The Role of Pension Fund Investments in Shaping Volatility Dynamics in South African Stock Markets

Abstract

This article examines the role of pension fund investments in shaping volatility dynamics in South African stock markets from 2010Q1 to 2022Q4. The article measures exchange rate volatility using the family of autoregressive conditional heteroskedasticity and Markov-switching dynamic regression. Results indicate that increased contributions by both individuals and employers are positively associated with higher stock market returns, suggesting a potential contribution to improved market performance. In addition, there are two distinct regimes when analyzing the impact of pension fund withdrawals on stock market volatility. In regime 1, pension fund withdrawals contribute to a decrease in stock market volatility, indicating a potential stabilizing effect on the market. However, in regime 2, higher withdrawals are associated with increased volatility, possibly reflecting market uncertainties and selling pressures. Given these, pension fund participants and investors should carefully consider the timing and amount of their withdrawals, taking into account the potential impact on stock market returns. Encouraging higher pension fund contributions by individuals and employers can play a crucial role in enhancing stock market performance and stability. Policymakers should closely monitor the impact of pension fund dynamics on market dynamics and ensure the integrity and stability of the stock market.

JEL Classifications G10, G18, G23

Keywords

Pension fund investments stock market volatility financial market stability pension fund withdrawals Markov-switching dynamic regression

Introduction

Stock market volatility has long been a subject of interest for researchers and policymakers alike. Understanding the factors contributing to volatility is crucial for investors, pension fund participants, and regulators seeking to maintain market stability and enhance financial performance. Among these factors, the role of pension fund investments in shaping stock market dynamics has received increasing attention in recent years. The South African stock market, as one of the largest and most influential in Africa, has witnessed significant growth and development over the past decade. With the increasing participation of pension funds in the market, it is imperative to examine the impact of their investments on stock market volatility in South Africa. Several studies have investigated the impact of pension fund investments on stock market development. Bayar et al. (2022) found that stock market development significantly affects the pension funds and insurance sectors in the short-term and that pension funds positively influence stock market development in Brazil, Chile, Hungary, Mexico, Peru, and South Africa. Similarly, Morina and Grima (2022) reported a positive link between pension fund investments and economic growth in non-OECD countries. Studies on pension funds and stock market volatility show mixed results: Fernandez (2014) found a mild effect of pension fund holdings on Chilean stock volatility; Thomas et al. (2014) observed a negative relationship in 34 OECD countries; Sanusi and Kapingura (2021) reported negligible effects on investment and growth in South Africa; and Xue et al. (2021) noted that while mutual funds reduced idiosyncratic volatility in developed markets, the effect was insignificant in emerging markets.

Despite the growing role of pension funds in South Africa, their impact on stock market volatility across different regimes remains unclear (Bayar et al., 2022; Sanusi & Eita, 2022; Sanusi & Kapingura, 2021). This gap constrains effective policy formulation and risk management. This article investigates the impact of expenditure (benefits paid and withdrawals) and income (member and employer contributions) of private pension and provident fund components on stock market volatility. Findings provide policymakers with insights to enhance market stability, protect investors, and ensure pension fund sustainability. Understanding these effects also helps investors make informed decisions and navigate market fluctuations confidently.

This article examines the impact of pension fund withdrawals, member contributions, and employer contributions on stock market performance and volatility across different regimes in South Africa. Using quarterly data from 2010Q1 to 2022Q4, the study employs autoregressive conditional heteroskedasticity (ARCH)-based volatility measures and a Markov-switching dynamic regression (MSDR) framework to capture regime-dependent dynamics and transition probabilities in both stock market returns and volatility. The results indicate that pension fund contributions from both members and employers are positively associated with stock market returns, underscoring their role in supporting market performance. In contrast, the effects of pension fund withdrawals on stock market volatility are regime-specific. In the first regime, withdrawals reduce volatility, suggesting a stabilizing effect, while in the second regime, higher withdrawals increase volatility, reflecting heightened market uncertainty and potential selling pressures. These findings highlight the importance of the timing and scale of pension withdrawals in shaping market outcomes.

The article is organized as follows: the following section presents the literature review, followed by the methodology and results of the study, and lastly the conclusion.

Literature Review

Various theories examine private pension funds, their strategies, and market impacts. Portfolio theory recommends diversifying across asset classes to balance risk and return and manage stock market volatility (Mittal et al., 2022). Lifecycle investing tailors asset allocation to age, reducing exposure to high-risk assets over time, potentially mitigating market volatility (Bodie, 2015). Behavioral finance highlights how psychological biases and herd behavior can amplify volatility, particularly during market stress (Fakhry, 2016).

Fernandez (2014) examined 42 Chilean firms from 2002 to 2008 using Arellano–Bond regressions, finding that pension fund holdings have a mild effect on stock return volatility, and higher volatility reduces pension fund holdings, suggesting a preference for safer stocks. Thomas et al. (2014) analyzed 34 OECD countries from 2000 to 2010 with panel dynamic regressions and reported a negative relationship between pension fund equity shares and market volatility. Gökçen and Yalçın (2015) applied factor models in Turkey from 2004 to 2011 to explain 76%–95% of fund returns variations. Bayar (2016) found that private pension funds positively influence the long-term growth of Turkey’s debt and stock markets and exhibit causal links among these markets. Alda García and Marco Sanjuán (2017) observed that domestic equity pension funds in eight European countries from 1995 to 2014 positively affect market size, returns, and stability, especially in the short-term. Staveley-O’Carroll and Staveley-O’Carroll (2017) showed that government-managed pensions in the USA and China influence private savings, capital allocation, and portfolio risk. Alda (2017) found that pension funds have become dominant in 13 European stock markets from 1999 to 2014, with effects varying over time due to adaptive managerial strategies. Coşkun et al. (2017) demonstrated in Turkey (2006–2016) that capital market development has a long-run cointegrating relationship with economic growth, with causality running from markets to growth.

De la Torre-Torres et al. (2018) examined Mexican public pension funds from 2005 to 2018 using a Markov-Switching model, which demonstrated that short-term underperformance in socially responsible equities does not persist long-term, with two regimes showing better outcomes during distress periods. Chovancova et al. (2019) analyzed pension fund portfolios from 2005 to 2015 and reported that bonds have a stronger impact on performance than stocks. Brzeszczyński et al. (2019) studied large capital inflows in a thin market from 2000 to 2018 and found that pension fund flows have negligible long-term effects on stock prices, questioning the hypothesis of information-based trades. Babalos and Stavroyiannis (2020) used panel vector autoregression (VAR) across 29 countries, showing that pension fund equity investments enhance stock market development, with significant bidirectional Granger causality between market development and pension fund investments.

Shen et al. (2020) examined Chinese pension funds (2007–2016) using a four-factor model and found that only 5 of the 16 entrusted funds generated positive risk-adjusted returns, suggesting a principal–agent problem. Daradkah and Al-Hamdoun (2021) used autoregressive distributed lag (ARDL) co-integration in Jordan from 1980 to 2017 and found a significant short-run relationship between pension funds and capital market development. Sanusi and Kapingura (2021) applied Bayesian linear regression in South Africa from 1990 to 2019 and reported negligible effects of pension funds on investment and growth. Holzner et al. (2022) demonstrated that public pension expenditures and fund assets reduce per capita consumption and macroeconomic volatility across 35 OECD countries from 1980 to 2018. Xue et al. (2021) showed that pension funds strongly stabilized country-specific market risk in 47 global markets, outperforming mutual funds, particularly in emerging markets. Babalos et al. (2021) analyzed US equity fund flows from 2000 to 2015 using VAR-GARCH and observed post-2008 feedback trading and volatility spillovers from stock returns to fund flows, though effects weakened during the crisis.

