According to StatsSA (2020), the sources of household income nationally in South Africa are salaries (62.2%), grants (46.2%), income from remittances (15.6%), income from businesses (16.3%) and pensions (4.0%). The South African DHET (Department of Higher Education and Training 2017) has set the limit for families that are considered wealthy at R600 000 per year, which disqualifies them from government financial assistance towards higher education. The so-called missing middle-income group falls between R350 000 and R600 000 per year (Garrod & Wildschut 2021:484). Considering these categories in further analysis could provide valuable input towards the financial experience of various income groups.

Brewer and Gardiner (2020:197) found that lower-income families in other parts of the world were less likely to have suffered a (probably forced) reduction in expenditure and loss of jobs because of the pandemic and were much more likely to have taken on new debt or borrowed from friends or family. These findings showed that the crisis has had a more pronounced negative effect on the living standards of lower-income working-age families than those of higher-income families (Brewer & Gardiner 2020:197).

In a similar vein, Almeida et al. (2021:413) noted that the COVID-19 pandemic was likely to affect households’ disposable income in the EU (European Union), and that lower-income households would be more severely hurt. Their results, however, showed that, because of government interventions, the impact of lost income was alleviated from an average reduction of 9.3% to a reduction of 4.3% in disposable income. Coibion, Gorodnichenko and Weber (2020:1) also reported that 50.0% of the participants in their USA (United States of America) study experienced a drop in income of between $5293 and $33482. Evidently, income has been affected by the pandemic and accompanying lockdowns, and various income groups have experienced the effect in different levels of severity.

Fouché and Barnard (2018:96) set out categories of expenditure as tax and salary-related expenses, housing (rent/bond payment, municipal fees/levies, etc.), travelling (car payments, fuel, etc.), groceries (groceries, food, takeaways/eating out), medical (medical aid, payments from pocket), insurance (short term, life and disability, etc.), personal expenses of adults, personal expenses of children (school, pocket money, etc.), entertainment and holiday/vacation, debt, giving and savings. As some of these expenses are fixed over the longer term (for example, rent) and others are variable (for example, travelling), the expectation is that the categories of expenditure would respond differently to the pandemic and reduction in income.

The initial reaction of the US consumers to the pandemic was to increase their retail, credit card and food spending, following by sharply decreasing their overall spending (Baker et al. 2020:1). Coibion et al. (2020:15) found that consumer spending with regard to recreation, travel and entertainment, clothing and footwear, housing including rent and maintenance, transportation and debt payments decreased the most, while utilities, food and education and childcare decreased moderately (Coibion et al. 2020). For Great Britain, Chronopoulos et al. (2020:16) reported a large increase in grocery expenditure and decreases in discretionary spending. In Canada, household spending decreased with an average of 5.5% (Achou et al. 2020:222) although spending on groceries and housing increased. The largest decrease was seen in transport and restaurant spending (Achou et al. 2020:223). Many of these spending changes can be linked to supply restrictions during lockdowns (Coibion et al. 2020:15).

Most importantly, though, Chronopoulos et al. (2020:179) concluded that the extent of the impact of the pandemic differed across demographic variables like age, gender and income level. For example, Achou et al. (2020:222) found that spending reported by females and older respondents increased.

Ngwakwe (2020:255) analysed the bearing of COVID-19 on selected stock exchange indexes, namely the Shanghai Composite Stock Index (for China), the Euronext 100 (for EU), the Dow Jones Industrial Average and S&P 500 (for the USA). The analysis used the market index data for 50 days before and during COVID-19. The findings revealed a significant effect on two stock indexes. The Shanghai Composite Stock Index showed significant gains, and the Dow Jones Industrial Average showed an adverse effect. Although not significant, the difference in stock values for Euronext 100 and the S&P 500 showed a reduction in value. By 16 March 2020, the Johannesburg Securities Exchange (JSE) closed down 32.6% since the beginning of 2020 (Claasen 2020).

