Forecasting the Impact of COVID-19 Epidemic on China Exports using Different Time Series Models

Abstract

Purpose: The primary objective of this paper is to identify the best forecasting model for China exports, especially during the spread of the COVID-19 pandemic.Methodology: We used the data of China exports to the United States and different economic regions from January 2014 to January 2021 to compare models using various criteria and selected the best exports forecast model. The hybrid model is employed to conduct the analysis.The combination of the hybrid model consists of six different models: ARIMA, ETS, Theta, NNAR, seasonal and trend decomposition, and TBATS model. Findings:Our results showed that the hybrid and ANN outperformed the remaining models in forecasting China exports to the world, considering the shock created by the ongoing coronavirus pandemic. This paper underscores the importance of using the specified models in forecasting exports during this period. The results also demonstrate that the magnitude of China exports to all groups decreased and will continue to decline for the next few months.Practical Implication: Forecasting of the export data is presented for the subsequent nine months, thereby providing insights to all policymakers, governments, and investors to be proactive in designing their strategies to avoid any delay/disruption in the imports from China, which could enhance the smooth flow of raw material and sustain industrial production.

A hybrid forecasting technique for infection and death from the mpox virus

“…They concluded that by using the appropriate methods that are suitable to the data, one can achieve consistent results in modelling and predicting any series of an epidemic. For more epidemic hybrid time series forecasting, see, for example, 34 – 42 .…”

Modeling COVID-19 Confirmed Cases Using a Hybrid Model

“…The root mean square error (RMSE), mean absolute error (MAE), and mean percentage errors (MAPE) are used to evaluate the predictive performance of the models. To the best of our knowledge, these are the most common measures used to assess the accuracy of nonlinear and hybrid models (Ravazzolo et al 2020;Safi et al 2022). Table 2 shows the accuracy of the prediction of ARIMA-NNAR model over the benchmark models when the 518 observed data are compared with the prediction data.…”

“…Moreover, COVID-19 has negatively affected the exportations. For example, Safi et al (2022), showed how the COVID-19 pandemic has disrupted the flow of goods that China exports to other parts of the world and forecast the China exports during the COVID-19 pandemic using seven different forecasting models.…”

“…In this context, this paper implemented the hybrid ANN-ARIMA proposed by Zhang (2003) to forecast the ten days ahead of the number of new confirmed cases for Pakistan, Tunisia, Indonesia, Malaysia, India, and South Korea. For simplicity, the Feed-forward Autoregression Neural Network with one single hidden layer is chosen among significant variants of neural networks since it is the most widely used model that brings fruitful results (Zhang, 2003;Safi et al 2022).Experiments showed that the proposed hybrid approach is better than single and hybrid ARIMA-WBF models (Chakraborty & Ghosh, 2020). This result is evident since the COVID-19 confirmed cases contain both linear and nonlinear patterns.…”

Does Bubble Still Exist after COVID-19? Evidence from Hong Kong Housing Market

“…After the emergence of the Global Financial Crisis, a large body of literature exists to discuss the detection of bubbles. (Safi et al, 2022;Foroni et al, 2020). In relation to the investors' behavior, Bourdeau-Brien and Kryzanowski (2020) found that "natural disasters cause a statistically and economically significant increase in risk aversion at the local level", Brown et al (2018) show that "being struck by an extreme event substantially changed individuals' risk perceptions as well as their beliefs about the frequency and magnitude of future shocks", and Sun et al…”