1 Introduction

The world was shocked when a novel coronavirus disease (COVID-19) struck China, specifically the capital of Hubei province, Wuhan in December 2019. Scientists have described it as a novel virus because of its uniqueness compared to previous coronavirus cases. Evidence from the World Health Organisation [1, 2] reveals that a few months after the outbreak, the virus had spread beyond Asia to multiple countries around the world, killing over 100,000 people. On March 11, 2020; WHO labelled it a pandemic. By May 11, 2020; the world had recorded 4,013,728 confirmed cases and 278,993 confirmed deaths in 215 countries, areas, or territories. By March 18, 2021; the number of global reported cases and deaths had reached 120,915,219 and 2,674,078 respectively, after 363,691,238 vaccine doses had been administered. Admittedly, the impact of the vaccines and some degree of compliance with COVID-19 protocols remain uncontested, leading to a zero number of cases as of August 7, 2023 [3].

Unlike previous infectious disease outbreaks such as Ebola that stood tall in Africa with a significant and devastating effect, this time around the impact has been less felt in Africa as compared to other regions such as Asia, Europe, and America. Several media speculations suggest that if care is not taken, Africa’s case will be worse than that of China. That is, Africa may not be able to contain the outbreak if it begins to experience China’s intensity and spread. This argument looks plausible owing to the poor health facilities, inadequate public health staff, low-income levels, poor living conditions, high levels of ignorance, corruption, etc., especially in most rural areas.

Against this background, several countries and governments in Africa have rolled out COVID-19 national management strategies in line with the WHO’s protocols. These include the closure of borders, halting international and domestic travel, the full or partial lockdown of residents, etc. These strategies have not been without associated socio-economic consequences for nations and their citizens. In most poor, lower-income, and middle-income African countries, as much as governments are concerned about the spread of the virus, they are also mindful of the economic impact of their decisions. For example, most African governments have started borrowing, and the average forecasted economic growth rate is likely to hit a record low of negative 5.1% by the close of the year 2020. Millions of people are predicted to start earning less than $1.90 a day, while the trade and agricultural sectors’ negative impact is likely to raise food security concerns [3].

In response to possible economic uncertainties, after 3 weeks of the lockdown, specifically April 20, 2020, Ghana is known to be the first African country to have lifted the lockdown in order to ease the economic burden on its citizens. Several other African countries, such as Nigeria, South Africa, etc., also followed suit. Since the lifting of the lockdown, confirmed cases and the number of deaths on record have increased in most African countries. Below is the trend of the COVID-19 confirmed cases in Africa.

As shown in Fig. 1, there was not a single confirmed case of COVID-19 in Africa until March 1, 2020. By March 2, Africa had recorded three confirmed cases. Since then, the number of cases kept rising by the day. To curtail the rising number of cases and deaths, on March 21, 2020; Rwanda was reported as the first African country to have imposed a lockdown. After Rwanda, several other African countries also imposed either a partial or complete lockdown. The enormous economic consequences saw most of these countries lift the lockdown, as earlier mentioned. Since the lifting of the lockdown, Africa has experienced a significant rise in the number of confirmed cases and deaths. On July 17, 2020; with Africa’s daily increase of 2440 confirmed cases, the highest number of confirmed cases stood at 19,719. Thus, Africa recorded a daily percentage change of 14.12%. As of July 23, 2020; Africa had recorded 642,387 confirmed cases, compared to 1,571,317 in Southeast Asia, 7,948,513 in the Americas, 3,147,860 in Europe, 1,429,084 in the Eastern Mediterranean, and 272,829 in the Western Pacific. Generally, except for the Western Pacific region, Africa would have been touted as having experienced the lowest number of confirmed cases. Nevertheless, what is worrying and calling for compliance is the rate at which confirmed cases keep rising by the day. As of October 31, 2021, the daily increase and total confirmed cases had reached 2065 and 6,151,145, respectively.

Fig. 1
figure 1

(Source: WHO [4])

Trend of COVID-19 confirmed cases in Africa (2020–2021)

Full size image

To halt the rising rate of COVID-19 cases, the development of vaccines and antivirals has become a priority consideration by scientists globally. Nonetheless, in an attempt to leave no stone unturned, classical public health and safety measures have been outlined by the WHO and adopted by most governments in Africa. These measures seek to prevent the spread of the virus while maintaining the livelihoods of the poor and ensuring the resilience of their economies. These measures include the compulsory wearing of a nose mask in public places, observing social distance, washing hands regularly with soap and running water, and so on. Wilder-Smith and Freedman [5] have argued that effective application of public health measures can contain the spread of the virus. Interestingly, while some people have purchased the certified nose masks and are adhering to the directives, for others, we observe a lack of compliance with social distancing, apathy in wearing nose masks, and sometimes ignorantly wearing unprescribed, self-made nose masks that pose health risks instead of providing safety.

