Evaluating and Predicting the Long-Term Impact of the COVID-19 Pandemic on Manufacturing Sales within South Africa

Abstract

Manufacturing sales forecasting is crucial for business survival in the competitive and volatile modern market. The COVID-19 pandemic has had a significant negative impact on the demand and revenue of firms globally due to disruptions in supply chains. However, the effect of the pandemic on manufacturing sales in South Africa (SA) has not been quantified. The progress of the country’s manufacturing sector’s recovery after the pandemic remains unknown or unquantified. This paper uses a Box–Jenkins approach to time series analysis to produce long-term forecasts/projections of potential manufacturing sales, thereby quantifying the effects of the pandemic shock when the projections are compared with actual manufacturing sales. The Box–Jenkins approach is chosen because of its credibility and ability to produce accurate forecasts. Long-term projections enable organisations to plan ahead and make informed decisions, develop successful recovery plans, and navigate through similar economic shocks in the future, thereby ensuring long-term business survival and sustainability of the manufacturing sector. The SARIMA (0,1,1)(0,1,1)₁₂ model best fits the SA manufacturing sales data according to the Akaike information criterion (AIC) and Bayesian information criterion (BIC), as well as the root mean square error (RMSE) and the mean absolute error (MAE). The results indicate that SA’s manufacturing sector was negatively impacted by the COVID-19 pandemic from about April 2020, but by November 2020 manufacturing sales had recovered to levels similar to projected levels had the COVID-19 pandemic not occurred. Long-term forecasts indicate that SA manufacturing sales will continue to increase. The manufacturing sector continues to grow, leading to increased employment opportunities and a boost to the gross domestic product (GDP).

1. Introduction

Sales forecasting in the manufacturing sector enables managers to ensure that product inventory does not significantly exceed anticipated demand, and that the organisation does not experience stockouts. It provides insights into future demands that are linked to various operations such as purchasing, production, and transportation. Sales forecasting techniques help manufacturing companies to achieve their objectives and optimise operational costs, meeting customer demand and minimising the mismatch between supply and demand [1]. Yenradeea et al. [2] noted that most manufacturing firms in developing nations base their production plans and projections of product demand on arbitrary and intuitive judgments, which can lead to production inefficiencies. Accurate manufacturing sales forecasting significantly reduces the likelihood of stockouts, increases sales, reduces costs, and improves customer satisfaction. According to [3,4,5], quantitative forecasting methods have demonstrated a higher degree of accuracy compared to judgmental or qualitative forecasts.

The COVID-19 pandemic has had a significant negative impact on the demand and revenue of firms globally due to disruptions in supply chains. However, the effect of the pandemic on manufacturing sales in South Africa (SA) has not been quantified. The progress of the country’s manufacturing sector’s recovery after the pandemic remains unknown or unquantified. The COVID-19 pandemic-induced economic crisis emphasised the importance of sales/inventory management optimisation, highlighting the need for businesses to reduce inventory buffers and implementation of just-in-time manufacturing methods [6]. Short- and long-term quantitative sales forecasting is the most effective strategy for reducing demand uncertainty in the retail or manufacturing sector [7,8,9]. The adoption of a quantitative forecasting method prevents, among other things, unreliable production plans, over- or understock issues [4,5], and can quantify the impact of the COVID-19 pandemic on SA manufacturing sales. The decline in sales caused by the pandemic was abrupt, but the recovery rate provided good information for manufacturers and marketing departments to now plan ahead to meet new demand for manufactured goods.

The spread of the COVID-19 virus disrupted and upended the supply chain dynamics of global production and trade through economic lockdowns. Most governments distinguished essential and non-essential sectors, suspending non-essential operations [10]. Economic lockdowns/shutdowns including movement restrictions were imposed to contain the deadly virus, but the measures ended up affecting production and marketing in both local and international sales markets [11]. Manufacturing businesses worldwide were directly and indirectly impacted by the economic movement restrictions, and a decline in demand and sales was noticed [12,13]. The result was a significant drop in production and employment in the world, including SA, surpassing some declines witnessed during the 2007/2008 global financial crisis (GFC) [14].

Beck et al. [15] assert that the COVID-19 pandemic had a negative impact on more than two-thirds of South African businesses that were assessed between April 2020 and June 2020. Omoruyi et al. [16] reported that the COVID-19 pandemic was responsible for disruptions in the food retail supply chain in SA, leading to a significant decline in food retail sales. This highlights some of the detrimental impacts of the COVID-19 pandemic on the SA economy. It is estimated that the COVID-19 pandemic caused an average decline in sales of 78% for all businesses in May 2020 globally [17].

Despite numerous studies on the COVID-19 pandemic, it remains unclear how the pandemic has affected SA’s manufacturing sales in particular. This study applies time series analysis techniques to quantify the impact of the COVID-19 pandemic on total manufacturing sales in SA and forecasts the expected sales going forward. Time series analysis techniques allow for the accurate quantification of the impact of the COVID-19 pandemic on total manufacturing sales in SA, and facilitate the development of reliable sales forecasts/projections into the future. The seasonal autoregressive integrated moving average (SARIMA) model is used in the quantification, so as to examine the effect of the COVID-19 pandemic on manufacturing sales in SA. The model is widely used for forecasting economic data, including manufacturing sales [18], and is considered fundamental and conventional. Quantifying the impact of the COVID-19 pandemic on manufacturing sales and gauging how long it took for the sector to recover is useful for policymakers, firms, and other stakeholders concerned with supply chain disruptions. It is important to manage the impact of the pandemic on businesses and the national economy as a whole. Indeed, such information is useful for managing future economic shocks or similar pandemics.

