Electricity Prices and Residential Electricity Consumption in South Africa: Evidence from Fully Modified Ordinary Least Squares and Dynamic Ordinary Least Squares Tests

Abstract

The sharp rise in electricity prices in South Africa has raised a growing concern over household electricity use, affordability, and the need for sustainable consumption patterns. This increasing cost of electricity has added financial pressure on South Africans already burdened by rising prices of water, food, and fuel. This study aims to determine the relationship between residential electricity consumption and electricity prices in South Africa, using annual time series secondary data spanning from 1975 to 2024. To determine the long-run relationship the study employed econometric techniques such as Fully Modified Ordinary Least Squares (FMOLS) and Dynamic Ordinary Least Squares (DOLS), then, for robustness, the Vector Error Correction Model (VECM) and diagnostics checks. The findings of the study revealed a negative relationship between electricity prices and residential electricity consumption. While disposable income showed a positive relationship with residential electricity consumption, the population growth revealed a negative relationship with residential electricity consumption. Based on the empirical findings of the study, South African policymakers should ensure the affordability of electricity and user-efficiency so that population growth does not worsen energy inequality. Hence, policymakers should ensure basic access for all households by supporting low-income groups and applying higher tariffs for higher consumption. These measures promote fairness, meet essential electricity needs, and encourage responsible use.

1. Introduction

Residential electricity consumption in South Africa has grown steadily over the years, reflecting changes in economic development and efforts to expand access to electricity, but rising electricity prices have become a growing concern for many households [1]. Following the end of apartheid, the government launched an electrification programme in 1995, electrifying around 2 million households by 2000, which is a total of 72.3% of the total population [2]. Due to the programme, electricity consumption increased, as the country believed that the more people are connected to the grid, the more it signals economic development. In 2010, around 82.8% of the population had access to electricity, and by 2020, it was approximately 90%; however, in 2023, there was a record of 87.7%, a slight decline of access of electricity as compared to 2020 [3].

In 1995, Eskom, the state-owned enterprise in South Africa responsible for generating, transmitting, and distributing over 90% of the country’s electricity, sold 3906 GWh of electricity to the residential sector, which rose to 6308 GWh in 2000, mainly due to the electrification programme and low tariffs [4,5]. By 2006, it stood at 8904 GWh compared to 10,423 GWh in 2008 [6]. However, in 2010, electricity sales dropped to 10,350 GWh from 10,392 GWh in 2009, a decline caused by rising electricity prices and supply constraints that pushed customers to seek alternatives [2,7]. By 2022, electricity consumption stood at 10,520 GWh, a decrease from 10,949 GWh in 2021 [8].

By 2023, residential electricity consumption was at 9177 GWh; meanwhile, in 2024, the power utility experienced a further decline of consumption, as it recorded 8559 GWh, indicating a deterioration in electricity usage by households compared to the pre-load shedding period [9]. For several years, as presented in Figure 1, Eskom customers consumed electricity without worry because of its affordability. The low prices made it easy and affordable for people to use as much electricity as they needed.

As presented in Figure 1, in 1995, the average cost of electricity was R11.15 c/kWh, compared to an average of R13.23 c/kWh in 2000; the price paid was still the lowest compared to the rest of the world and were kept at the lowest levels to protect the country’s poor population [10]. However, as the years progressed, electricity prices started to increase, causing a decline in the consumption of electricity. The South African residential sector has been hit hard by the steady rise in electricity prices, with an average rise from R16.04 c/kWh in 2004/05 to R62.81 c/kWh in 2014/15 and R160.42 c/kWh in 2024/25 [11]. According to [12], South African residential consumers now pay more for electricity than most households in Africa, Southeast Asia, and BRIC countries.

This rising cost of electricity has added pressure on South Africans already struggling with increases in water, food, and fuel prices. Urban residents rely heavily on electricity for daily needs such as cooking, lighting, and heating, while many rural households rely on alternative energy sources for cooking [13,14]. The study by [15] highlighted that electricity price hikes reduce household purchasing power, forcing consumers to cut back on other goods or reduce savings. The financial burden has also affected Eskom’s customer base, leading to affordability concerns and frustration [16,17].

Higher electricity prices not only limit access for low-income households [18], but also have broader economic implications, especially in a system where energy demand is inelastic in the short term. While in theory price increases could be offset through substitution [19], in practice, reduced income often results in lower electricity consumption, further highlighting the socioeconomic divide in energy access.