This present study, which commenced in July 2020, seeks to explain the determinants of COVID-19 public health and safety measures in sub-Saharan Africa. We first generated a unique index that focuses on the public health and safety measures of individuals, public places, and workplaces. Next, the index is used as the dependent variable to investigate its determinants. The paper’s contribution is examining factors explaining COVID-19 protocol compliance across 12 African countries. For example, Amuakwa-Mensah et al. [6], used the same dataset from Geopoll and the same number of countries to investigate COVID-19 and people’s handwashing behaviour with implications for water use in sub-Saharan Africa (SSA). The primary aim of their study was to examine the effect of the level of concern about COVID-19 and handwashing behaviour. The study provided evidence that concern about COVID-19 influences handwashing behaviour in SSA. Likewise, Khalatbari-Soltani et al. [7], have highlighted the need to consider socioeconomic factors to identify the disadvantaged socioeconomic cohorts in addressing potential COVID-19 risk factors. In this light, Philbin et al. [8] and Dwyer et al. [9], have shown a gender disparity gap in COVID-19 management practices. Similarly, Baye [10], used the Demographic and Health Survey of Ethiopia to investigate factors that drive COVID-19 prevention measures in Ethiopia. Using a descriptive approach, the study observed that, apart from the wealthy and mainly urban dwellers, most households are characterised by poor water quality, sanitation, and hygiene conditions. All else being constant, we can infer that wealth and urban factors are associated with a higher probability of complying with COVID-19 preventive protocols. This finding somewhat corroborates the conclusion by Qi et al. [11], that people with higher incomes are more likely to engage in preventive healthcare behaviour. Other studies have also focused on the role and trust in institutions in driving preventive behaviour and mortality. For instance, Oksanen et al. [12], posit that institutional trust is an important factor that is associated with lower rates of COVID-19 mortality in European countries. Likewise, Han et al. [13], used a representative sample of 23,733 from 23 countries to show that higher trust in government significantly influences the adoption of health behaviour. After the introduction of vaccines, a myriad of studies has been conducted to ascertain the potency and impact of the vaccines on patients and potential patients. In Watson et al. [14], they concluded that “COVID-19 vaccination has substantially altered the course of the pandemic, saving tens of millions of lives globally. However, inadequate access to vaccines in low-income countries has limited the impact in these settings, reinforcing the need for global vaccine equity and coverage” (p. 1). In the absence of equity and coverage, behavioural adjustment is a critical management option to consider. In the present study, we acknowledge the paucity of related studies and attempt to contribute to filling the existing gaps on protocol compliance in the literature.

2 Methods

2.1 Data and survey procedure

In such a pandemic, it will be highly unethical to constitute a team of fieldworkers to undertake fieldwork, irrespective of the motivation. Thus, several data-driven empirical studies have relied on online surveys using structured questionnaires during such periods. Given time and cost constraints, this study relied on Mobile Accord’s (GeoPoll, Incorporated) representative sample survey dataset, collected from April to May 2020. The first-round dataset was collected between April 2 and April 9, 2020; while the second-round dataset was collected between April 24 and May 2020. Twelve countries from sub-Saharan Africa, namely Benin, Côte d’Ivoire, Democratic Republic of the Congo, Ghana, Kenya, Mozambique, Nigeria, Rwanda, South Africa, Zambia, Tanzania and Uganda were used. The country selection is based on data availability in both rounds of the data collection process. Although SSA comprises 44 countries, 12 are used based on data availability. Hence, these countries are used to represent the entire SSA, which, based on the survey weighting, has been described by Geopoll as representative and can therefore be generalised. It is important to acknowledge the potential statistical skewness within the 12 countries; nevertheless, this skewness also reflects a reasonably diverse statistical distribution, thereby implying that conclusions drawn from these countries might effectively capture regional perspectives.