Ivanov [19] recognised the limited availability of literature on the COVID-19 pandemic’s impact on supply chains (SCs) in Germany and the corresponding adaptation strategies. The author performed a literature review analysis, a multiple case study approach, and a quantitative technique in the form of a generalised model to evaluate the COVID-19 pandemic’s impact on SCs. It was concluded that the COVID-19 pandemic posed an unparalleled challenge to SC ecosystems, networks, flows, and individual firms. The study resulted in a comprehensive model that gauges the effectiveness of various adaptation strategies, which offers valuable managerial insights.

The originality of this research should be viewed within the context of applying an already established statistical methodology to provide new insights into the complex relationship between global pandemics and local market conditions, and by demonstrating the impact of a shock on manufacturing sales in SA. The approach is used as a tool for understanding and predicting economic trends in SA based on existing manufacturing data. This study aims to fill an important gap in the existing scholarship, namely, investigating the effect of an an external shock, the COVID-19 pandemic, on manufacturing sales in SA. The study also serves as a frame of reference for similar research on the impact of the pandemic either on other industries in SA, or in other neighbouring countries for the same sector and other sectors.

Investigating the effect of an external shock, the COVID-19 pandemic on manufacturing sales in SA is important to investors and managers seeking to make data-driven decisions in the manufacturing sector, as well as to policymakers, businesses, and other stakeholders concerned with economic trends and stability. It also helps the development of more effective policies aimed at mitigating the negative impacts of economic shocks on the manufacturing sector, ultimately leading to improved or sustainable economic outcomes for businesses and communities.

The COVID-19 pandemic resulted in business interruptions, leading to consequential loss for the manufacturing companies as the businesses were not able to use their assets and revenue was lost, and all due to the disruptive events, e.g., economic lockdowns (shutdowns) necessitated by the pandemic. Evaluating and predicting the long-term impact of the COVID-19 pandemic on manufacturing sales within SA is crucial for achieving the sustainability goals of the sector. The pandemic has significantly disrupted global supply chains, leading to a decline in manufacturing sales and posing a threat to the sustainability of significant components of the manufacturing sector. By evaluating the impact of the pandemic on manufacturing sales, stakeholders can identify the areas that require support and prioritise sustainable measures to ensure the long-term viability of the sector. Additionally, predicting the long-term impact of the pandemic on manufacturing sales can enable stakeholders to plan ahead and implement sustainable strategies to mitigate the negative effects of future crises. Ultimately, evaluating and predicting the long-term impact of the COVID-19 pandemic on manufacturing sales within SA is essential for achieving sustainability goals and building a more resilient and sustainable manufacturing sector. When deemed necessary, implementing targeted measures such as funding, tax waivers, and relaxation of collateral requirements can help to sustain the manufacturing industry in the long run and promote economic sustainability in SA.

2. Theoretical Framework

As stated by [20,21], resilience refers to the capacity of organisations to take a proactive attitude towards supply chain disruptions, and withstand, adjust, and thrive in the presence of change and uncertainty, thereby recovering the organisation’s business. This points towards the need for manufacturing companies to have effective and efficient strategies to deal with supply chain and sales disruptions such as those caused by the COVID-19 pandemic. In addition, the resilience theory argues that it is not the nature of a shock, such as the COVID-19 pandemic, that is most important, but how manufacturing companies deal with the shock’s effects. When manufacturing companies face such a shock, with declining manufacturing sales, resilience helps them to bounce back. It helps organisations survive, recover their sales, and even thrive in the face and wake of such a pandemic.

The resilience theory sheds light on how manufacturing organisations in SA can effectively respond to and recover from the impact of the COVID-19 pandemic and other similar crises, enabling stakeholders to explore strategies, mechanisms, and processes that enhance resilience and build more robust systems and organisations. In analysing supply chain management (SCM) and understanding the causes of disruptions, managers can develop effective strategies to maintain operational continuity, as emphasised by [22]. According to [23], embracing the principles of resilience theory leads to the establishment of a resilient manufacturing sector, where companies can rapidly predict, adapt to, respond to, and recover from unforeseen disruptions. To further enhance resilience, the integration of predictive models, particularly SARIMA models, and the development of digital platforms play a crucial role, providing significant value and facilitating the transfer of digital knowledge from diverse sources, as highlighted by [24]. A resilient manufacturing organisation can overcome disruptions in its supply chain and sales operations, adapt to changing customer needs, and meet the expectations of stakeholders using information from models, such as the SARIMA model.

