestry, and fisheries). The economic

152

|

LUO and ZHANG

or financial uncertainty brought about by the green transformation will also seriously threaten the stability and

development of the economic and financial systems. Third, as mentioned earlier, there is plenty of research addressing

that investor behaviors are affected by environmental conditions, resulting in changes in trading activity and prices of

financial assets (Hirshleifer & Shumway, 2003; Li et al., 2021; Loughran & Schultz, 2004; Wu et al., 2018). In addition,

the news about climate change is also a good hedging indicator for investors to make investment decisions as suggested

by Engle et al. (2020), while attention to climate change is found to be increased when the local temperature is

abnormally high (Choi et al., 2020).

In this paper, we explore the potential effects of environmental factors on investors' attention and subsequently on

fluctuations in China's agricultural commodity futures markets. According to statistics released by the US Futures

Industry Association,

1

China's agriculture future market experienced remarkable growth since 2000, with commodity

futures trading volume surpassing that of any other nation worldwide. This lays the foundation for investigating the

relationship in this specific context. In addition, China provides a unique setting to investigate this relationship for

several reasons. First, the dominance of individual investors, as one of the distinct characteristics of Chinese futures

markets, enriches our comprehension of how investors' attention affects futures market volatility. Second, pollution is

severe on average in China and highly variable both across geographies and across time, which provides variation in

ambient circumstances that is of such magnitude as to plausibly have a causal impact on investors' attention. Third, the

Ministry of Environmental Protection of the People's Republic of China (MEPC) has provided several indicators since

2013 that measure the air‐pollution condition of China, which enables us to monitor environmental factors' variations

that could play a significant role in impacting market volatility.

With the increased availability of data and the rapid proliferation of machine learning techniques, it is essential to

adopt appropriate variable selection methods. There are two primary reasons for using the variable selection methods.

On the one hand, these techniques can eliminate unimportant predictors and simplify economic models, thereby

reducing the burden of information processing for forecasting purposes (Zhang et al., 2019). On the other hand, some

important exogenous predictors, such as air‐pollution and weather related variables or the climate and weather related

variables, are susceptible to overfitting, multicollinearity, and the curse of dimensionality (Zhang et al., 2023). By

implementing the variable selection methods to address these challenges, we can focus on the relevant predictors with

strong predictability, avoid the overfitting issue, and improve the forecasting performance (Campbell &

Thompson, 2008; Zhang et al., 2023).

Our study makes three contributions to the literature. First, we construct a new category of daily exogenous

predictors that are related to environmental factors by utilizing air pollution, weather, climate change, and investor

attention data. In the daily analysis which is central, we take particular care to isolate the role of climate change from

weather factors. Economic evaluations based on portfolio allocation across five agricultural commodity futures in

China demonstrate the superior performance of models with investors' attention to climate change or extreme weather

as predictors. Second, the new intuition is built on the literature's recent heuristic finding that the adoption of

appropriate variable selection methods to handle distinct sets of exogenous predictors can lead to better performance

than the HAR benchmark. We provide evidence that the least absolute shrinkage and selection operator (LASSO)

method is preferred at the weekly horizon, while the model without the variable selection method dominates at the

monthly horizon. Furthermore, our research shows that investors can further improve the performance of their

portfolio by replacing the standard HAR model with the alternative LASSO‐based, Bayesian model averaging (BMA)‐

based, or elastic net (Enet)‐based model when making investment decisions. These findings inform the design and

ranking of investment strategies for agricultural adaptation to environmental factors. Third, we find new evidence

regarding the causal influence of environmental factors by linking these factors to the RV forecasts of China's

agricultural commodity futures, in the context of increasing climate policies targeting carbon emission reduction and

environmental protection as well as growing investor interest in climate risk factors when making investment

decisions. The empirical results from out‐of‐sample analyses suggest that the HAR model incorporating all the

exogenous predictors is more likely to outperform other HAR‐type models in terms of both statistical and economic

criteria for the 1‐step‐ahead forecasts. Furthermore, with the inclusion of air pollution or weather factors in the HAR

model, a portfolio with an annualized average excess return of 16.2068% or a Sharpe ratio (SR) of 10.0431 can be

achieved for the wheat futures, respectively.

1

Website: https://www.fia.org/fia/etd-tracker.

LUO and ZHANG

|

153