Causality Relationship Between Producer and Consumer Price Indexes of Selected Meat Commodities in South Africa from 1991 to 2023

Abstract

The South African meat industry plays an important role in food provision, income generation, and economic stability. However, inflation-driven volatility in commodity markets has intensified the focus on the nexus between producer and consumer price indexes, as significant meat price changes affect both producers and consumers. Therefore, this study explored the association and the causality between consumer price index (CPI) meat and producer price index (PPI) beef, mutton, pork, and chicken from 1991 to 2023 using the Vector Error Correction Model (VECM) and Granger causality test. The findings of the study revealed a short- and a long-run relationship among the variables with an adjustment speed (ECT) of 67.67%. No causal relationship was found between PPI beef and CPI meat, while unidirectional relationships were identified between CPI meat and PPI chicken, pork, and mutton. These results suggest that consumer meat prices impact production prices for chicken, pork, and mutton, but not for beef, indicating that beef price changes are influenced by factors other than general meat price fluctuations. Thus, while price controls or subsidies may manage price changes in chicken, pork, and mutton, further research or targeted strategies are needed to address the unique factors influencing beef prices. Ultimately, understanding these dynamics will enhance price stability in the meat market, benefiting consumers, producers, and the overall stability of the South African economy.

1. Introduction

Favourable food prices in combination with other factors are seen by economists as an enabler for individuals to substitute cheap sources of calories such as cereals with relatively more expensive and tastier sources such as meat and fruits (Bogmans et al. 2024). Meat, being an important component of the South African diet (Erasmus and Hoffman 2017), serves as a key source of nutrition and a contributor to income generation (Pereira and Vicente 2013; Kumar et al. 2021; Parlasca and Qaim 2022). Bogmans et al. (2024) posit that reigning food inflation is the second most important strategy to tackle food insecurity following the fostering of economic growth. South African households from economically disadvantaged backgrounds tend to choose white meat over red meat as the primary protein source, influenced by factors such as disposable income, meat prices, and changes in taste and affordability (Bisschoff and Liebenberg 2017; Manyi-Loh and Lues 2023). This preference aligns with the fact that chicken is the most consumed and produced meat in South Africa, followed by beef, pork, and mutton (BFAP 2013), forming a substantial part of the average household’s diet. It has been long concluded decades ago in studies such as Taljaard et al.’s (2004) that chicken meat is a necessity good, and that beef and mutton are luxury goods while pork is almost regarded as a luxury. In recent years, the conclusions from Delport et al. (2017) have not changed the classification of chicken meat and mutton while pork and beef became categorised as normal goods. Furthermore, Delport et al. (2017) highlighted that, due to structural changes in price formation and demand dynamics in South Africa, these classifications may shift over time and may not be directly comparable with the findings of other authors. As a result, these four types of meat hold significant economic importance across various socio-economic groups, both in terms of production and consumption, compared to other types of meat (DAFF 2019, 2021a, 2021b). These meat types have always played a significant role in boosting the gross value of animal products through increased export revenues for South Africa (DAFF 2006). This makes chicken, beef, pork, and mutton more relevant for a comprehensive analysis of price dynamics than the less commonly consumed products such as game meat, rabbit meat, and ostrich within niche markets.

The four meat types selected in the present study have experienced price surges between 2011 and 2022 (Hyland 2015; DAFF 2022; NAMC 2019). Additionally, Sikuka (2022), in a report of USDA and GAIN, further indicates that food price inflation in South Africa increased from 3.4% in 2019 to 5.4% in 2021. These fluctuations signify the need to further understand the price changes and relationships between the selected meat types to assist in providing critical information at structural level into broader economic trends, such as inflation, consumer spending, and input production prices. Changes in meat prices triggers the food supply chains, affecting both the downstream market, including the consumer’s basket of goods, and the upstream markets at the producer level (Hobbs 2021). These changes, in turn, influence household food security, spending habits, and overall diet quality, while also affecting producers by impacting profitability, production decisions, and market stability (Charlebois et al. 2016; Vellinga et al. 2022; Kopytets et al. 2020). Moreover, if the meat price changes are unmonitored, this influence can lead to many households accumulating debt, food insecurity, and adding to the already existing highly persistent poverty rate (Shah 2022; Van Wyk and Dlamini 2018; Mkhawani et al. 2016; Faber and Drimie 2016). Thus, the monitoring of price changes provides a foundation and helps in informing targeted and necessary interventions to address food insecurity and poverty. It is therefore critical to comprehend if changes in the price of meat originate from producers and then flow to consumers, or vice versa, especially in the South African market. Knowing the direction and association of these changes is important since it can reveal how price fluctuations impact different points in the supply chain, providing insights into potential disruptions in the economy, informing policy and business decisions (International Monetary Fund 2004; Li et al. 2019).

Key measures for tracking price changes, such as the consumer price index (CPI) and producer price index (PPI), can assist in monitoring price fluctuations in consumer goods and services, as well as in producer input costs (Özpolat 2020). Previous research, such as (Akcay 2011; Topuz et al. 2018; Woo et al. 2019; Su et al. 2016; Ghazali et al. 2008; Oyeleke and Ojediran 2018; Galodikwe 2014; Tiwari et al. 2014; Ulke and Ergun 2014; Berat and Keskin 2021; Anggraeni and Irawan 2018) reviewed in detail in Section 2 of this paper, focused on the overall causality and relationship between the latter indexes. To the best of the authors’ knowledge, no studies have been identified that focus on the present study’s selected meat types when analysing the nexus between CPI and PPI as this research aims to. This paper builds upon the existing literature by examining sectoral inflation within the meat industry, focusing on both the aggregated CPI meat, representing the entire meat sector and the disaggregated PPI for specific components, namely, beef, chicken, pork, and mutton. This approach enables an analysis of how price changes in the broader meat sector relate to fluctuations in its individual sub-sectors. If such a relationship exists, it could indicate that price changes either originate from the broader meat sector and filter down to sub-sectors, or vice versa. Additionally, by analysing causality, the direction of these price changes can be identified in order to determine whether consumer price movements in the meat sector drive producer price changes, or if producer price fluctuations influence consumer prices.