Institutional theory highlights the crucial link between artificial knowledge and the resilience business model, as it asserts that the adoption of modern digitalisation effectively addresses stakeholder concerns and also effectively responds to external pressures [25]. The COVID-19 pandemic has introduced new institutional pressures, such as government regulations, health and safety guidelines, and evolving consumer expectations. To navigate these pressures, manufacturing companies can leverage artificial knowledge and digitalisation. Artificial knowledge (intelligence) (AI) aids in identifying the organisational resources that facilitate supply chain redesign during the management of disruptions [25]. As suggested by [26,27], artificial knowledge serves as an innovative avenue for manufacturing organisations to adapt and shape institutional frameworks that promote digital platforms.

Moreover, the COVID-19 pandemic has served as a catalyst for information and digital transformation, leading to improved operational performance characterised by cost reduction and increased strategic effectiveness. Simultaneously, it has created favourable circumstances for businesses to explore new markets and seize fresh opportunities [28] and accelerated the digital transformation of companies, impacting their processes and structures through digital platforms [25]. Digital platforms enable online order tracking, online marketing, and data provision with various other functionalities. This reduces disruptions in the supply chain and enhances manufacturing sales through the supply of relevant and timely information to aid in optimal decision making.

Manufacturing organisations may adopt artificial intelligence in their operations since it serves as a tool for analysing complex interrelationships within supply chains, enabling the identification of patterns, demand forecasting, inventory management optimisation, and enhancing operational agility, as highlighted by [19]. SARIMA models serve as a statistical tool to analyse and forecast time series data, including manufacturing sales data derived from digital platforms. SARIMA models are capable of uncovering patterns and trends within the data and provide valuable insights for planning and decision-making purposes.

In assessing the impact of COVID-19 on manufacturing sales, it is crucial to augment any forecasts/predictions with both institutional theory and resilience theory. This enables the exploration of both external institutional pressures and internal organisational capabilities, providing a complete understanding that helps in formulating effective strategies for sales management and business resilience in the face of ongoing challenges.

3. Literature Review

Fairlie and Fossen [29] conducted a study examining the initial effects of the COVID-19 pandemic on business sales in California. Their findings confirmed that businesses subjected to mandatory economic lockdowns, such as those offering accommodation services, drinking places, and arts, entertainment, and recreation businesses, experienced the most significant losses. Similarly, ref. [30] employed a conjunctural research approach to analyse the impact of the COVID-19 pandemic on sales of products and accompanying services in the Czech Republic’s electrical engineering sector. Their study revealed negative effects on sales; however, they did not propose specific mechanisms or strategies to address the challenges posed by the pandemic.

In China, ref. [31] assessed macroeconomic recovery after a natural hazard (the 2008 Wenchuan earthquake) using an ARIMA model. The model was deemed informative for planning in the event of new natural disasters happening.

In their study, ref. [16] examined the effects of the COVID-19 pandemic on food retail supply chain models in SA. To gain insight into the challenges facing the food retail supply chain, the authors conducted a comprehensive review of the academic literature. Although the study did not make any predictions, the authors concluded that investing in supply chains, developing local supply, and implementing technology to respond to the pandemic’s shocks paid off for large food retailers. These measures helped to enhance the food retail supply chain’s resilience and ensured that consumers had access to adequate supplies of food during the pandemic.

Kim et al. [18] predicted offline retail sales in South Korea during the COVID-19 pandemic, recognising the importance of accurate forecasts in understanding the pandemic’s impact on the retail sector. There were two models used: an exponential smoothing (ETS) model and an ARIMA model. Offline retail stores’ sales data (fashion, food and beverage, entertainment, cosmetics, and sport) were used. The COVID-19 pandemic was judged to have had a considerable and significant adverse/negative effect on the food and beverage, entertainment, fashion, and retail categories. The fashion retail sector was classified as progressively improving. The authors concluded that the pandemic had a minor influence on the cosmetics and sports retail categories, and by 2022, their retail sales had recovered.

Nigam and Shukla [32] conducted a study to assess the impact of the COVID-19 pandemic on the sales of transportation vehicles produced by the Eicher company in India. To predict the sales, the authors used the Box–Jenkins approach to time series analysis. Their findings revealed that the COVID-19 pandemic, which started around January 2020, had a negative impact on the manufacturing sector from March 2020 to April 2021, as revealed by the collected statistics. The authors suggested that the data obtained from the study could be useful in predicting sales patterns in similar future pandemics. They also recommended the use of the Box–Jenkins approach for forecasting in various scenarios due to its ability to provide accurate results.

Abe et al. [33] used time series analysis to evaluate the impact of the COVID-19 pandemic on ex-vessel pricing in Tokyo’s (Japan) fisheries industry. The study concluded that the COVID-19 pandemic had a significant impact on the fisheries industry due to its reliance on the domestic food service sector. There was also a substantial decrease in fish prices, resulting from a sharp decline in overseas demand. The primary cause of this decline was the stagnation of domestic demand, which arose from the rapid contraction of local demand. The reduction in fish demand was further exacerbated by a decrease in inbound tourist demand and economic lockdowns (restrictions on international travel), also impacting the export of high-grade fish. The authors emphasised the value of time series techniques in analysing quantitatively the effects of shocks on different fish species.