Bayar et al. (2022) examined 15 emerging markets from 2004 to 2019, which showed that stock market development significantly affects pension funds and insurance in the short-term, while pension funds positively influence stock markets in Brazil, Chile, Hungary, Mexico, Peru, and South Africa in the long-term. Sanusi and Eita (2022) analyzed South Africa from 1990 to 2022 using asymmetric cointegration and reported long-run correlations between pension fund investments and stock market variables, suggesting that secure pension funds can strengthen market depth. Bissoondoyal-Bheenick et al. (2023) observed that, over the period 1990–2016, investment risk increases as portfolios shift from moderate to more aggressive options, while balanced fund allocations do not influence long-term risk. Morina and Grima (2022) found that pension fund investments in global financial markets positively contribute to economic growth in non-OECD countries from 2002 to 2018.

Assefuah et al. (2023) examined 48 African countries using system generalized method of moments (GMM) and found that pension fund interactions negatively affect capital market development, recommending cross-listing, cross-border investment, and alternative asset allocation. Heusel and Mager (2023) studied corporate pension plan funding from 2000 to 2017 and found that lower funding levels are associated with higher unexpected earnings and stock returns, with weaker underreaction for firms with greater investor attention and transparency. Hwang and Cho (2023) analyzed the Korean stock market from 2002 to 2018 using multivariate regression, demonstrating that a 1% increase in pension fund activity raises the market by 0.168%, with trades of different investor types positively affecting newly added stock returns, indicating that pension fund trading influences index inclusion effects.

Methodology

This article quantitatively analyzes the role of pension fund investments in shaping South African stock market volatility from 2010Q1 to 2022Q4 (Table 1). To estimate volatility, the study employs generalized autoregressive conditional heteroskedasticity (GARCH), asymmetric power autoregressive conditional heteroskedasticity (APARCH), and nonlinear power autoregressive conditional heteroskedasticity (N-PARCH) models, which capture key features of financial time series, including volatility clustering, persistence, and asymmetry (Chinhamu & Chifurira, 2019; Engle, 1982; Jeribi & Ghorbel, 2021; Xiao & Koenker, 2009). APARCH and N-PARCH account for asymmetric responses to positive and negative shocks, while N-PARCH further incorporates nonlinearities, capturing complex patterns and interactions in volatility dynamics. These models enhance the accuracy of volatility estimation and forecasting. Similar approaches have been used by Babalos et al. (2021) and Aggarwal et al. (2021).

Table 1.

Economic Variables of the Study.

Variable	Description	Sourced
$l n_s e t r_j s e$	Secondary market—stock exchange transactions: total value (turnover) of shares traded on all exchanges	SARB (2023)
$l n_i n f l_e_f$	Inflation expectation by the financial sector	SARB (2023)
$l n_g d p$	The gross domestic product growth	SARB (2023)
$l n_w u i$	Smoothed world uncertainty index for South Africa	Fed-USA (2023) using Ahir et al. (2022) methodology
$l n_p p f_w t h d$	Private pension and provident funds: Expenditure—Benefits paid—Pension withdrawals and other benefits	SARB (2023)
$l n_p p f_c m b r$	Private pension and provident funds: Income—contributions from members	SARB (2023)
$l n_p p f_c e m p l$	Private pension and provident funds: Income—contributions from employers	SARB (2023)
Main Volatility Variables
$l n_s e t r_j s e_v_h t_g a r c h$	Stock exchange transactions volatility using the GARCH approach	Estimated
$l n_s e t r_j s e_v_h t_a p a r c h$	Stock exchange transactions volatility using the APARCH approach	Estimated
$l n_s e t r_j s e_v_h t_n p a r c h$	Stock exchange transactions volatility using the N-PARCH approach	Estimated

The article uses a MSDR model to assess pension funds’ impact on stock returns and volatility across regimes. MSDR model captures shifts in relationships under varying market conditions, including stable or volatile periods (Hansen, 1996). Similar applications appear in Coşkun et al. (2017) and De la Torre-Torres et al. (2018).

Theoretical Framework and Model Specification

The article used the framework for stock exchange transactions, expressed in Equation (1).

l n_s e t r_j s e = P \times Q

(1)

Here, $l n_s e t r_j s e$ represents the total value (turnover) of shares traded on all exchanges, with $P$ as the price and $Q$ as quantity. Equation (1) is extended using a factor-based model to capture the impact of macroeconomic variables, such as inflation and GDP growth, on stock market volatility (Huang, 2023), as shown in Equation (2).

\begin{matrix} l n_s e t r_j s e_{t} = l . l n_s e t r_j s e_{t} + l n_i n f l_e_f_{t} + l n_g d p_{t} + l n_w u i_{t} \\ + \sum_{k = 1}^{k} \{\begin{matrix} l n_p p f_w t h d_{t} \\ l n_p p f_c m b r_{t} \\ l n_p p f_c e m p l_{t} \end{matrix} \end{matrix}

(2)

Factor-based models can include multivariate ARCH models. Stock exchange volatility is estimated as $l n_s e t r_j s e_v_h t_g a r c h$ (GARCH); $l n_s e t r_j s e_v_h t_a p a r c h$ (APARCH), and $l n_s e t r_j s e_v_h t_n p a r c h$ (N-PARCH), as shown in Equations (3)–(5).

\begin{matrix} l n_s e t r_j s e_v_h t_g a r c h_{t} = l . l n_s e t r_j s e_{t} + l n_i n f l_e_f_{t} + l n_g d p_{t} + l n_w u i_{t} \\ + \sum_{k = 1}^{k} \{\begin{matrix} l n_p p f_w t h d_{t} \\ l n_p p f_c m b r_{t} \\ l n_p p f_c e m p l_{t} \end{matrix} \end{matrix}

(3)

\begin{matrix} l n_s e t r_j s e_v_h t_a p a r c h_{t} = l . l n_s e t r_j s e_{t} + l n_i n f l_e_f_{t} + l n_g d p_{t} + l n_w u i_{t} \\ + \sum_{k = 1}^{k} \{\begin{matrix} l n_p p f_w t h d_{t} \\ l n_p p f_c m b r_{t} \\ l n_p p f_c e m p l_{t} \end{matrix} \end{matrix}

(4)

\begin{matrix} l n_s e t r_j s e_v_h t_n p a r c h_{t} = l . l n_s e t r_j s e_{t} + l n_i n f l_e_f_{t} + l n_g d p_{t} + l n_w u i_{t} \\ + \sum_{k = 1}^{k} \{\begin{matrix} l n_p p f_w t h d_{t} \\ l n_p p f_c m b r_{t} \\ l n_p p f_c e m p l_{t} \end{matrix} \end{matrix}

(5)

Model Specification Autoregressive Conditional Heteroskedasticity and Other Models

Research on volatility modeling highlights that the ARCH(p) model, introduced by Engle (1982), captures the effect of past volatility on current and future conditional variance as represented in Equations (6)–(8).

r_{t} = μ + ε_{t}

(6)

