Marine insurance is the oldest practiced form of insurance and can be traced back to ancient Greece, where it became highly developed during the 15th century. During that time, it was a well-known insurance contract among maritime nations trading commercially with Greece (Encyclopaedia Britannica 2019). The earliest marine insurance policies in English were recorded in 1555, and in 1613, a Lloyd’s cargo insurance policy insured goods shipped on the vessel Tiger on terms that would exist in some form until 1982 (Dunt 2009).

During the 19th century, the global increase in trade led to the creation of the Marine Insurance Act (MIA) in 1906, and in 1912, the Institute of London Underwriters developed standardized clauses for cargo insurance called the Institute Cargo Clauses (ICC) (LMA 2009). Most companies worldwide took to underwriting risks subject to the ICC or a variation thereof. The core focus of the ICC is insurance against physical loss of or damage to the cargo, not financial losses or expenses following an insured event. Marine Insurance Act 1906 defines ‘marine insurance’ as:

The start and end of risk attachment are essential aspects of marine cargo insurance that need a clear definition. The ICC have a complex and comprehensive scheme that specifies the duration of the insurance and its continuation in case of events that affect the transit or alter the voyage. This scheme consists of three main clauses, but this article will only cover some key points from clause 8, the Transit Clause, that relate to the start and end of risk attachment. According to sub-clause 8.1, the risk starts ‘from the time the subject-matter insured is first moved … for the purpose of the immediate loading…’, and according to sub-clause 8.1.1, the risk ends at ‘unloading at final warehouse or place of storage named in the insurance’ (Dunt 2009; LMA 2009).

As is typical with other forms of property insurance following the principle of indemnity, it is a requirement for the insured to be able to substantiate that he or she has suffered a loss in the event of a claim. Marine Insurance Act 1906 stipulates: ‘The assured must be interested in the subject-matter insured at the time of the loss though he or she need not be interested when the insurance is effected…’ The principle of indemnity is also echoed in the ICC under clause 11, creating an express contractual requirement: ‘In order to recover under this insurance, the Assured must have an insurable interest in the subject-matter insured at the time of the loss’ (Dunt 2009:425).

Although research exists on cargo accumulation risks in maritime supply chains (Freichel et al. 2022), value loss in cargo transportation on water (Kotenko et al. 2022), cargo loss in logistics systems (Wu, Chen & Tsau 2017), and marine insurance claims analysis for accidents (Ching & Yip 2022), to the authors’ knowledge, there is no known research investigating weather-related marine cargo insurance claims in South Africa (or in general) (Du Plessis, Goedhals-Gerber & Van Eeden 2023).

South Africa is a critical player in the global logistics network, as it hosts several major ports that connect Africa with other continents. However, the country’s logistics network is vulnerable to disruptions caused by extreme weather events, such as floods, storms, droughts, and wildfires (Mutumbo 2017). On average, South African ports record one weather-related incident per day (Danladi 2020; Freight News 2014). These incidents are considered one of the main drivers of congestion at these ports (Nze & Onyemechi 2018; Potgieter, Goedhals-Gerber & Havenga 2020).

Weather-related effects that could impact South African logistics infrastructure include rising atmospheric and water temperatures, strong winds, strong waves, rising sea levels, strong ocean currents, and heavy rainfall (Phelp, Rossouw & Theron 2013). These events can damage infrastructure, disrupt operations, delay shipments, increase costs, and affect customer satisfaction. Furthermore, the likelihood and occurrence of these severe weather conditions are expected to rise, which will present significant challenges to various logistics infrastructure (Muyambo et al. 2023).

For instance, devastating floods hit the coastal city of Durban in 2017 and again in 2022. Experts have attributed the 2022 storm to climate change, which brought nearly a year’s worth of rain in 2 days (The Guardian 2022). At the Port of Durban and Cape Town, terminal equipment (cranes and gantries) cannot be operated safely in wind speeds of over 70 kph –100 kph. Adverse weather conditions and powerful winds affect the Port of Cape Town year-round, leading to substantial delays in marine and cargo handling operations (Potgieter et al. 2020). Nearly 15% of lost time in the Port of Cape Town is because of extreme weather conditions (Rosario 2019, 2020). In general, wind velocity is expected to increase year-round in South Africa as an effect of climate change (Theron & Rossouw 2008).

