Purpose

The purpose of this study is to examine the accuracy of combined models with the individual models in terms of forecasting Euro against US dollar during COVID-19 era. During COVID, the euro shows sharp fluctuation in upward and downward trend; therefore, this study is keen to find out the best-fitted model which forecasts more accurately during the pandemic.

Design/methodology/approach

The descriptive design has been adopted in this research. The three univariate models, i.e. autoregressive integrated moving averages (ARIMA), Naïve, exponential smoothing (ES) model, and one multivariate model, i.e. nonlinear autoregressive distributive lags (NARDL), are selected to forecast the exchange rate of Euro against the US dollar during the COVID. The above models are combined via equal weights and var-cor methods to find out the accuracy of forecasting as Poon and Granger (2003) showed that combined models can forecast better than individual models.

Findings

NARDL outperforms all remaining individual models, i.e. ARIMA, Naïve and ES. By applying a combination of different models via different techniques, the combination of NARDL and Naïve models outperforms all combination of models by scoring the least mean absolute percentage error value, i.e. 1.588. The combined forecasting of NARDL and Naïve techniques under var-cor method also outperforms the forecasting accuracy of individual models other than NARDL. It means the euro exchange rate against the US dollar which is dependent upon the macroeconomic fundamentals and recent observations of the time series.

Practical implications

The findings could help the FOREX market, hedgers, traders, businessmen, policymakers, economists, financial managers, etc., to minimize the risk indulged in global trade. It also helps to produce more accurate results in different financial models, i.e. capital asset pricing model and arbitrage pricing theory, because their findings may not be useful if exchange rate fluctuations do not trace effectively.

Originality/value

The NARDL models have been applied previously in different time series and only limited to the asymmetric or symmetric relationships. This study is using it for the forecasting exchange rate which is almost abandoned in earlier literature. Furthermore, this study combined the NARDL with univariate models to produce the accuracy which itself is a novelty. Moreover, the findings help to enhance the effectiveness of different financial theories as well.

中文翻译：

---

在 COVID-19 期间通过结合 NARDL 和单变量技术预测欧元兑美元

目的

本研究的目的是检查组合模型与单个模型在 COVID-19 时代预测欧元兑美元的准确性。疫情期间，欧元呈现大幅上下波动；因此，本研究热衷于找出在大流行期间更准确预测的最佳拟合模型。

设计/方法/方法

本研究采用描述性设计。选择了三个单变量模型，即自回归综合移动平均线（ARIMA）、朴素、指数平滑（ES）模型和一个多元模型，即非线性自回归分布滞后（NARDL）来预测欧元兑美元的汇率在 COVID 期间。上述模型通过等权重和 var-cor 方法组合起来，以找出预测的准确性，因为 Poon 和 Granger (2003) 表明，组合模型比单个模型可以更好地预测。

发现

NARDL 优于所有剩余的单个模型，即 ARIMA、Naïve 和 ES。通过通过不同技术应用不同模型的组合，NARDL 和 Naïve 模型的组合通过对最小平均绝对百分比误差值的评分（即 1.588）优于所有模型组合。var-cor 方法下 NARDL 和 Naïve 技术的组合预测也优于 NARDL 以外的单个模型的预测准确性。它是指欧元兑美元的汇率，取决于宏观经济基本面和最近对时间序列的观察。

实际影响

这些发现可以帮助外汇市场、对冲者、交易者、商人、政策制定者、经济学家、财务经理等，将沉迷于全球贸易的风险降至最低。它还有助于在不同的金融模型中产生更准确的结果，即资本资产定价模型和套利定价理论，因为如果汇率波动不能有效追踪，他们的发现可能没有用处。

原创性/价值

NARDL 模型以前已应用于不同的时间序列，并且仅限于不对称或对称关系。这项研究将其用于预测汇率，这在早期的文献中几乎被抛弃了。此外，这项研究将 NARDL 与单变量模型相结合以产生准确性，这本身就是一种新颖性。此外，这些发现还有助于提高不同金融理论的有效性。