Macroeconomic Imbalances and Financial Stress Among BRICS: Analysis of Frequency-Dependent and Asymmetric Causal Nexuses

Mohammed Armah; Ebenezer Bugri Anarfo; Emmanuel Numapau Gyamfi; Godfred Amewu

doi:10.1515/snde-2024-0045

Article

Macroeconomic Imbalances and Financial Stress Among BRICS: Analysis of Frequency-Dependent and Asymmetric Causal Nexuses

Mohammed Armah , Ebenezer Bugri Anarfo , Emmanuel Numapau Gyamfi and Godfred Amewu

Published/Copyright: April 16, 2025

Published by

Become an author with De Gruyter Brill

Submit Manuscript Author Information

From the journal Studies in Nonlinear Dynamics & Econometrics

Abstract

This study investigates the intricate and evolving causal relationship between macroeconomic imbalances and financial stress in BRICS nations. To analyse these dynamics across different time scales, thus enabling timely and policy intervention, we employ an asymmetric, noise-reducing-domain ICEEMDAN-based non-parametric model supplemented by a time-varying vector autoregressive model spanning from 1998 to 2021. Our findings reveal that financial stress and macroeconomic imbalance in BRICS exhibit both frequency-dependent and asymmetric linkages, shedding light on the mechanisms underlying financial contagion and market instability in BRICS nations. Specifically deteriorating financial conditions are linked to heightened macroeconomic instability, whereas periods of low stress correspond to improved macroeconomic condition. The results underscore the need to strengthen the macroeconomic frameworks with inflation serving as a critical nominal anchor and focal point of overarching policy guidelines. Proactive monitoring and responses to accumulation of financial imbalances are essential. The study recommends that policymakers should implement fiscal and monetary measures that prioritize sustainable economic growth while reducing reliance on debt and maintaining current account balance to mitigate financial vulnerabilities.

Keywords: financial stress; macroeconomic imbalance; ICEEMDAN; TVP-VAR; quantile-on-quantile; BRICS

JEL Classification: E44; F4; F32; F62

Corresponding author: Mohammed Armah, Department of Accounting and Finance, School of Business, Ghana Institute of Management and Public Administration (GIMPA), Achimota, Ghana, E-mail: mohammed.armah@st.gimpa.edu.gh

Acknowledgements

The authors would like to thank the editor and an anonymous referee for their valuable comments which led considerable improvement of the paper.

Author contributions: The authors has accepted responsibility for the entire content of this submitted manuscript and approved submission.
Conflict of interest: The authors declare no conflict of interest regarding this article.
Research funding: None declared.

Appendix A: Cross-Quantilograms

This appendix summarizes the nonparametric quantile-based method introduced by Han et al. (2016). As previously noted, this method is suitable for detecting extreme dependency, using arbitrary quantile, lags in detecting extreme dependency between variables (Abakah et al. 2023). CQ method takes into account heavy tail features of financial time series at various time lags and will enable us quantifies the strength of dependence between FS and MIs into short term medium and long term (Mensi et al. 2023). Assume u t , y t ; t ∈ z , where z is integer which is a static time series with u t = u 1 t u 2 t ∈ R d 1 x R d 2 and v t = y 1 t y 2 t T ∈ R 2 where R is the real number. u j t = u j t 1 ⋅ , x j t . , d j T ∈ R d j and d _j ∈ N for j = 1,2 ⋅ , F y j | u j ⋅ | u j represent the conditional distribution function of the time series v _1t given u _1t with F v j | u j ⋅ | u j as a density function. The conditional distribution can be written as q j , t τ j = j n f w : F y j | u j w | u j ≥ τ j for j = 1,2 ⋅ , τ j ∈ 0,1 . δ represent the quantile range assessing directional predictability. τ is the cartesian product of two consecutively closed interval in (0,1), i.e. τ = τ ₁ xτ ₂ where δ j = τ j ′ , τ j ′ ′ for some 0 < τ j ′ < τ j ′ ′ < 1 . To determine whether or not FS and MI are dependent: y 1 t ≤ q 1 t τ j and y 1 t − z ≤ q 2 t − z τ 2 . For 1,2 {1( y j t ≤ q 1 t ⋅ )} is refer as quantile hit. Following Han et al. (2016) we defined cross quantilogram as follows:

(A.1) ρ τ k = S ψ τ 1 y 1 , t − q 1 , t τ 1 ψ τ 2 y 2 , t − k − q 2 , t − k τ 2 S ψ τ 1 ( y 1 , t − q 1 , t ( τ 1 ) ) S ψ τ 2 y 2 , t − q 2 , t − k τ 2

For k = 0, ±1, ±2, …, where ψ a u ≡ 1 u < 0 − y 1 , t ≤ q 1 , t τ 1 represents the quantile hit process where 1(.) represent an indicator function. From equation (A.1) the quantile hit process is determine under at the time t − p and ρ _t−k is the correlation of the quantile-hit process. The unconditional correlation cross-quantile is define as follows;

(A.2) ρ ̂ τ k = ∑ t = k + 1 T ψ τ 1 y 1 , t − q ̂ 1 , t τ 1 ψ τ 2 y 2 , t − k − q ̂ 2 , t − k τ 2 ∑ t = k + 1 T ψ τ 1 2 y 1 , t − q ̂ 1 , t τ 1 ∑ t = k + 1 T ψ τ 2 2 y 2 , t − q 2 , t − k τ 2

where q ̂ ( a ) is the estimate for the a-quantile which is drived from quantile regression on a set of covariates. To perform any inference of ρ ̂ τ k , the asymptotic distribution T ρ ̂ τ k − ρ τ k cannot be derived analytically therefore stationary bootstrap is required since nuisance parameters are introduced. Following Han et al. (2016) bootstraped CQ is defined as follows:

(A.3) ρ ̂ τ * k = ∑ t = k + 1 T ψ τ 1 y 1 , t * q ̂ 1 , t * τ 1 ψ τ 2 y 2 , t − k * − q ̂ 2 , t * τ 2 ∑ t = k + 1 T ψ τ 1 2 y 1 , t * q ̂ 1 , t * τ 1 ∑ t = k + 1 T ψ τ 2 2 y 2 , t − k * − q ̂ 2 , t * τ 2

where 1 − α is the confidence interval for ρ ̂ τ k which is define as ρ ̂ τ k + c 1 , α * T ; ρ ̂ τ k + c 2 , α * T c 1 , α * ; c 2 , α * are critical value derived from α/2 and 1 − α/2 quantiles of T ρ ̂ τ * k − ρ ̂ τ k .

Appendix B

See Tables 1–3.

Table 1:

Summary statistics of FSI and macroeconomic imbalance.