Despite numerous studies on residential electricity consumption, most focus on general socioeconomic determinants or use obsolete data. Few studies [2,10,20,21] in South Africa have specifically examined the effect of electricity prices while controlling for household income. These studies have not fully considered the period where electricity consumption was affected by events such COVID-19, severe load shedding periods, and price hikes that have changed how South African households use electricity, especially those facing energy poverty. This study therefore seeks to address this gap by answering the following research questions: what is the relationship between residential electricity consumption and electricity prices in South Africa? How does household income affect the sensitivity of residential electricity consumption to changes in electricity prices? What is the role of population growth in shaping residential electricity consumption patterns?

To answer these questions, updated research using recent data is essential to better understand these changing consumption patterns. Hence, this study is significant, as it provides insight into how rising electricity prices influence household electricity consumption, particularly in South Africa, where electricity affordability remains a major challenge for many communities. This insight is crucial for designing fair electricity tariffs and policies that balance the need for a reliable power supply while protecting low-income consumers from financial hardship.

Therefore, the overall aim of this study is to examine the relationship between residential electricity consumption and electricity prices in South Africa. The outline of this study is as follows: Section 2 presents the literature review. Section 3 describes the method implemented. Section 4 discusses the empirical results. Section 5 provides the limitations of the study, while Section 6 provides the conclusion and policy implications of this study.

2. Literature Review

This study is anchored on Alfred Marshall’s Law of Demand, which states that, ceteris paribus, an increase in the price of a good leads to a decrease in quantity demanded, while a decrease in price leads to an increase in quantity demanded [22]. In the case of electricity consumption, this model implies that households respond to changes in electricity prices by altering their electricity consumption. Examining residential electricity consumption through this lens allows the study to analyze the extent to which electricity prices influence residential electricity consumption in South Africa.

Residential electricity consumption is influenced by a complex interplay of socioeconomic, demographic, and structural factors. Global evidence highlights household characteristics, income, and electricity prices as critical determinants of electricity consumption. Urbanization is increasingly recognized as an additional factor that affects both the level and composition of residential electricity consumption, particularly in rapidly developing countries. Understanding these dynamics using recent data that capture structural electricity changes and socioeconomic shocks is essential for contemporary energy policy.

2.1. International Perspective

Studies across different countries revealed consistent patterns in terms of factors shaping residential electricity consumption. Refs. [23,24,25,26,27,28] found that differences in lifestyle, household size, and higher employment levels contribute significantly to electricity consumption, as households with more space and income are likely to install and use more electrical appliances in countries such as Vietnam, Sweden, Dutch, Nigeria, Taiwan, and Hawaii. Collectively, these studies highlighted that, while income and appliance ownership are major drivers, broader socioeconomic and policy contexts also shape consumption patterns.

Contrasting these findings, Ref. [29] argued that in European Union countries, larger living spaces do not generally lead to higher consumption. The study revealed that electricity pricing structures, especially tariffs and taxes, play a stronger role in determining household expenditure, resulting in higher consumption, which may often lead to a lower cost per kWh, implying that affordability and policy frameworks determine consumption more than space alone. In support, Refs. [30,31] outlined that, in China, knowledge about electricity pricing policies influence households to adjust their consumption more effectively.

In contrast, middle- and low-income groups showed limited behavioural changes. With similar conclusions drawn, Ref. [32] observed that in Spain, medium- and high-income households respond more to price increases, while low-income groups are more sensitive to income changes than to price fluctuations. Ref. [33] highlighted that a Time-of-Use tariff policy in Korea helps households use less electricity during peak hours so that they can save on electricity costs. Likewise, Ref. [34] suggested that maintaining high electricity prices is an effective strategy to manage demand where growth in household size and income levels also influence consumption trends, mainly in Brazil.

Studies conducted in India, Japan, Tunisia, and China emphasized the role of behaviour by revealing that, while electricity usage is often habitual, Refs. [35,36,37,38] concluded that dramatic price increases can alter consumption patterns and that habits are more influential than prices in shaping consumption, thus limiting the effect of price hikes. Ref. [39] argued that price increases had minimal impact on Ethiopia’s urban household electricity use, indicating low price elasticity; however, the study’s urban focus may miss variations in rural areas.

From a welfare perspective, Ref. [40] revealed that, in Zambia, price increases raise general living costs, disproportionately affecting low-income households. Supporting the findings, Ref. [41] found that rural and low-income households in India are the most affected by rising electricity prices.

Population growth was shown to also play a supplementary yet meaningful role in shaping residential electricity consumption. Studies focusing on Malaysia and OECD countries showed that population growth drives energy consumption. Ref. [42] found that the rising population and GDP in Malaysia increased electricity consumption, while Ref. [43] shows that the total population and urbanization in OECD countries affect both non-renewable and renewable energy consumption. These studies highlight the central role of population in electricity consumption. Despite its growing recognition, many studies do not fully explore how population growth interacts with household electricity consumption, leaving a partial understanding of its effects. The mixed findings from the international perspective makes it unclear whether there is a positive or negative relationship between residential electricity consumption and electricity prices, highlighting the need for updated research, especially for countries undergoing structural electricity changes and broader socioeconomic shifts.

