1. Introduction

Trust may be defined as dependence or faith on the integrity, competence, or certainty of an individual or entity. [1] asserts that trust is an expectation derived from a group’s cooperative behaviour grounded in shared norms. Trust connects citizens to their representatives and state institutions established to serve them, making its attainment challenging in practical terms. These points of view highlight several critical aspects of the acts, actors, expectations, and the demand and supply that must all be aligned for trust to be established. This relationship demands a thorough understanding of both supply and demand in order to effectively predict and engage in trust-related activities.

Globally, citizens obtain services from institutions- mostly state institutions. In Africa, non-state institutions such as religious organizations, traditional leaders, and neighbours play a significant role in shaping individual and social outcomes. As a result, it is acceptable for these institutions to command trust, and for their service recipients to have confidence in them. Trust signifies democracy, and democratic governments encourage economic success, stability, peace, civic responsibility, and lawful behaviour. This position resonates with [2] who opine that trust is perhaps the fundamental component of social capital, essential for social integration, economic efficiency, and a predictable, democratic existence. Consequently, we may all agree that increasing trust in institutions, particularly governmental institutions, has various benefits for both individuals and society. Initiatives to build trust are crucial, but so are appropriate assessments, which reveal strengths and weaknesses, aiding in the assessment and prediction of political and social outcomes for specific groups or nations.

Relatedly, governments and non-state entities have increasingly sought to assess the overall level of trust in institutions. Since the 1950s, institutional trust has been assessed and analyzed [3] as a crucial element of citizens' well-being. In Europe, various authors have assessed trust in European institutions utilizing Eurobarometer data. [4] have constructed a formula that assesses net trust, by subtracting the percentage of those who trust from those who do not trust the institution. This still yields distinct elementary indicators. The Afro-barometer conducts cross-sectional surveys throughout Africa, serving as a significant source of various metrics on trust. However, measuring trust holistically is still complex, as current metrics fail to fully encapsulate the concept's elusive, multi-dimensional, and dynamic nature. Additionally, distinguishing between active distrust and mere lack of trust is challenging [5] as it is also difficult to ascertain whether citizens in various countries interpret the underlying concepts of trust similarly. This presents significant issues in contextualizing trust and the impact that autocratic or democratic administrations may exert on survey response patterns. As such, [2] contends that metrics may be skewed when citizens lack the requisite knowledge to assess the functioning of public institutions and incumbent leaders. In our view, trust is a multidimensional construct that is influenced by both exogenous and endogenous factors, as well as cultural norms, normative perspectives, and judgments of institutional efficacy. As a result, trust functions as a component within a system of interrelated aspects targeted at the effective operation of governance, and we quantify it using a composite indicator. Individual, cultural, state, or non-state variables, as well as foreign influences, may be included in the numerous components inside a system, and their individual and collective measures are critical to understanding the multidimensional nature. This is justifiable as composite indicators (CIs) are frequently employed for longitudinal and cross-sectional performance monitoring, comparisons, communication, and policy analysis as previously observed in [6,7,8,9].

This paper offers a unique contribution to the body of knowledge by demonstrating a comprehensive assessment of a multi-faceted construct—trust—while effectively capturing the diverse sources of authority within African contexts. This holds significant importance in Africa, where formal state institutions frequently exist alongside informal or traditional ones, and are in a state of ongoing development. A composite indicator of institutional trust in Africa offers a thorough, context-aware, and comparable measure that encompasses both formal and informal institutions, thereby providing policymakers and scholars with a more nuanced perspective on governance and legitimacy throughout the continent.

This study builds a composite indicator of trust in Africa utilizing the Afro-barometer round 9 dataset gathered from 39 nations. We correlate the scores with established country metrics, like the corruption perception index, as a part of external validation of the findings. This article is structured into five sections. Section two outlines the conceptual framework for the composite indicator, whilst Section three outlines the methods utilised in this study. Section four presents the results, whereas section five is the discussion and conclusion of the article.