GeoPoll reports that the respondents self-administered the questionnaire. It presupposes that recruitment was not based on compulsion but on choice. Hence, voluntary responses were obtained during the survey. According to the number of questions reported on their website, about 27 were administered. These included questions on concerns about coronavirus, preventive practices (management strategies), food and security issues, consumer behaviour, trust in the government’s ability to handle the virus's spread, and the respondent's demographic characteristics. The main source of data for this study is GeoPoll.Footnote 1 Details of the variables are presented in subsequent sections.

2.2 Empirical strategy

This study uses preventive practices as management strategies or safety protocols/practices. This is done by aggregating the six preventive strategies identified in the dataset to form a unique variable called the preventive practices (strategies or protocols) index. The index's components are social distance, hand washing, working from home, increasing personal hygiene, avoiding public transport and avoiding public places. Given that the dependent variable is binary, we present a pooled cross-sectional logit model. This is estimated together with its corresponding predicted margins (delta-method). In line with Amoah and Addoah [15] and Amoah et al. [16], the logit model is represented as Eq. 1:

$$p\left( {preventive\;behaviour\;index\left[ y \right] = 1{|}X} \right) = f\left( {\alpha_{0} + \alpha_{1} {\varvec{X}}} \right)$$
(1)

where the outcome variable represents the probability of engaging in COVID-19 preventive behaviour. This preventive behavioural variable does not measure what one seeks to do or intends to do, but what the respondent actually does. If the respondent engages in COVID-19 preventive behaviour, it is coded as 1; otherwise, 0. All six possible preventive measures were first aggregated and re-constructed as an index to represent a composite measure of preventive management protocols or practices. Following Galasso et al. [17], and Hughes and Amoah [18], the average index is computed as Eq. 2:

$$\overline{X} = \frac{1}{n}\mathop \sum \limits_{i = 1}^{n} x_{i} = \frac{1}{n}\left( {x_{i} + \cdots x_{n} } \right)$$
(2)

This average composite index appears as a dummy variable, where one (1) means adhering to preventive management practices while zero (0) means otherwise. Again, \({\varvec{X}}\) represents a vector of independent variables. This includes the age of the respondent, which is expected to influence behaviour. In social science literature, this is commonly used to measure experience. Older cohorts with much experience in life may have different expectations relative to younger cohorts, who can easily influence their behaviour. A priori, we expect that older people with much experience regarding the impact of various pandemics and epidemic outbreaks in Africa and their impact on life will be very compliant with COVID-19 preventive measures. Generally, older people have been observed to be more vulnerable and are prone to a higher risk of severe COVID-19 disease relative to younger people.

Again, this study included a measure of economic concern as an independent variable. The variable measures the extent to which people are concerned about the expected economic impact of the COVID-19 pandemic on businesses and lives (eco). The responses were ranked from not concerned to very concerned. As expected, respondents who are very concerned that the impact of the pandemic would be devastating are expected to change their behaviour to influence members of their households. On the other hand, respondents who have not yet felt the impact or expect any impact on their households may not be concerned; hence, they may choose to be indifferent or not comply with the preventive measures.

Gender differences matter when making choices; hence, gender is included as an independent variable. Gender plays an important role in shaping preventive healthcare behaviours (Philbin et al. [8]). The empirical literature (see, for example, Dwyer, et al., [9]) has shown that, on average, females or women are more risk-averse in their decisions while males or men are more risk-loving. This variable is important because engaging in preventive behaviours shows that how a person is concerned about the risk of being infected or dying because of exposure to coronavirus. Based on the disparity in gender tendencies, we expect that women will engage more in preventive behaviours than men. Here, respondents were voluntarily made to indicate their gender, of which males and females were coded as one (1) and zero (0), respectively.

Also, practical experience is one of the ideal ways to acquire first-hand knowledge about behavioural differences. Lessons like this cannot ordinarily be taught except through practical experience. For example, it will be extremely difficult for a patient’s pain to be holistically appreciated without experience. This variable was obtained after the respondent indicated that he or she had tested positive for COVID-19 or otherwise. Those who tested positive were coded as 1, while those who tested negative, not sure, or not tested were treated as otherwise and coded 0. We expect that those who have tested positive will engage more in preventive behaviour as compared to the others.

Another variable of interest is the interaction between those who have tested positive and gender. Thus, we expect that risk averse respondents who have tested positive will be highly inclined to engage in preventive behaviour as compared to risk-loving respondents who have tested positive.