2.2. South African Perspective

South African studies consistently emphasized income, electricity prices, and household characteristics as key determinants of electricity consumption. An earlier study by [44] predicted that demand would fall by over 25% by 2030 due to tariff restructuring, identifying disposable income and electricity prices as the primary long-term determinants of demand. Ref. [2] concluded that, in South Africa, rising economic growth raises household income and thus electricity demand, particularly in urban areas where households tend to have more rooms and appliances. Electricity prices, however, were negatively associated with household consumption. Ref. [20] identified income as a key driver, noting that price reforms need to be sensitive to socioeconomic conditions. Ref. [21] highlighted that low-income households are particularly sensitive to electricity price increases, and Ref. [45] showed that middle-income households are most vulnerable due to their limited capacity to invest in energy-saving technologies.

Overall, the literature demonstrates that income, electricity prices, and household characteristics are critical determinants of residential electricity consumption, with population growth emerging as an additional factor. However, there is a lack of recent South Africa-specific studies that examine these interactions, using data reflecting electricity price restructuring and pandemic-related effects. Many studies, particularly those focusing on urban populations, overlooked variations in consumption arising from differences in income, affordability, and everyday usage patterns. The results from the literature are also inconsistent and mixed, underscoring the need for an updated investigation. This study addresses these gaps by examining the interactions between electricity prices, income, and population growth, offering new insights into contemporary residential electricity consumption patterns in South Africa under evolving socioeconomic and structural conditions.

3. Methodology

3.1. Model Specification

This study aims to examine the relationship between residential electricity consumption and electricity prices in South Africa. The relationship was investigated in South Africa, as seen in the works of [2,20,21,45]. Therefore, based on the previous literature review within this study, variables such as residential electricity consumption, average electricity prices, per capita disposable income, and population growth are included, and the study period is adjusted. This modification allows the model to capture the current economic challenges faced by South African households, such as fluctuating electricity prices and income disparities, providing a more contemporary representation of consumption patterns in South Africa.

The econometric model of the study is given as follows:

$${LREC}_t=a_1+a_{LAEP}{LAEP}_t+a_{LPCDI}{LPCDI}_t+a_{POPG}{POPG}_t+{\mathsf{\varepsilon }}_t$$

(1)

where $a_1$ is an intercept, LREC_t represents the natural logarithm of residential electricity consumption, LAEP_t represents the natural logarithm of average electricity prices, LPCDI_t represents the natural logarithm of per capita disposable income, POPG_t represents population growth, ${\mathsf{\varepsilon }}_t$ represents the error term, and _t represents the period.

Conceptual Framework

To guide the analysis of the relationship between electricity prices and residential electricity consumption, a conceptual framework is developed in Figure 1. The framework visually represents the hypothesized relationships among the variables considered in the study, with arrows indicating the direction of influence toward residential electricity consumption. Electricity prices are treated as the primary explanatory variable, while per capita disposable income and population growth are included as control variables.

This framework presented in Figure 2 provides a visual guide for the study design and the relationships among the variables. Building on this framework, the next step involves a systematic data processing procedure to prepare the variables for econometric analysis.

3.2. Data Processing Procedure

3.2.1. Data Collection and Variable Definition

This study follows a quantitative approach utilizing secondary annual time-series data covering the period 1975 to 2024, collected from reputable sources, as shown in

Table 1. Variable description and source.

Variable	Description	Measurement	Source
REC	Residential electricity consumption	Gigawatt hour	Eskom
AEP	Average electricity prices	Nominal value	Eskom
PCDI	Disposable income per capita	Nominal value	SARB
POPG	Population growth	Annual percentage	The World Bank

Source: Authors’ own computation.

. The study offers variable definitions and links to related studies that used the same variables. For computational analysis, the study make use of EViews 13.

Residential electricity consumption refers to the amount of electricity households use for various purposes, including lighting, heating, cooling, appliances, electronics, and other household activities [21].

Average electricity prices are the costs associated with the consumption of electrical power, typically measured in monetary units, such as cents per kilowatt-hour [20].

Per capita disposable income is the average amount of money that each person in a population has available to spend or save after paying taxes [38].

Population growth refers to the increase in the number of people in a specific area over a period of time [43,46].