From a policy standpoint, it is important for central banks to understand the persistence and source of global shocks, particularly those impacting food prices, as noted in Tiedemann et al. (2024). This is especially important in countries such as South Africa, that seeks to maintain inflation within a target range of 3–6% through its monetary policy instruments (SARB 2013). However, maintaining this target can be challenging due to the ongoing increase in food prices, which are among the primary drivers of inflationary pressures in South Africa (Tshiakambila and Chisasa 2017; Louw et al. 2018; Rangasamy 2011). Nevertheless, while the South African Reserve Bank (SARB) plays a significant role in addressing inflationary pressures driven by rising food prices, it lacks direct mechanisms to control specific food markets, such as meat. Instead, these interventions would fall under the purview of the government, which possesses broader regulatory tools to address issues such as exchange rate volatility, import costs, supply chain disruptions, and agricultural productivity challenges (BFAP 2022) that further complicate the efforts by the SARB to maintain inflation within the targeted range. The 9.3% food price inflation recorded in 2005/06 and the 12.1% food price inflation at the end of 2016 in South Africa contributed to the overall rise in inflation during those periods (Iddrisu and Alagidede 2020). This highlights the need for studies such as this one to understand the underlying drivers of inflation.

This present study analyses the relationship between CPI meat and PPI beef, PPI pork, PPI mutton, and PPI chicken meat to provide insights into how fluctuations in producer prices can affect consumer prices. Such knowledge is critical for developing targeted policies that address specific inflationary pressures within the agricultural sector, specifically the meat sector, ensuring that policies are well aligned with actual market conditions. This type of knowledge may assist policymakers developing more precise interventions to support sustainable agricultural practices, and ultimately contribute to the overall stability of food prices and maintain the inflation target range within the economy.

Following this introduction, empirical literature is reviewed in Section 2. Section 3 describes methodological techniques and Section 4 presents the results of the study while Section 5 discusses the policy implications of the results. Conclusions are presented in Section 6.

2. Literature Review

The seminal literature from the 1990s embedded in cost-push and demand-pull theories has been used to explain the causal relationships between CPI and PPI with claims of the PPI causing the CPI and vice versa for the cost-push and demand-pull approaches, respectively (Cushing and McGarvey 1990; Ceylan 2020). Exploring such relationships continues to be an area of interest for policymakers given the varied results that depend on the analysis techniques used, the level of aggregation and separate inflation measures. The likelihood of examining the nexus between CPI and PPI indices for specific individual commodities is relatively new in the literature is high, due to the limited number of studies in this area. However, studies such as that of Anggraeni and Irawan (2018) have analysed the relationship between these two indices for broader commodity groups (such as the foodstuff group; clothing group; food, beverage, cigarette, and tobacco group) in the Indonesian market economy using Granger causality. The findings revealed that a bidirectional relationship exists between PPI and CPI for the foodstuff group with a unidirectional relationship between PPI and CPI for the clothing group. There was no causality between PPI and CPI for processed food, beverage, cigarette, and tobacco group. The general relationship and/or causality between PPI and CPI was analysed using various econometric techniques such as Granger causality test, Trivariate Structural Vector Autoregression (TSVAR), Multiple regression, Autoregressive Distributed Lag (ARDL), Error Correction Model, Impulse response and Variance decomposition techniques, Engle–Granger cointegration and Ordinary Least Square (OLS), Scatter plots, Toda–Yamamoto causality, Vector Error Correction Model (VECM), and Wavelet approach reviewed in detail in this section. Granger causality and VECM were the most prevalent techniques used in previous research, and they are also applied in this present study to examine the direction of causality and both short- and long-run associations between CPI and PPI.

In understanding the causality and short- as well as long-run association between CPI and PPI in Turkey, Ulke and Ergun (2014) and Berat and Keskin (2021) used Johansen cointegration, VECM, Engle–Granger cointegration, and Granger causality tests. Both studies found that there is a presence of short- and long-run relationships suggesting that the series move together while Ulke and Ergun (2014) found a unidirectional long-run causality with no causality from CPI to PPI in the short run. Berat and Keskin (2021), using a different time series data set, found that the relationship between CPI and PPI in Turkey was bidirectional in the long run. Such ambiguity in results is expected given the different data sets used and that the economy can change due to the presence of demand-pull inflation in the long run as indicated by Ulke and Ergun (2014). Alemu (2012), in a similar manner, analysed the causal relationship between CPI and PPI using the Error Correction framework and incorporated the Granger causality test in order to ascertain causality in South Africa. Parallel to the findings of Ulke and Ergun (2014), the study by Alemu (2012) found that there exists a dynamic relationship between CPI and PPI characterised by unidirectional causality from PPI to CPI. By using monthly data, Hakimipoor et al. (2016) explored the relationship between CPI and PPI using Granger causality and Johansen cointegration tests. In contrast to the study by Alemu (2012) and that of Ulke and Ergun (2014), it was found that in the long run, there is no relationship between these series, implying no causality relationship between CPI and PPI in Iran. Even so, in the short run, there was bidirectional causality between two indices. The findings suggest that the PPI is capable of forecasting the CPI in the short run rather than the long run.

Previous research (Ghazali et al. 2008; Galodikwe 2014; Tiwari et al. 2014; Su et al. 2016; Oyeleke and Ojediran 2018) analysed the relationship and the causality between CPI and PPI using various methods that are not used in the present study. For example, using a rolling bootstrap approach, Su et al. (2016) found evidence of a bidirectional causal relationship between CPI and PPI. In contrast, Ghazali et al. (2008) found a unidirectional causality relationship from PPI to CPI using Engle–Granger and Toda–Yamamoto causality tests. Similarly, Tiwari et al. (2014) found that there was causality running from PPI to CPI using the Wavelet Transform Method (WTM) approach. Oyeleke and Ojediran (2018), using the Johansen, Engle–Granger, and Vector Autoregressive (VAR) methods, found that there was no cointegration and no causality between CPI and PPI. Galodikwe (2014) used regression analysis and correlation analysis to quantify the relationship and simple linear regression to define the linear relationship between CPI and PPI in South Africa. The study found that there is existence of a positive long-run relationship between the indexes, and unlike other reviewed studies, the direction of the relationship was not clear, which leaves a gap in the literature. To fill this gap in the literature, Meyer and Habanabakize (2018) used ARDL model, ECM, and Granger causality analysis to analyse long- and short-run relationships between the CPI, the PPI, and the purchasing manager’s index. Differentiating the study by Meyer and Habanabakize (2018) from other studies is the additional variable, namely, the purchasing manager’s index. Still, the study found that these variables exhibited cointegration over the long term, with observable causal relationships explained. Specifically, the findings indicate that the CPI exerts influence on the purchasing manager’s index, while conversely, the PPI influences the purchasing manager’s index. Furthermore, Meyer and Habanabakize (2018) emphasised the notable inconsistency between the empirical findings and the predicted results derived from the existing literature, highlighting the unique characteristics of each nation’s economic environment and the distinctive interrelationships among economic variables within it.