South Africa	FSI	OGAP	CA	CGD
Mean	−0.007	0	−2.140***	1.589***
Variance	1.068***	0.000***	6.214***	0.017***
Skewness	1.190***	−5.218***	0.737***	0.384
JB	24.170***	6,776.081***	9.416***	4.407
ERS	−0.938	−2.682***	−1.493	−0.674
Q(20)	159.057***	19.791**	217.832***	648.362***
Q2(20)	47.297***	0.254	181.251***	639.913***
Correlation
Kendall	FSI	OGAP	CA	CGD
FSI	1.000***	0.152**	0.234***	0.071
OGAP	0.152**	1.000***	−0.268***	0.114
CA	0.234***	−0.268***	1.000***	0.152**
CGD	0.071	0.114	0.152**	1.000***
Brazil	FSI	OGAP	CA	CGD
Mean	−0.008	0	−2.287***	1.831***
Variance	1.067***	0.000***	4.140***	0.003***
Skewness	−0.562**	−1.762***	0.498**	0.741***
Ex.Kurtosis	1.495**	6.921***	−1.022***	−0.301
JB	13.985***	241.260***	8.141**	9.143***
ERS	−2.556**	−3.666***	−1.863*	0.581
Q(20)	213.349***	61.092***	286.332***	432.473***
Q2(20)	87.252***	2.12	140.764***	432.756***
Correlation
Kendall	FSI	OGAP	CA	CGD
FSI	1.000***	0.075	0.071	−0.246***
OGAP	0.075	1.000***	−0.290***	−0.282***
CA	0.071	−0.290***	1.000***	0.129
CGD	−0.246***	−0.282***	0.129	1.000***
India	FSI	OGAP	CA	CGD
Mean	0.004	0.019	−1.214***	1.697***
Variance	1.092***	13.285***	3.584***	0.002***
Skewness	0.128	−0.900***	0.116	0.348
Ex.Kurtosis	−0.635	29.379***	0.532	−1.374***
JB	1.874	3,465.475***	1.346	9.485***
ERS	−0.546	−5.828***	−1.893*	−1.143
Q(20)	187.235***	22.896***	164.374***	554.308***
Q2(20)	64.501***	25.469***	112.402***	554.797***
Correlation
Kendall	FSI	OGAP	CA	CGD
FSI	1.000***	0.045	−0.254***	−0.104
OGAP	0.045	1.000***	0.011	−0.051
CA	−0.254***	0.011	1.000***	0.329***
CGD	−0.104	−0.051	0.329***	1.000***
China	FSI	OGAP	CA	CGD
Mean	−0.429***	0.058	3.200***	1.450***
Variance	1.572***	4.908***	7.485***	0.007***
Skewness	−0.253	0.379	1.303***	−2.568***
Ex.Kurtosis	0.098	22.576***	0.942*	6.819***
JB	1.059	2040.994***	30.725***	291.496***
ERS	−1.494	−4.996***	−1.426	−0.002
Q(20)	235.704***	32.904***	372.818***	261.918***
Q2(20)	74.252***	20.769**	347.974***	278.706***
Correlation
Kendall	FSI	OGAP	CA	CGD
FSI	1.000***	−0.098	−0.124	−0.001
OGAP	−0.098	1.000***	−0.036	−0.113
CA	−0.124	−0.036	1.000***	0.099
CGD	−0.001	−0.113	0.099	1.000***
Russia	FSI	OGAP	CA	CGD
Mean	−0.008	0	5.949***	1.202***
Variance	1.624***	0.001***	9.938***	0.090***
Skewness	1.303***	−0.094	0.044	0.973***
Ex.Kurtosis	2.377***	−0.844***	−1.404***	0.078
JB	49.777***	2.989	7.920**	15.179***
ERS	−0.65	−3.266***	−1.467	−0.85
Q(20)	150.728***	195.655***	393.065***	542.648***
Q2(20)	62.435***	54.375***	397.626***	515.746***
Correlation
Kendall	FSI	OGAP	CA	CGD
FSI	1.000***	0	−0.022	0.363***
OGAP	0	1.000***	−0.06	−0.048
CA	−0.022	−0.06	1.000***	0.149**
CGD	0.363***	−0.048	0.149**	1.000***

Table 2:

Principal components analysis: proportion of eigenvalue components.

Eigenvalue components	PC1	PC2	PC3	PC4	PC5	PC1-PC3
South Africa	0.2862	0.202	0.1965	0.1812	0.1342	0.6847
China	0.322	0.2574	0.1862	0.1296	0.1048	0.7656
Brazil	0.2475	0.2149	0.2059	0.1809	0.1508	0.6683
Russia	0.3594	0.2877	0.1528	0.1191	0.0809	0.7999
India	0.2800	0.2527	0.2147	0.1402	0.1123	0.7474

Source: Authors computation.

Table 3:

The number of IMFs and the selected IMFs from the ICEEMDAN algorithm.

Date series	Time scale	Countries	Selected IMF
Signals
IMF.1	Short term	Brazil	1
IMF.2	Short term	Russia
IMF.3	Medium term	India	3
IMF.4	Medium term	China
Residual	Long term	South Africa	Residual

Appendix C

See Figures 4a, 4b(i)-4b(xv).

Figure 4a:

Identification of financial stress event of BRICS market.