Beyond Financial Market Dualism: An Empirical Analysis of Variations in Use of Financial Services in South Africa

Abstract

This paper empirically analyses variation in use of formally, semi-formally, and informally regulated finance using the South African National Income Dynamics Study longitudinal data. The logistic regressions indicate that many individuals use a combination of services across all levels of regulation depending on age, gender, education, population group, religiosity, and social trust. Widespread use of informally regulated finance in South Africa is particularly evident on the savings side through savings groups/stokvels. The originality of the paper lies in its use of nationally representative longitudinal data to disentangle and analyze the variations in the use of different financial mechanisms, moving beyond the conventional formal–informal dichotomy. In doing so, it contributes to ongoing debates on financial inclusion by demonstrating that informally regulated finance represents a rational, adaptive response to the limitations of formally regulated services rather than a residual or inferior alternative. Depicting the market as dualistic is therefore misleading, ignoring the need for a more nuanced understanding and official recognition of the drivers of financial services’ use.

1. Introduction

Financial inclusion in Sub-Saharan Africa has made significant strides since the financial sector reforms of the 1990s, with account ownership increasing from 23.2% in 2011 to 42.6% in 2017 (Simatele, 2021). Despite this progress, informal financial systems remain vibrant, serving over 50% of the region’s adult population (Shuaib, 2018). In South Africa, where formal financial inclusion reached 80% in 2018, informally regulated savings mechanisms like stokvels continue to thrive, with usage increasing by 6% between 2018 and 2019 (Mashigo, 2020).

This paradox, where improved access to formally regulated financial services coincides with increased use of informally regulated financial mechanisms, forms the central focus of this study. Although this phenomenon is particularly pronounced in South Africa, similar patterns have been observed globally, raising broader questions about the interplay between financial inclusion, regulation, and user preferences (Demirgüç-Kunt et al., 2022; Banerjee & Duflo, 2019).

Traditional financial theories, such as the McKinnon-Shaw hypothesis (1973) and Stiglitz and Weiss’s imperfect information theory (1981), attribute the reliance on informally regulated finance to formal financial exclusion (Aliber et al., 2015; Zondi, 2016; Allen et al., 2019). These frameworks suggest that low-income and asset-poor individuals and small businesses turn to informally regulated finance due to the inadequacy of formally regulated financial services (Chepkogei Sile & Bett, 2015). However, evidence from South Africa and other developing countries indicates that the use of informally regulated finance extends beyond these excluded groups, encompassing middle- and high-income individuals, as well as those with access to formally regulated finance (Ezzahid & Elouaourti, 2021; Lujja, 2006; B. Nguyen & Canh, 2020). This raises critical questions about the drivers of the use of informally regulated financial services in the context of increasing formal inclusion.

South Africa presents a paradoxical financial landscape that makes it a critical case for analyzing the dynamics of formally and informally regulated finance. On the one hand, the country hosts one of the most advanced and regulated banking systems on the African continent, with widespread penetration of formal savings accounts and digital financial services. On the other hand, reliance on informally regulated mechanisms, such as stokvels, burial societies, and informal moneylenders, remains widespread, cutting across income, gender, and social divides. This coexistence is not marginal but central to how South Africans navigate credit, savings, and insurance, raising fundamental questions about the assumptions underpinning financial inclusion policy and the regulation of informally regulated finance.

Despite the prominence of informally regulated finance in South Africa, two important research gaps persist. First, most existing studies rely on cross-sectional or small-scale survey data, which limit the ability to capture the evolving and overlapping use of different forms of finance across different levels of regulation. Second, the bulk of the existing literature conceptualizes financial behavior through a dualistic lens, formal versus informal, without sufficiently theorizing the complementarities, substitutions, and rational strategies that individuals adopt across regulatory levels. This oversimplification risks misclassifying financial services that possess characteristics of both formal and informal services, leading to inaccurate assessments of usage patterns. A more nuanced understanding is necessary to distinguish the roles of formally regulated, semi-formally regulated, and informally regulated financial services in South Africa’s financial landscape. Addressing these gaps is particularly timely given the global debates on financial inclusion, resilience, and the role of regulation in shaping household financial choices (Demirgüç-Kunt et al., 2022; Meagher, 2021).

While the paper draws primarily on South Africa, this country offers a uniquely illustrative case for studying global patterns of financial behaviour in contexts of high formal inclusion and persistent informality. South Africa’s advanced financial infrastructure, combined with enduring informal mechanisms, mirrors broader tensions faced in both developing and emerging economies. By treating South Africa as a case study, this research contributes to a wider understanding of how financial systems evolve under hybrid regulatory landscapes, providing insights that are applicable to other settings where formal and informal financial sectors coexist (Banerjee & Duflo, 2019; Mbiti & Weil, 2016; Aryeetey & Udry, 1997).

We address these gaps and anomalies by making three contributions. First, we integrate insights from demographic, socioeconomic, and behavioral factors to shed light on the continued appeal of informally regulated finance and its implications for financial inclusion policies. Second, we argue that the dichotomization of finance into formal and informal finance in the extant literature is insufficient and does not give a clear and complete picture of the financial landscape as it omits or misclassify financial services that have characteristics of both formally and informally regulated finance. Thus, this paper contributes by classifying the financial landscape into three levels of regulation namely informally regulated, semi-formally regulated and formally regulated finance. Third, we examine whether the factors that influence the use of formally regulated finance differ from those that influence informally and semi-formally regulated finance. We provide a granular breakdown of financial services, identifying whether they are primarily utilized for asset accumulation through savings or debt accumulation.

To achieve these, this paper applies nationally representative longitudinal data from the South African National Income Dynamics Study (NIDS) to analyze patterns of financial service use across formally, semi-formally, and informally regulated institutions. By employing multinomial logistic regression, the study demonstrates that the use of informally regulated finance is not merely a residual option for the excluded but rather a rational and adaptive strategy in response to structural constraints in the formal sector. In doing so, the paper contributes both empirically and theoretically to debates on the regulation of informally regulated finance and the need for a more nuanced framework beyond the formal–informal dichotomy. Following the introduction, Section 2, Section 3, Section 4 and Section 5 of this paper present the literature review, material and methods, results, discussion and conclusion, respectively.