Geographical location influences a person’s level of exposure, perspective, taste and preference or choice. From a theoretical perspective, we expect people who have had much exposure and a broader perspective of activities and opportunities to make choices that are different from those with a narrower perspective. In this study, respondents who are geographically located in urban areas are contrasted with those who reside in rural areas. Generally, rural residents are believed to have a narrower perspective on modernity and current developmental issues as compared to urban residents. In this study, we expect urban respondents to be associated with a higher probability of engaging in pro-preventive behaviour relative to rural respondents.

The role of institutions in behavioural choices cannot be overemphasised. According to North [19], institutions shape human interactions and their behavioural strategies or choices. We define institutional quality as those systems that seek to improve effectiveness and efficiency in society. Strong institutions are expected to hold humans illegitimate or unprofitable choices in check for the betterment of the whole society. So, we considered trust in financial institutions (Banks) as a driver of COVID-19 preventive practices. That is, if the financial sector demonstrates their support to both the government and the private sector by complying with the moratorium request as well as financially supporting the private sector in the production and supply of PPE kits and other needs, these essentials will be readily available on the market for easy access. However, where the financial institutions are not supportive, access to PPE kits and other needs will be difficult and expensive due to scarcity, which can easily discourage most Africans because of their lower income characteristics.

The role of government and political will is very important in making societal decisions that seek to influence individuals’ behaviour. In addressing COVID-19 challenges, most of the decisions made by government appear like a double-edged sword. In one breadth, it may end up undermining the rights and liberties of individuals, while in another breadth, it has the overarching aim of seeking the interests of the entire society. Such crossroads decisions require a bold and courageous government that, regardless of political pressure from opposition political parties and voters, will be consistent and push the entire society’s interests over the interests of the teary few. Here, we expect that respondents who have seen the achievements of the government, in addition to what it is capable of, will repose their trust in him. Such a trusted government can easily influence the preventive behaviour of the populace. Besides, through the government’s role, subsidies can be made available to the private sector to increase the supply of PPE kits, which will promote access and compliance with COVID-19 preventive protocols.

Another institution that plays a key role in ensuring the protection of the minority (for example, the poor and marginalised in society) is the Non-Governmental Organisations (NGOs). In this study, we included this variable in the model to ensure that no one is left out in the spread of COVID-19 preventive measures. By expectation, the more NGOs we have, the more people will be educated through advocacy to use PPE kits and comply with preventive protocols.

Lastly, in as much as the public sector is observed to play an important role in Africa, averagely, the private sector is relatively bigger as it employs about 55% of working age individuals, 90% of available jobs, and accounts for 80% of total consumption (Stampini et al. [20]). We expect the private sector to play a key role in promoting demand and supply of PPE kits, which will facilitate compliance with preventive measures.

2.3 Summary statistics of modelling variables

From Table 1, the outcome variable (index) is an average measure of all six preventive management protocols or practices. The index is constructed as the average of the six sets of dummy variables following Galasso et al. [16], and Hughes and Amoah [17]. This average composite index was defined to equal one (1) if the index is above 0.5 and zero (0) if otherwise. This appears as a dummy variable where one (1) means adhering to the preventive management practices while zero (0) means otherwise. The mean value of 0.72 and a median of one suggest that a significant proportion of the respondents religiously abide by the protocols. An estimated 46%, 42% and about 88% of respondents trust that the NGOs, private sector and Government, respectively, are responsible for providing assistance during the COVID-19 pandemic. About 71% do not trust the role of the financial sector. Eco has a mean of 4.29, which suggests that on average, respondents are highly concerned about the economic impact of the COVID-19 pandemic. The sample constitutes about 55% males and 45% females. Out of this, about 10% indicated that they tested positive, while the majority, constituting about 67% are urban dwellers. As one would have expected, the sample looks very youthful, with an average age of about 31 years.

Table 1 Descriptive statistics
Full size table

2.4 Diagnostic test

Given that the variables used for our modelling contain several demographic, institutional and country-specific variables it is expected that some of the variables, will themselves correlate. Now, the presence of such correlations may lead to multicollinearity. This problem makes it difficult to ascertain the effect of an individual variable on the outcome variable. To test for the presence of this problem, we used the pairwise correlation test (see Table 2) and found that the highest degree of correlation coefficient of approximately 72% exists between NGO and private, while the lowest correlation coefficient (in absolute terms) of 0.04% also exists between gender and private. Against this background, NGOs and private companies are not estimated in the same model. Model 1 of Table 3 included NGO, while Model 2 included private companies. That is, we do not expect severe multicollinearity challenges in our modelling; hence, the results are valid on this score.