3.2.2. Data Cleaning and Preparation

The datasets were checked for missing values, unit inconsistencies, and anomalies. Where necessary, adjustments were made to ensure consistency across years and variables.

3.2.3. Data Transformation

Key variables such as residential electricity consumption, average electricity prices, and per capita disposable income were converted into logarithmic form to stabilize the variance and facilitate elasticity interpretation.

3.2.4. Data Integration

The different datasets were merged and aligned by year to produce a uniform annual time series covering the study period.

3.3. Data Analysis

The study employs a multi-step econometric approach to investigate the relationship between residential electricity consumption and electricity prices. The econometric approach is summarized in Figure 3.

3.3.1. Descriptive Statistics

The descriptive statistics analysis provides essential information about the variables of the study. It briefly summarizes the descriptive coefficients with central tendency measures, which include the mean, median, minimum, maximum, and standard deviation of the variables [47]. The test will provide an overview of the trends and variability in residential electricity consumption, disposable income, electricity prices, and population growth.

3.3.2. Stationarity

After analyzing the descriptive statistics of the variables, the study will employ the augmented Dickey–Fuller unit root by [48] and the Phillips–Perron unit root test by [49] to check for stationarity and determine the order of integration of the variables, addressing potential spurious regression. Additionally, the Zivot–Andrews (ZA) unit root test introduced by [50] is employed, as it incorporates a structural break on the time series. A structural break may occur in the time series data because of changes in large economic shocks, economic policies, or even institutional and legislative changes [51].

3.3.3. Lag Length Selection

Once stationarity has been achieved, the VAR optimal lag length criterion will be applied to select the appropriate number of lags. Ref. [52] outlines the criteria for selecting lag length in econometric models, including AIC, SIC, HQ, FPE, and sequential modified LR test statistics. For this study, the AIC is chosen, owing to fewer than 60 observations. The next section discusses the Johansen approach, which focuses on testing for cointegration.

3.3.4. Johansen Cointegration

Thereafter, to examine the long-run relationships among the variables, the Johansen [53] cointegration test will be employed. This method determines whether two or more non-stationary time series share a long-run equilibrium relationship. It allows for testing and estimating multiple cointegrating vectors within a Vector Autoregression (VAR) framework. It identifies the number of cointegrating relationships among the variables using the Trace statistic test for the overall number of cointegrating relationships and the Maximum Eigenvalue statistic test, which tests each cointegrating relationship individually [53].

3.3.5. Fully Modified Ordinary Least Squares and Dynamic Ordinary Least Squares

The study then employs the Fully Modified Ordinary Least Squares (FMOLS) to estimate the long-run relationship between variables while addressing endogeneity and serial correlation that often bias the ordinary least squares (OLS) estimates in time series data [54]. This method provides robust and efficient estimates, making it preferable for policy-relevant long-term analysis. DOLS complement the FMOLS by including the leads and lags of differenced regressors, addressing small-sample bias, and confirming the robustness of the long-run relationship [55]. Together, these techniques ensure a comprehensive and accurate analysis in determining the long-run relationship between residential electricity consumption, electricity prices, disposable income, and population growth.

3.3.6. Vector Error Correction Model

The study also applies the Vector Error Correction Model (VECM), as it simultaneously captures short-run dynamics and long-run equilibrium relationships among cointegrated variables [56]. It is particularly suitable for this study, as it allows deviations from the long-term relationship between residential electricity consumption and electricity prices to be corrected over time, providing a more accurate and policy-relevant understanding.

A VECM of the study is represented as follows:

$$∆LREC={\beta }_0+{\beta }_1{\sum }_{i=1}^{n-1}∆{LAEP}_{t-i}+{\beta }_2{\sum }_{i=1}^{n-1}∆{LPCDI}_{t=i}+{\beta }_3{\sum }_{i-1}^{n-1}∆{POPG}_{t=i}+\phi {ECT}_{t=1}+{\mu }_t$$

(2)

where Δ is the first difference, and the joint consequences of the lags are ${\beta }_1$, ${\beta }_2$, and ${\beta }_3$, while the error correction term is specified by ${ECT}_{t=1}$, and φ is the adjustment speed of the model towards equilibrium. The error term is represented by $\mu$ and _t represents the time period. The error term must be negative and statistically significant to achieve equilibrium. The study then performs the diagnostic and stability tests, as explained in the section below.

3.3.7. Diagnostic Tests

This section outlines the tests that are used to assess the reliability, stability, and validity of the results. To achieve these objectives, this study focuses on the serial correlation Lagrange multiplier test, heteroskedasticity test, normality test, and AR autoregressive root graph to check the stability of the model.