1.1. Review of Concepts

The distinction between formal, semi-formal, and informal financial services has long been central to debates on financial inclusion in developing economies (Ezzahid & Elouaourti, 2021; Geraldes et al., 2022; Uitto, 2020; Wabwire, 2020; Yimer, 2025). Formally regulated finance typically refers to services provided by regulated institutions such as commercial banks, licensed insurers, and registered pension funds, which operate under national financial regulatory frameworks (Sakyi-Nyarko et al., 2022; Turkson et al., 2022). Semi-formally regulated finance encompasses institutions and services that may have some degree of legal recognition or regulatory oversight but operate outside the full ambit of the mainstream financial system, for example, cooperatives, microfinance institutions, mobile money platforms, and village banks (de Shalit, 2024; Turkson et al., 2022). Informally regulated finance, by contrast, denotes financial transactions occurring outside state regulation, such as rotating savings and credit associations (ROSCAs), burial societies, moneylenders, and interpersonal borrowing between family or friends (Aryeetey & Udry, 1997; Beck et al., 2015).

Although this tripartite classification is widely accepted, scholars have debated whether such distinctions oversimplify financial behavior. One strand of literature frames informally regulated finance as the residual domain of the “excluded poor” (Collins et al., 2009), while another highlights its centrality across income groups as a rational complement to formally regulated finance (Demirgüç-Kunt et al., 2022). In South Africa, concepts such as stokvels (rotating savings groups) and mashonisas (informal moneylenders) illustrate how informality is deeply embedded in social and cultural practices rather than being purely an economic fallback. Similarly, burial societies represent a form of risk pooling that is both financial and social, reflecting norms around dignity, solidarity, and community reciprocity (Khavhagali, 2019).

Importantly, regulation does not map neatly onto these conceptual categories. Mobile money, for instance, occupies a semi-formal space in many African economies but is tightly regulated in Kenya while still relatively under-regulated in South Africa. This suggests that finance should not be conceptualized in binary or static terms but instead as existing along a continuum of regulation and informality (Meagher, 2021).

Moreover, theoretical and empirical contributions such as those by Zondi (2016), Aliber et al. (2015), and Allen et al. (2019) have emphasized the structural role that informal and semi-formal finance plays in the financial behavior of households. These works underscore that informality is not merely a coping mechanism but part of a broader adaptive strategy shaped by social norms, regulatory constraints, and access barriers. Their insights are central to the conceptual framework of this study, particularly in understanding household-level decision-making across regulatory levels.

1.2. Review of Empirical Studies

The use of informally regulated financial services predates the use of formally regulated financial services by decades, if not millennia (Tengeh & Nkem, 2017). The empirical literature on informally regulated finance is extensive, spanning Asia, Latin America, Africa, and South Africa. Globally, studies show that informally regulated finance persists even in contexts of deepening financial liberalization. In India, ROSCAs coexist with microfinance institutions (Gupta, 2021), while in Mexico, tandas remain popular alongside digital banking (Banerjee & Duflo, 2019). These findings suggest that informality is not merely transitional but structurally embedded.

In Africa, similar trends emerge. In Ghana and Nigeria, susu collectors and moneylenders remain prevalent (Aryeetey & Udry, 1997). In Kenya, informal borrowing interacts with mobile money to form hybrid systems (Mbiti & Weil, 2016). In Zambia, burial societies and savings groups are vital for resilience in rural areas (Tembo, 2021).

South Africa presents an interesting case. Research consistently documents the resilience of stokvels, burial societies, and mashonisas despite the country’s sophisticated formal banking system. Collins et al. (2009) showed households juggling multiple instruments to smooth consumption, while more recent work highlights the continued popularity of informal groups across income levels (Khavhagali, 2019). Payday lending and mashonisas continue to grow despite regulation (James, 2021). Yet most South African studies are cross-sectional or ethnographic, with little use of longitudinal data to capture financial dynamics over time. New dynamics are also emerging: Cele and Gumede (2024) note how fintech platforms are reshaping access, while the National Treasury’s (2023) Financial Inclusion Policy Framework recognizes informal systems as part of the broader financial ecosystem. Recent reports show the persistence of informality despite technological innovation: (Yimer, 2025) finds growing reliance on mashonisa lenders amid stricter formal credit, while (Bernal et al., 2023; Zhakata & Wayi-Mgwebi, 2023) documents the endurance of stokvels and burial societies alongside fintech and crypto innovations.

Traditional theories attempt to explain this persistence. The McKinnon-Shaw Hypothesis (1973) and Stiglitz and Weiss Imperfect Information Theory (1981) attribute the widespread use of informally regulated finance to the inadequacies of formal systems. These theories posit that people use informal finance due to financial exclusion stemming from financial repression and imperfect information, respectively (Alhassan et al., 2019; Eschenbach, 2004). Rural and township-based individuals and small businesses, whose poor financial positions condemn them to this exclusion, therefore gravitate towards informal systems. These postulations ignite expectations that improved financial inclusion would catalyze a seamless transition from informal to formal mechanisms. Indeed, Ky et al. (2021) show that digital payment mechanisms such as mobile money may help to shift a fraction of deposits from informal to formal channels.

Yet, despite these theoretical expectations, the past few decades have seen an anomalous growth of informally regulated services, not only in their persistence but also in their intensity and variety. Studies document the contemporaneous growth of formal and informal finance, showing that people continue to use informal services despite access to formal ones. Importantly, this is not limited to the exclusion. Lujja (2006) demonstrates that middle- and high-income groups in South Africa, who are not financially excluded, also use stokvels and other informal services. This supports the Neo structuralist assumption that informal finance is competitive, attracting users across income groups and even influencing formal financial agents, some of whom adopt practices from the informal sector.