Additionally, other studies such as Mpofu (2011) and Ocran (2010) have conducted research that is different from the context of the present study and those of the available literature reviewed. Nevertheless, a thorough review of findings in those studies remains crucial as they offer valuable insights into variables and outcomes pertinent to this research. For instance, Mpofu (2011) investigated the relationship between the CPI and several macroeconomic variables, including money supply, prime overdraft interest rate, exchange rate, and oil using the Multiple regression approach. As far as targeting inflation is concerned, the study revealed that the collective influence of these macroeconomic variables’ accounts for roughly 97% of the CPI’s movement. Ocran (2010) examined the effect of exchange rate changes on domestic prices by using Impulse response and Variance decomposition techniques in an unrestricted Vector Autoregression (VAR) framework. The study found that a 1% shock to the nominal effective exchange rate led to a 0.125% rise in the CPI level. Additionally, it was observed that the pass-through elasticity of the PPI is 20% after 24 months. This indicates that positive changes in PPI inflation may greatly reduce CPI inflation.

On the comparative level, the causal relationship between CPI and PPI have been explored (Akcay 2011; Topuz et al. 2018; Woo et al. 2019). Topuz et al. (2018) employed VAR models, Impulse response analysis, Variance decomposition, and Granger causality tests to examine this relationship in a two-country study involving Turkey and the UK. The study found a bidirectional causality between PPI and CPI in both countries. In contrast, Akcay (2011) used VAR models across various European Union countries; the study identified a unidirectional causality from PPI to CPI in Finland and France, and bidirectional causality between the two indexes in Germany. However, no significant causality was detected in the case of the Netherlands and Sweden. Woo et al. (2019) used the Momentum-threshold Autoregressive (MTAR) cointegration model for empirical analysis. It was found that the CPI and the PPI are cointegrated, with a bidirectional long-run Granger causality between CPI and PPI in the UK, France, and Germany from 1997 to 2013.

The current study focuses on the relationship between two indexes with meat commodities as a case. Of all the studies reviewed, none was able to ascertain the relationship between CPI and PPI with a focus on specific commodities. There is not enough information on whether individual commodities’ CPI and PPI can Granger cause each other, whether the direction of the causality is unidirectional, bidirectional, independent, or not existing. In addition, whether in the short or long run exists a relationship between two indexes of selected specific commodities is still unknown. This information is crucial for policymakers for the purposes of maintaining the specific inflation target range, and most importantly, to minimise the producer and consumer price inflation of those specific goods. Thus, the present study aims to build upon the existing literature and methodologies utilised in previous studies. Additionally, it aims to contribute to the existing literature by assessing the association between CPI and PPI following the Granger causality and Vector Error Correction Model (VECM), providing a more comprehensive exploration of the relationship between these indexes while focusing on specific meat commodities such as beef, chicken meat, mutton, and pork.

3. Methodology

3.1. Data Sources

Publicly available secondary annual time series data from 1991 to 2023 accessible from the Food and Agriculture Organization (FAO) and Statistics South Africa (STATS SA) have been used in the study, covering a 32-year period that captures significant meat price fluctuations that occurred in 2011, 2013, and 2017, as noted by DAFF (2022), Hyland (2015), and NAMC (2019). This timeframe includes the period leading up to the global economic disruption caused by COVID-19 and the Ukraine–Russia war. In South Africa, CPI data are provided in aggregated commodity groups, such as CPI meat, while the PPI is available both in aggregated groups and as single-item indexes, such as PPI beef, PPI chicken meat, PPI pork, and PPI mutton (STATS SA 2016). The components of CPI meat include beef, chicken, lamb, dried, salted, smoked meats, and other preserved or processed meats (STATS SA 2019). The CPI measures average prices for frequently purchased commodities that may be established or modified at regular periods, such as annually, while the PPI measures the average fluctuations in the prices received by domestic producers for their goods and services, and these price fluctuations are assessed at the producer level (Anggraeni and Irawan 2018; Akcay 2011; Stewart 2008). This study uses the broader (aggregated) CPI meat category while analysing its relationship with the disaggregated PPI data for individual meat types. The approach follows the methodologies of Nakajima et al. (2010) and Apaitan et al. (2020), which combine aggregated and disaggregated data sets in the same model to better understand the association between the variables. The base year for all the chosen variables was 2016, therefore 2016 = 100. EViews 12 was used to conduct data analysis in this research.

3.2. Analytical Techniques

Stationarity testing holds significant importance in the field of research due to the widespread presence of trend or non-stationarity in financial and economic time series data (Mushtaq 2011). The study used the Augmented Dickey–Fuller test to assess stationarity of the selected variables in accordance with the general model proposed by Dickey and Fuller (1979) shown below:

$$\mathrm{General}\mathrm{model}:\Delta Y_t=\alpha +{\beta }_t+\rho Y_{t-1}+{\rho }_n\Delta Y_{t-n}+{\epsilon }_t$$

(1)

$$\mathrm{Specific}\mathrm{models}:\Delta {CPI}_t=\alpha +{\beta }_t+\rho {CPI}_{t-1}+{\rho }_n{\Delta CPI}_{t-n}+{\epsilon }_t$$

(2)

$$\Delta {PPI}_t=\alpha +{\beta }_t+\rho {PPI}_{t-1}+{\rho }_n{\Delta PPI}_{t-n}+{\epsilon }_t$$

(3)

$$\mathrm{Model}\mathrm{hypotheses}:H_0=Non-stationary\mathrm{and}{H}_A=stationary$$

where $\Delta CPIt\mathrm{a}\mathrm{n}\mathrm{d}\Delta PPIt$ are the changes in CPI meat and PPI beef, pork, mutton, and chicken meat at time t; ${PPI}_{t-1}$ $\mathrm{a}\mathrm{n}\mathrm{d}{CPI}_{t-1}$ are the lagged values of CPI meat and PPI beef, pork, mutton, and chicken meat; ${\Delta CPI}_{t-n}\mathrm{a}\mathrm{n}\mathrm{d}{\Delta PPI}_{t-n}$ are the changes in lagged values of CPI meat and PPI beef, pork, mutton, and chicken meat; $\beta$ and $\rho$ are the coefficients; $\alpha$ is the intercept; and ${\epsilon }_t$ is the error term.