ε_{t} = \sqrt{h, z_{t}}

(7)

h_{t} = \sum_{j = i}^{q} α_{j} ε_{t - 1}^{2}

(8)

where, $r_{t}$ is the return of $g_s e t r_j s e$ , with $μ$ as the mean and $ε_{t}$ the residual, $h$ denotes conditional variance, and $z_{t}$ is a standard normal variable following an autoregressive process (Engle, 1982). Bollerslev (1986) extended ARCH to GARCH $(p, q)$ , incorporating past squared errors and past conditional variances, with zero-mean normality shown in Equations (9)–(10).

y_{t} = ε_{t} = e_{t} \sqrt{h_{t}}

(9)

y_{t} | ψ_{t - 1} ~ N (0, h)

(10)

where $ψ_{t - 1}$ denote the past information $t - 1$ , and the variance $h_{t}$ is the variance (Bollerslev, 1986) in Equation (11).

h_{t} = \bar{ω} + \sum_{j = i}^{m} α_{j} ε_{t - j}^{2} + \sum_{j = i}^{s} β_{i} h_{t - i}

(11)

where $h_{t} = l n_s e t r_j s e_v_h t_g a r c h_{t}$ denotes variance, with $p \geq 0; q > 0$ $q$ as GARCH moving-average and ARCH autoregressive orders, $α$ and $β$ as parameters (Bollerslev, 1986). APARCH extends ARCH (Engle, 1982) by modeling time-varying volatility and allowing asymmetry, with conditional variance $σ_{t}^{2}$ as a function of past squared and absolute residuals (Jeribi & Ghorbel, 2021; Xiao & Koenker, 2009) as in Equation (12).

σ_{t}^{2} = ω + \sum_{j = i}^{q} α_{i} {(|ε_{t - 1}| γ_{j} ε_{t - j})}^{2} + \sum_{i = i}^{p} β_{t - i} {(σ_{t - 1})}^{2}

(12)

where $σ_{t}^{2} = l n_s e t r_j s e_v_h t_a p a r c h_{t}$ , $ε$ is the residual, ω the constant, $α$ the effect of past absolute residuals, $β$ is the asymmetry, and $γ$ the effect of squared residuals (Engle, 1982; Jeribi & Ghorbel, 2021; Xiao & Koenker, 2009). N-PARCH extends APARCH by adding nonlinearities to the conditional variance as in Equation (12), allowing more flexible modeling of volatility responses to positive and negative shocks, as in Equation (13).

σ'_{t}^{2} = ω + \sum_{j = i}^{q} α_{i} {(|ε_{t - 1}| γ_{j} ε_{t - j})}^{δ} + \sum_{i = i}^{p} β_{t - i} {(σ_{t - 1})}^{δ}

(13)

where $σ'_{t}^{2} = l n_s e t r_j s e_v_h t_n p a r c h_{t},$ the N-PARCH model includes nonlinear terms $\sum_{j = i}^{q} α_{i} {(|ε_{t - 1}| γ_{j} ε_{t - j})}^{δ}$ and lagged variance $\sum_{i = i}^{p} β_{t - i} {(σ_{t - 1})}^{δ}$ , allowing flexible modeling of volatility dynamics (Engle, 1982; Jeribi & Ghorbel, 2021; Xiao & Koenker, 2009). Positive $δ$ amplifies past volatility, causing clustering, while negative $δ$ dampens it, promoting mean reversion.

Model Specification Markov-Switching Dynamic Regression Model

In a MSDR, the series moves between different hidden regimes according to a Markov process, with the time spent in each regime varying randomly. This framework allows the behavior of the series to adapt depending on the regime it is in, capturing changes in dynamics over time (Hansen, 1996, 2000). If given an economic data series denoted by $\begin{matrix} l n_s e t r_j s e_v_h t_g a r c h_{t} \\ l n_s e t r_j s e_v_h t_a p a r c h_{t} \\ l n_s e t r_j s e_v_h t_n p a r c h_{t} \end{matrix}\} = x_{s_{t}}$ , where $t = 1, 2, ..., T$ , is characterized by two regimes, such an economic data series can be presented in Equations (14) and (15) as follows.

S t a t e 1 : \begin{matrix} l n_s e t r_j s e_v_h t_g a r c h_{t} \\ l n_s e t r_j s e_v_h t_a p a r c h_{t} \\ l n_s e t r_j s e_v_h t_n p a r c h_{t} \end{matrix}\} = ψ_{s_{t}} + φ l n_s e t r_j s e_v_h t_{t - 1} + ϵ_{s}

(14)

S t a t e 2 : \begin{matrix} l n_s e t r_j s e_v_h t_g a r c h_{t} \\ l n_s e t r_j s e_v_h t_a p a r c h_{t} \\ l n_s e t r_j s e_v_h t_n p a r c h_{t} \end{matrix}\} = ψ_{s_{t}} + β \sum_{k = 1}^{4} X_{t} + Z_{t} λ_{s} + ϵ_{s}

(15)

Equations (14) and (15) present the Markov-switching autoregressive model with constant and time-varying transition probabilities, respectively, allowing regime shifts over time. In Equation (15), $λ_{s}$ denotes regime-dependent coefficients, $β$ is regime-invariant, and the control variables enter with $ϵ_{t}$ . Regime-specific parameters are defined in Equations (16)–(18).

ϵ_{t} \sim N (0, δ_{s_{t}}^{2})

(16)

δ_{s_{t}}^{2} = \sum_{s = 1}^{k} δ_{s}^{2}, \sum_{s = 1}^{k - 1} δ_{s}^{2} > 0

(17)

ψ_{s_{t}} = \sum_{s = 1}^{k} ψ_{s}

(18)

In this specification, $ψ_{1}$ and $ψ_{2}$ represent the intercepts for regimes 1 and 2, respectively, while $ϵ_{t}$ is a white noise error with variance $δ_{s_{t}}^{2}$ . The model accounts for changes in the intercept across regimes, following the framework proposed by Hamilton (1989, 1990). If the points at which regime switches occur are known, the model can be rewritten as in Equation (19).

\begin{matrix} l n_s e t r_j s e_v_h t_g a r c h_{t} \\ l n_s e t r_j s e_v_h t_a p a r c h_{t} \\ l n_s e t r_j s e_v_h t_n p a r c h_{t} \end{matrix}\} = ψ_{s_{t}} ψ_{1} + (1 - ψ_{s_{t}}) ψ_{2} + ϵ_{t}

(19)

The subscript $ψ_{s_{t}}$ is 1 if the process is in regime 1 and 0 otherwise. Markov-switching regression models allow the parameters to vary over the unobserved regimes. The MSDR model with a regime-dependent intercept term is reflected in Equation (20).

\begin{matrix} l n_s e t r_j s e_v_h t_g a r c h_{t} \\ l n_s e t r_j s e_v_h t_a p a r c h_{t} \\ l n_s e t r_j s e_v_h t_n p a r c h_{t} \end{matrix}\} = ψ_{s_{t}} ψ_{2} + ϵ_{t}

(20)

The parameter of interest, $ψ_{ψ_{s_{t}}}$ , takes the value µ when the system is in regime $ψ_{ψ_{s_{t}}} = μ$ when $ψ_{s_{t}} = 1$ and µ₂ when in regime $ψ_{s_{t}} = 2$ . The probability of being in a particular regime is determined through transition probabilities. For a two-regime model, the one-step transition probability p₁₁ represents the chance of remaining in regime 1 in the next period, given the current period is in regime 1, while $p_{22}$ denotes the probability of staying in regime 2 (Hansen, 1996, 2000). More generally, the likelihood that $ψ_{s_{t}}$ equals a specific regime $j = \in (1, ... k)$ depends on the previous state $ψ_{s_{t - 1}}$ , as formalized in Equation (21).

