Statistical Approach to Examining the True Status of Long Memory and Volatility Persistence in PM10 Air Pollutant at Different Regions of Malaysia: A Methodical Methodology

Abstract

Air pollution continues to be an international problem that endangers both human health and the environment. Over the past few decades, air pollution in Malaysia has emerged as a serious potential risk due to accelerated economic expansions and seasonal transnational pollution. Particulate matter atmospheric air pollutants in Malaysia have been identified as the most rampant and dominant in the air pollution index (API) amongst other criteria pollutants. The aim of this study is to investigate the statistical issues of long memory and volatility persistence in the level of particulate matter emission from 1 January 2011 to 31 December 2021 in fourteen continuous air monitoring stations of industrial, urban, and suburban categories using the main and partitioned series before and after the regimes of break. The Ordinary Least Square Cumulative Sum (OLS-based CUSUM) test was employed to partition the original series in each monitoring station based on its estimated break dates. The long memory parameter d alongside its standard error was estimated through three techniques namely, Geweke and Porter-Hudak, Fractionally Differenced Sperio, and Exact Local Whittle estimation. The issue of volatility persistence was investigated using the hybrid of the Autoregressive Fractionally Integrated Moving Average (ARFIMA) and Generalized Autoregressive Conditional Heteroskedastic (GARCH) model. The results confirm evidence of a mean-reverting form of long memory with a higher degree of persistence in the main series and volatility persistence in both the main and partitioned series that encountered structural break. This confirms that the data-generating process of particulate matter pollutant in Malaysia possesses true long memory and volatility persistence not spurious due to neglected structural break problem. Maximum emissions in all monitoring sites were observed during the pre-break regime except for Kota Kinabalu station where it occurred during the post-break regime. Most series were characterized by higher values of kurtosis and skewness implying the significant fluctuation and non-Gaussian behavior in the affected series.