Table 2 Pairwise correlation test results
Full size table
Table 3 Cross-sectional pooled logit regression results
Full size table

Next, we included country-specific dummies to control for the differences in the random error term across the different countries adhering to COVID-19 protocols. Then again, to deal with community-level correlations in the random error term, we clustered the standard errors at the primary administrative level. This seeks to ensure that inherent correlations in the random error term of the rate at which respondents complied with the safety protocols within the same community are addressed. That is, we do not expect severe heteroskedasticity concerns in our modelling, and we are confident that our estimates are valid on this score.

In order to further ensure that our model is properly identified, we used the Wald chi-test to find out whether the explanatory variables significantly explain the outcome variable. With highly statistically significant results, we conclude that the explanatory variables together explain the COVID-19 management protocol index and that their coefficients are not equal to zero.

3 Results and discussion

A traditional factor that promotes preventive healthcare utilisation is the respondent's experience over time. According to Amoah and Addoah [15], a person’s age can affect his cognitive capacity to recall experiences. All other things being equal, older cohorts are more likely to recall many past life-ending epidemics than younger cohorts. Similarly, older people, especially those with multimorbidity, appear more vulnerable and stand a higher risk of severe COVID-19 disease than younger people (Zaninotto et al. [21]). Based on this, we expect older cohorts to engage positively in COVID-19 preventive protocols. However, the contrary is found. We have evidence of a negative and statistically significant relationship between the age of the respondent and COVID-19 preventive protocols. That is, relative to the younger cohort, older people are less likely to engage in COVID-19 preventive protocols. The associated coefficient of the marginal effect obtained is 0.0020. This suggests that, if age increases by one year, the probability that a respondent will engage in COVID-19 preventive protocols will decrease by 0.20%. This finding is counterintuitive yet plausible. By explanation, first, given that older people are more vulnerable to severe COVID-19 disease and its associated consequences, they are more likely to stay away from COVID-19 hotspots or highly infectious environments and activities than the younger cohort. Second, it can also be argued that access to credible information from credible institutions (for example, the website of the WHO) regarding the data, causes, effects, and preventive measures of COVID-19 is very pronounced on technologically driven platforms. Available data shows that in 1999, as few as 11% of Africans had mobile phone coverage (a medium to access information regularly); however, this had risen to 60% by 2008 (Aker & Mbiti, [22]). Indeed, with technology, the younger, technologically savvy cohorts, the youth, or ‘digital natives’, are privy to more information that drives their behaviour than the older cohorts. Relative to the younger cohort, information is asymmetric among the older cohort, hence their lower likelihood to engage in preventive practices.

One of the impacts of COVID-19 has been on the output of economies (Kugbey et al. [23]), and an import-driven continent like Africa has not been spared. That is, all else held constant, the negative shocks to the manufacturing world will trickle down faster than expected to hurt the economies of Africa. The OECD [24] explains that, for the global economy, the impact of the COVID-19 pandemic will cause a negative shock to the supply of goods and services, which in effect will force many factories to close down, thereby unsettling global supply chains. During the lockdown, most African countries relaxed the lockdown for economic reasons. Against this background, this study seeks to ascertain whether individuals' concerns regarding economic repercussions will exert a constructive influence on their adoption of preventive behaviours. Here, the respondents were asked to rank the extent to which they are concerned about the economic impact of COVID-19. We expected that those who have been severely affected would indicate that they are ‘very concerned’, while those who have been less affected would indicate that they are ‘not concerned'. Using ‘not concerned’ as our reference category, we have evidence of a positive relationship for all categories, with concerned and very concerned’ being statistically significant at 1% and 5% levels, respectively. In much more detail, relative to those who are ‘not concerned’, those who are ‘concerned and very concerned’ are associated with a 4–9% probability of engaging in COVID-19 preventive management practices. This evidence is consistent with Amuakwa-Mensah et al. [6], who posited that concern influences behaviour.