Explanations for this persistence extend beyond exclusion. B. Nguyen and Canh (2020) shows that informal services remain attractive due to their relative speed, lower transaction costs, and absence of collateral requirements. Schemes such as RoSCAs and ASCAs are participatory, locally responsive, and convenient, making them accessible to low-income and asset-poor individuals often deemed unbankable (Demirgüç-Kunt et al., 2022; B. Nguyen & Canh, 2020). Similarly, informally regulated services benefit from intimate client knowledge, flexibility, and social embeddedness (Abrahams, 2015; Ayyagari et al., 2010; Li & Hua, 2023). These advantages are less available in formal systems, which continue to ration out small borrowers even when donor schemes or quotas encourage inclusion (Duarte et al., 2012; Fisman et al., 2017).

This evolving landscape also highlights overlaps between categories. Semi-formal providers such as microfinance institutions, crowdfunding platforms, and digitally enabled groups operate in the space between informally and formally regulated finance. Their hybrid models blur definitions based solely on registration or intermediation (Tengeh & Nkem, 2017). Evidence further shows that users diversify, mixing informal, semi-formal, and formal mechanisms within their portfolios (Madestam, 2014). Thus, these forms of finance are not mutually exclusive but instead operate in complementarity.

Despite extensive scholarly literature, three gaps remain. First, most studies reinforce a formal–informal binary, overlooking semi-formal providers. Second, few apply panel data to track financial behavior over time. Third, little attention is paid to the regulatory implications of recognizing informal systems within national financial frameworks. This study seeks to address these gaps by using nationally representative panel data to examine how South Africans diversify across formal, semi-formal, and informal finance, contributing both empirically and theoretically.

2. Materials and Methods

2.1. Data and Measurement

This study uses secondary data collected from four waves of the South African National Income Dynamics Study (SA-NIDS) spanning from 2011 to 2017. The NIDS longitudinal data which follows the lives of the same individuals and their households’ members every two to three years, provides a baseline data for at least 28,000 individuals from at least 7000 households across South Africa. The NIDS data set is more suitable because it provides rich data on different sources of informally, semi-formally and formally regulated finance at a micro level, which is not the case with other national surveys.

Although the most recent wave of the NIDS panel was collected in 2017, it remains the only nationally representative longitudinal dataset that provides systematic information on household financial behavior, including informally regulated finance. The use of data from this period is methodologically justified by the absence of more recent nationally representative datasets that capture informal financial mechanisms with comparable depth and consistency. More recent surveys, such as the National Income Dynamics Study—Coronavirus Rapid Mobile Survey (NIDS-CRAM), were explicitly designed to assess short-term labour market and welfare impacts of the COVID-19 pandemic and do not include detailed modules on informal saving and borrowing practices. As a result, they are unsuitable for analysing variations in financial service use across different regulatory levels. Furthermore, the patterns examined in this paper, such as the coexistence of formal and informal finance amid high levels of formal inclusion, are structural rather than transitory, reflecting long-standing features of financial systems in South Africa and other emerging economies (Demirgüç-Kunt et al., 2022; Meagher, 2021). Consequently, the insights derived from this dataset remain analytically and theoretically relevant despite the time lapse. This paper adopts a case study approach, using South Africa as a representative context to explore globally observed tensions between formal financial inclusion and persistent informal financial practices (Demirgüç-Kunt et al., 2022; Mbiti & Weil, 2016).

The dependent variables capture the type of financial services used, distinguishing between saving and borrowing options. Saving was coded as use of (i) formal banks, (ii) informal groups such as stokvels, or (iii) both (blended use). Borrowing was coded as (i) formal unsecured credit (credit/store cards, microlending), (ii) informal lenders (mashonisas), (iii) family/friends, or (iv) blended use. This multinomial categorization follows prior studies on household financial service diversification (Porteous, 2003; Karlan & Morduch, 2009).

Independent variables include income, age, gender, education, population group, religiosity, and social trust. Income was grouped into six categories of monthly net earnings, consistent with its strong predictive role in financial access (Beck et al., 2015). Age was grouped into six life-cycle bands, reflecting systematic variation in financial behavior (Allen et al., 2016). Gender (male = 0, female = 1) captures documented gender gaps in finance (Dupas & Robinson, 2013). Education was divided into four levels, consistent with evidence linking schooling to financial behavior (Lusardi & Mitchell, 2014). Population group (Black African, Colored, Indian/Asian, White) controls for structural inequalities in South Africa’s financial system (Collins et al., 2009; James, 2021). Religiosity (importance of religious activities, four categories) captures the role of faith in economic behavior (Barro & McCleary, 2003). Social trust (likelihood of wallet return, four categories) measures generalized trust, a known determinant of informally regulated finance participation (Guiso et al., 2004; Nunn & Wantchekon, 2011). A full summary of variable definitions and measurement is provided in

Table 1. Summary of Variables.

Variable Type	Category	Variable	Definition	Measurement
Dependent	Saving	Formally Regulated	Use of formal bank savings	1 = Bank
		Informally Regulated	Use of informal savings groups (e.g., stokvels)	2 = Stokvel/Saving Group
		Blended Use	Use of both formal and informal savings	3 = Combined Use
	Credit/Borrowing	Formally Regulated	Use of unsecured credit (e.g., credit/store cards)	1 = Credit Card, Store Card, Microlending
		Semi-Formally Regulated	Use of semi-formal credit (e.g., microfinance)	2 = Microfinance/Unsecured Credit
		Informally Regulated	Borrowing from informal lenders (mashonisa)	3 = Informal Money Lender (Mashonisa)
		Informally Regulated	Borrowing from friends and relatives	4 = Friends and Relatives
		Blended Use	Use of multiple credit sources across categories	5 = Combined Use
Independent	Socioeconomic	Income	Individual’s net income per month	0 = No income; 1 = R1–R5000; 2 = R5001–R15,000; 3 = R15,001–R30,000; 4 = R30,001–R60,000; 5 = >R60,000
	Demographic	Age	Number of years lived	0 = 16–25; 1 = 26–34; 2 = 35–44; 3 = 45–54; 4 = 55–64; 5 = >64
		Gender	Gender of respondent	0 = Male; 1 = Female
		Education	Level of formal education completed	0 = No education; 1 = Primary; 2 = Secondary; 3 = Tertiary
		Population Group	Racial or ethnic identity	0 = Black African; 1 = Coloured; 2 = Indian/Asian; 3 = White
	Social-Cultural	Religiosity	Importance of religious activity in respondent’s life	0 = Very important; 1 = Important; 2 = Unimportant; 3 = Not important at all; 4 = Refused/Don’t know
		Social Trust	Likelihood of lost wallet being returned	0 = Very likely; 1 = Somewhat likely; 2 = Unlikely; 3 = Not likely at all; 4 = Refused/Don’t know