This trend is further supported by the Annual State of the Climate of South Africa 2022 report. The South African weather service (SAWS) highlighted eight extreme climate events during 2022 in their report, ranging between January (2 events), April (1 event), November (4 events), and December (1 event) (SAWS 2022). Furthermore, the report highlighted that 2022 was hotter than usual for South Africa, especially in the middle of the country. The data from 26 weather stations show that the average yearly temperature was about 0.4 degrees Celsius higher than the average from 1991 to 2020, which makes 2022 one of the four warmest years since 1951. According to SAWS, the country has been warmer by 0.16 degrees Celsius every 10 years from 1951 to 2022 (SAWS 2022).

The increasing frequency and intensity of extreme weather events in South Africa have been observed by other studies as well. The Long-Term Adaptation Scenarios Flagship Research Programme (LTAS) specified that South African infrastructure is more at risk from the impacts of floods and storms (see Table 1) (Munzhedzi et al. 2016). They observed that floods in KwaZulu-Natal destroyed thousands of kilometres of roads and 14 bridges, blocked all the roads, and left 68 000 people homeless and 388 people dead back in 1987. Furthermore, the Western Cape had floods in March 2003 and April 2005, caused by cut-off lows, that cost up to R260m in damages. Storm surges have threatened the coastal provinces of South Africa many times and caused massive damage to infrastructure such as sea walls, railway lines, harbours, and coastal properties. Moreover, weather experts state that the frequency and intensity of extreme weather events in South Africa are increasing (Mpungose 2021; Munzhedzi et al. 2016).

As discussed here, weather-related incidents are becoming more frequent in South Africa, causing disruptions to the logistics network. These disruptions can significantly impact supply chains, leading to shipment delays and increased costs. By focusing on weather-related risks and taking appropriate measures to address them, companies can ensure that their operations continue running smoothly despite the challenges of changing weather conditions.

In 2015, Ho et al. (2015) defined supply chain risk management (SCRM) as:

This definition considers the impact and frequency of potential risks and includes concepts from the risk management process (Ho et al. 2015).

The key steps to managing supply chain risks are: (1) risk identification, (2) risk assessment, (3) risk mitigation, and (4) risk monitoring (Grant, Wong & Trautrims 2017; Ho et al. 2015). The mechanisms for mitigating risks include prevention of risk (e.g., stopping activities), reduction of risk (e.g., training and education), transfer of risk (e.g., insurance), or bearing the risk by the company itself (Blecker & Kersten 2006; Tarei, Thakkar & Nag 2020).

With reference to a recent publication by Tarei et al. (2020), where they examined the relationship between supply chain risk mitigation strategies and practices, most research has focused on identifying and evaluating potential risks and very few mitigating them. Therefore, the explicit focus of this research is to evaluate the ‘Transfer of risk’ strategy, specifically concerning marine cargo insurance (see Figure 1) as it relates to the research questions (RQs) of this research. In addition to risk transfer, it is also important to consider the supply chain’s resilience to certain impacts.

While there is a lack of consensus on the definition of supply chain resilience (SCR), Yang, Tian and Gao (2023) proposed the following framework: (1) the absorptive capacity – refers to the preparedness for crises and/or catastrophes; (2) the adaptive capacity – indicates the response to crises and/or catastrophes, and; (3) the restorative capacity (the ability to restore) determines the recovery from the crises and/or catastrophes and the competitive advantage achieved afterward. The three capacities correspond to temporal aspects before, during, and after an emergency and/or disaster, respectively, and form a valuable framework for assessing SCR (Yang et al. 2023). Shekarian and Mellat Parast (2021) emphasized the holistic way of managing SCRM and SCR regarding supply chain disruptions (Shekarian & Mellat Parast 2021). They further observed that by finding ways to respond to and recover from a disruption, either by restoring or improving its original function, a supply chain could enhance its resilience and reduce the adverse effects of the disruption, which is in line with Step 3 of the SCRM framework.

Therefore, according to Yang et al. (2023), measuring the SCR could be defined as quantifying the three capacities of SCR. For instance, the absorptive capacity could be quantified by diversification of suppliers, various sourcing strategies, inventory positioning, and multiple logistic networks; the adaptive capacity could be quantified based on the ability of reserve suppliers, rerouting strategies, communication, and input replacement; and the restorative capacity could be quantified by the suppliers’ reinstatement budgets (insurance) and technology reinstatement sources (Wedawatta, Ingirige & Amaratunga 2010; Yang et al. 2023). One can, therefore, conclude that marine cargo insurance is not only a risk transfer strategy through the SCRM framework but also presents a restorative capacity through the SCR definition.

