A new look at asymmetric effect of oil price changes on inflation: Evidence from Malaysia

Abstract

The existing literature that examined the effect of oil price on domestic price inflation only focused on consumer price at an aggregate level. The studies that focused on producer and production levels, and based on disaggregated data are very lack. Besides, previous studies also mainly applied the linear regression approach in studying the effect of oil price. This study seeks to explore the mentioned issues by focusing on sectoral consumer (CPI), industrial (IPI), and producer (PPI) prices of Malaysia. The Markov-switching (MS) regression technique is applied. The models are innovated by incorporating the asymmetric effects of oil price changes. The results reveal different reactions of sectoral domestic price inflation to oil price changes. The oil price has asymmetric effects on domestic price inflation with higher impacts on industrial and producer prices than the consumer price. The effect is larger in sectors that have higher linkages with oil/ energy resources. These sectors are oil-intensive and are sensitive to oil price changes. Among these sectors are the CPI transportation sector, IPI manufacturing, and electrical sectors as well as PPI fuel, chemicals, and manufacturing sectors. However, oil is not the main factor causes to domestic inflation. The main determinants of inflation are real exchange rate, aggregate supply, and demand. Besides, the policy decisions are also influential on price stability. The sectors of CPI transportation, PPI animals & vegetable oils, and PPI fuel have a high tendency to increase prices and should be well-monitored.

Keywords

Oil price shock inflation asymmetric effect markov-switching regression energy intensity

1 Introduction

For decades, economists and researchers are alarmed with the negative impacts of oil price shock through historical data and empirical findings. Yet, oil appears as an important factor to explain and predicts economic conditions such as recession and high episodes of inflation. Despite the long literature studies and historical evidence, there are still arguments on the inconclusive results. According to Salisu et al.¹ the heterogeneous results in the oil-inflation relation might be due to some factors. Among the possible factors include the selected countries, the utilized models, the data samples, and estimation techniques. This study seeks to highlight the main issues/ lack from previous studies and further explore the impact of oil price changes on inflation using the data of Malaysia. Our study contributes to the literature on oil-inflation three ways:

Firstly, the relationship between oil and inflation could be asymmetric but was neglected by the symmetric method (Hamilton,² and Valcarcel & Wohar.³⁾ A nonlinear modeling approach could be a better choice to model the effect of oil price due to theoretical and empirical reasons. Theoretically, the effect of oil price can be explained in two ways, i.e. symmetric versus asymmetric. The symmetric effect is captured through the direct effect transmitted from supply and demand channels while the asymmetric effect is through indirect channels such as sectoral reallocation, precautionary saving, monetary policy response, and uncertainty (Herrera, Lagalo & Wada).⁴ In terms of empirical reasoning, it is not reasonable to assume a relationship is linear over time. The oil-inflation relation might change over time or be asymmetric due to factors like market structure, public regulations, and cost structures (Ibrahim).⁵ In particular, policy measures such as price controls that limit the price movement may cause price asymmetry behavior (Ibrahim).⁵ The asymmetric price-setting behavior could also be arisen due to the interaction between market power and the cost structure of firms (Karantininis, Katrakylidis, & Persson).⁶ Restricting the relationship to be linear in the presence of nonlinearities might lead to inaccurate estimates. The nonlinear model provides more precise and more accurate results (Sek).⁷

The other limitation from previous research on the oil-inflation nexus is the use of aggregated data. Previous studies mainly applied aggregated data of CPI to study the effect of oil price changes on inflation. The use of aggregated data leads to a very general result on the overall CPI inflation which is less accurate and may not be able to represent the result for different sectors. This is because the effect of oil price changes on inflation might differ across sectors as revealed by some studies. For instance, Choi et al.⁸ Ibrahim and Said,⁹ Hansman et al.¹⁰ and Ibrahim.⁵

Thirdly, previous studies mainly focused on the direct effect of oil price on consumer price index (CPI) inflation. According to Alvarez et al.¹¹ many studies are focused on the direct effect of oil price shock but the studies on the indirect effect and second–round effect were very lack. Theoretically, oil may affect the economy through different channels. For instance, the boost of oil price is transmitted into higher production cost, resulting in higher energy refined products (direct effect), triggering higher consumer price inflation (indirect effect). Higher inflation may cause a drop in real income, hence lower demand for goods. This further causes lower production/ output which induces policy responses. Such transmission effect is the first round effect (direct and indirect effects). The direct effect is transmitted to the consumer price on the oil refined products consumed by consumers while the indirect effect is felt as consumers use the final goods which use oil as inputs in the production process.