Consistent with expectations, gender is negative and statistically significant. The associated coefficient of the marginal effect obtained is − 0.0169. This suggests that being a male relative to a female decreases the probability of engaging in COVID-19 preventive protocols by 1.69%. The plausibility of this finding is underscored by the fact that, generally, females are relatively risk-averse and are more likely to engage in preventive measures than males (Harrant and Vaillant, [25]). This evidence contradicts some of the risk-taking-related studies that found males to be relatively more risk-averse and more willing to spend to avert risk [15, 26, 27]. Admittedly, these authors explain that their results could be explained by the household social and economic responsibilities of the males in the study areas. Interestingly, these opposing views have been upheld by Nelson [28] who argues that gender effects are ambiguous and overlap. It is also important to mention that, in some instances, no statistically significant gender effect is found (see Amoah [29]).

Contrary to expectations, those who tested positive were found to be statistically insignificant. Indeed, people who reported to have tested positive were perhaps under quarantine, and thus they did not consider the outlined preventive practices necessary under quarantine.

To better understand the gender differences with respect to respondents who have tested positive, we found a negative and statistically significant relationship. We report a marginal effect of − 0.0539 for the interaction between gender and test. This means that, relative to females, if positive-tested males increase by 1%, COVID-19 preventive management practices will decrease by approximately 5%. That is, relative to females who have tested positive, males who have tested positive are associated with a lower probability of adhering to COVID-19 preventive practices. This is justified by the fact that females are observed to be risk-averse and would want to comply with the preventive practices more religiously than their male counterparts.

Several studies (e.g., Casey et al., [30]) have shown that urban residents are more likely to be associated with certain preventive health services than rural residents. That is, geographical location plays a crucial role in one’s ability to engage in preventive healthcare practices. The results show a positive and statistically significant relationship between urban location and COVID-19 preventive management protocols. The reported marginal effect is 0.0226 for urban geographical locations. This suggests that, relative to rural areas, being in an urban area is associated with a 2.26% increase in the probability of engaging in COVID-19 preventive management protocols.

Institutions and their readiness to provide the needed support are very important for quality preventive healthcare delivery. Given the absence of data on access to quality preventive healthcare delivery, we used other measures of institutional quality in the respective countries. The first institution considered is the financial sector because of its expected supportive role in the PPE supply chain. Respondents were asked if they trusted the commercial banks to assist during the outbreak of COVID-19. This is used as a proxy for institutional quality, following Wang and Gordon [31]. Our finding shows a positive and statistically significant relationship between trust in commercial banks’ ability to assist during the period of the COVID-19 outbreak and COVID-19 preventive management practices. With a marginal effect of 0.1095, it suggests that those who trust in commercial banks’ ability to assist are associated with an approximately 11% probability of engaging in COVID-19 preventive management practices. By implication, this study has revealed the impact of commercial banks activities in promoting COVID-19 preventive practices in Africa.

Another institutional variable included in the model is the role of government. This is considered relevant as it examines the role of government in societal decisions towards engaging in preventive management practices. We have evidence of a positive and statistically significant relationship between trust in the government’s assistance to society during the period of the COVID-19 outbreak and COVID-19 preventive management practices. We report a marginal effect of 0.0097, which implies that those who trust in the government’s ability to provide the needed assistance during the pandemic are associated with approximately a 1% probability of engaging in COVID-19 preventive management practices. This finding acknowledges the impact of the government’s role on human behaviour towards promoting COVID-19 preventive management practices in Africa.

Most NGOs in Africa can be described as the bridge between the high/rich and the low/poor. Their role is important in promoting equitable distribution of scarce PPE kits and other essentials for survival. This study finds a positive and statistically significant relationship between trust in NGOs for providing the needed assistance during the COVID-19 pandemic and preventive management practices. From the marginal effect estimates of 0.0461, it implies that those who trust in the NGO’s ability to provide the required assistance during the pandemic are associated with an approximately 5% probability of engaging in COVID-19 preventive management practices. Our result provides evidence that those who trust the NGOs believe that they have generally been involved in promoting preventive management practices in Africa.

In Africa, most of the NGOs are privately owned. Thus, we suspected a possible collinearity between NGOs and the private sector. A pairwise correlation test undertaken shows a correlation coefficient of approximately 72%. Hence, we estimated model 1 by including the NGOs and model 2 with the private sector's assistance. Similar to the NGO results, we have a positive and statistically significant relationship between trust in the private sector’s assistance during the pandemic and COVID-19 preventive management practices. The results show a marginal effect estimate of 0.2511. That is, those who trust in the private sector’s role in providing the expected assistance during the pandemic are associated with an approximately 25% probability of engaging in COVID-19 preventive management practices. This result also lends support to the expectation that the private sector has positively impacted preventive management practices in Africa.