The Johansen cointegration method was utilised to investigate the long-run relationships among the selected variables. The series are considered cointegrated when each individual economic time series becomes stationary after differencing (i.e., they are integrated of order one, I (1)), and a linear combination of these series is also stationary (Bhatta et al. 2020). This approach involves two tests: the trace test and the maximum eigenvalue test (Asari et al. 2011), both of which were introduced by Johansen (1988) as a multivariate extension of the Dickey–Fuller test. These tests use the maximum likelihood principle to determine the presence of cointegrating vectors in time series data.

When conducting cointegration tests, the null hypothesis assumes that there is no evidence of cointegration in the data, and this is tested against the alternative hypothesis, which suggests the presence of cointegration with an unspecified number of cointegrating vectors (Horvath and Watson 1995). The trace test evaluates the hypothesis of no cointegration against the alternative of multiple cointegrating vectors, while the maximum eigenvalue test examines the null hypothesis of no cointegrating vectors against the alternative of exactly one cointegrating vector (Bilgili 1998). This study employed the Johansen cointegration approach, following the general model established by Johansen (1988), which is specified as follows:

$$\mathrm{General}\mathrm{model}:∆Y_t=\mathrm{ʯ}+\phi Y_{t-1}+\sum \partial i\Delta Y_{t-i}+{\epsilon }_t$$

(4)

where $∆Yt$ is the first difference of the vector of endogenous variables, $\mathrm{ʯ}$ is the constant term, $\phi$ and $\partial$ are the adjustment coefficients, $Y_{t-1}$ is the value of $Y$ at the previous time period, $\Delta Y_{t-i}$ is the lagged value of endogenous variables, and ${\epsilon }_t$ is the error term.

Johansen (1988) further proposes the two likelihood tests as follows:

$$J_{trace}=-T{\sum }_{i=r+1}^{n}In(1-{\lambda }_i)$$

(5)

$$J_{max}=-TIn(1-{\lambda }_{r+1})$$

(6)

where $T$ is the number of observations, λ is the i th largest canonical correlation, and ${\lambda }_{r+1}$ is the ($r+1$)-th eigenvalue obtained from the estimation of the cointegration matrix.

The Vector Error Correction Model (VECM), introduced by Engle and Granger in 1987, is generally applied to accounts for short-run fluctuations and divergence from equilibrium in multivariate time series (Dalina and Liviu 2015; Liang and Schienle 2019). The application of the VECM requires determining the appropriate lag length. According to Dalina and Liviu (2015), this is typically achieved using the automated lag selection provided by the statistical software package, which helps guide how the variables should be lagged. Once the results are obtained, a negative and statistically significant coefficient suggests that short-term fluctuations between the independent and dependent variables lead to a stable long-term relationship between them (to Asari et al. 2011). Hendry (1995) defined the general paradigm of Vector Error Correction utilised in the study as depicted below:

$$\mathrm{General}\mathrm{model}:{△X}_t=\alpha +{\lambda }_i{ECT}_{t-1}^1+\sum {\beta }_i{△X}_{t-1}+\sum {\delta }_j{△Y}_{t-1}+{\epsilon }_t$$

(7)

$${△Y}_t=\alpha +{\lambda }_i{ECT}_{t-1}^1+\sum {\beta }_i{△Y}_{t-1}+\sum {\delta }_j{△X}_{t-1}+{\epsilon }_t$$

(8)

$$\mathrm{Specific}\mathrm{models}:{△CPI}_t=\alpha +{\lambda }_i{ECT}_{t-1}^1+\sum {\beta }_i{△CPI}_{t-1}+\sum {\delta }_j{△PPI}_{t-1}+{\epsilon }_t$$

(9)

$${△PPI}_t=\alpha +{\lambda }_i{ECT}_{t-1}^1+\sum {\beta }_i{△PPI}_{t-1}+\sum {\delta }_j{△CPI}_{t-1}+{\epsilon }_t$$

(10)

where ${△CPI}_t$ is the change in CPI meat; ${△PPI}_t$ is the change in PPI pork, beef, chicken meat, and mutton at time t; $\alpha$ is the constant term; ${\delta }_j$ and ${\beta }_i$ are the coefficients of the lagged values of CPI meat and PPI pork, beef, chicken meat, and mutton; ${\lambda }_i$ is the coefficient estimates for error correction term; $ECT$ is the error correction term; and ${\epsilon }_t$ is the error term.

Granger causality measures whether one variable causes or predicts another variable and helps in prediction (Sorensen 2005). Yii and Geetha (2017) stated that it is common for two economic time series to be either Granger causing or non-Granger causing each other. The study used the Granger causality defined by Granger (1969) as depicted below:

$$\mathrm{General}\mathrm{model}:X_t={\sum }_{i=1}^P{\mathrm{ʯ}}_i{X}_{t-i}+{\sum }_{j=1}^P{\beta }_j{Y}_{t-j}+{\epsilon }_t$$

(11)

$$Y_t={\sum }_{i=1}^P{\mathrm{ʯ}}_i{Y}_{t-i}+{\sum }_{j=1}^P{\beta }_j{X}_{t-j}+{\epsilon }_t$$

(12)

$$\mathrm{Specific}\mathrm{models}:{CPI}_t={\sum }_{i=1}^P{ʯ}_i{CPI}_{t-i}+{\sum }_{j=1}^P{\beta }_j{PPI}_{t-j}+{\epsilon }_t$$

(13)

$${PPI}_t={\sum }_{i=1}^P{ʯ}_i{PPI}_{t-i}+{\sum }_{j=1}^P{\beta }_j{CPI}_{t-j}+{\epsilon }_t$$

(14)

where ${CPI}_t$ and ${PPI}_t$ are the endogenous variables at time t, p is the maximum number of lags included in the model, ʯ and β are the coefficients, and ${\epsilon }_t$ is the error term.