Ardolino et al. [34] employed a systematic literature review (SLR) approach to evaluate the primary effects of the COVID-19 pandemic on the manufacturing industry. Their study concluded that production in the manufacturing sector was significantly reduced due to economic lockdowns and shutdowns resulting in changes in the balance of supply and demand, social distancing requirements, remote work, and shifts in consumer behaviour patterns.

Shang et al. [35] explored how the COVID-19 pandemic affected various capital structures across the global economy. Their study used data from both liberal market economies (LMEs), such as Australia, New Zealand, the United Kingdom, and Germany, and coordinated market economies (CMEs), including Sweden, Germany, and Japan. The study revealed that the LMEs were more susceptible to the pandemic’s effects than the CMEs. Nevertheless, both LMEs and CMEs faced negative impacts from the pandemic, although the impact on LMEs was relatively less severe.

Makurumidze and Mpofu [36] conducted a case study on female entrepreneurs in Harare to investigate the impact of COVID-19 on their businesses in Zimbabwe. The study used regression and correlation analysis, and it was concluded that small manufacturing and SMEs led by women were among the hardest hit financially by the crisis as compared to large companies. However, the study did not assess whether these sectors had recovered or not.

Fairlie [37] conducted a study on the initial impact of the COVID-19 pandemic on small businesses in the United States in April 2020. A qualitative approach was used, and the study found that there was a significant decline in the number of business owners and almost every industry, including manufacturing, suffered losses. Particularly, African-American businesses were affected and experienced a sharp drop (41%) in business activity.

Ivanov’s [19] research involved a comprehensive review of the current literature on the COVID-19 pandemic, which revealed several shared characteristics of the adaptation strategies used by businesses during the COVID-19 pandemic. A quantitative technique was used in the development of a generalised model that demonstrated the interconnectedness between the COVID-19 pandemic’s effects and the efforts required for the deployment and evaluation of the adaptation strategies. They recommended the use of quantitative methods for measuring the impact of the COVID-19 pandemic on businesses.

The demand for beverage products in Serbia was predicted by [38]. The SARIMA (5,0,1)(1,0,0)₅₂ model gave a good fit to the weekly data.

The current study will adopt the SARIMA model to analyse the COVID-19 pandemic’s impact on SA’s manufacturing sales. The current study has an added focus of determining the duration the manufacturing industry took to recover from the pandemic’s negative impact. This aspect was not explored in any previous study.

4. Materials and Methods

The Box–Jenkins approach is an iterative procedure which entails model identification using the autocorrelation function (ACF), partial autocorrelation function (PACF), and extended autocorrelation function (EACF). Parameter estimation is performed using, for example, the maximum likelihood estimator (MLE), followed by diagnostic testing of the model. Forecasting is performed as part of the diagnostic process to check model performance, and as the final product of the best model. The R programming language (TSA and forecast packages) is used in the data analysis.

4.1. SARIMA Model

The SARIMA model is an adaptation of the widely used ARIMA model that also includes a seasonal component. A SARIMA (p,d,q)(P,D,Q)_s model can be expressed as:

$${\mathsf{\Phi }}_P\left(B^s\right)\varphi \left(B\right){\nabla }_s^D{\nabla }^d{Y}_{\mathrm{t}}={\mathrm{c}+\mathsf{\Theta }}_Q\left(B^s\right)\theta \left(B\right){\mathsf{\epsilon }}_{\mathrm{t}},$$

(1)

where ${\mathsf{\epsilon }}_{\mathrm{t}}$ is a white noise process and $\mathrm{c}$ is a constant term. ${\nabla }_s^D$ is the seasonal difference order and ${\nabla }^d$ is the non-seasonal difference order. The moving average (MA) [$\theta \left(B\right)$] and autoregressive (AR) [$\varphi \left(B\right)$] polynomials are defined as follows, using backshift operators (B):

$$\theta \left(B\right)=1+{\theta }_{1}B+{\theta }_2{B}^2+...+{\theta }_q{B}^q,$$

(2)

and

$$\varphi \left(B\right)=1-{\varphi }_{1}B-{\varphi }_2{B}^2-...-{\varphi }_p{B}^p,$$

(3)

where ${\mathsf{\Phi }}_i(i=\mathrm{1,2},\dots ,P)$ are the seasonal AR coefficients and ${\mathsf{\Theta }}_{\mathrm{j}}(j=\mathrm{1,2},\dots ,Q)$ are the seasonal MA coefficients. The seasonal AR [${\mathsf{\Phi }}_P\left(B^s\right)$] and the seasonal MA [${\mathsf{\Theta }}_Q\left(B^s\right)$] polynomials are defined, respectively, as follows:

$${\mathsf{\Phi }}_P\left({\mathrm{B}}^{\mathrm{s}}\right)=1-{\mathsf{\Phi }}_1{B}^s-{\mathsf{\Phi }}_2{B}^{2s}-...-{\mathsf{\Phi }}_P{B}^{Ps},$$

(4)

and

$${\mathsf{\Theta }}_Q\left(B^s\right)=1+{\mathsf{\Theta }}_1{B}^s+{\mathsf{\Theta }}_2{B}^{2s}+...+{\mathsf{\Theta }}_Q{B}^{Qs},$$

(5)

where ${\mathsf{\Phi }}_P\left(B^s\right)$ and ${\mathsf{\Theta }}_{\mathrm{Q}}\left(B^s\right)$ are the seasonal AR and MA components of order P and Q, respectively.