The method of computing a simple arithmetic average of a binary response may not capture the overall weight that respondents put on each individual component. Hence, an alternative approach is to use Principal Component Analysis (PCA), which is widely used for generating indices. Specifically, the PCA is a method used for generating a composite index from a set of different components (variables). For example, with all the various preventive measures for COVID in the dataset, it is possible to use PCA to generate a composite index that represents the overall preventive strategies of a respondent in the dataset. As part of validating the pooled cross-sectional logit results of the study, we used the PCA results in Tables 4 and 5 and found that, following the eigenvalues, the highest value, which is component one, has a proportion of about 0.24. The next highest component is component two, with a proportion of about 0.21, followed by component three in that order. This implies that about 24% and 21% of the variations are accounted for by components one and two, respectively. We admit that these are not as high variations as one would have expected. Indeed, they are not high yet informative enough for our purpose. So, we further estimated the predicted values of the component to obtain a continuous variable. Now, the predicted variable is used as our dependent variable, while the same covariates are used to investigate the robustness of our model. Unlike the logit model, this time we investigate the relationship using the pooled ordinary least squares (POLS) method. The results, as presented in Table 5, provide evidence to support our earlier findings vis-à-vis the signs and significance that our results are robust across the different techniques.

Table 4 Eigenvalues from the principal component analysis
Full size table
Table 5 Principal component analysis predicted scores
Full size table

3.1 Robustness checks

We start with age, as in the earlier case. The results in Table 6 show evidence of a negative and statistically significant relationship between the respondent’s age and COVID-19 management practices. The results based on the concept of experience are counterintuitive because our study suggests that as people age, their willingness to engage in COVID-19 preventive practices declines. This is consistent with our earlier results.

Table 6 Pooled ordinary least squares (POLS) regression results
Full size table

Again, in line with our earlier results, respondents who are “very concerned and concerned” about the economic impact of the COVID-19 pandemic are more likely to engage in COVID-19 management practices as compared to those who are unconcerned.

Similar to our earlier results, an interaction between gender (males) and those who have tested positive was found to be negative and statistically relevant in driving COVID-19 management practices. That is, relative to females, male respondents who have tested positive are less likely to engage in COVID-19 management practices.

Also, akin to our earlier finding, urban areas exhibit a positive relationship with engaging in COVID-19 management practices, though irrelevant in this case. Unfortunately, the study is unable to explain why the robustness evidence has the right sign yet has failed to converge in statistical significance.

Moreover, we expect the financial sector to support PPE businesses during such pandemics. Analogous to our earlier evidence, trust in financial institutions has a positive and statistically significant effect on COVID-19 management practices.

Likewise, the role of government in such difficult times is very crucial. This consistently exhibits a positive and statistically significant effect on COVID-19 management practices.

The results from NGOs and the private sector align with what we had estimated before. Thus, both results show a positive and statistically significant estimates. Thus, the role of NGOs and private firms is found to drive COVID-19 management practices in Africa.

Overall, we show overwhelming evidence that the results in Table 3 have been validated to a very large extent by the results in Table 6.

3.2 Heterogeneous effect of aggregated COVID-19 management practices index

The paper further investigates the heterogeneity of the drivers of our constructed COVID-19 management practices Index. In Table 14 (see Appendix), with the index as the dependent variable, the drivers are again estimated for males and females, respectively. Generally, the results are very similar to the earlier results, as reported in Table 3. To begin with, we have evidence that NGO and Trust are positive and statistically significant for both males and females, with marginal changes in the coefficients. This implies the absence of gender heterogeneity for NGO and Trust. Thus, both males and females believe that NGO and Trust are more likely to provide assistance towards COVID-19 management practices (index). Further, we observe gender heterogeneity in relation to Government.

Similarly, age is negative and statistically significant. That is, the younger cohort is more inclined to practise the protocols than the older cohort. However, this is not sensitive to gender disparity. Again, while the number of people who tested positive (Tested) is found not to be relevant in both models, being in an urban area relative to a rural area is found to be positive and statistically significant for females but not for males. That is, females in urban areas are more likely to comply with the index, however, this is immaterial in the case of males. Also, relative to lower levels, higher levels of economic impact are found to drive the index. This evidence is statistically significant and corroborates our earlier results. Furthermore, the Wald chi-test is used to examine gender disparity across variables in the model for the Index. The results as presented in Table 7 show that, given the independent variables, gender disparity alone is irrelevant in explaining the COVID-19 management protocols (Index).