P_{r} (ψ_{s_{t}} = j | ψ_{s_{t - 1}} = i) = P_{i j}

(21)

The transition probabilities $k \times k$ from one regime to another can be presented in matrix (22).

P = (\begin{matrix} p_{11} & p_{12} & \dots & p_{k 1} \\ p_{21} & p_{22} & \dots & p_{k 2} \\ \cdot & \cdot & \dots & \cdot \\ \cdot & \cdot & \dots & \cdot \\ \cdot & \cdot & \dots & \cdot \\ p_{k 1} & p_{k 2} & \dots & p_{k k} \end{matrix})

(22)

This mechanism governs the evolution of the Markov chain by determining the probability of transitioning between regimes over time. The MSDR framework, which captures regime-dependent dynamics, is formally specified in Equations (23)–(25). In addition, Equations (1)–(5) establish the core theoretical structure of the model. These equations are subsequently extended and embedded within the Markov-switching framework, allowing the baseline relationships to vary across regimes and thereby providing a more flexible representation of the underlying economic dynamics.

s l n_s e t r_j s e_v_h t_g a r c h_{t} = \{\begin{matrix} β_{11} + γ_{12} l . s e t r_j s e_{t - 1} + γ_{13} l n_i n f l_e_f_{t} + γ_{14} l n_g d p_{t} \\ + γ_{15} l n_w u i_{t} + \sum_{k = 1}^{k} γ_{16} \{\begin{matrix} l n_p p f_w t h d_{t} \\ l n_p p f_c m b r_{t} \\ l n_p p f_c e m p l_{t} \end{matrix} + ϵ_{1, t} \\ β_{21} + γ_{22} l . s e t r_j s e_{t - 1} + γ_{23} l n_i n f l_e_f_{t} + γ_{24} l n_g d p_{t} \\ + γ_{25} l n_w u i_{t} + \sum_{k = 1}^{k} γ_{26} \{\begin{matrix} l n_p p f_w t h d_{t} \\ l n_p p f_c m b r_{t} \\ l n_p p f_c e m p l_{t} \end{matrix} + ϵ_{2, t} \end{matrix}

(23)

l n_s e t r_j s e_v_h t_a p a r c h_{t} = \{\begin{matrix} β_{11} + γ_{12} l . s e t r_j s e_{t - 1} + γ_{13} l n_i n f l_e_f_{t} + γ_{14} l n_g d p_{t} \\ + γ_{15} l n_w u i_{t} + \sum_{k = 1}^{k} γ_{16} \{\begin{matrix} l n_p p f_w t h d_{t} \\ l n_p p f_c m b r_{t} \\ l n_p p f_c e m p l_{t} \end{matrix} + ϵ_{1, t} \\ β_{21} + γ_{22} l . s e t r_j s e_{t - 1} + γ_{23} l n_i n f l_e_f_{t} + γ_{24} l n_g d p_{t} \\ + γ_{25} l n_w u i_{t} + \sum_{k = 1}^{k} γ_{26} \{\begin{matrix} l n_p p f_w t h d_{t} \\ l n_p p f_c m b r_{t} \\ l n_p p f_c e m p l_{t} \end{matrix} + ϵ_{2, t} \end{matrix}

(24)

l n_s e t r_j s e_v_h t_n p a r c h_{t} = \{\begin{matrix} β_{11} + γ_{12} l . s e t r_j s e_{t - 1} + γ_{13} l n_i n f l_e_f_{t} + γ_{14} l n_g d p_{t} \\ + γ_{15} l n_w u i_{t} + \sum_{k = 1}^{k} γ_{16} \{\begin{matrix} l n_p p f_w t h d_{t} \\ l n_p p f_c m b r_{t} \\ l n_p p f_c e m p l_{t} \end{matrix} + ϵ_{1, t} \\ β_{21} + γ_{22} l . s e t r_j s e_{t - 1} + γ_{23} l n_i n f l_e_f_{t} + γ_{24} l n_g d p_{t} \\ + γ_{25} l n_w u i_{t} + \sum_{k = 1}^{k} γ_{26} \{\begin{matrix} l n_p p f_w t h d_{t} \\ l n_p p f_c m b r_{t} \\ l n_p p f_c e m p l_{t} \end{matrix} + ϵ_{2, t} \end{matrix}

(25)

Results

Appendix Table A1 (available online as supplemental material) presents the descriptive statistics of the main variables. Inflation expectations $l n_i n f l_e_f$ exhibit a mean of 1.658% with a narrow range of 1.361–1.85%, suggesting relatively stable short-term expectations over the sample period. GDP growth $l n_g d p$ averages 0.06% with a standard deviation of 0.76, reflecting modest fluctuations in economic activity. The World Uncertainty Index $l n_w u i$ has a mean of −0.567, indicating periods of below-average global uncertainty as measured for South Africa. Private pension and provident fund activity shows substantial average flows: withdrawals $l n_p p f_w t h d$ 9.734%, member contributions $l n_p p f_c m b r$ 9.944%, and employer contributions $l n_p p f_c e m p l$ 9.694%, highlighting the importance of these funds in the domestic financial system.

The correlation analysis in Appendix Table A2 (in supplemental material) reveals several economically meaningful relationships. Inflation expectations are negatively correlated with stock market volatility −0.347, suggesting that higher expected inflation may introduce uncertainty in financial markets, consistent with empirical evidence that rising inflation expectations often elevate risk perceptions and market volatility (Andersen et al., 2003; Bekaert & Wang, 2010). GDP growth shows a positive correlation of 0.338 with volatility, indicating that periods of stronger economic expansion may coincide with heightened financial market fluctuations, potentially due to adjustments in monetary policy or investor repositioning, which aligns with findings by Campbell et al. (2001). The World Uncertainty Index positively correlates with volatility (0.34), confirming that macroeconomic and geopolitical uncertainty amplifies stock market fluctuations, as documented by Ahir et al. (2022). Pension fund variables exhibit exceptionally strong positive correlations with volatility: withdrawals $l n_p p f_w t h d$ 0.93, member contributions $l n_p p f_c m b r$ 0.92, and employer contributions $l n_p p f_c e m p l$ 0.87. These relationships underscore the close linkage between pension fund flows and financial market dynamics. This finding is in line with studies showing that large institutional flows, such as those from pension funds, can amplify market volatility due to their size and portfolio rebalancing effects (Chen et al., 1986; Fama & French, 1993). The strong correlations suggest that movements in pension fund activity not only reflect investor behavior but may also serve as a conduit through which macroeconomic and policy shocks transmit to the stock market.

Appendix Table A3 (in supplemental material) presents the Dickey–Fuller (DF) unit root test results for the key macroeconomic, financial, and public finance variables used in the study. The test was applied to the second-differenced series $Δ ²$ , with critical values at 1%, 5%, and 10% significance levels reported alongside p values. All DF test statistics are substantially lower than the 1% critical value −3.58, and all p values are below .001, indicating strong rejection of the null hypothesis of a unit root.

Figure 1.

Source: Data source from SARB (2023), figure compiled by the author.