The above-mentioned decreases are expected to result in a decrease in the value of investments and specifically retirement savings. Nearly 20% – 30% of human life happens post-retirement, and, therefore, requires good planning (Janetius 2020:47). Any reduction in retirement savings would impact the savers and, ultimately, the future generations on which they may become dependent.

The term ‘savings’ is described as the portion of income not spent and which should help a household become financially secure (Makasudede 2022:759). According to Martin et al. (2020:453), the COVID-19 pandemic has forced households in the San Francisco Bay area (USA) to use their savings, which is likely to result in them becoming more vulnerable to future shocks. A Polish study revealed that savings levels decreased because of a reluctance among consumers to incur debt during the pandemic (Szustak, Gradoń & Szewczyk 2021:166), and Achou et al. (2020:233) indicated that savings were used during the pandemic to smooth consumption. Again, demographic variables were found to play a role in the use of savings. For example, Phillipson et al. (2020:5) indicated that female-owned businesses were less likely to withdraw household savings.

The last main element of personal finance affected by the pandemic is debt. One way in which consumers/households adjusted to lower income during lockdown was to defer payments on debt. Cherry et al. (2021:1) reported that 60 million borrowers in the USA were expected to miss debt payments by the end of 2021. Achou et al. (2020:224) indicated that 5.4% of homeowners missed a mortgage installment and 13.4% elected to defer mortgage payments. With interest still being charged, the total outstanding debt only keeps increasing. However, the results showed that women, older individuals and those with higher income were less likely to miss payments (Achou et al. 2020:226).

Another way to deal with a reduction in income was to take one new debt. Those who were laid off increased credit card debt, but, again, women and older people were less likely to incur new debt (Achou et al. 2020). Phillipson et al. (2020:5) similarly indicated that female-owned businesses in the United Kingdom were less likely to take on new debt during the pandemic. In contrast, Cherry et al. (2021:4) noted that, in the USA, this relief, offered by banks and other creditors to people with debt, was more prevalent among higher-income households because of their higher debt balances.

The study was cross-sectional and exploratory in nature as past research on the topic is limited. The study followed a quantitative non-experimental design from a positivist paradigm. The target population comprised individuals in South Africa over the age of 18 years. The South African population in 2019 was estimated at 58.5 million (StatsSA 2019).

The sampling frame set for this convenience sample was mainly social media (Facebook) users, seeing that the data was gathered during the pandemic when lockdown regulations applied. A study in Great Britain found that no social media platform is totally representative of the general population and that social media data cannot be generalised to any population other than itself (Blank & Lutz 2017:741). While the use of social media for data collection has potential pitfalls, well-known platforms have millions of users and should provide a sound basis for a sampling frame that would otherwise have been difficult to construct in exploratory research (Mirabeau, Mignerat & Grange 2013). However, this was the most practical solution at the time; that is, during the pandemic with lockdown regulations in place.

According to Statista (2020), the number of South African Facebook users in 2020 was 17.6 million. From the start, the author was aware that complete representativeness would probably not be obtained. Consequently, focused social media advertising was employed. The main purpose was to collect demographic information to compare with the general population and to target underrepresented variables (Shaver et al. 2019), for example, certain age groups. According to SurveyMonkey’s sample size calculator, a sample of 385 was appropriate for the study at a 95% confidence level with a 5% margin of error (SurveyMonkey 2020).

For this study, a new survey instrument was developed consisting of two sections. The first included two questions on the consequences of lockdown on income, expenditure, investments and retirement savings, outstanding debt and bank balances (including savings and emergency funds), with various subcategories, based on the work of Fouché and Barnard (2018). This section used a 5-point Likert scale: 1 – Large increase (> 20%), 2 – Increase (< 20%), 3 – Stayed more on less the same, 4 – Decrease (< 20%), 5 – Large decrease (> 20%).

The second section pertained to demographic data that were needed to describe the sample and compare the variables (including age, gender, relationship status, employment status, household income). The Statistical Consultation Services at the North-West University provided advice regarding the validity of items for statistical purposes and the interpretation of the results. The instrument was administered using QuestionPro, which provided the option of sharing the link on many social media platforms. Data were gathered and social media advertisements were placed in the third quarter of 2020, paid for by the author. The criteria were the South African population, all genders, aged 18 years and older. The QuestionPro results were monitored for the intended number of responses.