4. Results

This section provides a concise overview of the study’s findings, divided into three (3) subsections. Section 4.1 discusses the results of the Augmented Dickey–Fuller (ADF) test, followed by the Johansen cointegration test, the Vector Error Correction Model (VECM), and the Granger causality test in Section 4.2. Concluding this section, Section 4.3 discusses the model diagnostics.

4.1. Augmented Dickey–Fuller Unit Root Test

The ADF unit root test results are presented in

Table 1. Augmented Dickey–Fuller unit root test results.

Variables	With intercept at levels
	ADF stats	5% level	Prob	Results
CPI meat	−1.192384	−2.971853	0.6633	Non-stationary
PPI Beef	−0.360333	−2.957110	0.9044	Non-stationary
PPI Chicken meat	−1.527583	−2.971853	0.5052	Non-stationary
PPI Pork	−2.445905	−2.971853	0.1390	Non-stationary
PPI Mutton	−1.173023	−2.963972	0.6727	Non-stationary
	Trend with intercept at levels
	ADF stats	5% level	Prob	Results
CPI meat	−4.483151	−3.557759	0.0060	Stationary
PPI Beef	−3.548711	−3.557759	0.0510	Non-stationary
PPI Chicken meat	−4.191225	−3.562882	0.0124	Stationary
PPI Pork	−3.900911	−3.212361	0.0237	Stationary
PPI Mutton	−4.365410	−3.212361	0.0080	Stationary
	With intercept at first difference
	ADF stats	5% level	Prob	Results
CPI meat	−4.946506	−2.971853	0.0004	Stationary
PPI Beef	−7.450471	−2.960411	0.0000	Stationary
PPI Chicken meat	−4.825832	−2.971853	0.0006	Stationary
PPI Pork	−4.133884	−2.971853	0.0034	Stationary
PPI Mutton	−6.221388	−2.963972	0.0000	Stationary
	Trend with intercept at first difference
	ADF stats	5% level	Prob	Results
CPI meat	−4.988927	−3.580622	0.0021	Stationary
PPI Beef	−4.098878	−3.622033	0.0193	Stationary
PPI Chicken meat	−5.094606	−3.580622	0.0016	Stationary
PPI Pork	−5.015334	−3.580622	0.0020	Stationary
PPI Mutton	−3.152916	−3.603202	0.1165	Stationary

, showing that all variables were not integrated of order (0) at intercept and suggesting that CPI meat, PPI beef, PPI chicken meat, PPI pork, and PPI mutton were non-stationary. However, at trend and trend with intercept, all variables were integrated of order (0) except PPI beef. The non-stationarity of the PPI beef series aligns with Mushtaq (2011)’s observation that many economic and financial time series are often non-stationary. The null hypothesis is rejected at 5% level of significance after first differencing the series, showing that all variables are integrated of order (1) at both trend as well as trend with intercept. This implies that there was absence of temporal fluctuations assuring the authenticity of subsequent statistical tests, eliminating the possibility of inconsistent results.

4.2. Johansen Cointegration Test, VECM, and Granger Causality Results

The LR, FPE, AIC, SC, and HQ were five (5) tests used to automatically determine the Vector Autoregressive (VAR) lag order selection. The criteria with the smallest value and asterisk (from

Table 2. Maximum lag order (p) results for all variables.

Lags	LogL	LR	FPE	AIC	SC	HQ
0	−518.7239	NA	3.25 × 108	33.7886	34.0199	33.8640
1	−388.7345	209.6603	380,967.4	27.1513	28.4029 *	27.4675 *
2	−358.5371	38.9644 *	312,302.0 *	26.6798 *	29.2230	27.5092

Where * indicates lag order selected by the criterion, LR: sequential modified LR test statistic (each test at 5% level), FPE: Final prediction error, AIC: Akaike information criterion, SC: Schwarz information criterion, HQ: Hannan–Quinn information criterion.

) are chosen to establish the maximum lag order for use (Zhang et al. 2020). Hence, the study chose lag 2 based on the AIC, since it has the lowest value.

The null hypothesis is rejected at 5% level at both Trace and Maximum Eigenvalue statistics of the Johansen Cointegration test, as shown in

Table 3. Johansen cointegration test results.

Trace Test
Hypothesised No of CE(s)	Eigenvalue	Trace Statistics	0.05% Critical Value	Prob. **	Decision
None *	0.696120	92.32662	69.8189	0.0003	Cointegrated
At most 1 *	0.648563	55.40187	47.8561	0.0083	Cointegrated
At most 2	0.317122	22.98443	29.7970	0.2468	Not cointegrated
At most 3	0.288450	11.15981	15.4947	0.2019	Not cointegrated
At most 4	0.019491	20.610189	3.8414	0.4347	Not cointegrated
Maximum Eigenvalue Test
Hypothesised No of CE(s)	Eigenvalue	Maximum eigenvalue Statistics	0.05% Critical Value	Prob. **	Decision
None *	0.696120	0.696120	33.87687	0.0209	Cointegrated
At most 1 *	0.648563	0.648563	27.58434	0.0010	Cointegrated
At most 2	0.317122	0.317122	21.13162	0.5651	Not cointegrated
At most 3	0.288450	0.288450	12.26460	0.1783	Not cointegrated
At most 4	0.019491	0.019491	3.841465	0.4347	Not cointegrated

* indicates rejection of the hypothesis at the 0.05 level; ** Mackinnan–Haug–Michelis p values, Trace and Max Eigen test indicates r cointegrating model (s) at 5% significance level.

. This suggests that there is presence of a long-run cointegration relationship with at least two cointegrating equations between CPI meat, PPI beef, PPI chicken meat, PPI pork, and PPI mutton.

The results in

Table 4. VEC model results for all variables.