2. Conceptual Framework for a Composite Indicator of Trust

We categorized trust into three components: trust in state institutions, trust in non-state organizations, and individual-level trust. State institutions are defined as entities that derive financial resources from the national budget and are obligated to provide services to all citizens, irrespective of political affiliation, creed, beliefs, or ethnicity. They constitute the government and are authorized to govern a nation. These institutions are anticipated to deliver services consistently and impartially, in accordance with established national laws and mandates, demonstrating a degree of permanence and universality in minimal operational criteria. In our analysis, we classify the presidency, parliament, judiciary, police, military, local governments, and election commission as state institutions. We view non-state institutions as entities that do not receive financial support from the government but are established organizations that enhance and impact governmental operations by significantly contributing to service delivery, social cohesion, and development. They are not included in the official governmental framework and may consist of traditional leaders, religious authorities, the ruling party, and the opposition. At the individual level, we examine persons themselves (interpersonal trusts), which typically manifest as naturally occurring structures such as ethnicity, religion, neighborhood, and familial relations. Incorporating them into the measurement model is essential, as the primary unit of a state is a household comprised of individuals. Events occurring at the family level mirror those at the collective national level. Researchers such as Arrow and Rudziewicz, as cited by [10] assert that numerous developing countries and enterprises are unable to expedite economic development due to a deficiency of mutual trust, thereby justifying the inclusion of its variables as a dimension in the composite indicator. The conceptual framework is shown in Figure 1.

Figure 1. Conceptual Framework for the Composite Indicator of Trust

Dimensions are sub-groups of the CI that resemble its layered structure, hence facilitating comprehension of the underlying forces driving the CI [11]. The significance of establishing dimensions lies in facilitating the assessment of relative weights among various components. A crucial factor in validating this selection is that alterations in dimensions must exhibit a discernible, linear correlation with the CI. The elementary indicators encompassed under each of the three dimensions is depicted in Figure 1. Whereas, a change in the elementary indicators causes a change in the dimensional sub-index and ultimately a change in the CI, it is important to acknowledge the independence among the elementary indicators but also the dimensions. This implies that a change in one elementary indicator does not create a change in another, likewise as for the dimensions.

3. Methods

3.1. Source of Data

We utilized the Afro-barometer Round 9 data set to construct the CI. Afro-barometer is a Pan-African, non-partisan survey research network that delivers dependable statistics on African experiences and assessments of democracy, government, and quality of life. The survey comprises a nationwide probability sample of 1,200 to 2,400 persons (aged 18 and older) that produces country-level estimates. Data are gathered via in-person interviews and are adjusted to provide national representativeness. The data for the 39 countries is derived from 53,444 interviews. The data employed to create this composite indicator were previously analyzed and obtained from MS Excel applications through the Afro-barometer analysis tool available on the Internet (https://www.afrobarometer.org/data/data-sets/. The survey included the question: “To what extent do you trust each of the following, or are you insufficiently informed to provide an opinion?” We queried and combined the nationwide percentage of respondents who answered either “somewhat” or “a lot” to derive the values utilized in this study. Seventeen (17) elementary indicators, classified into three dimensions, were utilized in the initial stages of this investigation. In the inevitable instances of addressing missing data, we allocated the median value to the respective country for that variable.

3.2. Measurement Model

The measurement model provides a linkage between the phenomenon to be measured (latent variable) and its indicators (elementary indicators) [12]. The CI construction is interpreted as an exploration for a latent variable, initiated from the observed variables. The application of an appropriate measurement model enhances the content validity of the CI and facilitates the accurate depiction of the structural relationships in which the constructs are situated [13]. Maggino, as referenced in [12], contends that a proper comprehension of the link between indicators and the latent variable enables the accurate identification of the technique for aggregating individual indicators. Accordingly, we employed a formative measurement model in which causality flows from the elementary indicators to the composite indicator, and a change in the phenomenon does not necessarily entail modifications in all its measures. The formative model is as presented in Equation 1.