Figure 1 shows clustering of economic variables into the first and second volatility waves. Graphs (a) and (b) indicate that volatility in the first wave persists into the second, suggesting clustering. Such patterns are effectively modeled using ARCH models, which capture time-varying volatility. Graphs c–d show the data approximate a normal distribution, supporting the use of ARCH models for accurate volatility estimation and prediction.

Table 2.

Autoregressive Conditional Heteroskedasticity (ARCH) Effect Test.

White’s Test							Breusch–Pagan or Cook–Weisberg Test
Heteroskedasticity	3.18	2	0.2041	χ²(2)	=	4.82	χ² (2)	=	11.77
Skewness	1.26	1	0.2616	Prob > χ²	=	0.0282	Prob > χ²	=	0.0028
Kurtosis	6.41	1	0.0114
Total	10.84	4	0.0002

Note: Ha: Unrestricted heteroskedasticity; Ho: Homoskedasticity or constant variance; Variables: Fitted values of $s e t r_{j s e}$ .

Table 2 reports White’s and Breusch–Pagan tests for heteroskedasticity and ARCH effects. White’s test prob > χ² = 0.0282 and Breusch–Pagan prob > χ² = 0.0028 indicate volatility clustering, where high-volatility periods tend to follow high-volatility periods, and low-volatility periods follow low-volatility periods. Figure 2 shows the conditional variance of $l n_s e t r_j s e_v_h t_{g a r c h}; l n_s e t r_j s e_v_h t_a p a r c h$ ; and $l n_s e t r_j s e_v_h t_n p a r c h$ stock exchange volatility. All three follow similar trends but differ in magnitude. Volatility spikes in 2012 were driven by global uncertainty, including the European debt crisis. In 2015, weak domestic growth, high unemployment, and electricity constraints increased volatility (van der Westhuizen et al., 2022). Political uncertainty and credit rating downgrades in 2017 further affected the market (Chang et al., 2023). From 2018 to 2021, a weak economy, Eskom’s crisis, and COVID-19 contributed to continued fluctuations, with markets sharply declining in early 2020 before gradually recovering (van der Westhuizen et al., 2022).

Before estimating the conditional variance for the GARCH family models, the appropriate mean equation for the stock exchange series $l n_s e t r_j s e$ was determined. Appendix Table A4 (in supplemental material) shows that several ARMA(p,q) specifications were tested to capture the autocorrelation structure of the series. Model selection was guided by the Akaike Information Criterion and Bayesian Information Criterion, and the residuals were checked for white noise using the Ljung–Box Q test. The ARMA(1,1) specification was selected as the optimal mean equation because it minimizes both Akaike Information Criterion and Bayesian Information Criterion while producing residuals that are consistent with white noise. The residuals from this mean equation were subsequently used as input for the conditional variance equations in the GARCH, APARCH, and N-PARCH models, ensuring that volatility dynamics are modeled on properly filtered innovations.

Appendix Table A5 (in supplemental material) reports the estimation results for GARCH(1,1), APARCH, and N-PARCH models applied to JSE secondary market transactions $s e t r_j s e$ . The lagged dependent variable is highly significant across all models (~0.91, t > 11), indicating strong persistence in transaction values. The intercept is not significant, suggesting that past values largely determine current transactions. In the volatility equations, the GARCH (1,1) model shows weak persistence with an insignificant ARCH term of 0.303 and a negative GARCH term of −0.0668. APARCH and N-PARCH models, allowing for asymmetry and nonlinearity, also produce positive but insignificant volatility coefficients, indicating a limited impact of past shocks. The power parameter is highly significant in all models, supporting the use of transformed volatility. Overall, while transaction values are strongly persistent, conditional volatility is only moderately persistent, and more complex ARCH specifications offer little improvement over the standard GARCH in this sample.

Figure 2.

Note: $l n_s e t r_j s e_v_h t_g a r c h$ ; $l n_s e t r_j s e_v_h t_a p a r c h$ ; and $l n_s e t r_j s e_v_h t_n p a r c h$ show volatility from GARCH, APARCH, and N-PARCH models, while $s e_s e t r_j s e_g a r c h$ , $s e_s e t r_j s e_a p a r c h$ , and $s e_s e t r_j s e_n p a r c h$ show the corresponding residuals.

Table 3, the stock exchange in regimes 1 and 2, where estimation 1 reflects the impact of $l n_p p f_w t h d$ . The coefficient of $l n_p p f_w t h d$ is found to be statistically significant at the 1% level. The results indicate that a 1% increase in $l n_p p f_w t h d$ is associated with an increase of 0.468% in $l n_s e t r_j s e$ in regime 1 and 0.980% in regime 2. The results indicate that higher pension fund withdrawals $l n_p p f_w t h d$ positively affect stock market returns $l n_s e t r_j s e$ in both regimes, suggesting that withdrawals may increase market activity. Participants should consider the timing and amount of withdrawals, while fund managers and advisors can guide strategies to mitigate adverse effects. Regulators should monitor withdrawals to ensure market stability. Appendix Table A5 (in supplemental material) shows that both member $l n_p p f_c m b r$ and employer contributions $l n_p p f_c e m p l$ significantly boost stock market returns $l n_s e t r_j s e$ across regimes. A 1% increase in member contributions raises returns by 0.540% in regime 1 and 1.014% in regime 2, while employer contributions increase returns by 0.260% and 1.072%, respectively. The stronger effects in regime 2 suggest that pension inflows stabilize markets during higher volatility. These results align with prior studies. Thomas et al. (2014) and Alda García and Marco Sanjuán (2017) report that pension funds enhance market returns and stability, while Bayar (2016) highlights their role in long-term market growth. The findings also resonate with De la Torre-Torres et al. (2018), who observed that pension strategies mitigate losses in distressed periods. In contrast, studies in thin or less-developed markets (Brzeszczyński et al., 2019; Sanusi & Kapingura, 2021) found limited effects, highlighting the importance of market depth and fund management.

Table 3.

The Stock Exchange in Regimes 1 and 2.

Regime 1
Estimation	(1)	(2)	(3)
Variables	$l n_s e t r_j s e$	$l n_s e t r_j s e$	$l n_s e t r_j s e$
$L . s e t r_j s e$	0.000000301^*	0.000000251	0.000000518^*
	(2.45)	(1.18)	(1.97)
$l n_i n f l_e_f$	−0.0443	0.00284	−0.155
	(−0.27)	(0.02)	(−0.35)
$l n_g d p$	0.00877	0.00999	0.00550
	(0.34)	(0.54)	(0.12)
$l n_w u i$	0.0268	0.0175	0.0572^*
	(1.23)	(0.89)	(2.41)
$l n_p p f_w t h d$	0.468^***
	(4.95)
$l n_p p f_c m b r$		0.540^***
		(3.72)
$l n_p p f_c e m p l$			0.260^*
			(2.52)
$_c o n s$	9.119^***	8.267^***	11.10
	(11.14)	(6.90)	(2.24)
Regime 2
$L . s e t r_j s e$	−0.000000504^***	−0.000000355^***	−0.000000353^*
	(−4.52)	(−3.74)	(−1.99)
$l n_i n f l_e_f$	−0.00688	0.316	−0.0212
	(−0.05)	(1.92)	(−0.06)
$l n_g d p$	0.0638^**	0.0391	0.107^*
	(2.75)	(1.38)	(2.01)
$l n_w u i$	0.0801^**	0.119^***	0.101
	(2.82)	(4.49)	(1.22)
$l n_p p f_w t h d$	0.980^***
	(10.07)
$l n_p p f_c m b r$		1.041^***
		(28.15)
$l n_p p f_c e m p l$			1.072^***
			(16.45)
$_c o n s$	5.114^***	3.638	4.303
	(5.49)	(4.21)	(9.46)
$N$	48	48	48

Note: t-statistics in parentheses and ^*p < .05, ^**p < .01, ^***p < .001.