The main purpose of this study is to examine the asymmetric effects (increases and decreases) of oil price changes on domestic prices (CPI, PPI, and IPI) of Malaysia at disaggregated levels. In particular, the study seeks to close the gaps or limitations in previous studies as discussed above. In filling these limitations, the study applies several procedures. First, the original (net) oil price index is decomposed into increases and decreases series. Second, the Markov-switching model is innovated by incorporating the oil price increases and decreases series to replace the net oil data into the original model. The innovative model can capture the asymmetric effects of oil price changes rather than the symmetric effect of oil price under the original symmetric relation. Third, the study applies four different models by considering three specifications (cases), and the models are estimated using time-invariant versus time-varying effect. The best model for each sector is selected using forecast error indicators. The results are discussed and compared based on the best models.

This study is focused on Malaysia due to several reasons. First, there are very limited studies examining the impacts of oil price changes on inflation for a small country like Malaysia. Besides, the examination on inflation is mainly focused on CPI at an aggregate level which provides limited information. Second, Malaysia is an oil-importer and oil-exporter. According to the U.S. Energy Information Administration (EIA) updates of 2021,¹² Malaysia is the second-largest producer of oil and natural gas among the Southeast Asian countries. In 2019, Malaysia appeared as the 24^th largest exporter of crude petroleum in the world ranking. In the same year, it is ranked as the 21^st largest importer of crude petroleum globally.¹³ Oil is highly consumed in transportation and manufacturing. Based on such a scenario, it is interesting to study how oil price changes might affect different economic sectors in Malaysia.

The results have evident that oil price changes have asymmetric effects on domestic prices in which the effects differ across sectors. The oil-intensive sectors received larger impacts from oil price changes while the sectors that are less oil-intensive and the products that are under price control and subsidies are less reactive to oil price changes. Oil price changes have larger impacts on IPI and PPI sectors than CPI sectors as they are at production level compared to consumer prices as a final good level. Besides oil, other factors are found to have much larger effects on domestic prices, they are real exchange rate, aggregate supply, and demand. In addition, the policy decisions are also influential on price stability.

The remaining of the paper is organized as follows: Section 2 provides a review on literature; Section 3 reviews the inflation of Malaysia; Section 4 describes the data; Section 5 discusses the methodology; Section 6 interprets the results and the last section concludes the findings.

2 Literature review

The effect of oil price changes on the global economy, in particular, inflation has long been studied but results are ambiguous and might vary across sectors, countries and change over time. There are different views and claims about the inconclusive findings. The theoretical reasons include the oil intensity, effective monetary policy, the change in the global oil demand, and the change of flexibility in wages (De Gregorio, Landerrectche & Neilson,)¹⁴ and Chen.¹⁵ For instance, the earlier studies on the effect of oil price mainly focused on the inflationary or oil-supply driven shock. Some studies found a strong impact of the oil shock on economic growth and inflation prior to the 1980s but the effect was declined or limited since the mid-1980s (for instance, Hooker;¹⁶ and Alvarez et al.)¹¹ Besides, oil intensity may determine the magnitude of oil impact. Industries that are more oil-intensive are more sensitive to oil price changes. Hence these industries may experience a larger oil price pass-through effect relative to non-oil intensive sectors. Among the sectors that received significant oil price pass-through effect are food/ agricultural (Ibrahim & Said),⁹ transportation (Hansman et al.)¹⁰ rent, fuel, and power (Ibrahim).⁵ Also, the same condition applies to oil-intensive countries. Oil importing countries may receive a larger impact from the oil price shock (Sek).⁷ Some studies found that oil-importing countries tend to receive a negative impact while oil-exporting countries experience a positive effect from oil price increases (Filis & Chatziantoniou).¹⁷ Gelos and Ustyugova¹⁸ found that those economies exhibited a larger share of food in their CPI basket, in addition to fuel intensities and pre-existing inflation levels were more affected by commodity price shocks.