(1)

Where: is the coefficient capturing the effect of indicator X_i on R and is an error term.

This model resulted in aggregation occurring in two phases, as the empirical behavior of the trust phenomenon indicated that alterations in the elementary indicators (according to the dimensions) would result in modifications to the trust CI values. However, a change in one dimension does not inherently induce a change in another dimension, which is a critical assumption of the formative model.

Notation

The initial dataset is organized as records representing countries, with fields corresponding to indicators or variables, and their combination yielding observations or values. This study employs the following notation: Let the observation for indicator j (j=1, …, 17) belonging to country i (i=1, …, 39) be denoted as X_ij which may be measured on ratio scale. Let w_k denote a weight attached to a dimension k (k=1, …, 3) where, .

3.3. Normalization, Weighting, and Aggregation

Normalization was performed to render the data comparable prior to the application of weights. The variables were normalized via the minimax method, the formula for which is presented in Equation 2. Minimax normalization adjusts the variable to a range of [0,1] by removing the minimum value from the indicator value and dividing the result by the indicator range. This method does not reshape the distribution of the original variable.

(2)

Weights were assigned to the normalized variables in two phases. Initially, equal weighting was assigned to the variables, followed by additive aggregation to derive dimension sub-indices. Equal weighting was adopted due to its simplicity, and the assumption that all elementary indicators within a dimension are equally important. In the second stage, budget allocation (BAL) weights were added to the sub-indices, with state institutions receiving a weight of 0.4, non-state institutions 0.4, and the individual level 0.2, which were then aggregated multiplicatively to derive the final composite indicator of trust. In composite indexing, weights represent value judgments; hence, state and non-state institutions were allocated greater weight than individual-level weight, as these institutions are required to establish and demonstrate trust in their operations more than people. The formula for multiplicative aggregation is presented in the Equation 3.

(3)

Where D_ik is the dimension sub-index for country i and dimension k. The choice of multiplicative aggregation was informed by the fact that it is less compensatory, thus less affected by outliers within a statistical unit. A heat map was created to depict the composite indicator values and ranks, as well as the driving force for the dimension sub-indices.

Additional national-level indicators were sourced from various international authorities for the years 2021 and 2023 to validate but also illustrate their association with trust levels. The factors encompass population size, poverty headcount rates, GDP per capita, unemployment rate for individuals aged 15-24, duration of the current president or head of government in office, the former colonial power or country, and the geographical location of the nation.

3.4. Robustness Tests

Uncertainty analysis (UA) and sensitivity analysis (SA) were performed to assess the robustness of the CI. According to [14], UA focuses on how uncertainty in the input factors propagates through the structure of the composite indicator and affects the values of the CI while SA is an extra step (7) and studies how much each individual source of uncertainty contributes to the output variance. In this paper a comparison of conceivable scenarios at normalization (minimax and z-score), weighting (equal weights and budget allocation) and aggregation (additive and geometric) stages was performed and the absolute shifts in ranks noted.

4. Results

In Africa, individual trust reaches a maximum of 0.55, followed by non-state institutions at 0.46, and state institutions at 0.40. This indicates that while individuals may have confidence in their community members, they exhibit greater skepticism towards formal institutions, particularly those operated by the state.

Table 1 indicates that the top five performers are Tanzania (0.92), Niger (0.79), Mali (0.69), Senegal (0.62), and Ethiopia (0.62), respectively. These nations exhibit widespread trust across all three dimensions of institutions. Tanzania exhibits remarkably high ratings across all three dimensions: 0.99 (state), 0.85 (non-state), and 0.96 (individual), although Niger likewise demonstrates a well-balanced and robust trust profile. This performance suggests that nations with consistently high ratings, such as Tanzania, Niger, and Mali, may possess more cohesive societies and competent governance.