Note: financial services included in this study are only those that could be accessed through the NIDS data set.

.

2.2. Methods

This paper models financial service use across formally, semi-formally, and informally regulated systems using the Random Utility Model (RUM), which posits that individuals select the option that maximizes their utility (McFadden, 1974). To this end, we assume that each potential user is faced with a variety of different forms of financial services options, which may be formally regulated, semi-formally regulated, or informally regulated.

Given these options, we assume that users make rational choices in a quest to maximize utility. According to the utility maximization theory, a user chooses one form of financial service if the utility derived from that service exceeds the utility that is derived from other financial services available and accessible. This is premised on McFidden’s Random Utility Model (RUM) which postulates that, when faced with different alternatives, people tend to choose alternatives that yield the highest utility. The utility of a given alternative is determined by the attributes of that alternative as well as the characteristics of the individual making a choice. The utility (U) of a given financial option is a function of observable (X) and unobservable (Z) individual characteristics:

$$\mathit{U}=\mathit{f}(\mathit{X}\mathbf{;}\mathit{Z})$$

(1)

where X represents the observable individual characteristics, while Z represents the unobservable individual characteristics. This means that the utility (U) is a function of these characteristics. This utility function can also be expressed more specifically as

$${{\mathit{U}}_{\mathit{i}\mathit{j}}=\mathit{X}}_{\mathit{i}\mathit{j}}\mathbf{;}{\mathit{Z}}_{\mathit{i}\mathit{j}}={\mathit{V}}_{\mathit{j}}\left({\mathit{X}}_{\mathit{i}\mathit{j}}\mathbf{;}\mathsf{\beta }\right)\mathbf{,}\mathbf{i}=\mathbf{1}\mathbf{;}\mathbf{2}\mathbf{;} \mathbf{\dots }\mathbf{,} \mathbf{M}$$

(2)

where i represents individuals, while j represents financial services; ${\mathit{U}}_{\mathit{i}\mathit{j}}$ represents the utility that is derived by individual 1 from the financial service j chosen; ${\mathit{X}}_{\mathit{i}\mathit{j}}$ represents the observed characteristic of the individual and the financial service in question; ${\mathit{Z}}_{\mathit{i}\mathit{j}}$ represents the unobserved characteristic of the individual and the financial service in question; and ${\mathit{U}}_{\mathit{j}}$ represents the deterministic component of the utility function.

Considering the Random Utility Model, equation two can be rewritten as follows:

$${\mathit{U}}_{\mathit{i}\mathit{j}}{(\mathit{X}}_{\mathit{i}\mathit{j}}\mathbf{;}{\mathit{Z}}_{\mathit{i}\mathit{j}})={\mathit{V}}_{\mathit{j}}\left({\mathit{X}}_{\mathit{i}\mathit{j}}\mathbf{;}\mathsf{\beta }\right)+{\mathsf{\epsilon }}_{\mathit{i}\mathit{j}}$$

(3)

where ${\mathit{V}}_{\mathit{j}}$ and $\mathsf{\beta }$ represent the estimated deterministic component and coefficient, respectively. Meanwhile, ${\mathsf{\epsilon }}_{\mathit{i}\mathit{j}}$ represents the unknown utility derived by an individual from the financial service. Considering this theoretical framework as well as the existing empirical literature, an individual i is faced with a variety of financing options, which may be formally regulated, semi-formally regulated, and informally regulated. Thus, the dependent variable Y is made up of multiple financing options that are available to each individual i. For example, the different financing options may include banks, microfinance institutions, store cards, informal moneylending (ML), friends and relatives. This modeling approach supports a broader theoretical inquiry into global patterns of financial pluralism, using South Africa as a critical site for empirical investigation.

In a case such as this one where the dependent variable is made up of more than two different categories that are not ordinal, a maximum likelihood estimator such as multinomial logit should be used (Kwak & Clayton-Matthews, 2002). The multinomial logistic regression model is an extension of the binary logistic model that allows for more than two categorical dependent or outcome variables. Like binary logistic regression, the multinomial logistic regression model uses the maximum likelihood estimation to evaluate the probability of categorical membership (Kwak & Clayton-Matthews, 2002). This paper primarily focuses on the demand side of the financial services landscape. Therefore, factors that potentially influence the use of informally regulated finances are a critical component of the model. This study assumes that people’s choice of different forms of financial services is determined by the associated costs and benefits, which in turn determine the expected utility from the use of each form. In line with Mcffiden’s Random Utility model, the literature suggests that both institutional factors and individual characteristics influence people’s choice of different forms of financial services. The use of informally regulated finance may be influenced by factors such as age, gender, education, geographic area, ethnicity, and financial literacy. The analysis determines which ones of these factors are associated with the use of which form of informally regulated finance.