Serial Correlation Langrage Multiplier Test

In econometrics, serial correlation occurs when error terms are correlated across time periods [57]. The study employs the Breusch–Godfrey serial correlation LM test to assess whether residuals are independent [58]. The null hypothesis posits no serial correlation, whereas the alternative hypothesis suggests that there is the existence of a serial correlation. If the calculated LM serial correlation exceeded the 5% significance level, the null hypothesis is not rejected; if it is lower, the null hypothesis is rejected.

Heteroskedasticity Test

In classical linear regression, constant variance is crucial for model reliability. Testing for heteroskedasticity ensures that the error variance remains consistent [59]. To address this, the study uses the white heteroskedasticity test to detect and correct any issues. The null hypothesis indicates no heteroskedasticity, whereas the alternative posits the existence of heteroskedasticity. The null hypothesis is accepted if the probability value exceeds 5%, indicating that the model does not exhibit heteroskedasticity. Conversely, the null hypothesis is rejected if the probability value is <5%.

Normality Test

The central limit theorem states that a sufficiently large sample follows a normal distribution for joint sampling [40]. The study will use the Jarque–Bera test by [60], which is commonly used to assess normality and evaluates both kurtosis and skewness in the sample data. Kurtosis measures the peak probability of the variable, whereas skewness assesses the asymmetry of the probability distribution around the mean. The null hypothesis indicates that residuals are normally distributed, whereas the alternative hypothesis suggests that residuals are not normally distributed.

The Jarque–Bera test statistics are defined as below:

$$JB={\displaystyle \frac{N-K}{6}}[S^2+{\displaystyle \frac{(K-{3)}^2}{4}}]$$

(3)

where N is the sample size, K is the number of estimated parameters, S is the skewness of a variable, and K is the kurtosis of a variable. The null hypothesis is accepted if the probability value is greater than the 5% level of significance, concluding that the residuals are normally distributed. The null hypothesis is rejected if the probability value is less than the 5% level of significance.

4. Results and Analysis

The empirical results presented in this section are descriptive statistics, unit root tests, lag selection criterions, and Johansen cointegrations. Then, to achieve the study’s objectives and to answer the research questions, the Fully Modified Ordinary Least Squares (FMOLS) and Dynamic Ordinary Least Squares (DOLS) models are employed to determine the long-run relationship between residential electricity consumption and electricity prices. To ensure the robustness of the results, the study employed the Vector Error Correction Model (VECM) to capture short-run dynamics and long-run equilibrium relationships among the variables. The study also presents the diagnostics tests such as serial correlation, normality, heteroskedasticity, and stability checks.

4.1. Descriptive Statistics Results

Descriptive statistics describe the fundamental features of the data in a model.

Table 2. Descriptive statistics results.

	REC	AEP	PCDI	POPG
Mean	6111.620	31.64144	24,171.50	1.912520
Median	6266.500	12.83500	15,251.00	1.635870
Maximum	11,917.00	160.4200	75,977.00	3.493390
Minimum	906.0000	0.800000	769.0000	0.653837
Std. Dev.	4294.140	39.86520	23,260.78	0.890114
Skewness	−0.048906	1.575346	0.769867	0.469686
Kurtosis	1.291082	4.571639	2.255399	1.718772
Jarque–Bera	6.104103	25.82690	6.094195	5.258263
Probability	0.047262	0.000002	0.047497	0.072141
Observations	50	50	50	50

Source: Authors’ own computation.

represents the descriptive statistics of the study.

The descriptive statistics presented in Table 2 show a notable difference across variables. Residential electricity consumption (REC) averages 6111 GWh, close to the median of 6266 GWh, suggesting a fairly symmetric distribution. However, slight deviations from normality indicate fluctuations in usage patterns, which may reflect structural and policy changes in the electricity sector. The average electricity price (AEP) is highly skewed, with a mean (31.64 cents/kWh) more than double the median (12.83 cents/kWh) and extreme spikes (maximum = 160.42) that drive non-normality. These sharp increases highlight the volatility of South Africa’s electricity pricing, with potential affordability challenges for households. Per capita disposable income (PCDI) also shows right skewness, with a mean (24,171) above the median (15,251), reflecting income growth over time but also periods of inequality. Variability in disposable income is economically important, as it influences households’ abilities to absorb rising electricity costs. Population growth (POPG) is comparatively stable, averaging 1.91%, with no strong deviations from normality.

4.2. Unit Root Results

Table 3. Unit root of ADF, PP, and ZA tests at levels.