Table 1. Heat map of the CI of Trust

The lowest performers are Gabon (0.13), Eswatini (0.14), Sao Tome and Principe (0.17), Nigeria (0.20), and South Africa (0.23), respectively. These nations exhibit persistently low levels of trust, especially in state institutions. For instance, Gabon and Eswatini attained a mere score of 0.08 in state trust. Nigeria exhibits comparatively higher individual-level trust (0.48), although experiences significantly low institutional trust (state institutions: 0.09). Notable trends are evident in Guinea, Botswana, Mauritius, Morocco, and Cape Verde. Guinea exhibits exceptionally high individual trust (0.93) while demonstrating lower institutional trust, signifying a disparity in trust between the populace and institutions. Botswana, historically seen as a robust performer in institutional governance, is unexpectedly ranked low (30) due to an exceedingly low level of individual trust (0.10), indicating a disconnection between residents and institutions. Mauritius and Morocco exhibit mid-to-low rankings despite their status as upper-middle-income nations, suggesting that economic advancement does not necessarily correlate with elevated institutional trust. Cabo Verde exemplifies a successful small-island nation, achieving balanced ratings across all dimensions and ranking sixth on the continent.

These findings identify several observations: i) Discrepancies between institutional and individual trust indicate that grassroots trust may not convert into overarching legitimacy, ii) Low ratings of state institutions in numerous countries underscore a continental governance issue, and iii) bolstering non-state actors (e.g., NGOs, religious organizations, local leaders) may be a crucial strategy in nations with diminished state legitimacy. This composite indicator of institutional trust in Africa demonstrates significant diversity throughout the continent. The findings underscore both successful cases (such as Tanzania and Niger) and areas of worry (such as Gabon and Nigeria). Comprehending the drivers of trust — and its absence — is crucial for efficient governance, development, and civic participation throughout the continent.

The histogram and density graphs in Figure 2 indicate that the CI is not uniformly distributed. The histogram exhibits two peaks, with one peak being higher, indicating that the CI results are bimodal (the data possesses two separate modes). Each peak (mode) signifies a data cluster centered around a specific value, indicating polarized beliefs. A distinct group of countries exhibits a score of approximately 0.3, indicating poor trust, while another group congregates around a score of 0.5, reflecting moderate trust. Multi-variable analysis should treat these results as a classification problem to more effectively elucidate the patterns.

Figure 2. Histogram, and density plot for the composite indicator

As shown by the box plots in Figure 3, regionally, East Africa averages 65%, followed by West Africa at 45%, South Africa at 39%, and North Africa at 35%. Regarding classification by colonial power or language, there are no substantial disparities among Anglophone (43%), Francophone (45%), and Portuguese-speaking populations (45%).

Figure 3. Boxplots for regional location and former colonial power

Figure 4 illustrates the extent to which the rankings of countries fluctuate in relation to various plausible scenarios, as part of the uncertainty analysis. The uncertainty interval of the rank of the African countries from the least trusting to the most trusting is shown, and the large dot shows the median of the rank uncertainty distribution. The length of the uncertainty interval provides significant insights. A short interval indicates that the ranking is less reliant on the index formula: the shorter the interval, the more robust the country ranking becomes. The differences in rankings can be caused by outliers, uneven performance among elementary indicators, or correlations.

Figure 4. Uncertainty intervals of the African countries' ranks

From Figure 4, the most uncertain performance is observed in Guinea, Tunisia, Angola, South Africa, etc. It is also important to note that the top performing countries have more robust rankings, an indication of even performance across all dimensions and elementary indicators. Generally, these rankings should be considered stable and the CI stable.

The results of SA are shown in Figure 5. Here, the aggregation stage accounts for the most sensitivity in the constructed CI of trust, accounting for more than 60% of the variations in rankings. There is also observed sensitivity (close to 15%) arising out of the interaction between the aggregation and normalization stages of CI construction. The weighting stage is the least sensitive accounting for about 20% of the variations in rankings. Therefore, country rankings are more sensitive to the aggregation method (linear or geometric) than to the choice of data normalization or the weighting scheme.