Therefore, the logistic regression model is described by the following function:

$$\mathbf{L}\mathbf{o}\mathbf{g}\mathbf{i}\mathbf{t} (\mathsf{\pi }\mathbf{i}\mathbf{t})=\mathbf{L}\mathbf{o}\mathbf{g} ({\displaystyle \frac{{\mathsf{\pi }}_{\mathit{i}\mathit{t}}}{\mathbf{1}-{\mathsf{\pi }}_{\mathit{i}\mathit{t}}}})={\mathit{a}}_{\mathbf{0}}+{\mathit{a}}_{\mathbf{1}}{\mathit{X}}_{\mathit{i}\mathit{t}\mathbf{1}}+{\mathit{a}}_{\mathbf{2}}{\mathit{X}}_{\mathit{i}\mathit{t}\mathbf{2}}+\mathbf{\cdots }+{\mathit{a}}_{\mathit{p}}{\mathit{X}}_{\mathit{i}\mathit{t}\mathit{p}}+{\mathit{a}}_{\mathit{p}}{\mathit{Z}}_{\mathit{i}\mathit{t}\mathit{p}}+{\mathsf{\mu }}_{\mathit{i}\mathit{t}}$$

(4)

Let

$${{\mathit{Y}}_{\mathit{i}\mathit{t}}=\mathit{a}}_{\mathbf{0}}+{\mathit{a}}_{\mathbf{1}}{\mathit{X}}_{\mathit{i}\mathit{t}\mathbf{1}}+{\mathit{a}}_{\mathbf{2}}{\mathit{X}}_{\mathit{i}\mathit{t}\mathbf{2}}+\mathbf{\cdots }+{\mathit{a}}_{\mathit{p}}{\mathit{X}}_{\mathit{i}\mathit{t}\mathit{p}}+{\mathsf{\mu }}_{\mathit{i}\mathit{t}}$$

(5)

Equation (3) can be transformed into probabilities as follows:

$${\mathsf{\pi }}_{\mathit{i}\mathit{t}}={\displaystyle \frac{{\mathit{ϵ}}^{\mathit{f}\mathit{i}\mathit{t}}}{\mathbf{1}-{\mathit{ϵ}}^{\mathit{f}\mathit{i}\mathit{t}}}}$$

(6)

where ${\mathsf{\pi }}_{\mathit{i}\mathit{t}}$ the probability that the ith person is using a certain type of financial service at time t; $\mathit{f}\mathit{i}\mathit{t}$ is the multinomial outcome, comprising options shown in Table 1, for the ith person at time t; ${\mathit{X}}_{\mathit{i}\mathit{t}\mathbf{1}},{\mathit{X}}_{\mathit{i}\mathit{t}\mathbf{2}}\mathbf{\dots }{\mathit{X}}_{\mathit{i}\mathit{t}\mathit{p}}$ is a set of explanatory variables, including those that are summarized in

Table 2. Descriptive statistics for credit.

Variables	Mean	Standard Deviation	Variance	Skewness	Kurtosis	Minimum	Maximum
Credit	3.115	1.190	1.416	−0.098	2.558	1	5
Income	2.267	0.918	0.843	−0.338	1.867	0	5
Age	3.061	1.279	1.636	0.483	2.588	1	6
Gender	1.610	0.488	0.238	−0.449	1.202	1	2
Education	1.793	0.492	0.242	−2.173	7.319	0	3
Population Group	1.281	0.667	0.445	2.800	10.817	1	4
Religiosity	1.640	0.775	0.599	1.356	5.183	1	5
Social trust	3.263	1.216	1.479	−0.844	2.191	1	5

, with p being the number of explanatory variables; e is the exponential term.

The dependent variable in this study captures household participation across multiple, unordered categories of financial mechanisms, including formal, semi-formal, and informal saving and credit facilities. Given the nominal structure of the outcome, the multinomial logistic regression model (MNL) represents the most appropriate empirical strategy. The MNL framework allows for the estimation of relative risk ratios, providing clear interpretation of how socio-economic, demographic, and institutional factors shift the likelihood of households selecting one financial mechanism over another. While alternative approaches such as the multinomial probit or mixed logit exist, they either impose computational and identification challenges that are difficult to address with the NIDS dataset, or require repeated choice data across multiple time periods, which are not available in the current panel structure. The MNL thus strikes an optimal balance between empirical rigor, interpretability, and feasibility, and it has been widely employed in similar studies on household financial inclusion and credit market participation (e.g., Beck & Brown, 2015; Allen et al., 2016). To test the robustness of our findings, we further estimated a binary logit model contrasting informal versus formal participation, and the results were broadly consistent with the multinomial specification, reinforcing the validity of our empirical strategy. We used Stata 14 software for all the estimations.

3. Results

3.1. Descriptive Statistics

Table 2 and

Table 3. Descriptive statistics for saving.

Variables	Mean	Standard Deviation	Variance	Skewness	Kurtosis	Minimum	Maximum
Saving	0.241	0.428	0.183	1.210	2.465	0	1
Income	2.187	0.929	0.863	−0.125	1.867	0	5
Age	3.008	1.209	1.461	0.515	2.708	1	6
Gender	1.590	0.492	0.242	−0.367	1.135	1	2
Education	1.859	0.412	0.170	−2.645	10.453	0	3
Population Group	1.302	0.702	0.492	2.684	9.909	1	4
Religiosity	1.633	0.769	0.591	1.370	5.272	1	5
Social trust	3.240	1.227	1.505	−0.812	2.123	1	5

present the descriptive statistics which provide basic information about the characteristics of all the variables in the data set being used herein.

The descriptive statistics indicate that all the variables do not conform to normal distribution. Normality is not a must or a critical requirement for panel study. As such, the assumption of normality can be relaxed especially in cases where N is larger. This assumes that the sampling distribution of the estimates tends to gravitate towards normal distribution as the sample size N increases to infinity (Ozili, 2023). Nonetheless, more diagnostic checks were performed, and the suitable models of estimation were applied accordingly.

3.2. Multinomial Logistic Regression Results

This section encompasses the interpretation of the multinomial logistic regression results from

Table 4. Multinomial Logistic Regression Results for Credit.