Variables	Coefficient	Std. Error	t-Statistic	Prob.
ECT (−1)	−0.6767	0.2560	−2.6432	0.0165 *
LCPI_Meat (−1)	0.1895	0.4273	0.4434	0.6628
LCPI_Meat (−2)	0.2953	0.2755	1.0721	0.2978
LPPI_beef (−1)	−0.3488	0.2162	−1.6139	0.1240
LPPI_beef (−2)	−0.2909	0.1404	−1.9299	0.0695 *
LPPI_chicken meat (−1)	0.6620	0.3045	2.1741	0.0433 *
LPPI_chicken meat (−2)	0.2727	0.2475	1.1018	0.2851
LPPI_mutton (−1)	−0.1544	0.1850	−0.8346	0.4149
LPPI_mutton (−2)	−0.1671	0.1663	−1.0047	0.3283
LPPI_pork (−1)	−0.4218	0.1712	−2.4645	0.0240 *
LPPI_pork (−2)	−0.2124	0.1258	−1.6887	0.1085
C	0.0985	0.0267	3.6936	0.0017 **
R-squared			0.5375
Adj. R-squared			0.2549
F-statistic			1.9019
Log likelihood.			62.0563
Akaike AIC			−3.3371

Notes: ** 5%, * 10% level of significance, L refers to the logarithm of the CPI meat and PPI beef, PPI chicken meat, PPI pork, and PPI mutton.

are based on the analysis of both the short- and long-run equilibrium dynamics using the automated VAR model with a lag of 2. The VEC model was applied to test the equilibrium relationship between CPI meat and the PPI indices for beef, chicken, mutton, and pork.

The significant and negative coefficient of the ECT (−0.6767) indicates that when the system deviates from its long-term equilibrium, it adjusts back towards equilibrium at a rate of approximately 67.67% per period. This strong adjustment rate suggests a long-term equilibrium relationship between CPI meat and the PPI for various meat types. The system self-correction mechanism indicates that negative deviations tend to persist longer, while positive deviations are corrected more quickly (Mallick et al. 2020). The above-average rate of correction, as indicated by the error correction term, implies that CPI meat consistently corrects any disequilibrium from the previous period, steering towards long-term equilibrium. Whenever variables such as PPI beef, PPI chicken meat, PPI mutton, and PPI pork deviate from this equilibrium, the ECT acts to correct these short-run disequilibria. Moreover, the lagged values of PPI chicken meat, PPI beef, and PPI pork were found to be statistically significant, highlighting their notable impact on CPI meat in the short run. Specifically, a 1% increase in the past value of PPI chicken meat is associated with a 0.66% increase in CPI meat in the current period, demonstrating a positive relationship at the 10% significance level. This finding suggests that changes in production costs for chicken meat (as reflected in the PPI) are likely to affect consumer prices for meat, contributing to price fluctuations in the sector. The positive relationship between CPI meat and PPI chicken meat indicates that an increase in the average price of chicken meat leads to a rise in overall meat prices, and vice versa. These findings align with those of Galodikwe (2014), which identified a positive relationship between the general CPI and PPI in South Africa, though this study specifically found such a relationship between the CPI for meat and the PPI for chicken meat. This positive relationship could further help to explain why South African consumers tend to prefer white meat over red meat, as observed by Bisschoff and Liebenberg (2017). It also aligns with the conclusions of Taljaard et al. (2004) made decades ago and recent conclusions of Delport et al. (2017) that chicken meat is considered a necessity good, meaning that its demand remains steady despite price changes in the short run. On the other hand, PPI mutton exhibits a statistically insignificant negative impact on CPI meat, suggesting that changes in PPI mutton from previous periods do not influence the current CPI meat. This could be due to limited consumer demand for mutton compared to other meats and the relatively smaller share of mutton in the overall meat market, leading to a weaker influence on the broader CPI for meat. Moreover, the first lag of PPI pork shows a significant negative effect on CPI meat, implying that a 1% increase in PPI pork from the previous period results in a 0.42% decrease in the current CPI meat. This implies that when pork average prices increase, there may be a substitution effect, where consumers shift away from pork to other meats (such as chicken, beef, or mutton), driving down consumer meat prices. The impact of PPI pork from two periods prior, PPI beef from one period prior, and PPI mutton from both one and two periods prior are all statistically insignificant, indicating they do not significantly affect the current CPI meat in the short run. This suggests that fluctuations in pork, mutton, and beef producer prices beyond one period do not influence consumer meat prices, highlighting the short-term nature of price adjustments in the meat market. However, after two periods, a 1% increase in PPI beef results in a 0.29% decrease in CPI meat, indicating a delayed negative effect. This suggests that while immediate fluctuations in beef production costs may not impact consumer prices, longer-term changes exert downward pressure on meat prices. The negative short-run relationship between PPI beef and CPI meat contradicts the findings of Berat and Keskin (2021), that identified a positive relationship between CPI and PPI on a monthly basis. This difference could be due to the fact that the present study focuses specifically on meat-related CPI and PPI, with the analysis conducted on a yearly basis.

To further understand the interrelationships and support the VECM model results, impulse response functions were utilised. These functions illustrate how economic variables react to shocks within the system (Lütkepohl 2008). Figure 1 displays the impulse response functions for the PPI of different meat types in response to shocks from the CPI for meat. The results indicate that it takes approximately eight months for the variables to stabilise after the shocks occur.

The findings under the 95% confidence interval (+2 S.E and −2 S.E) represented by the red dotted line and the black solid line illustrate impulse movements of the variables. These results show that the PPI for beef initially experiences a sharp decline in response to a CPI meat shock, followed by a slight increase around the third period, and gradually stabilises by the eighth period, indicating that the initial impact of the CPI shock diminishes over time. Similar patterns are observed in the PPIs for mutton (Panel C) and pork (Panel D), where there is an initial decline followed by a gradual return to stability. The response in mutton prices is slightly less volatile than in beef, suggesting a smoother adjustment process. In contrast, the PPI for chicken meat (Panel B) shows a more volatile reaction to CPI shocks, with an initial increase followed by a decline and a slight upward adjustment before stabilisation. This indicates that the impact of the CPI shock on chicken meat prices is somewhat more persistent but eventually levels out after about 8 months. These findings suggest that different meat types exhibit varying degrees of sensitivity to changes in consumer prices of meat in South Africa, with beef, mutton, and pork showing more predictable stabilisation patterns, while chicken meat demonstrates greater volatility. These findings are consistent with the findings of Li et al. (2019), who made similar observations when analysing the general relationship between CPI and PPI.

After establishing the relationships among the variables, it is crucial to understand the direction of these associations. The results in

Table 5. Granger causality test results.