Apart from the theoretical views, the heterogeneous effect of oil price could also be due to estimation approaches or the data composition and sample used. According to Salisu et al.¹ the heterogeneous results in the oil-inflation relation might be due to possible factors like selected countries, utilized models, data samples, and estimation techniques. Some researchers claimed that the relationship between oil and inflation could be asymmetric but it was neglected by the symmetric method (Hamilton,² and Valcarcel & Wohar.³) Previous studies mainly applied linear modeling approaches in examining the effect of oil price on inflation. The main limitation of linear modeling is to restrict a linear relationship between dependent and explanatory variables which remain no change over time. This assumption does not not truly reflect the real-world situation in which the economic relationship may deviate from the linear relationship in reacting to external influences or shocks. If the real relationship is nonlinear, the application of a linear regression might lead to biases and misleading conclusions. Hence, there are increasing recent studies to apply nonlinear regression approaches in examining the effect of oil price changes. Nonlinear modeling approaches are applicable to examine the nonlinearity relationship in two ways. On the first hand, nonlinear models are used to capture the switch in the relationship due to structural breaks or switching effects. To name a few, the nonlinear models applied include threshold autoregressive model, smooth transition regression, and Markov-switching regression. These nonlinear models help to detect breaks or threshold values that trigger changes in the relationship and provide the results in different regimes. For instance, Bala et al.¹⁹ (threshold autoregressive and momentum threshold autoregressive models), Yang et al.²⁰ (panel smooth transition regression), Donayre and Wilmot²¹ (threshold vector autoregression), and Ozdemir and Akgul²² (MS-VAR model). These studies have evidence on the nonlinear/ asymmetric effect of oil price on domestic variables. For instance, Pagliacci and Barraez²³ applied a two-regime MS regression to estimate the New Keynesian Phillips curve using the consumer price data of Venezuela. They found that the Phillips curve with backward-looking and rational expectation regimes showed consistency between inflation with the episodes of high uncertainty. Output gap and money supply are the main determinants of inflation. Besides, Ozdemir and Akgul²² applied a MS-VAR model to examine the effects of oil prices and domestic gasoline prices on domestic CPI. They detected a pass-through effect from both oil prices and gasoline prices on CPI and found that a sudden increase in gasoline price is more influential than that of crude oil prices on CPI in Turkey. More recent studies that applied Markov-switching models to examine the trend of inflation include Bojanic²² and Behera and Patra.²⁴

Besides, the second strand of nonlinear regression models is applied to examine the asymmetric effect of oil price changes. For this purpose, the nonlinear autoregressive distributed lags (NARDL) model is applied by decomposing the oil price data into two series, the oil price increases and decreases series. Some studies found that the negative effect caused by oil price increases is relatively larger than the positive effect due to oil price decreases. For instance, Ibrahim⁵ analyzed the effect of oil price changes on food price inflation in Malaysia by decomposing oil price into oil increases and decreases series. The results showed that the effect of oil price is asymmetric, i.e. oil prices changes induce different responses to increases and decreases. Oil price increases led to food price inflation, whereas oil price decreases did not lead to a decrease in food price inflation in the long run. Other studies that reported the asymmetric effect of oil include Chou and Lin,²⁵ Bala and Chin²⁶ and Long and Liang,²⁷ Babuga and Mohd Naseem.²⁸

While the application of nonlinear modeling approaches in two strands leads to more informative findings in detecting the change in the oil price-inflation relationship over time/ regime (first strand) and the asymmetric effect of oil price increases versus decreases (second strand), the combination of both nonlinearity effects are still not yet explored and utilized. This study hence takes the initiative to combine the two nonlinearity features by incorporating the asymmetric oil price elements into the standard two-regime-Markov-switching model. The modified model enables the capture of the asymmetric oil price effect and the threshold (two-regime) effect.

A number of studies were conducted using the data of Malaysia. Among these studies include, Ibrahim,⁵ Razmi et al. Sek,⁷ Lily et al.²⁹ and Xuan and Chin.³⁰ Among them, Sek⁷ and Lily et al.²⁹ applied the Nonlinear ARDL model to compare the effect of oil price increases and decreases on inflation in Malaysia. They found evidence that the effect of oil price changes is asymmetric. Razmi et al.³¹ applied the SVAR model on ASEAN4 revealed that Malaysia and the Philippines are mostly affected by oil shocks compared to other countries. Besides, oil shocks could transmit through monetary channels (indirect effects) which were found in the post-crisis period. Xuan and Chin found that the effect of actual diesel price is much more prominent with no subsidy given compared to that of the retail diesel price with subsidies. On the other hand, most of these studies applied the aggregate data of the consumer price index. The application of sectoral data and the comparison of different price indices (CPI, PPI, and IPI) is hardly found. Hence, this study seeks to fill the gaps mentioned by applying the modified MS regression model with asymmetric oil price effects using the sectoral price indices of Malaysia.