Variables	Formula	Levels
		ADF	PP	ZA		Conclusion
		t-Values	t-Values	t-Values	Breakpoint
LREC	Intercept	−0.608	−0.682	−2.935	1994	Non-Stationary
	Trend and Intercept	−1.456	−1.686	−3.030	2007	Non-Stationary
	None/Both	1.238	1.084	−2.876	2010	Non-Stationary
LAEP	Intercept	−0.387	−1.462	−2.977	1996	Non-Stationary
	Trend and Intercept	−1.908	−2.840	−2.371	2007	Non-Stationary
	None/Both	2.350	3.228	−2.534	1996	Non-Stationary
LPCDI	Intercept	−6.434 **	−5.236 **	−1.096	1983	Non-Stationary
	Trend and Intercept	0.473	0.303	−3.420	1990	Non-Stationary
	None/Both	0.539	5.371	−3.221	1990	Non-Stationary
POPG	Intercept	−1.237	−1.202	−3.918	2004	Non-Stationary
	Trend and Intercept	−1.476	−1.460	−3.470	1994	Non-Stationary
	None/Both	−1.180	−1.174	−3.495	2004	Non-Stationary

Source: Authors’ own computation. Note: ** indicate significance at 5% level.

presents the results of ADF, PP, and ZA unit root tests at level form, while

Table 4. Unit root of ADF, PP, and ZA tests at first difference.

Variables	Formula	First Difference
		ADF	PP	ZA		Conclusion
		t-Values	t-Values	t-Values	Breakpoint
LREC	Intercept	−6.055 **	−6.096 **	−8.437 **	1983	Stationary
	Trend and Intercept	−3.527 **	−5.988 **	−8.032 **	1989	Stationary
	None/Both	−1.459	−5.986 **	−7.345 **	2004	Stationary
LAEP	Intercept	−4.173 **	−3.996 **	−6.332 **	2008	Stationary
	Trend and Intercept	−4.106 **	−3.916 **	−4.616 **	1999	Stationary
	None/Both	−2.876 **	−2.446 **	−6.350 **	2009	Stationary
LGDI	Intercept	−3.866 **	−2.930 **	−6.809 **	1983	Stationary
	Trend and Intercept	−5.683 **	−7.512 **	−6.201 **	1990	Stationary
	None/Both	−0.851	−1.031	−6.753 **	1990	Stationary
POPG	Intercept	−5.940 **	−6.459 **	−5.364 **	1985	Stationary
	Trend and Intercept	−5.896 **	−6.401 **	−6.346 **	1993	Stationary
	None/Both	−6.463 **	−6.465 **	−5.304 **	1997	Stationary

Source: Authors’ own computation. Note: ** indicate significance at 5% level.

provides the results of the unit root tests at first difference.

According to the ADF, PP, and ZA unit root test results presented in Table 3, residential electricity consumption, electricity prices, disposable income, and population growth are non-stationary for all formulas. However, if it happens that the tests produce different results, then trend and intercept results are recommended over intercept-only or none to avoid misspecification and ensure more reliable results [46,47]. Therefore, the study concludes that all variables are non-stationary at level form for ADF, PP, and ZA unit root tests, and, therefore, the series is differenced in Table 4.

According to Table 4, the series was first differenced using ADF, PP, and ZA unit root tests; the results indicated that residential electricity consumption, electricity prices, disposable income, and population growth are stationary for all formulas. The ZA unit root test results are preferred over the ADF and PP unit root tests, as the test also reports a single structural break, concluding that the null hypothesis of non-stationarity is rejected, as stationarity is obtained in order I (1).

4.3. Lag Length Selection Criterion Results

Table 5. Lag length selection criterion results.

Lag	LogL	LR	FPE	AIC	SC	HQ
0	−111.5652	NA	0.001450	4.815215	4.971148	4.874142
1	195.4132	550.0029 *	7.89 × 10−9 *	−7.308884 *	−6.529217 *	−7.014247 *
2	210.4579	24.44761	8.34 × 10−9	−7.269079	−5.865678	−6.738732

Source: Authors’ own computation. Note: * indicates the optimal lag order selected by the respective criterion

provides the results of the lag length selected by various criterions.

After achieving stationarity, the study performed a VAR optimal lag criterion to indicate the number of lags to be used in the model. According to Table 5, the LR, FPE, AIC, SC, and HQ criterion recommend lag 1 to be used by the study.

4.4. Johansen Cointegration Results

The study employed the Johansen cointegration to determine the existence of the long-run relationship among variables, as presented in

Table 6. Cointegration results.

Hypothesis No. of CE (s)	Trace Statistic	0.05 Critical Value	Max-Eigen Statistic	0.05 Critical Value
None *	54.83458	47.85613	28.93248	27.58434
At most 1	25.90210	29.79707	16.43728	21.13162
At most 2	9.464817	15.49471	8.026759	14.26460
At most 3	1.438058	3.841466	1.438058	3.841466

Source: Authors’ own computation. Note * indicates rejection of the null hypothesis at the 5% significance level

.