Figure 5. Sensitivity Analysis results of the CI of trust

The effect of each dimension onto the CI is estimated using a non-variance-based sensitivity analysis by removing a dimension from the input model, and observing the mean absolute shifts in ranks. The results are shown in Table 2. This information can be useful when a decision needs to be taken to reduce the number of dimensions of the CI.

Table 2. Non-variance based sensitivity analysis

A higher value means that the ranks are changed more when the dimension is removed. Therefore, the sub-index of trust in non-state institutions produces more stable rankings than the rest of the dimensions. Therefore, in instances necessitating the removal of some dimensions, the individual level trust may be removed first.

A correlation analysis utilizing Pearson's product-moment correlation was employed to depict the relationship between the constructed composite indicator and known cross-country metrics as shown by Table 3.

Table 3. Correlation analysis of the CI with selected variables

A correlation coefficient is a statistical measure ranging from -1 to +1 that indicates the strength and direction of a linear relationship between two variables, albeit correlation doesn’t imply causality. The results indicate a weak positive relationship between the composite indicator of trust and the poverty head-count and the corruption perception index. A negative correlation exists among GDP per capita, population size, years of the incumbency, and unemployment rate. Generally, the relationship is weak. The correlation between the CI of trust scores and the unemployment rate is moderate and also statistically significant at the 95% confidence interval.

5. Discussion and Conclusions

Institutional trust in Africa is shaped not only by political institutions but also by customary authorities, religious bodies, NGOs, and international organizations. Accordingly, in this article we present a composite indicator to assess trust across three dimensions: state institutions, non-state institutions, and individual-level trust. This is based on proposals gathered through literature to integrate individual-level trust into comprehensive assessments of trust on the continent [15]. This work enhances the expanding literature on the utilization of composite indicators to assess multi-dimensional phenomenon such as trust. The accurate assessment of trust is crucial for the ambitions of a developing nation, as it indicates the legitimacy of the democratic transformations that countries undertake. This study conducts inter- and intra-country comparisons to identify potential reasons contributing to variations in trust levels. In contrast to numerous global governance indicators that frequently neglect informal institutions, a CI of trust for Africa encapsulates the hybrid governance systems present on the continent, encompassing both state and non-state authorities. This analysis recognizes the multiple sources of legitimacy that influence the lived experiences of citizens across the continent.

This study revealed several fundamental issues, particularly the conceptual framework linking individual trust to both state and non-state institutions. In contrast to Afro-barometer reports that just give discrete performance indicators for state and non-state institutions [16], we assert that individual-level trust is equally significant and should be integrated into the measurement paradigm. Trust, being multi-dimensional, necessitates a meticulous clarification of its basic elements to enable thorough assessment. We originate from a mindset that posits, "if citizens cannot trust themselves, how can they trust institutions?" Consequently, institutions are established or dismantled by citizens (individually or collectively). The findings from this investigation can be generalized, as individual-level trust far surpasses institutional trust in most countries. For instance, nations such as Congo, Gabon, Guinea, and Burkina Faso exhibit high levels of individual trust but demonstrate low institutional trust. This situation indicates that, in the African context, interpersonal trust precedes institutional trust. This performance at individual level may be attributed to improved interactions and communications regarding the obligations of each participant, as well as interfaces that promote psychological safety and facilitate risk-taking. Whereas most citizens may possess sufficient information while interacting with other individuals, this is not necessarily true when engaging with institutions. Furthermore, in certain communities—particularly rural ones—the deficiency of sufficient information regarding institutional commitments and explicit experiences in addressing them influences their judgments, a phenomenon commonly observed in Africa.