To improve the validity of the Internet survey, the study applied measures suggested by the University of Rochester (2016). For instance, QuestionPro has the functionality of tracking computer IP addresses to prevent respondents from completing the survey more than once. Also, minors were screened out. This was done by asking the participants their age and restricting the advertisements on Facebook to persons 18 years and older.

The final number of participants was 431. The captured data were analysed using the statistical package JASP 0.16.3.0. Descriptive statistics are reported on later in the article. The income and expenditure items were reduced by using PCA (principal component analysis). The KMO (Kaiser-Meyer-Olkin) measure indicated that the sample was meritorious (adequate) to allow PCA, with the KMO of 0.85 being larger than 0.80 (Kaiser 1974:35). Bartlett’s test of sphericity was also significant (Χ² = 2453.87, df = 78.00 and p < 0.001); thus, distinct and reliable factors could be extracted (Field 2013). Cronbach’s α was calculated for each of the factors, yielding an α of 0.88 for factor 1 and 0.72 for factor 2. To explore the differences among the clusters, t-tests (for two groups) and one-way ANOVA (analysis of variance) tests (for more than two groups) were conducted. Tukey’s post hoc tests were performed to determine pairwise differences among groups.

Table 1 presents the summarised demographic statistics of the participants. Regarding the age variable, those aged 50–60 years accounted for over 29.77% of the participants, followed by those aged 40–49 years (21.12%) and 30–39 years (20.36%). For the South African population, people between 50 and 60 years represent 7.78% of the population, those aged 40–49 years, 11.25% and those 30–39 years, 17.52% (StatsSA 2020).

In relation to gender, the female share of the study was larger, at 55.86%. This compares well with the 2020 General Household Survey in South Africa, which indicated 51.00% female representation of the population (StatsSA 2020). Only two participants indicated a gender other than male or female, and were therefore excluded from further analysis. With regard to relationship status, most participants were married (61.60%), followed by single persons (27.68%), which include those divorced, widowed or separated. In the General Household Survey, 28.50% of South Africans indicated that they were legally married (StatsSA 2020).

The majority of participants in the study were full-time employed (53.00%), followed by self-employed (23.00%), while the household income of most participants was less than R350 000 (43.96%). Most participants (77.19%) lived in urban areas. According to O’Neill (2021), 67.85% of South Africans lived in urban areas in 2021.

Despite the limitations of the data collection method, some form of representation of the South African population was obtained, as is evident from the above.

Participants were asked to indicate the perceived effect that COVID-19, and the lockdown had on various financial aspects, ranging from 1 – Large increase (> 20%) to 5 – Large decrease (> 20%). The percentage indication was added to give guidance as to the meaning of, for example, ‘large’. A Likert scale was used to not complicate and necessitate complex mathematical calculations.

Table 2 presents the association between lockdown and income, expenditure, investments and retirement savings, bank balance and savings, as well as outstanding debt (sorted from the highest decrease to highest increase). Entertainment and holiday expenses showed a large decrease (more than 20%), whereas bank balance and savings, value of investments and retirement savings, salary, other income and regular savings showed a decrease of up to 20%. Personal expenses of adults, giving and personal expenses regarding children stayed the same or decreased up to 20%. Debt repayments, insurance expenses, housing expenses and medical expenses stayed more or less the same. The amount of outstanding debt and grocery expenses increased the most (up to 20%).

Factor analysis was performed on the income and expenditure sub-categories that were affected by COVID-19. Factor analysis is a statistical method used to determine that items in a questionnaire were answered in a similar way and measure the same dimension (factor) (ed. Maree 2019). The results are shown in Table 3.