4.2. Stationarity and Model Selection

Before fitting a S/ARIMA model, the augmented Dicky–Fuller (ADF) test is used to check for the presence of a unit root on the manufacturing sales series, while the Box–Cox transformation plot is used to identify the best data transformation needed to achieve normality of the dataset. The Akaike information criterion (AIC) and Bayesian information criterion (BIC) are used to select the best model and they can be expressed as:

$$\mathrm{A}\mathrm{I}\mathrm{C}=-2\mathit{log}\left(L\right)+2\left(p+q+k+1\right)\mathrm{a}\mathrm{n}\mathrm{d} \mathrm{B}\mathrm{I}\mathrm{C}=-2\mathit{log}\left(L\right)+2\left(p+q+k+1\right)log\left(n\right),$$

(6)

where L represents the likelihood function, $k=1$ if $c\ne$ 0 (c = constant term), $k=0$ if $c=$ 0. Additionally, p represents the order of the AR part and q represents the order of the MA part, while n is the total number of observations.

4.3. Model Adequacy Testing

Accurate forecasts can greatly reduce errors in forecasting, and to evaluate the quality of the fitted models, both the root mean square error (RMSE) and the mean absolute error (MAE) are used. The use of both RMSE and MAE allows for a comprehensive assessment of forecasting accuracy, which helps to identify and reduce errors, ensuring the reliability of the forecasted results. The RMSE and the MAE are given as:

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}=\sqrt{\frac{1}{n}\sum _{t=1}^n{{(Y}_t-\widehat{Y_t})}^2}\mathrm{a}\mathrm{n}\mathrm{d} \mathrm{M}\mathrm{A}\mathrm{E}=\frac{1}{n}\sum _{t=1}^n|Y_t-\widehat{Y_t}|,$$

(7)

where the original manufacturing sales are represented by $Y_t$, while $\widehat{Y_t}$ denotes the projected manufacturing sales. The projected period spans $n$ total months.

5. Results

The study uses data on SA’s monthly total manufacturing sales from January 2009 to November 2022, obtained from Statistics SA’s Manufacturing: Production and Sales reports (https://www.statssa.gov.za/ accessed on 25 February 2023). The data is divided into training data covering January 2009 to February 2020, and testing data spanning March 2020 to November 2022.

5.1. Descriptive Statistics and Model Identification

The descriptive statistics for the monthly total manufacturing sales $\left(Y_t\right)$ in rand are shown in

Table 1. SA’s total manufacturing sales $\left(Y_t\right)$ descriptive statistics.

Minimum	Maximum	Mean	Std. Deviation	Skewness	Kurtosis
100,153,945	243,365,406	166,324,113	34,555,197	0.09	−0.82

.

The average monthly manufacturing sales amount in SA is ZAR 166,324,113 million. ZAR 100,153,945 million and ZAR 243,365,406 million are the minimum and maximum monthly sales, respectively, indicated in Table 1. The total manufacturing sales time series plot is shown in Figure 1.

Sales of all manufactured goods exhibit a seasonal pattern, a rising trend, and increased volatility, which point to a non-stationary series. To determine the key features of the data, the manufacturing sales are broken down into their time series components.

The top series in Figure 2 is the manufacturing sales series showing fluctuations. The second plot isolates the distinct increasing trend. The third plot isolates the strong seasonal component, suggesting a SARIMA model may be well-suited to analysing/explaining this dataset. The final plot shows the random component of the time series.

The Box–Cox technique is used to identify any potential data modifications that may be required.

As seen in Figure 3, where the maximum log-likelihood of the transformation parameter lambda $\left(\mathsf{\lambda }\right)$ is extremely close to 0; a logarithm transformation is necessary to reduce the volatility and smoothe the manufacturing sales data. The total manufacturing sales $\left(Y_t\right)$ for SA are log-transformed through the equation

$$Z_t=\log \left(Y_t\right)$$

(8)

The plot of the logarithm transformed manufacturing sales ($Z_t)$ is illustrated in Figure 4.

The plot of $Z_t$ in Figure 4 exhibits a more homogeneous variation in the data set with time. The ADF test is used to test the null hypothesis that a unit root exists in $Z_t$, and the results are shown in

Table 2. ADF test results of $Z_t$.

Dickey–Fuller	Lag Order	p-Value
−3.1304	5	0.1065

.

The data in Table 2 show that $Z_t$ has a unit root at the 5% level of significance, as shown by a high p-value of 0.1065. To make the series stationary, $Z_t$ is subjected to both seasonal and non-seasonal differencing. Figure 5 depicts both the seasonal and non-seasonal differenced $Z_t$ series.

Figure 5 depicts a series whose characteristics are independent of the observational time. The ADF test is employed to further verify the data’s stationarity.

Table 3. Seasonal and non-seasonal differenced $Z_t$ series ADF test results.

Dickey–Fuller	Lag Order	p-Value
−8.1103	4	0.02

provides a summary of the ADF results.