	Semi-Formally Regulated	Informal Regulated		Blended Use
Variable	Unsecured Credit (RRR)	Informal Money Lender (RRR)	Friends & Relatives (RRR)	More Than One Service (RRR)
Income: 1–5000	0.465	0.620	0.271 **	0.747
5001–15,000	0.158 **	0.336 *	0.042 ***	0.924
15,001–30,000	1.270	0.804	0.730	0.877
30,001–60,000	0.286	0.486	0.087 ***	1.623
600,001>	2.517	0.340	1.150	0.883
Age: 25–34	0.363 ***	0.492 ***	0.118 **	0.841
35–44	0.249 ***	0.273 ***	0.077 ***	0.715 *
45–54	0.208 ***	0.208 ***	0.050 ***	0.663 **
55–64	0.198 **	0.209 ***	0.044 ***	0.542 ***
65>	0.552 **	0.480 **	0.129 ***	0.540 **
Gender: Women	3.264 ***	4.617 ***	2.856 ***	2.253 ***
Education: Primary	0.498 **	1.178	0.461 ***	1.265
Secondary	0.101 ***	1.576 **	0.132 ***	1.893 **
Tertiary	0.068 **	3.024 **	0.176 **	1.539
Population Group: Colored	0.573 ***	0.921 ***	0.429	0.851 *
Asian/Indian	0.359 *	0.862 **	0.434	0.575 *
White	0.152 ***	0.491 ***	0.2383	0.408 ***
Religiosity: Important	1.116	0.921	1.018	0.907
Unimportant	2.649 ***	1.117	1.746 **	1.255
Not important at all	1.606 **	0.801	1.711 ***	0.886
Don’t know/refused	1.474	0.704	0.616	0.507
Social trust: Somewhat likely	1.019	0.966	0.794	0.866
Unlikely	0.282 **	0.543 **	0.416 *	0.423 ***
Not likely at all	1.149	0.968	0.998	1.040
Refused/don’t know	1.501 *	0.995	1.142	1.023
Constant	4.321 ***	3.929	30.280	0.742

Note: The Relative Risk Ratios (RRR) with no aesthetics (*) show that the results are statistically insignificant, thus not interpreted in this study. Meanwhile, the RRR values with one aesthetic, two aesthetics and three aesthetics show that the results are statistically significant at 0.1, 0.05 and 0.01 significant levels, respectively.

and

Table 5. Multinomial Logistic Regression Results for Savings.

Variable	Savings Group/Stokvel (RRR)	Blended Use (RRR)
Income	1.245	1.650
Age: 25–34	1.086	1.949 ***
35–44	1.133	2.692 ***
45–54	1.131	2.514 ***
55–64	1.308	2.5197 ***
65>	2.698 **	1.696 **
Gender: Women	4.894 ***	2.257 ***
Education: Primary	0.439 ***	0.984
Secondary	0.111 ***	1.252
Tertiary	0.135 *	2.107 *
Population Group: Coloured	0.417 ***	0.377 ***
Asian/Indian	−0.953	0.431 **
White	0.064 ***	0.386 ***
Religiosity: Important	1.342 **	0.935
Unimportant	1.137	1.091
Not important at all	1.365	0.858
Refused/don’t know	4.959	1.212
Social trust: Likely	0.897	1.002
Unlikely	0.681	0.471 **
Not likely at all	1.492 **	0.999
Refused/don’t know	1.333	0.988
Constant	0.056 ***	0.073

Note: The Relative Risk Ratios (RRR) with no aesthetics (*) show that the results are statistically insignificant, thus not interpreted in this study. Meanwhile, the RRR values with one aesthetic, two aesthetics and three aesthetics show that the results are statistically significant at 0.1, 0.05 and 0.01 significant levels, respectively.

. Table 4 presents the factors explaining the use of different forms of informally and semi-formally regulated credit relative to formally regulated credit, while Table 5 presents the factors explaining the use of different forms of informally and semi-formally regulated savings relative to formally regulated savings. We interpret the Relative Risk Ratio (RRR) associated with each type of financial service where relative risk is the risk of an event in an experimental group relative to that in a control group. For example, the RR refers to the risk of one type of semi-formally or informally regulated finance relative to formally regulated finance by a person falling into one category or grouping relative to the other. A relative risk of 1 indicates that the risk is comparable in the groups. A relative risk value that is greater than 1 indicates an increased risk, while a relative risk value that is less than 1 indicated decreased risk of using a particular form of semi-formally or informally regulated finance relative to formally regulated finance.

3.2.1. Income

Table 4 shows that people earning 0 to R5000 are at lower risk of borrowing from friends and relatives than those whose income level is zero. Meanwhile, people earning between R5001 and R15,000 are less likely than those who earn zero income to borrow through unsecured lending, informal money lending as well as friends and relatives. Moreover, this means that although low-income groups earning below R15,000 a month may borrow from informal or semi-formal and/or unsecured credit sources, they tend to switch to more formal sources such as the banks as their income levels increase beyond R15,000. This could mean that borrowing from unsecured and informal credit facilities is generally not the people’s first preference. Moreover, this could be due to the notion that credit from informal money lenders is generally regarded as “underground” or “dirty” finance which is characterized by high interest rates and aggressive collection measures in the event of default. Thus, these results affirm the notion that the credit facilities from informal money lenders are largely reserved for the poor whose low-income status precludes them from accessing the formally regulated credit facilities from the banks (Zondi, 2016).

Table 4 also shows that people earning between R30,001 and R60,000 are less likely to borrow from friends and relatives compared to banks as their income levels increase. Informally regulated financial transactions are premised on interpersonal relationships between the parties involved. Generally, friends and relatives are the closest and thus more accessible credit sources and although they are both non commercially trivial and less costly (Johnson, 2017; B. Nguyen & Canh, 2020; N. Nguyen et al., 2015), loans from friends and family can put the relationship with the lender at risk. Thus, it makes sense that people tend to borrow less from these sources as their income levels improve.