In their article on data-driven analytics for cargo loss, Wu et al. (2017) proposed a ‘business analytics for cargo loss severity’ framework for processing claims data. This framework was adapted to explore the RQs of this study (see Figure 2).

This research analysed claims statistics from Companies A & B (which are among the top five marine insurers in South Africa, accounting for ±35% of the non-life insurance sector). The datasets pertain to the marine insurance portfolios of these two companies (as a case study) and did not include specific locations of the loss, so it is possible that some of the identified claims occurred outside the borders of South Africa, although it still impacts the company as a claim. The combined dataset covers 17 727 unique claims over 10 years from 2013 to 2022 and offers a comprehensive market view. The data are specific to marine cargo insurance products (owner of the goods) and does not include claims that would normally be found under marine liability products (third party services). The first objective was to create a master dataset for investigation. The datasets received contained: (1) the date of loss, (2) a description of the loss, and (3) the gross claim amount.

Initially, datasets were cleaned by removing duplicates, correcting the date format, and applying value filters. The next step was to analyse the description of the claims to identify weather and non-weather-related claims. As a result of inconsistencies in the descriptions of claims (open text field where each claims handler has textual claim description freedom), a set of keywords was developed to filter the dataset with a keyword formula. Weather-related claims and non-weather-related claims were coded as 1 and 0, respectively. Finally, a master dataset was obtained and used for further analysis. A sample dataset is presented in Table 2.

The master dataset was analysed using Excel^® to answer RQ2 and RQ3, and, with the assistance of Stellenbosch University’s Centre for Statistical Consultation, Statistica 14^® was used to explore RQ1. The master dataset was further processed for the RQ1 analysis to include a percentage (%) weather-related column to provide a ratio. This ratio is obtained by dividing the weather-related claims by the total number of recorded claims for the month. It is worth noting that the datasets from both Company A and B were reviewed separately, and the results were comparable, which provided reliability and validity for a combined master dataset.

Ethical clearance to conduct this study was obtained from the Stellenbosch University Social, Behavioural and Education Research Ethics Committee (REC: SBE).

The number of claims was plotted for 2013 until 2022 (see ‘Weather claims’ in Figure 3 and ‘Non-weather claims’ in Figure 4). As a result of the significant size of the outliers in weather claims (compared with the rest of the data), the situation will distort the subsequent analysis of trends. As a result, the data have undergone a winsorisation procedure to eliminate the impact of significant outliers. Winsorisation is a general approach to constructing robust statistics (Cheng & Young 2023). This procedure involves altering statistics by restricting extreme variation in the data to minimize the influence of outliers, as is detailed in the procedure below.

The master dataset was analysed by filtering weather and non-weather-related claims per year, including: (1) the gross value related to each claim and (2) the corresponding number of claims. It is important to note that the dataset for this analysis was not winsorised and outliers (as shown in Figure 5) will impact the overall statistics. The authors and consulting statistician agreed with this approach as the four extremes identified in 2017, 2019, 2021 and 2022 are crucial for presenting the actual state. The results from this analysis are shown in Table 6. Averages over 10 years were also calculated and are used in the discussion.

This section concludes that there was a considerable statistical difference between the average value of weather-related and non-weather-related claims in South Africa for the period investigated.

Firstly, the master dataset was analysed by filtering weather-related claims by month and keyword (keywords noted in Figure 2), including the corresponding number of claims (see Figure 9). It is important to notice that: (1) some claims trigger more than one keyword and (2) that the authors decided not to winsorise the dataset for the analysis of RQ3, despite the presence of outliers (as depicted in Figure 5) that could affect the overall statistics. This decision was made because the extreme values identified are considered essential for accurately representing the current situation, showing that these outlier events are happening more often. Secondly, the gross value related to each keyword was analysed (see Figure 10).

Lastly, the dataset was filtered by annual weather-related claims per month to identify whether the cargo is more prone to loss during certain months of the year (see Figure 11 and Figure 12). The years 2017 and 2022 were excluded as extreme outliers. The discussion of this analysis is in the next Discussion section. This section concludes that there is statistical seasonality in the dataset and specific events have a more significant gross impact than others for the years 2013–2022.