Null Hypothesis	Lags	F Statistic	Prob	Decision
LPPI (BEEF) does not Granger cause LCPI (MEAT)	2	1.57258	0.2266	Reject
LCPI (MEAT) does not Granger cause LPPI (BEEF)	2	1.04701	0.3653	Reject
LPPI (CHICKEN MEAT) does not Granger cause LCPI (MEAT)	2	2.39902	0.1106	Reject
LCPI (MEAT) does not Granger cause LPPI (CHICKEN MEAT)	2	9.53566	0.0008	Accept
LPPI (MUTTON) does not Granger cause LCPI (MEAT)	2	2.13644	0.1383	Reject
LCPI (MEAT) does not Granger cause LPPI (MUTTON)	2	3.04270	0.0650	Accept
LPPI (PORK) does not Granger cause LCPI (MEAT)	2	0.81939	0.4518	Reject
LCPI (MEAT) does not Granger cause LPPI (PORK)	2	6.42058	0.0054	Accept

Notes: L refers to the natural logarithm of the CPI meat and PPI beef, PPI pork, PPI mutton, and PPI chicken meat.

indicate that there is no directional flow from PPI beef to CPI meat, from CPI meat to PPI beef, from PPI mutton to CPI meat, or from PPI pork to CPI meat. However, the study did identify a directional flow from CPI meat to PPI chicken, from CPI meat to PPI mutton, and from CPI meat to PPI pork.

The findings suggest that there is no causal relationship between CPI meat and PPI beef, indicated by the statistically insignificant p values of 0.2266 and 0.3653. However, there is a unidirectional relationship between CPI meat and PPI chicken meat, PPI mutton, and PPI pork indicated by the statistically significant p values of 0.0008, 0.0650, and 0.0054, respectively, which will be explained in detail in Section 5.2 of this study.

4.3. Diagnostic Tests

The study used Breusch–Godfrey to test for serial correlation through the LM test shown in

Table 6. Breusch–Godfrey serial correlation LM test results.

F-statistic	0.05362
Obs* R-squared	0.1998
Prob. F (2,16)	0.9480
Prob. Chi-square (2)	0.9049

. The null hypothesis was that there is no serial correlation tested against the alternative hypothesis that there is serial correlation. The study fails to reject the null hypothesis at 5% level of significance. This implies that there is no evidence of serial correlation between CPI meat and PPI beef, PPI chicken meat, PPI mutton, and PPI pork. The F-statistic is greater than 0.005 together with the probability of F and its chi-square value.

Heteroskedasticity was tested under the null hypothesis that there is no heteroskedasticity against the alternative hypothesis stating the opposite using the Breusch–Pagan–Godfrey test. The study fails to reject the null hypothesis because the F-statistic and probability chi-squares are greater than 0.005. This implies that there was no presence of heteroskedasticity in the models used in this study.

Youness et al. (2021) supports the results on serial correlation and heteroskedasticity in Table 6 and

Table 7. Breusch–Pagan–Godfrey heteroskedasticity test results.

F-statistic	1.0288
Obs* R-squared	15.72984
Prob. F (4,23)	0.4813
Prob. Chi-square (4)	0.4002
Prob. Chi-square (4)	0.9480
Scaled explained SS	7.3246

, as they found that there was no serial correlation and heteroskedasticity and concluded that the model was valid.

5. Further Discussion of Results and Policy Implications

5.1. VECM and Cointegration Policy Implications

Studies reviewed earlier in this paper focused on the general relationship between CPI and PPI with some authors such as Anggraeni and Irawan (2018), analysing the broader commodity groups associated with these two metrics.

The Johansen cointegration results shown in Table 3 proved that there is existence of a long-run cointegration relationship between the variables, thus suggesting the presence of a stable and equilibrium relationships among these variables over time. The VECM results in Table 4 and impulse response function in Figure 1 align with the Johansen cointegration findings by further demonstrating the short-run dynamics and the ability of the variables to return to equilibrium in the long run given any deviation. Ulke and Ergun (2014) made emphasis on the CPI as a primary indicator for price changes in which these changes in the short run do not affect the PPI in the long run. A similar dynamic may be observed in the case of South Africa, where the average producer prices of mutton are not significantly influenced by short-term fluctuations in consumer meat prices. This contrasts with the prices of other meats, such as chicken, pork, and beef, where producer prices may be more responsive to changes in consumer prices within the short term. This could be because of the excess demand of chicken meat, pork, and beef by consumers, which exceeds the supply of these meat types, resulting in demand-pull inflation. These results suggest that average prices of chicken meat, pork, and beef are more sensitive to changes in consumer demand or production cost as compared to mutton average prices in the short run, potentially leading to excess supply and demand-pull inflation. This finding is expected and is in line with the explanation of Berat and Keskin (2021), highlighting the difficulty of increasing the supply in the short run as compared to demand that can change at any point in time. These findings further highlight the importance of closely monitoring both consumer meat prices and producer prices for beef, chicken, mutton, and pork, as they play a critical role in maintaining a long-run equilibrium. Significant deviations in these prices could trigger short-term disruptions, necessitating corrective measures. While the ECT suggests that the meat market has inherent mechanisms to restore equilibrium, unmonitored and persistent deviations could overwhelm these mechanisms, leading to price instability in the agricultural commodity market. Therefore, the South African Reserve Bank should consider the relationship of the selected meat type’s CPI and PPI as important when formulating inflation targeting strategies. The detection of deviations within meat categories and their quick corrections suggests that the market can respond efficiently, but also that these deviations have the potential to impact general food prices. Addressing these deviations promptly can prevent them from escalating into more significant inflationary issues.

5.2. Granger Causality Policy Implications

As noted in Akcay (2011), three possible relationships can be found in the literature: unidirectional, bidirectional, and no causality relationships. This is in line with the findings of this study, where two relationships, namely the no causality and the unidirectional relationship, were found and explained in Section 4.2.