The results presented in Table 6 show that the Trace test has one cointegration equation at a 0.05 level of significance. The Max-eigen test showed similar results, as it also indicated one cointegration equation at a 0.05 level of significance. According to the results, the study concludes that there is the existence of a long-run relationship among the variables.

4.5. DOLS and FMOLS Results

To achieve the study’s objectives, the long-run results from the Dynamic Ordinary Least Squares and Fully Modified Ordinary Least Squares techniques are presented in

Table 7. Dynamic Ordinary Least Squares and Fully Modified Ordinary Least Squares results.

Variables	DOLS		FMOLS
	Coefficient	Probability	Coefficient	Probability
LAEP	−0.483	0.001	−0.386	0.001
LPCDI	0.784	0.001	0.741	0.001
POPG	−0.289	0.001	−0.362	0.001
	R2 = 0.983		R2 = 0.964
	Adj R2 = 0.977		Adj R2 = 0.962
	(lead = 1, lag = 1)

Source: Authors’ own computation. FMOLS: Fully Modified Ordinary Least Squares, DOLS: Dynamic Ordinary Least Squares.

.

The results of the Dynamic Ordinary Least Squares (DOLS) and Fully Modified Ordinary Least Squares (FMOLS) are presented in Table 7, and they reveal that there is a long-run negative and statistically significant relationship between electricity prices and residential electricity consumption. The DOLS shows that a one percentage increase in electricity prices leads to a decrease in residential electricity consumption by −0.48. FMOLS revealed that a one percentage increase in electricity prices decrease residential electricity consumption by −0.39. This implies that as electricity prices increase, households use less electricity. This suggests the need for balanced pricing policies that encourage efficiency without limiting access, especially for low-income households. The results support those of [30,32,41], as they postulated that an increase in electricity prices reduces the electricity consumption by households.

More so, the results of both DOLS and FMOLS reveal that there is a positive and statistically significant long-run relationship between disposable income and residential electricity consumption. The DOLS shows that a one percentage increase in disposable income increases electricity consumption by 0.78. While the FMOLS indicates that an increase in disposable income increases electricity consumption by 0.74. This implies that as household income increases, the more electricity is consumed. The results are in line with the studies by [27,61], as these studies concluded that income is the main driver of electricity consumption.

Lastly, the results of the DOLS and FMOLS reveal that there is a negative and statistically significant long-run relationship between population growth and residential electricity consumption. The DOLS shows that a one percentage increase in population growth leads to a −0.29 decrease in residential electricity consumption. Based on FMOLS, population growth was found to decrease residential electricity consumption by −0.36. The results are consistent with those of [62], as the study highlighted that population growth has a detrimental effect on electricity consumption. This implies that South African households may lack access to electricity, especially in rural areas, or simply cannot afford it. This point to growing energy poverty and infrastructure strain. Policymakers should focus on expanding electricity access, improving affordability, and investing in infrastructure to ensure that population growth does not worsen energy inequality. However, the results of the current study contradict those of [42,43], as they concluded that as the population grows, overall electricity consumption rises.

4.6. Vector Error Correction Model Results

To verify the long-run relationship in a multivariate setup and estimate short-run elasticities and the adjustment of speed, the study employed a Vector Error Correction Model. The results of the long run and short run are presented in

Table 8. Vector Error Correction Model long-run results.

CointEq1	Coefficient	Standard Error	t-Statistics
Constant	1.459	-	-
Variables
LREC	1.000	-	-
LAEP	0.919	0.320	2.869
LPCDI	−0.915	0.259	−3.101
POPG	0.105	0.161	0.651

Source: Authors’ own computation.

and

Table 9. Vector Error Correction Model short-run results.

Variables	Coefficient	Standard Error	t-Statistics
ECT	0.116	0.054	−2.120
D(LAEP)	0.001	0.030	0.024
D(LPCDI)	−0.056	0.012	−4.341
D(POPG)	−0.015	0.116	−0.126
R2	0.5210	-	-
Adj R2	0.3158	-	-
F-statistics	2.231	-	-

Source: Authors’ own computation.

, respectively.

The long-run results presented in Table 8 confirm a long-run relationship between electricity prices and residential electricity consumption. The results revealed that a one percent increase in electricity prices decreases residential electricity consumption by 0.92, as shown by the negative and significant relationship. The results verify that, in South Africa, while higher electricity prices can promote efficient consumption and support a more reliable electricity supply, they also increase costs for households, potentially affecting their overall household spending and electricity consumption, supporting the findings of [45].