In the bi-variate association to validate the CI, our findings indicate a weak positive correlation (ρ=0.07) between the corruption perception index and our composite indicator of trust. These findings are compelling and contradict the positions of [17,18,19,20] that the degree of corruption significantly undermines public' trust in democratic institutions. Compared to the methods used in these studies, these results could be attributed to the aggregation of various components to form a singular estimate. The substitutability effect inherent in CIs implies that weak performance in one dimension is compensated for by strong performance in another. As a result, people may perceive formal institutions as corrupt but still trust them when mediated by informal authorities. This overlap may sustain institutional trust despite corruption perceptions. Nonetheless, these contentious outcomes may be ascribed to the data generation methodologies employed, wherein Afro-barometer recruits a diverse range of respondents through probabilistic selection, whereas Transparency International engages experts equipped with sufficient information to render accurate assessments of both direct and indirect corruption. Additionally, corruption has been normalised in some developing countries with weak institutions- as citizens perceive corruption as a necessary mechanism to swift service delivery. At that point, corruption may be ascribed to individuals rather than institutions as a whole. Likewise, the correlation between poverty head-count and trust levels is weak but positive. This outcome is rather unexpected; nonetheless, potential explanations may be linked to cultural, political, and measurement dynamics. Typically, individuals in lower socio-economic strata exhibit a pronounced dependence on institutions for their survival, which may result in a lack of objective assessments. Their expectations from governmental entities tend to be modest, hence influencing their trust levels. Furthermore, in less developed economies, poverty can exist alongside a robust sense of communal or institutional trust, as individuals often gravitate towards collective frameworks rather than market solutions.

Additionally, we observe that the CI exhibits a negative correlation with the unemployment rate, GDP per capita, and population size of the countries. This indicates that nations characterized by elevated population levels, wealth per individual and unemployment rates tend to have diminished levels of trust overall. This is anticipated, particularly in regions where population increase does not correspond with the expansion of resources necessary to sustain the populace and generate work possibilities for the youth. In many instances, the limited chances present are not directly allocated to residents based on merit, hence fostering possible nepotism, psychological distress, and diminished trust in institutions. Notably, GDP per capita, as an indicator of economic progress, exhibits a negative correlation with the CI of trust, contrary to the assertions of writers such as [21] and [10], particularly in Western nations. The African nations excelling in GDP per capita include Seychelles, Botswana, South Africa, and Mauritius, among others, and they also rank highly in literacy rates. Informed citizens are said to understand the responsibilities of their institutions, thereby affecting their service-seeking behaviours and enabling them to make sound evaluations of institutional performance. Conversely, [22] assert that economic growth and trust are inversely correlated, attributing this to income inequality typically found in high-income countries. Growing wealth inequality heightens the likelihood of societal comparisons, which replace trust in individuals' utility functions. Moreover, wealth inequality undermines collaboration and social cohesion in favour of competitiveness, hence heightening the likelihood of social instability. In consistence with [4], our results show that the unemployment rate is the sole significant factor correlated with aggregated trust levels. Also, this finding aligns with several research, including those by [23]; [24]; and [25], indicating that individuals facing unemployment generally exhibit diminished faith in government and political institutions. This can be ascribed to sentiments of economic distress, employment instability, and a perception that the government is insufficiently meeting their needs or offering suitable assistance. Furthermore, a significant level of unemployment among the populace might engender a perception of crisis and instability, prompting residents to scrutinize the government's competency and efficacy.

In conclusion, the low levels of trust across the continent necessitate the engagement of both state and non-state actors to reassess their operational methods and duties towards a people-centered governance. This diminished level of trust is self-sabotaging, resulting in decreased internal legitimacy and consequently restricted participation in governmental matters. Externally, it contributes to the increasing prevalence of afro-pessimistic attitudes that hinder investment, overall economic development, and misjudgment of Africans. Trend analysis of composite indicator values from the Afro barometer survey rounds as well as correlating the trust CI with more country variables awaits future research.