A unidirectional causal relationship was identified between CPI meat and the PPI for chicken meat, pork, and mutton. These findings are consistent with those of Akcay (2011), Alemu (2012), and Ghazali et al. (2008), who observed that changes in producer prices typically drive consumer prices in a unidirectional manner. However, while the previous studies identified a causal flow from PPI to CPI, this study revealed contrary results in South Africa, showing that changes in consumer prices of meat drive changes in producer prices for chicken meat, pork, and mutton, rather than the other way around. This aligns with Meyer and Habanabakize (2018), who found a different direction of causality between general CPI and PPI in the South African economy, showing that, whether at the general level or in specific sectors such as meat, the CPI–PPI relationship in South Africa follows a unique direction, different from the traditional flow from PPI to CPI. It is therefore evident that meat producer pricing strategies for pork, mutton, and chicken are influenced by consumer prices, which are likely driven by high demand and the strong purchasing power of consumers. This suggests that chicken, pork, and mutton prices can be used by producers to predict meat price changes in South Africa. By implementing measures to monitor and regulate producer prices for these meats, stability in the CPI meat index can be better ensured. Moreover, using these metrics for decision making could help shield households from significant price fluctuations through the implementation of targeted measures such as producer subsidies, import tariff adjustments, or the release of strategic reserves, which help stabilise retail prices and maintain affordability, ultimately assisting in maintaining affordability and reducing the impact of price changes on household budgets.

The study found that no causal relationship exists between CPI meat and PPI beef in South Africa. This implies that the averages producers’ prices of beef do not influence the average consumer meat prices. As such, PPI beef is not useful in predicting CPI meat in South Africa. These findings suggest that the changes in the prices of beef do not trigger the food supply chains and the consumers basket of goods, as Hobbs (2021) indicated. This is economically plausible and may indicate that beef producers, retailers, or intermediaries in the supply chain absorb price changes instead of passing them on to meat consumers, possibly due to competitive pressures, pricing strategies, or contracts. Additionally, these findings suggest that price changes in beef have a smaller impact on disrupting the overall meat economy in South Africa, due to the strong segmentation in the market, particularly in chicken, pork, and mutton. As such, the CPI for meat may be influenced by various factors, including prices of other meats, as indicated in this study. Moreover, broader economic factors like interest rates, oil prices, exchange rates, and money supply, as highlighted in Mpofu’s (2011) study on general CPI, could also contribute to increases in the CPI for meat. These findings, showing no causal relationship between CPI meat and PPI beef, further suggest that there is a stronger presence or absence of government policies such as subsidies, price controls, or import/export regulations in the beef market compared to other meat types during the study period, which could possibly stabilise consumer prices despite the fluctuations from the producer prices. Consequently, policymakers should focus on other factors that might influence consumer meat prices, such as supply chain dynamics, consumer demand, and the prices of other meat types. Since PPI beef does not directly impact CPI meat, beef prices at the producer level are not the primary drivers of changes in consumer meat prices, as proven in this study. These findings are consistent with the results of the VECM, which also found that PPI beef of one period prior is not statistically significant in explaining the relationship with CPI meat. This implies that policies targeting the South African beef market and its prices should continuously be precise and effective, avoiding unnecessary or ineffective interventions that do not achieve the desired outcomes for consumer meat prices in the long run, like interventions aimed at stabilising beef producer prices such as input cost subsidies, drought resilience programs, strategic reserves, and improved supply chain infrastructure to address production inefficiencies and disruptions. On the demand and beef meat producer side, measures such as market transparency, import/export adjustments, and temporary price caps adopt practices to improve efficiency, such as investing in better breeding, feed management, and sustainable farming practices, while also improving supply chain logistics to reduce costs, and they should be handled separately from those targeting consumer meat prices in the South African market. Although this study focused on specific meat commodities, the findings of no causality still align with those of Oyeleke and Ojediran (2018) that focused on the general relationship between the two indexes in Nigeria. The study found no evidence of bidirectional causality between the CPI for meat and the PPI for different kinds of meat, including beef, chicken, pork, and mutton. These results contradict the findings of previous studies, such as Hakimipoor et al. (2016), Berat and Keskin (2021), Anggraeni and Irawan (2018), and Su et al. (2016). The reasons for this could be due to the structure of the meat market, including the degree of competition and the role of intermediaries, which may vary across regions and over time. In this study, the meat market might be characterised by more efficient supply chains or greater market power for retailers, allowing them to absorb producer price changes without passing them directly to consumers in the short run. Additionally, macroeconomic conditions such as inflation, economic shocks, or changes in consumer demand during the time of this study could differ from those in the previous studies due to shocks such as COVID-19, the Ukraine–Russia war, and government elections, amongst others, which may have dampened the direct transmission of the CPI and PPI of different meat types.

6. Conclusions

This paper adds to the existing literature on the importance of monitoring the individual commodity price indexes in the meat industry and agricultural commodity markets at large. The study successfully fulfilled its objectives by providing empirical evidence of the linkages between producer and consumer prices of individual meat commodities. The study findings showed that changes in meat prices originate from consumer prices of meat (broader meat sector) to producer prices of mutton, pork, and chicken meat (meat sub-sectors) as a result of the unidirectional relationship that exists. Beef producer price changes originate from factors other than CPI meat since there was no causality relationship. Despite this, a short- and long-run cointegration and equilibrium relationship was found to exist between CPI meat and PPI beef, chicken meat, mutton, and pork. It has been proven in the present study that it takes approximately 67.67% adjustment speed for CPI meat to return to equilibrium after the deviations from the PPI meat types. In addition, using PPI beef to predict the prices that the consumers will pay in the market for meat is not a useful solution in this case; PPI chicken meat, PPI mutton, and PPI pork offer a better solution in monitoring the prices changes for attaining price stability. Existing interventions by the South African Reserve Bank and Statistics South Africa should be further extended to the specific commodity groups’ PPI and CPI to assist in attaining the specific range of inflation targets. Additionally, in the absence of effective input price control mechanisms, they should be implemented or, if present, enhanced. This will not only protect the producers against supply inflation but also consumers against demand inflation. Moreover, meat average prices that already play a significant role in formulating the general CPI and PPI together with other specific commodity groups will be controlled over time. The fluctuation of meat prices can be reduced by implementing targeted policy measures and regulatory frameworks, such as price stabilisation programs or strategic reserves, resulting in more predictable and beneficial prices that benefits both consumers and producers while contributing to overall economic stability. This can potentially lead to diminished levels of inflation within the meat sector. Consequently, consumers together with beef, chicken meat, mutton, and pork producers will be secured against price volatility, thereby fostering investment opportunities for businesses in the meat sector. The study recommends that when making pricing decisions, chicken meat, mutton, and pork producers should utilise the CPI for meat to forecast the possible supply inflation. In addition, beef producers should determine possible predictive metrics that can enable them to forecast inflation in the average prices of beef products they produce.