Appendix Table A6 (in supplemental material) reports the impact of pension fund activity on stock market volatility across regime 1 (lower-volatility periods) using GARCH, APARCH, and N-PARCH models. Pension fund withdrawals $l n_p p f_w t h d$ significantly reduce volatility across all model specifications. Specifically, a 1% increase in withdrawals leads to a decline in ln_setr_jse_v_ht_garch; ln_setr_jse_v_ht_aparch; and ln_setr_jse_v_ht_nparch by 1.437%, 0.594%, and 1.237%, respectively. Moreover, the stabilizing effect is stronger under certain specifications, where a 1% increase in pension fund withdrawals reduces ln_setr_jse_v_ht_garch by 2.010%, with statistical significance at the 1% level (p < .01). This indicates that withdrawals reduce market exposure, dampening demand, and stabilizing returns, a finding consistent with Shen et al. (2020), who report that pension flows can stabilize market risk and enhance liquidity. Member contributions $l n_p p f_c m b r$ have a modest effect, significant only in the APARCH specification (0.090% reduction). This aligns with lifecycle and portfolio theories, suggesting that stable contributions reduce the need for frequent reallocations, thereby lowering volatility (Bodie, 2015; Mittal et al., 2022).

Employer contributions $l n_p p f_c e m p l$ exert a stronger stabilizing effect, with a 1% increase reducing $l n_s e t r_j s e_v_h t_g a r c h$ by 2.010% with p < .01. This reflects the signaling role of employer contributions, indicating financial stability and commitment, which can improve investor confidence and reduce uncertainty. These results echo findings from Bayar (2016) and Alda García and Marco Sanjuán (2017), who observed that institutional inflows enhance market stability and efficiency. Appendix Table A7 (in supplemental material) reports the impact of pension fund activity on stock market volatility in regime 2 (high-volatility periods) using GARCH, APARCH, and N-PARCH models. Unlike the stabilizing effects in regime 1, pension fund withdrawals $l n_p p f_w t h d$ significantly increase volatility in GARCH and APARCH models, with a 1% rise leading to 1.255% and 1.091% higher $l n_s e t r_j s e_v_h t$ , respectively. This suggests that withdrawals under stressed conditions may signal liquidity needs or financial strain, creating selling pressure and amplifying market fluctuations, consistent with behavioral finance insights on herd behavior and panic selling (Fakhry, 2016).

Member contributions $l n_p p f_c m b r$ also increase volatility, with 1% contributions raising $l n_s e t r_j s e_v_h t_g a r c h$ by 0.756% and $l n_s e t r_j s e_v_h t_a p a r c h$ by 1.914%. Higher contributions may spur trading activity, enhance liquidity, and heighten price fluctuations, reflecting the dual role of pension inflows in improving market participation while temporarily increasing short-term volatility (Bodie, 2015; Xue et al., 2021). Employer contributions $l n_p p f_c e m p l$ show the strongest effect, raising volatility across all specifications by 1.237%–1.400%. This highlights that significant institutional inflows can influence short-term market dynamics, emphasizing that regulators should monitor contribution-induced volatility while promoting pension fund growth. In high-volatility regimes, pension fund activity can temporarily increase market fluctuations, contrasting with the stabilizing role observed in lower-volatility periods and reinforcing the context-dependent nature of pension fund impacts reported in Shen et al. (2020) and Bayar (2016).

Figure 3 presents the regime 1-to-2 transition probabilities for $l n_s e t r_j s e,$ capturing secondary market turnover. Regime 1 averages 13.68%, reflecting a stable post-crisis period 2010Q1–2015Q1 characterized by moderate growth, low inflation, and subdued volatility. Regime 2 averages 14.14%, corresponding to 2015Q4–2022Q4, a period marked by political uncertainty, credit downgrades, policy instability, and global shocks, leading to higher market volatility and turnover. The higher probability of transitioning from regime 1-to-2 signals increased volatility, providing investors with insights for risk management and portfolio adjustments. These patterns align with the Markov-Switching literature, where shifts in macroeconomic and political conditions induce regime-dependent changes in market dynamics (De la Torre-Torres et al., 2018; Xue et al., 2021).

Figure 3.

Note: Variables: $l n_s e t r_j s e$ is total share turnover; Rgm_1: Regime 1, Rgm_2: Regime 2.

Figure 4 shows regime 1-to-2 transition probabilities for stock exchange rate volatility. Regime 1 has a mean volatility of 24.91%, while regime 2 has 45.66%. The shift between regimes highlights the dynamic nature of market risk. In low-volatility regime 1, investors may favor stable returns, whereas in high-volatility regime 2, strategies that account for greater uncertainty, such as diversification or hedging, are required. Understanding these transitions is essential for investors, policymakers, and regulators to manage risk and maintain market stability.

Figure 4.

Note: Rgm_1 = Regime 1, Rgm_2 = Regime 2.

Table 4 reports the expected duration of each regime for $l n_s e t r_j s e$ stock market volatility. For market turnover $l n_s e t r_j s e$ , regime 1 persists for 4.120 quarters, while regime 2 lasts 4.100 quarters, indicating roughly equal durations for stable and high-volatility periods. In contrast, volatility measures show a stark asymmetry: in regime 1, $l n_s e t r_j s e_v_h t_g a r c h$ ; $l n_s e t r_j s e_v_h t_a p a r c h$ ; and $l n_s e t r_j s e_v_h t_n p a r c h$ persist for 5.150 quarters, reflecting prolonged periods of low volatility, whereas regime 2 lasts only 1.458 quarters, suggesting that volatility spikes are short-lived. These findings highlight the transient nature of extreme market fluctuations and underscore the importance of pension fund strategies and investor risk management in adapting to temporary high-volatility episodes.

Table 4.

Expected Duration to Stay in the Regime

Estimation	$l n_s e t r_j s e$				$l n_s e t r_j s e_v_h t_g a r c h$ $l n_s e t r_j s e_v_h t_a p a r c h$ $l n_s e t r_j s e_v_h t_n p a r c h$
Regime	Estimate	Std. Err.	[95% Conf. Interval]		Estimate	Std. Err.	[95% Conf. Interval]
Regime 1	4.120	2.039	1.866	12.235	5.150	2.329	2.381	13.468
Regime 2	4.100	2.320	1.715	4.436	1.458	0.409	1.080	3.635

Number of observations = 48

Table 5 reflects the matrix of transition probabilities of $l n_s e t r_j s e$ ; $l n_s e t r_j s e_v_h t_g a r c h$ ; $l n_s e t r_j s e_v_h t_a p a r c h$ ; and $l n_s e t r_j s e_v_h t_n p a r c h$ in different regimes. Where regime 1 is characterized by the volatility mean rate of 24.9113% for $n_s e t r_j s e_v_h t_g a r c h;$ $l n_s e t r_j s e_v_h t_a p a r c h$ ; and $l n_s e t r_j s e_v_h t_n p a r c h$ . In regime 1, it is found to have a transition probability of 0.806. This reflects that there is an 80.6% chance that the economy will move from regime 1 and return to regime 1. Regime 2 is characterized by a volatility mean rate of 45.658% for $l n_s e t r_j s e_v_h t_g a r c h$ ; $l n_s e t r_j s e_v_h t_a p a r c h$ ; and $l n_s e t r_j s e_v_h t_n p a r c h$ . In regime 1, it is found to have a transition probability of 0.314. This reflects that there is a 31.4% chance that the economy will move from regime 2 and return to regime 2.