Table 7 A Test of COVID-19 management practices index by gender across variables
Full size table

3.3 A test of disaggregated COVID-19 management practices by gender

Here, we use both parametric and non-parametric tests (see Tables 8, 9, 10, 11, 12, 13, 14 in Appendix) to examine how the effect of disaggregated COVID-19 management practices vary across gender. The evidence from Philbin et al. [8] and Dwyer et al. [9] suggests that we should expect gender-disparity in COVID-19 management practices. However, what is unknown is whether the protocols together as a composite or individually exhibit gender-disparity. First, in Table 8, our results show that there is a statistically significant relationship between females and males in avoiding public transport in line with COVID-19 management practices. This implies that differences in gender play a role in the decision to patronize or avoid public transport during the COVID-19 pandemic. Averagely, males are more likely to avoid public transportation relative to females. This is plausible because, for most African countries with poor pedestrian access to motorable road, perhaps males can easily take the risk by using unconventional means of transport as compared to females. Although further investigation may be needed to explain this further, it is important to point out that males are more likely to take road risk than females (Pawlowski [32]).

Also, in Table 9, our results indicate that there is no statistically significant difference between females and males in avoiding public places. Thus, the decision to avoid public places is not gender sensitive. This is most likely the case, if your source of socio-economic or health survival is within the context of a public place. In such cases, gender disparity is immaterial.

Similarly, Table 10 shows no statistically significant difference between females and males in their choice to work from home. This is intuitively justified on the grounds that, if a firm mandates its workers to work from home, whether you are a male or female, such a decision is made at the firm level regardless of gender. Thus, gender disparity is irrelevant in a firm’s decision for their workers to work from home.

Besides, in Table 11, we find evidence of a statistically significant difference between females and males in practicing good hygiene. Our results show that, on average, males are marginally more likely to practice good hygiene than females. This is counterintuitive; however, it may sound plausible if the man is the head of his household and bears the cost of practicing bad hygiene (ill-health), then the man will demonstrate leadership by practicing what he preaches.

Again, in Tables 12 and 13, our study finds a statistically significant difference between females and males in practicing social distance as well as regularly washing their hands. The evidence suggests that, on average, females are more likely to practice social distance and handwashing than males. The latter result corroborates Amuakwa-Mensah et al.’s [6] finding that women wash their hands more regularly than men. Again, females are more risk-averse than males and would want to socially keep their distance as a risk-averting mechanism.

4 Conclusion

This empirical study shows that despite the rising trend of COVID-19 cases, the fear surrounding the spread and its impact across the world, especially in sub-Saharan African countries, this empirical investigation reveals a noticeable tendency towards complacency in strictly adhering to all the COVID-19 management protocols as prescribed by governments. Thus, the study sought to investigate the factors that may account for this unusual behavioural response. To achieve this, we used a unique secondary dataset of 12 sub-Saharan African countries and generated a unique index that aggregates six COVID-19 management protocols. The study found that younger people were more inclined to comply with the safety protocols than their older cohorts. Again, people who were more concerned about the economic impact of the pandemic were associated with complying with the protocols. Females who have tested positive were more inclined to comply with the protocols, unlike males. The urban respondents were also found to be more compliant with the protocols than the rural respondents. All institutional variables were observed to play significant roles in influencing the behaviour of respondents in complying with the COVID-19 pandemic management protocols.

Further, we investigated gender differences across the disaggregated COVID-19 pandemic management protocols and found that with respect to avoiding public transport, hygiene, handwashing, and social distance, females behave differently in complying with the protocols than males. However, in relation to avoiding public places and working from home, we did not observe significant differences in the behaviour of females and males.

In as much as some efforts have been made towards complying with the COVID-19 pandemic management protocols, especially regarding individual protocols, much needs to be done in the call for all the protocols to be adhered to. That is, we acknowledge that a big gap still exists in Africa when it comes to complying with all six COVID-19 pandemic management protocols used in this study. We recommend policies that focus on demographic factors, institutional measures, and country-specific factors to drive compliance with COVID-19 management protocols.

Although this study produces valid estimates relevant for policy decisions, it must be acknowledged that it is not without its limitations. For example, the use of an online survey by Geopoll and the number of sub-Saharan African countries used are not without their unique limitations. Nonetheless, it goes without saying that the survey tool used and the number of countries reached at the height of the pandemic are reasonable and could not have been overemphasised.