The low p-value of 0.01911 from the ADF test results demonstrates that the series is now stationary at the 5% level of significance. Figure 6 displays the ACF and PACF plots used to pinpoint the potential models.

The ACF cuts-off after lag 1, with a significant spike at lag 12, and the PACF is decaying, suggesting a SARIMA (0,1,1)(0,1,1)₁₂ model. To strengthen the model’s validity, the EACF is applied and is shown in

Table 4. EACF of the seasonal and non-seasonal differenced $Z_t$ series.

AR/MA

0

1

2

3

4

5

6

7

8

9

10

11

12

13

0

x

o

o

o

o

x

o

o

x

x

x

x

o

x

1

x

x

o

o

o

x

o

o

o

o

o

x

o

x

2

o

x

x

o

o

o

o

o

o

o

o

x

o

o

3

x

o

x

o

o

o

o

o

o

o

o

x

o

o

4

x

o

x

o

o

o

o

o

o

o

o

x

x

o

5

o

o

x

o

o

o

o

o

o

o

o

x

o

o

6

o

o

x

o

o

o

o

o

o

o

o

x

o

o

7

x

o

x

o

o

o

o

o

o

o

o

x

o

o

.

The EACF suggests the SARIMA (0,1,1)(0,1,1)₁₂ model, which is the same model that the ACF and PACF have advised. The AIC, BIC, and RMSE are used to determine which model is the best after comparing the SARIMA (0,1,1)(0,1,1)₁₂ model to other competing models.

Table 5. The fitted models’ AIC, BIC, and RMSE.

Model	AIC	BIC	RMSE	MAE
SARIMA (0,1,1)(0,1,1)12 model without drift	−483.45	−475.06	0.1606	0.0046
SARIMA (1,1,1)(0,1,1)12 model without drift	−481.52	−470.33	0.1611	0.0046
SARIMA (0,1,1)(0,1,0)12 model without drift	−446.64	−441.05	0.1777	0.0061
SARIMA (1,1,1)(0,1,0)12 model without drift	−444.95	−436.56	0.1764	0.0059

Note: The final model considered is in bold.

displays the fitted models’ AIC, BIC, RMSE, and MAE. The RMSE and MAE are calculated using the validation data (March 2020 to November 2022).

According to Table 5, the SARIMA ${\left(\mathrm{0,1},1\right)(\mathrm{0,1},1)}_{12}$ model is chosen as the best model for the data since it has the lowest AIC, BIC, RMSE, and MAE values.

5.2. Model Parameter Estimation

The coefficients of the SARIMA ${\left(\mathrm{0,1},1\right)(\mathrm{0,1},1)}_{12}$ model estimated through the MLE are shown in

Table 6. SARIMA ${\left(\mathrm{0,1},1\right)(\mathrm{0,1},1)}_{12}$ model coefficients.

Parameter	Coefficient/ Parameter Estimate	Standard Error (SE)	Test Statistic	p-Value
${\widehat{\mathsf{\theta }}}_1$	−0.6724	0.0767	−8.7624	<0.001
${\widehat{\mathsf{\Theta }}}_1$	−0.7860	0.1030	−7.6335	<0.001

.

The small p-values indicate that all of the model parameters shown in Table 6 are statistically significant at the 5% significance level. The SARIMA (0,1,1)(0,1,1)₁₂ model is written as:

$${\widehat{Z}}_t={\widehat{Z}}_{t-1}+{\widehat{Z}}_{t-12}-{\widehat{Z}}_{t-13}+{\widehat{\epsilon }}_t-0.6724{\widehat{\epsilon }}_{t-1}-0.7860{\widehat{\epsilon }}_{t-12}+0.5285{\widehat{\epsilon }}_{t-13}.$$

(9)

5.3. Model Residual Analysis

The SARIMA ${\left(\mathrm{0,1},1\right)(\mathrm{0,1},1)}_{12}$ model residuals are checked for autocorrelation, but the hypothesis is rejected according to the Ljung–Box test results (X-squared = 2.2465, df = 3, p-value = 0.5229). The model residuals are found to be normal according to the Jarque–Bera test (X-squared = 1.4603, df = 2, p-value = 0.4818). Figure 7 displays linearity on the Q–Q plot, and the histogram also suggests normality of the residuals.

Figure 7 demonstrates that the model residuals are reasonably close to a normal distribution, the null hypothesis of normality of residuals can be accepted. The model is a good fit to the dataset.

5.4. In-Sample Prediction

The SARIMA (0,1,1)(0,1,1)₁₂ model is examined to see if there is any substantial discrepancy between the observed and the in-sample predictions, and the findings are shown in Figure 8.

According to Figure 8, the SARIMA (0,1,1)(0,1,1)₁₂ model gives a good fit to the data because there are no significant differences between the fitted and actual values. Thus, the model has the ability to generate good and reliable long-term out-of-sample forecasts.

5.5. Long-Term Out-of-Sample Forecasting

Long-term out-of-sample forecasts for the future 70 months (March 2020 to December 2025) are generated using the SARIMA (0,1,1)(0,1,1)₁₂ model. The logarithm transformation is first reversed, and Figure 9 shows a comparison of the total manufacturing sales that are expected/projected and the observed data.