A positive long-run relationship was found between disposable income and residential electricity consumption, postulating that a one percent increase in disposable income increases residential electricity consumption by 0.92. This positive relationship boosts revenue for Eskom and supports investment in generation and maintenance for a more reliable electricity supply. However, rising consumption can strain the grid, increase electricity costs, and widen disparities in access, highlighting the need for policies that ensure affordability and promote efficiency. The results support the conclusion made by [2,20], as they alluded that a high income increases electricity consumption. These results further confirm the ones found in Table 7 while estimating long-run coefficients using FMOLS and DOLS.

Lastly, the study found a negative long-run relationship between population growth and residential electricity consumption, highlighting that a one percent increase in population growth reduces electricity consumption by 0.11. This implies a lower per capita consumption of electricity in South Africa, which may ease pressure on the grid and reduce the need for immediate investment in a new capacity. However, it could also indicate limited access to electricity for some households, potentially constraining living standards and slowing economic growth, confirming the results found in Table 7 and further supporting the findings by [63]. Table 9 below represents the short-run coefficients of VECM.

The VECM short-run results are presented in Table 9, and they reveal that the coefficient of error correction term is negative −0.116 and statistically significant, with t-statistics −2.120. This implies that it will take 12% speed for the systems of residential electricity consumption to adjust to equilibrium in a single year. The R² value of 0.521 shows that the independent variables explain the 52% change in the dependent variable. In addition, the results indicate that in the short run, electricity prices and population growth do not have a significant relationship with residential electricity consumption, while disposable income has a positive and significant relationship with residential electricity consumption.

4.7. Diagnostic Tests Results

To ensure reliability of the results for policy recommendations, this section covers the diagnostic tests employed by the study. The tests include serial correlation, heteroskedasticity, normality, and stability, as shown in

Table 10. Diagnostic tests results.

Test	H0	Probability	Conclusion
Serial Correlation	No serial correlation	0.512	No serial correlation.
White (CH-sq)	No heteroskedasticity	0.172	No heteroskedasticity.
Jarque–Bera	Residuals are normally distributed	0.367	Residuals are normally distributed.

Source: Authors’ own computation. Reject if p < 0.05.

and Figure 4, respectively.

Serial Correlation: The model indicates that there is no serial correlation, as the probability value is 0.512, which is greater than the 5% significance level.

Heteroskedasticity: The model does not suffer from the heteroskedasticity problem, as the probability value of 0.172 is greater than 5%.

Normality: The residuals are normally distributed, as shown by the probability value of 0.367 of Jarque–Bera, which is greater than the 5% significance level.

The stability of the model is presented in Figure 4. The AR root test confirms stability if all the roots are inside the unit circle with a modulus of less than one. The results indicate that the model is stable, as all inverse roots of the AR are within the unit circle.

Overall, the findings of the study reveal a significant relationship between electricity prices, disposable income, population growth, and residential electricity consumption in South Africa. By using recent data, the study provides fresh insights into how these dynamics continue to shape residential electricity consumption, offering evidence that is both timely and relevant for economic analysis and policy design.

5. Limitations

The study relies on annual time series data, which may fail to capture important short-term fluctuations in the variables of interest. Additionally, constraints in the availability of secondary data limited the scope and number of variables that could be included in the analysis.

6. Conclusions

The study’s main aim focuses on examining the relationship between residential electricity consumption and electricity prices in South Africa, using annual time series secondary data from the year 1975 to 2024. The study performed ADF, PP, and ZA unit root tests to confirm stationarity. Then, the Johansen cointegration test was used for cointegration, and the study found that there is a long-run relationship among variables. With the employment of FMOLS and DOLS and the verification of the results with VECM, the findings of the study revealed a negative relationship between electricity prices and residential electricity consumption. While disposable income showed a positive relationship with residential electricity consumption. Population growth revealed a negative relationship with residential electricity consumption.

Based on the empirical findings, it is recommended that South African policymakers should make electricity more affordable while promoting efficient household use. Since population growth tends to reduce electricity use per household, policies should guarantee that every household has access to a basic electricity supply. Measures such as providing a basic electricity allowance for low-income households, applying higher tariffs for increased consumption, and offering targeted support to vulnerable groups can help achieve this balance. These strategies not only improve fairness and access but also ensure households can meet essential electricity needs, encourage responsible use, and allow resources to be directed toward those who need them the most. Future studies should continue to explore the relationship in-depth by using more frequent or household-level data for greater detail. Additional variables such as weather patterns or appliance usage should also be considered.