3.2.2. Age

Table 4 shows that as people grow, they are more likely to use formally regulated credit sources such as banks than as they grow old such as the banks than the unsecured informal or semi-formal sources such as credit cards, store cards, microcredit, informal money lenders and friends or relatives. This might be due to people’s changing perception of the unsecured semi-formal and informal sources of credit as they become more matured and responsible, and their inclination to borrow more secured and formal credit sources.

Further, Table 4 also shows that people older than 35 years relative to the younger ones are less likely to borrow from multiple sources of credit. This could be because, generally, people younger than 35 are either still studying, unemployed or underemployed and thus require multiple sources of credit to fund their needs. This could also mean that people’s borrowing habits reduce as they age. Table 5 shows that people aged above 65 years are more likely to save through savings groups or stokvels relative to banks. Although this could reflect the attractive social character of the savings groups or stokvels for retired elders, this could also mean that when people age and become less able to accumulate income and wealth, they find it less difficult and more convenient to save through informal savings facilities with less stringent requirements than the banks. Further, Table 5 shows that people aged above 25 years are more likely to use more than one saving facility. This could mean that as people get old and more financially literate, they tend to diversify between different types of savings facilities to minimize the risks and costs while maximizing the expected benefits and returns.

3.2.3. Gender

Table 4 shows that women are most likely than men to borrow from informally and semi-formally regulated sources of credit relative to the formally regulated sources such as the banks. This finding is not surprising and could mean that women have relatively limited access to formal banking finance due to lack of collateral and cash income from formal employment in South Africa. This could reflect the traditional and cultural kinship relations which restrict women’s access to property ownership and cash income. Similarly, Table 5 shows that women are relatively more likely than men save their money through stokvels or savings groups compared to banks. This finding also mirrors the high inequalities and unemployment rate in South Africa which are gender bias, as well as the social norms that restrict women’s rights and opportunities.

Women in different localities, especially villages and townships, pool resources together and create such opportunities for themselves through savings groups. This finding could also reflect the popularity, amongst women, of “stokvels” which take different forms and serve diverse needs of the formally unserved and/or underserved women especially in villages and townships. The widespread use of “stokvels” by women could reflect the strong bond among women who are more adept than men at managing cohort dynamics. The sense of community, socializing and women empowerment might also form part of the attractive features propelling the popularity of these groupings amongst women. Similarly, and probable for the reasons discussed above, Table 5 also shows that the women are more likely than men to use multiple sources of finance.

3.2.4. Education

Table 4 shows that people with primary, secondary and tertiary education as less likely than those who have no education to borrow through unsecured credit and friends and relatives, compared to banks. This means that people tend to borrow less through unsecured semi-formal credit sources, compared to formal banks, as their educational levels improve. Moreover, people with secondary and tertiary education are more likely than those with no education to borrow from informal money lenders relative to banks. This finding is contrary to the hypothesized expectation that people tend towards more formally regulated sources and aways from informally regulated sources of credit as they improve their education. This could reflect limited access to formal banking services due to the reduced per capita income as well as high unemployment rate which have worsened not only among the least educated but also graduates in South Africa since the 2008/09 global recession.

Table 5 shows that people with primary, secondary and tertiary education are less likely than those with no education to use savings groups or stokvels relative to banks to save their money. Table 5 also shows that people with tertiary education are more likely to use a combination of bank and savings groups or stokvels to save their money. Although this might reflect people’s appreciation of the value of diversification as their educational levels improve, it could also mirror the attractiveness of the flexible and diverse nature of the savings groups even to the educated individuals who might have access to the banks. The educated individuals might use bank and saving groups or stokvels to complement each other. For example, one of the unique features of the savings group is its social character, such as the ability to network and form groups as students and colleagues, which is lacking from the banking services. Consistent with the finding of Lujja (2006), this also suggests that there are high budget stokvels that consist of highly educated individuals, which are formed along the gender, workplace/colleagues and kinship lines.

3.2.5. Population Group

Table 4 shows that coloured, Indian and white population groups are less likely than black Africans to borrow from unsecured credit sources and informal money lenders relative to banks. This relationship mirrors the uneven distribution of wealth which is skewed to the non-black racial groups. The low income and asset poor population groups such as black Africans are largely excluded from the formally regulated financial services landscape and thus tend to resort to less formally regulated sources of finance. This could be mirroring the legacy of the previous regime of racial segregation against the non-white population groups whose exclusion from the formally regulated economy compels them to rely on informally regulated services. Table 5 also shows that coloured, Indian and white population groups are less likely than black Africans to save through savings groups or stokvels relative to banks, also emphasizing the formal financial exclusion and skewed income inequality. Further, Table 4 and Table 5 show that the other population groups are less likely than black Africans to use a combination of different type of both credit and savings facilities.

3.2.6. Religiosity

Table 4 shows that people who believe that religion is not important and not important at all are more likely than those who believe that it is very important to borrow from unsecured credit sources and friends and relatives compared to banks. These results are inconsistent with (Cao et al., 2019) which shows that religiosity promotes positive thinking and conduct, which in turn improves the lender’s willingness to grant credit and borrower to repay it. However, these results might reflect the teachings of some religions such as Christianity, Islamic and Judaism against indebtedness, especially interest-bearing debt. This is in line with (Sipon et al., 2014) who found that high religiosity leads to low financial borrowing. Table 5 further shows that people who view religiosity as important are more likely than those who view it as very important to use savings groups or stokvels vis a vis bank account. This evidences that savings groups are also based on interpersonal relationships formed through church or religious gatherings.

3.2.7. Social Trust

Table 4 shows that people who believe that a neighbor or stranger is unlikely to return a lost wallet with money inside are less likely to borrow through unsecured credit sources and friends or relatives or to borrow from different sources relative to banks, than those who believe that a stranger or neighbor would return a wallet. This means that those who lack social trust are less likely than those who have social trust to borrow from unsecured semi-formal and informal, as well as a combination of different sources relative to banks. This is in line with (Aliber et al., 2015; Allen et al., 2019) which shows that importance of family values and interpersonal or community trust as an important enabler of informally regulated financial activities.