Table 5.

Matrix of Transition Probabilities.

Estimation	$l n_s e t r_j s e$				$l n_s e t r_j s e_v_h t_g a r c h$ $l n_s e t r_j s e_v_h t_a p a r c h$ $l n_s e t r_j s e_v_h t_n p a r c h$
Transition probabilities	Estimate	Std. Err.	[95% Conf Interval]		Estimate	Std. Err.	[95% Conf Interval]
$p_{11}$	0.757	0.120	0.464	0.918	0.806	0.088	0.580	0.926
$p_{12}$	0.243	0.120	0.082	0.536	0.194	0.088	0.074	0.420
$p_{21}$	0.244	0.138	0.069	0.583	0.686	0.192	0.275	0.926
$p_{22}$	0.756	0.138	0.417	0.931	0.314	0.192	0.074	0.725

Number of observations = 48

Conclusion

This study examined the relationship between pension fund activities, specifically withdrawals and contributions, and stock market performance in South Africa. It addresses a critical gap in the literature regarding the dynamic impact of pension fund flows on stock market returns and volatility under different market regimes, providing nuanced insights into investor behavior and market stability. The research question focused on how pension fund withdrawals and contributions influence stock market returns and volatility, and whether these effects vary across distinct market conditions.

The empirical analysis reveals several key findings. First, both pension fund withdrawals and contributions significantly affect stock market returns. Higher withdrawals are positively associated with returns across both regimes, highlighting the importance of participant behavior in shaping market outcomes. Similarly, contributions from both individuals and employers are positively linked to returns, suggesting that inflows into pension funds can reinforce stock market performance. Second, the study uncovers regime-dependent effects on volatility. In regime 1, withdrawals reduce volatility, implying a stabilizing effect, possibly due to counter-cyclical investment behavior during periods of market distress. Conversely, in regime 2, withdrawals increase volatility, reflecting liquidity pressures and market imbalances from excessive outflows. Employer contributions exhibit a similar pattern: they decrease volatility in regime 1 but increase it in regime 2, highlighting the role of market conditions and investor sentiment in mediating the impact of pension fund flows.

These findings have both theoretical and policy implications. Theoretically, the study contributes to the literature on institutional investors by demonstrating that pension fund activities can have dual effects on market stability depending on prevailing market regimes. This reinforces the need to consider nonlinearities and behavioral responses in modeling the relationship between institutional flows and market dynamics. From a policy perspective, regulators and policymakers should develop mechanisms to guide withdrawal and contribution behavior to align with market conditions. Measures could include monitoring excessive fund outflows, establishing risk-based withdrawal guidelines, and promoting investor education to mitigate volatility shocks. Pension fund managers and financial advisors play a crucial role in equipping participants with knowledge about the potential market impact of their actions, thus enhancing overall financial stability.

This study is subject to certain limitations. The analysis relies on available market and pension fund data, which may not fully capture informal or unreported flows. The regime classification assumes stability over the examined periods, which may not account for abrupt market changes outside the identified regimes. Additionally, the findings are specific to South Africa, and generalization to other emerging markets should be made cautiously. Future research could expand the analysis by incorporating high-frequency data, enabling a more precise examination of short-term market responses to pension fund activities. Investigating the interaction between pension fund flows and other institutional investors, such as mutual funds and hedge funds, could further illuminate market dynamics. Moreover, exploring behavioral factors influencing participant decisions, such as risk perception and financial literacy, would provide a deeper understanding of the mechanisms driving the observed effects.

Footnotes

Data Availability Statement

The data that support the findings of this study are derived from publicly available sources. These data are available from the corresponding author upon reasonable request.

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

Supplemental Material

ORCID iD

Eugene Msizi Buthelezi

References

Aggarwal

, Doifode

, & Tiwary

M. K.

(2021). Volatility spillover impact of FII and MF net equity flows on the Indian sectoral stock indices: Recent evidence using BEKK-GARCH. International Journal of Indian Culture and Business Management , 22(3), 350–363. https://doi.org/10.1504/IJICBM.2021.114084

Ahir

, Bloom

, & Furceri

(2022). The world uncertainty index (Working Paper No. 29763, pp. 114). National Bureau of Economic Research. https://doi.org/10.3386/w29763

Alda

(2017). The relationship between pension funds and the stock market: Does the aging population of Europe affect it? International Review of Financial Analysis , 49, 83–97. https://doi.org/10.1016/j.irfa.2016.12.008

Alda García

, & Marco Sanjuán

(2017). The importance of domestic equity pension funds on stock market. Spanish Journal of Finance and Accounting/Revista Española de Financiación y Contabilidad , 46(2), 227–248. https://www.jstor.org/stable/26597858

Andersen

T. G.

, Bollerslev

, Diebold

F. X.

, & Labys

(2003). Modeling and forecasting realized volatility. Econometrica , 71(2), 579–625. https://doi.org/10.1111/1468-0262.00418

Assefuah

K. A.

, Abor

J. Y.

, Kuttu

, & Amoah

(2023). Pension funds and capital market development in Africa: The role of institutional quality. Cogent Economics & Finance , 11(1), 2172809. https://doi.org/10.1080/23322039.2023.2172809

Babalos

, & Stavroyiannis

(2020). Pension funds and stock market development in OECD countries: Novel evidence from a panel VAR. Finance Research Letters , 34.

Babalos

, Caporale

G. M.

, & Spagnolo

(2021). Equity fund flows and stock market returns in the USA before and after the global financial crisis: A VAR-GARCH-in-mean analysis. Empirical Economics , 60, 539–555. https://link.springer.com/article/10.1007/s00181-019-01783-5

Bayar

(2016). Individual pension funds and capital market development in Turkey. Review of Economic and Business Studies , 9(2), 95–109. https://doi.org/10.1515/rebs-2016-0036

10.

Bayar

, Gavriletea

M. D.

, Danuletiu

D. C.

, Danuletiu

A. E.

, & Sakar

(2022). Pension funds, insurance companies and stock market development: Evidence from emerging markets. Mathematics , 10(13), Article 2335. https://doi.org/10.3390/math10132335

11.

Bekaert

, & Wang

(2010). Inflation risk and the inflation risk premium. Economic Policy , 25(64), 755–806. https://www.jstor.org/stable/40929921

12.

Bissoondoyal-Bheenick

, Brooks

, & Do

H. X.

(2023). Risk analysis of pension fund investment choices. Abacus , 59(3), 872–898. https://doi.org/10.1111/abac.12264

13.

Bodie

(2015). Thoughts on the future: Life-cycle investing in theory and practice. Financial Analysts Journal , 71(1), 43–48. https://doi.org/10.2469/faj.v71.n1.2

14.

Bollerslev

(1986). Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics , 31(3), 307–327. https://doi.org/10.1016/0304-4076(86)90063-1

15.

Brzeszczyński

, Gajdka

, & Schabek

(2019). Pension funds, large capital inflows and stock returns in a thin market. Journal of Pension Economics & Finance , 18(3), 347–387. https://doi.org/10.1017/S147474721800001X

16.