In Figure 9, the forecasted (blue line) total manufacturing sales eventually closely match the actual sales data (red line), including the seasonal patterns, which are becoming more pronounced over time. The 80% confidence limit is depicted by the dark grey colour, whereas the stricter 95% confidence limits are represented by the grey colour. By comparing the red line (actual sales during and after the pandemic) to the blue line (forecasted/projected sales if the pandemic had not occurred), it is evident that the COVID-19 pandemic had a significant impact on manufacturing sales for a given period. SA declared various levels of economic lockdowns/shutdowns starting in March 2020 in an effort to reduce/stop the spread of the COVID-19 virus. After the large dip in manufacturing sales of April 2020, the gap between the two lines has been narrowing, indicating a recovery in manufacturing sales. As of November 2020, manufacturing sales had recovered to levels similar to the projected levels had the COVID-19 pandemic not occurred.

Figure 9 confirms that SA’s manufacturing sales were negatively affected by the COVID-19 pandemic, but there has been a good and relatively fast recovery. The trendline has recovered to its original trajectory before the onset of the pandemic.

6. Discussion

The results of the study reveal a significant decline in South African manufacturing sales after the start of the COVID-19 pandemic and the economic lockdowns that followed. This finding aligns with those of [16], which concluded a decline in SA’s food retail sales due to the COVID-19 pandemic. Figure 10 proposes a supply chain resilience framework to guide the transformation/recovery and adaptation efforts.

The integration of e-commerce, online marketing, online sales, online order tracking, and the potential for manufacturing repurposing presents a significant opportunity for the manufacturing sector in SA to recover/enhance its resilience and address challenges arising from the COVID-19 pandemic. Manufacturers can enhance their agility, improve decision making, streamline operations, and effectively respond to evolving market dynamics by utilising digital technologies and employing, for example, SARIMA models for time series analysis.

According to [19], the majority of manufacturing firms globally turned to manufacturing repurposing, a temporary tactic that calls for the manufacturing of goods unrelated to the primary company. Some manufacturers, in SA as well, began to produce some of the products that were in short supply, including COVID-19 virus testing kits, respirators and face shields/masks [39,40], oxygen, sanitisers and cleansers [41], and other personal protective equipment (PPE) including gloves [42]. All of this helped SA’s manufacturing companies to recover from the negative impact of the COVID-19 pandemic. In SA, the negative impacts of the COVID-19 pandemic were possibly lessened by all of these activities, which aided in the manufacturing sales’ recovery. Reductions in or relaxation of some of the lockdown rules also aided in the recovery of all other sales.

According to this study’s findings, manufacturing sales started to fall quickly after the economic lockdowns (shutdowns), including social distancing/isolation and remote work introduction. Baroroh et al. [43] and de Giorgi et al. [44] recommended the adoption of advanced digital technologies, including machine learning algorithms and augmented reality systems, to help the manufacturing industry gain information timeously and to overcome the lack of information needed to meet challenges brought about by the COVID-19 pandemic. The recommended adoption of advanced digital technologies and information usage in the manufacturing industry aims to address the challenges of the COVID-19 pandemic to enhancing operational efficiency, and improve resilience, remote collaboration, workforce safety, and adaptability. In leveraging time series analysis, machine learning algorithms, and augmented reality systems, manufacturers can optimise their processes, overcome disruptions, and navigate the evolving landscape in a more efficient and effective manner through the use of the information provided. Thus, manufacturing production lines can become more flexible in responding to changes, and be capable of addressing the skills gaps of workers in need of training on the production of any new products introduced or in demand.

The negative impact of the COVID-19 pandemic on SA’s manufacturing sales is consistent with the findings of [36]. This study noted that the pandemic had a detrimental effect on multiple industries, including manufacturing, agriculture, ICT and stationery, tourism, fashion, arts, and entertainment. The authors emphasised the urgent need for financial institutions to relax collateral requirements, particularly for women-led and small businesses, to access loans for reviving their operations or restocking. Additionally, the authors highlighted that entrepreneurs were procuring supplies at an increased cost, and suggested that governments implement affirmative procurement principles for goods and services related to the COVID-19 pandemic response. In doing so, governments could support local entrepreneurs and businesses, create more sustainable supply chains, and facilitate faster economic recovery in the wake of the pandemic.

Faggioni et al. [45] reported that the COVID-19 pandemic had caused a significant disruption, affecting the internal and external aspects of the supply chains in many organisations. These findings highlight the detrimental effect that global shocks, such as the COVID-19 pandemic, could have on economic growth. Ardolino et al. [34] acknowledged the significant negative impact of the COVID-19 pandemic on manufacturing companies worldwide. They identified lockdowns, shutdowns, social distancing, remote work, and changes in consumer behaviour as the major factors contributing to this detrimental effect. However, they concluded that companies that adopted strategies such as manufacturing repurposing, remote work, layout and workplace reconfiguration, workforce reorganisation, and business model innovation were able to recover from some of the pandemic’s negative impacts a lot quicker.