4. Discussion

4.1. Theoretical Contributions

The findings of this study offer compelling empirical evidence that challenges the binary framing of South Africa’s financial sector, traditionally split into formal and informal systems. Rather than operating within a rigid dualist framework, the use of financial services among South Africans appears to be fluid and multidimensional, as shaped by intersecting factors such as income, age, gender, education, race, religiosity, and social trust. This complexity aligns with and extends prior arguments by Zondi (2016), Aliber et al. (2015), and Allen et al. (2019), who emphasize the socioeconomic and cultural embeddedness of informal finance rather than its marginality.

The study contributes conceptually by reframing the financial landscape through a tripartite structure, formal, semi-formal, and informal, thereby moving beyond the dominant formal/informal dichotomy. It also challenges assumptions in financial inclusion theory by demonstrating that informally regulated finance persists not merely due to exclusion, but as part of an intentional and rational strategy across different income groups and demographics. These insights deepen our understanding of financial behavior as contextually embedded, adaptive, and layered.

4.2. Practical and Policy Implications

The empirical patterns observed in this study hold several important implications for policymakers, regulators, and practitioners. First, income gradients emerged as a crucial determinant. As income levels increase, individuals progressively shift from informal and semi-formal credit options toward formal banking services, confirming the hypothesis that informal credit is often a default rather than a preferred option. This pattern echoes existing literature (Zondi, 2016; Johnson, 2017), suggesting that underground lending is driven by economic necessity rather than cultural preference.

Second, while wealthier individuals tend to use formal institutions for saving, the appeal of savings groups (e.g., stokvels) endures across income brackets, especially when informal and formal tools are blended. This suggests a need for inclusive financial products that account for the social embeddedness of informal mechanisms. Gender differences are particularly striking. Women are substantially more likely than men to rely on informal and semi-formal mechanisms, both for credit and savings. This reinforces the need for gender-sensitive financial inclusion strategies, particularly those that address property rights, informal employment, and social capital constraints.

Educational attainment showed a complex relationship with informal borrowing, complicating the assumption that financial literacy and education automatically translate into formal financial usage. This calls for policy interventions that go beyond literacy and address structural constraints in the labor and credit markets. Race and ethnicity continue to shape financial access, with Black South Africans disproportionately represented among informal finance users. This underscores the need for intersectional policy interventions that address historical inequities and lingering systemic barriers such as credit scoring biases and asset inequality.

Finally, religiosity and social trust, often overlooked in financial inclusion, were found to meaningfully influence financial behavior. Informal savings mechanisms thrive in communities with high religiosity and trust, suggesting that policy design should consider cultural and community values as key components of effective inclusion.

4.3. Study Limitations

This study is subject to several limitations. First, the data used were collected between 2011 and 2017. Although this remains the most recent panel dataset that captures informal, semi-formal, and formal financial usage in South Africa, it may not fully reflect recent changes in the financial landscape, particularly those influenced by the COVID-19 pandemic or fintech innovations. Second, while the findings provide deep insights into the South African context, the generalizability to other countries is limited by unique regulatory, cultural, and historical factors. Further comparative studies would be needed to extend the applicability of these findings. Third, while this study identifies broad patterns of financial service use, it does not capture qualitative dimensions such as trust dynamics, power relations, or user satisfaction with specific financial services.

4.4. Directions for Future Research

Future research could benefit from more recent data that captures the evolving role of digital platforms, mobile money, and fintech-driven savings groups. There is also room to investigate how informal and formal systems interact under stress, such as during economic shocks or public health crises. Additionally, comparative studies across African or emerging market countries could further validate the tripartite regulatory framework introduced in this paper. Mixed-methods studies could enrich our understanding of motivations and perceptions driving blended financial service use. Finally, further exploration of semi-formally regulated institutions, such as cooperatives, micro-lenders, and mobile-based saving platforms, may offer valuable insight into the evolution of hybrid financial ecosystems in contexts of partial regulation.

5. Conclusions

This study investigated the coexistence and nuanced usage patterns of informally, semi-formally, and formally regulated financial services in South Africa, going beyond the dichotomous lens of financial market dualism. While formal financial inclusion has improved over the years, the continued and widespread use of informal finance, particularly for savings rather than credit, suggests that such services serve not merely as substitutes but as complements that address gaps left by the formal system. The strong preference for stokvels and other group-based savings mechanisms highlights an intentional strategy to build assets and not accumulate debt, emphasizing social embeddedness, accessibility, and trust.

Consistent with prior literature (e.g., Zondi, 2016; Aliber et al., 2015; N. Nguyen et al., 2015), this study confirms that users of informal and semi-formal finance are often those structurally excluded from the formal sector: low-income earners, women, less-educated individuals, and non-white population groups. However, the findings also complicate this narrative, revealing that even socioeconomically advantaged individuals continue to engage with informal mechanisms due to their flexibility, cooperative nature, and relational advantages. These dynamics call into question the sufficiency of canonical theories like the Stiglitz–Weiss model and McKinnon–Shaw hypothesis, which link informal finance solely to market failure and exclusion. Instead, informal financial practices may also reflect cultural logic, community cohesion, and strategic financial diversification.

The blended and context-sensitive usage patterns identified here suggest that informal finance should not be dismissed or over-regulated. Rather, policies should recognize its developmental potential, particularly in promoting saving behaviors, enhancing social resilience, and supporting financial inclusion from the bottom up. Formal institutions may benefit from integrating elements of informal finance, such as flexibility, group support, and interpersonal trust, into their product design and outreach strategies.

Areas of Future Research

Future research should investigate the evolving interplay between formal and informal finance as digital financial services expand across urban and rural South Africa. Further exploration is also needed into how gender, trust, and religiosity shape long-term financial decision-making. Importantly, interdisciplinary research combining financial economics, sociology, and development studies could provide deeper insights into the adaptive logic of informal financial ecosystems.