Campbell

J. Y.

, Lettau

, Malkiel

B. G.

, & Xu

(2001). Have individual stocks become more volatile? An empirical exploration of idiosyncratic risk. The Journal of Finance , 56(1), 1–43. https://doi.org/10.1111/0022-1082.00318

17.

Chang

H. W.

, Chang

, Ling

Y. H.

, & Yang

Y. L.

(2023). Dynamical linkages between the Brent oil price and stock markets in BRICS using quantile connectedness approach. Finance Research Letters , 54, 103748. https://doi.org/10.1016/j.frl.2023.103748

18.

Chen

N.-F.

, Roll

, & Ross

S. A.

(1986). Economic forces and the stock market. The Journal of Business , 59(3), 383–403. https://doi.org/10.1086/296344

19.

Chinhamu

, & Chifurira

(2019). Evaluating South Africa’s market risk using asymmetric power auto-regressive conditional heteroscedastic model under heavy-tailed distributions. Journal of Economic and Financial Sciences , 12(1), 1–11. https://hdl.handle.net/10520/EJC-1d074e160e

20.

Chovancova

, Hudcovsky

, & Kotaskova

(2019). The impact of stocks and bonds on pension fund performance. Journal of Competitiveness , 11(2), 22–37. https://doi.org/10.7441/joc.2019.02.02

21.

Coşkun

, Seven

Ü.

, Ertuğrul

H. M.

, & Ulussever

(2017). Capital market and economic growth nexus: Evidence from Turkey. Central Bank Review , 17(1), 19–29. https://doi.org/10.1016/j.cbrev.2017.02.003

22.

Daradkah

, & Al-Hamdoun

(2021). Pension funds, capital market development, and macroeconomic variables: Evidence from Jordan. Journal of Public Affairs , 21(2), Article e2215. https://doi.org/10.1002/pa.2215

23.

De la Torre-Torres

O. V.

, Galeana-Figueroa

, & Álvarez-García

(2018). Efficiency of the public pensions funds on the socially responsible equities of Mexico. Sustainability , 11(1), Article 178. https://doi.org/10.3390/su11010178

24.

Engle

R. F.

(1982). Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica , 50(4), 987–1007. https://doi.org/10.2307/1912773

25.

Fakhry

(2016). A literature review of behavioural finance. Journal of Economics Library , 3(3), 458–465.

26.

Fama

E. F.

, & French

K. R.

(1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics , 33(1), 3–56. https://doi.org/10.1016/0304-405X(93)90023-5

27.

Fed-USA. (2023). Federal Reserve Bank, FRED: Federal Reserve Economic Data . https://fred.stlouisfed.org/

28.

Fernandez

(2014). Stock volatility and pension funds under an individual capitalization-based system. Journal of Business Research , 67(4), 536–541. https://doi.org/10.1016/j.jbusres.2013.11.010

29.

Gökçen

, & Yalçın

(2015). The case against active pension funds: Evidence from the Turkish private pension system. Emerging Markets Review , 23, 46–67.

30.

Hamilton

J. D.

(1989). A new approach to the economic analysis of nonstationary time series and the business cycle. Econometrica , 57(2), 357–384. https://doi.org/10.2307/1912559

31.

Hamilton

J. D.

(1990). Analysis of time series subject to changes in regime. Journal of Econometrics , 45(1–2), 39–70. https://doi.org/10.1016/0304-4076(90)90093-9

32.

Hansen

B. E.

(1996). Inference when a nuisance parameter is not identified under the null hypothesis. Econometrica , 64(2), 413–430. https://doi.org/10.2307/2171789

33.

Hansen

B. E.

(2000). Sample splitting and threshold estimation. Econometrica , 68(3), 575–603. https://doi.org/10.1111/1468-0262.00124

34.

Heusel

, & Mager

(2023). Pension funding and the cross section of stock returns—The case of Germany. Journal of Banking & Finance , 150, Article 106816. https://doi.org/10.1016/j.jbankfin.2023.106816

35.

Holzner

, Jestl

, & Pichler

(2022). Public and private pension systems and macroeconomic volatility in OECD countries. Scottish Journal of Political Economy , 69(2), 131–168. https://doi.org/10.1111/sjpe.12278

36.

Huang

(2023). The more, the better? Forecasting gains from high-frequency housing prices in a Markov-switching dynamic factor model. Macroeconomic Dynamics , 27(1), 93–110. https://doi.org/10.1017/S1365100521000304

37.

Hwang

, & Cho

(2023). Pension fund trading behavior and the index inclusion: Evidence from the Korean stock market. Emerging Markets Finance and Trade , 59(3), 605–622. https://doi.org/10.1080/1540496X.2022.2109960

38.

Jeribi

, & Ghorbel

(2021). Forecasting developed and BRICS stock markets with cryptocurrencies and gold: Generalized orthogonal generalized autoregressive conditional heteroskedasticity and generalized autoregressive score analysis. International Journal of Emerging Markets , 17(9), 2290–2320. https://doi.org/10.1108/IJOEM-06-2020-0688

39.

Mittal

, Bhattacharya

, & Mandal

(2022). Characteristics analysis of behavioural portfolio theory in the Markowitz portfolio theory framework. Managerial Finance , 48(2), 277–288. https://doi.org/10.1108/MF-05-2021-0208

40.

Morina

, & Grima

(2022). The impact of pension fund assets on economic growth in transition countries, emerging economies, and developed countries. Quantitative Finance and Economics , 6(3), 459–504.

41.

Sanusi

K. A.

, & Eita

J. H.

(2022). Pension funds and stock market performance in South African economy: Asymmetric cointegration analysis. International Journal of Economics and Finance Studies , 14(1), 202–217. https://doi.org/10.34109/ijefs.202220008

42.

Sanusi

K. A.

, & Kapingura

F. M.

(2021). Pension funds as fuel for overall investment level and economic growth: An empirical insight from South African economy. Cogent Business & Management , 8(1), Article 1935661. https://doi.org/10.1080/23311975.2021.1935661

43.

SARB. (2023). Online statistical query (historical macroeconomic time series information) . https://www.resbank.co.za/Research/Statistics/Pages/OnlineDownloadFacility.aspx

44.

Shen

, Chen

, & Lin

(2020). The stock investment performance of pension funds in China. Emerging Markets Finance and Trade , 56(12), 2732–2748. https://doi.org/10.1080/1540496X.2018.1558053

45.

Staveley-O’Carroll

, & Staveley-O’Carroll

O. M.

(2017). Impact of pension system structure on international financial capital allocation. European Economic Review , 95, 1–22.

46.

Thomas

, Spataro

, & Mathew

(2014). Pension funds and stock market volatility: An empirical analysis of OECD countries. Journal of Financial Stability , 11, 92–103.

47.

van der Westhuizen

, van Eyden

, & Aye

G. C.

(2022). Contagion across financial markets during COVID-19: A look at volatility spillovers between the stock and foreign exchange markets in South Africa. Annals of Financial Economics , 17(1), 2250002.

48.

Xiao

, & Koenker

(2009). Conditional quantile estimation for generalized autoregressive conditional heteroscedasticity models. Journal of the American Statistical Association , 104(488), 1696–1712. https://doi.org/10.1198/jasa.2009.tm09170

49.

Xue

, He

, & Hu

(2021). The stabilizing effects of pension funds vs. mutual funds on country-specific market risk. Journal of Multinational Financial Management , 60, 100691.