Two factors could be identified: 1) household expenditure and 2) income and directly related expenditure. Savings and giving (especially religious) are often expressed as a percentage of income. The tax deductibility of retirement savings is expressed as a percentage of taxable income (SARS 2022), and benchmarks on how much individuals need to save are often also expressed as a percentage of income (Biggs 2019). The concept of ‘tithing’ (giving money [sometimes literally a 10th of income] to the church) is well known in Christianity (Rakotsoane 2021), while Zakat (payment to solve poverty and help those in need) is practised in Islam (Dawam et al. 2021). This grouping (income, savings and giving) therefore makes sense. Further analysis was performed using these factors, along with the association between lockdown and investments and retirement savings, bank balances and savings, as well as outstanding debt.

Firstly, independent t-tests were performed to compare the association between lockdown and gender (male vs female) and geotype (urban vs rural). Table 4 indicates that males and females differed significantly (p < 0.5) only with respect to lockdown and their expenditure, where males (mean = 3.29) reported a larger decrease than females (mean = 3.09). A statistically significant difference was also found between rural and urban areas, with people in rural areas (mean = 3.89) reporting a larger decrease in income than people in urban areas (mean = 3.62). According to Cohen (1988), an effect size of 0.2 is small, 0.5 is medium and 0.8 is large. The effect sizes in all instances were small.

ANOVA tests were performed next, with reference to relationship status, employment status, and household income group. The ANOVA with regard to relationship status indicated no significant difference for any of the factors: that is, being married, single or in a relationship. This factor is, therefore, not discussed further.

An ANOVA was performed on the five different factors with regard to employment status. Significant differences were found with regard to expenditure (p < 0.05), income and related expenditure (p < 0.001), bank balances and savings (p < 0.001) and outstanding debt (p < 0.001) (see Table 5). Post hoc tests were used to analyse the variances further.

Regarding expenditure, there was a significant difference (p < 0.05) between self-employed persons (mean = 3.39) and unemployed (mean = 2.94) persons, where self-employed reported a larger decrease in expenditure. With regard to income and related expenditure, part-time (mean = 3.96), self-employed (mean = 4.28) and unemployed (mean = 3.84) persons reported a significantly larger decrease in income levels and related spending compared with full-time employed (mean = 3.37) persons. Retired persons (mean = 3.58) did, however, show a significant smaller decrease in income and related expenditure in comparison with self-employed (mean = 4.28) persons.

No significant difference was found with regard to the change in the value of investments. With regard to the value of bank balances and savings, full-time employed (mean = 3.65) persons reported a lower decrease than self-employed (mean 4.49) persons, who also reported a larger decrease than retired (mean = 3.69) and unemployed persons (mean = 3.79). For outstanding debt, full-time employed (mean = 2.70) persons reported a smaller increase than part-time employed (mean 1.88), self-employed (mean = 2.15) and unemployed persons (mean = 1.77). Retired persons (mean = 2.87) reported a significant smaller increase in outstanding debt compared to part-time (mean = 1.88) employed and unemployed persons (mean = 1.77).

The ANOVA performed on income levels revealed significant differences between the groups with respect to expenditure (p < 0.05), income and related expenditure (p < 0.001), bank balances and savings (p < 0.001) and outstanding debt (p < 0.001) (see Table 6). The results were, again, further analysed using post hoc tests.

Table 6 indicates that, with regard to expenditure by the various income groups, there was a significant difference between people earning less than R350 000 (mean = 3.119) and those earning more than R600 000 (mean = 3.40), with the higher-income earners experiencing a larger decrease in expenses. With regard to income, the same two groups reported a significant difference, this time with lower-income earners (mean = 3.93) experiencing a larger decrease in income compared to higher-income earners (mean = 3.46). There was no significant difference between the groups with regard to the value of investments and retirement savings.

When it comes to bank balance, savings and emergency funds, the groups differed significantly. The lower-income group (earning less than R350 000 per year) (mean = 4.17) reported the largest decrease in savings, while the higher-income group (earning more than R600 000 per year) (mean = 3.50) reported the lowest decrease in savings. The groups also differed significantly with regard to the change in their outstanding debt. For the higher-income group (mean = 2.92), there was a marginal increase in debt. The lower-income group (mean = 2.02) reported the largest increase of up to 20% in outstanding debt.