Long-term forecasts indicate that SA’s manufacturing sales will increase due to, among other things, the implementation of new technologies and increased demand from other emerging markets. The projected increase in manufacturing sales aligns with a previous report [46] that forecasted a growth in e-commerce retail sales from 2010 to 2024. This suggests that SA’s manufacturing sector may experience economic growth due to the expected increase in sales, potentially leading to job creation and an overall positive impact on the economy, despite the COVID-19 pandemic. These results suggest that there is a need for the manufacturing sector to adopt new strategies going forward to cope with the current and future pandemics’ impacts. Such strategies may include implementing cost-cutting measures, diversifying product offerings, and exploring new markets to offset any decline in domestic sales. Pandemics also produce an opportunity for new products.

7. Conclusions

This study used the Box–Jenkins approach to time series analysis to assess the impact of the COVID-19 pandemic on manufacturing sales in SA, and to forecast future sales. The findings of this study are valuable to all manufacturing companies in SA and around the world, particularly to investors, managers, and stakeholders who are seeking to develop survival strategies during challenging economic times as presented by the COVID-19 pandemic. It lays the groundwork for further research in the manufacturing sector by providing valuable insights and explanations of certain trends, such as recovery rates, repurposing manufacturing to meet new demands posed by the pandemic. The COVID-19 pandemic had a significantly negative impact on manufacturing sales in SA, with sales declining sharply around April 2020. However, there was a positive sign of recovery in the sector, with manufacturing sales showing an upward trend and indicating a gradual return to pre-pandemic expected levels by November 2020. This suggests that with the right strategies in place, the manufacturing sector rebounded from the effects of the COVID-19 pandemic and contributed to the production of solution-oriented products and the overall sustainability of the sector and economic recovery of SA.

8. Recommendations

To make informed decisions and develop successful plans, manufacturing organisations could utilise quantitative analysis techniques to determine the potential impact of any economic shocks on their sales and profitability. This helps in making more informed decisions about how to allocate resources, adjust pricing strategies, and manage supply chains to minimise the impact of economic shocks on businesses. Repurposing manufacturing, targeting products needed to meet challenges posed by such an economic shock, could aid in the recovery and the sustainability of SA’s manufacturing sectors from future shocks and disruptions. It is also imperative for the manufacturing sector to develop various marketing communication methods, particularly digital/online and social media marketing, and take advantage of the growing trend of “localism”, to achieve corporate success in the uncertain post-COVID-19 times.

Albertzeth et al. [47] recommended that organisations should adopt an informed and agile supply chain approach to enable them to quickly react and adapt to erratic changes in supply and demand. This allows manufacturing organisations to effectively respond to unexpected disruptions and mitigate the negative impact of such disruptions on their operations and the overall supply chain. This approach can improve the resilience of the supply chain and lead to better economic recovery and sustainability.

Embracing time series modelling and digital technologies and e-commerce can provide numerous benefits for businesses, particularly during times of external shocks, such as the COVID-19 pandemic. By leveraging these technologies, businesses can not only increase revenue and reach new customers, but also foster stronger relationships with local suppliers and create a more efficient and resilient supply chain.

E-commerce platforms allow businesses to expand their customer base, reaching beyond physical limitations and attracting customers from different geographical locations, thereby tapping into new markets. Additionally, implementing e-commerce platforms enables businesses to diversify their sales channels, providing customers with multiple options for making purchases, such as online marketplaces, mobile apps, or social media platforms. Through digital platforms, businesses can easily exchange information, place orders, and track inventory, ensuring a smoother supply chain process. Furthermore, digital technologies enable businesses to optimise their inventory management through real-time monitoring and demand forecasting. This helps to reduce the risk of stockouts or overstocking, enhances operational efficiency, and minimises costs.

Ensuring supply chain resilience is crucial for the manufacturing industry’s sustainability and economic recovery, as highlighted by [48]. By implementing initiatives to make supply chains more resilient, organisations can improve their production and sales capabilities, minimise disruptions caused by external shocks such as the COVID-19 pandemic, and ultimately achieve long-term economic sustainability.

The manufacturing industry in SA requires a stimulus to reignite further growth, and the government has an important role to play in this effort. One possible strategy is to reduce levies and taxes on raw material imports, which would lower manufacturing costs and make locally produced goods more competitive. In addition, international financing, for example, from the World Bank and International Monetary Fund (IMF), could provide crucial liquidity to sustain the industry as it further recovers. However, it is important that this funding is designed to address the specific needs of the manufacturing industry and that it benefits all players/or intended players within the industry. This will help to counteract the negative impact of COVID-19 on the manufacturing sector and contribute to a more robust and sustainable economy.

The scope of this study is restricted to the economic crisis in the manufacturing sector of SA specifically caused by the global COVID-19 pandemic. The findings cannot necessarily be extrapolated to other economic crises as each crisis can have varying impacts. Moreover, this study is limited to manufacturing firms that remained functional during the COVID-19 pandemic era. However, the methods used in the analysis and some of the solutions to increase sales after the impacts would largely be the same for other economic shocks. Although there are sales from various sectors, this study focused specifically on manufacturing sales in SA. Future studies would look at other sector’s sales.