2.1 Markov-switching regression – A review

The switching feature is useful in handling data for three main purposes: time series data with multiple breaks, time series data with outliers and enables changes in both mean and variance. The early study could traceback to Hamilton³² who proposed the Markov-switching autoregressive (MS-AR) model to model the structural changes based on the autoregressive processes. The Markov-switching with the autoregressive term (MS-AR) and Markov-switching with dynamic regression (MS-DR) are applied broadly in economics and finance. However, when the MS model was introduced by Lindgren³³ and Baum et al.³⁴ the model was without any AR term. Later, the AR term was incorporated in some studies such as Poritz³⁵ and Rabiner.³⁶ These models were termed as “hidden Markov-models’ which were used in speech recognition algorithms. The early MS applications did not consider the time series components. Hamilton³² was the first to develop the algorithms for estimating the MS regression function (see Hamilton and Raj,³⁷ for detailed discussions). Hamilton³¹ in his seminar paper applied the MS-AR model to the U.S. real gross national product growth data in explaining the business cycle fluctuation by allowing a change in mean with a two-regime switching model. Since then, the MS-AR model was applied broadly in many areas. The MS-AR model applied so far assumed a constant transition probability. There were efforts to extend the model to the time-varying transition probability. Among them, Diebold et al.³⁸ and Filardo³⁹ criticized the constant transition probability to be too restrictive for empirical settings. Later studies have developed the application of the MS model to the time-varying transition probability function. These applications are based on a mean change but the constant variance.

Apart from the MS-AR model, there were studies to apply the MS regression to the dynamic linear models. For instance, Shumway and Stoffer⁴⁰ applied a dynamic linear model that switch endogenously based on the independent random process. Other studies that combined the dynamic effect into the MS framework include Kim and Yoo,⁴¹ Chauvet,⁴² and Xaba et al.⁴³

Next, considering the data with heteroscedasticity such as stock prices, there are studies to apply a change in variance with no AR effect. For instance, Kim and Nelson⁴⁴ demonstrated a three-stateregime-switching regression with regime heteroskedasticity (switching of variances) to a stock market data. Later, there were attempts to apply MS regression to the volatility models such as ARCH (Hamilton and Susmel,⁴⁵⁾ GARCH (Klaassen⁴⁶ and Brunetti et al.,⁴⁷⁾ VAR models (Mandilaras and Bird.⁴⁸⁾ More recently, a limited number of studies extend the MS regression to the nonlinear models such as threshold autoregressive (TAR), self-exciting TAR (SETAR) and smooth transition model (STAR). For instance Giordani et al.⁴⁹ Clements and Krolzig⁵⁰ and Elliott et al.⁵¹ These studies involve more complicated estimation procedure and there is still space forimprovement. For instance, Elliott et al.⁵¹ faced the convergence issue as certain datasets did not converge.

From the above review, MS regression is applied broadly and is useful in modeling structural breaks and handling outliers. It has been applied to different models and specifications. However, there is not yet any attempt to incorporate the nonlinearity/ asymmetric effect in the MS regression through the data decomposition approach which enables the estimation to capture the asymmetric effects, for example the effect of oil price increases versus its decreases. This study will contribute to this way.

3 Background study – inflation in Malaysia

Domestic prices (CPI, PPI, and IPI) of Malaysia are sensitive to internal and external influences. The main internal influences could be in terms of changes in economic structure and the business cycle, governmental policy/ economic plan, and the behavioral of economic agents. The economy of Malaysia went through evolutions from agricultural to manufacturing, and service-based. Such change in the economic structure affects the weighting of CPI and PPI baskets. The weighting on food has declined substantially while that of housing, transport, and communication has risen. For instance, the weighting for CPI food has declined from 34.9% (1994) to 30.2% (2010) while housing has risen from 21.1% (1994) to 23.8% (2010). Also, the weighting for PPI food and crude materials experienced large declines but that of fuels, chemical, manufacturing, and transports are among the sectors that experienced increases as observed from the press release of the Department of Statistics Malaysia for multiple years. Such reallocation of weighing may affect the measurement of price indices.

The governmental policies are also very influential. Among the main governance policies include the administered price mechanism, petrol subsidy, and taxes. The administered price mechanism helps to control the prices of necessary goods to ensure stable living cost do not burden the low and middle-income groups. According to the Bank Negara Malaysia Annual Report (2010),⁵² the scheme covered 29.3% of the CPI basket in Malaysia. Apart from this, subsidies for vehicle fuel, sugar, and cooking gas are implemented to reduce the burden of cost living. Among all subsidies, fuel subsidy is the main portion which covered 8.9% and 3.65% of total government expenditure and GDP respectively in 2008 (Hamid & Rashid).⁵³ The subsidies on fuel were ended on the 1^st of December 2014 and the price of fuel was replaced by a managed float mechanism. However, when the new government took over the lead, there were changes in the fuel subsidy since May 2018. The new government also re-introduced the Sales and Service Tax (SST) to replace the former Goods and Services Tax (GST). Before GST, Malaysia levied a SST. The change of taxes/ governmental policies caused uncertainties and price fluctuations.

Apart from the internal influences, the domestic prices of Malaysia are also highly affected by the global shocks/ crises including oil price shock. The global oil price has experienced sharp increases and sudden declines dues to episodes of the oil crisis. Figure 1 shows the plots of CPI sectoral indices in Malaysia. All CPI sectors do not show a very close pattern to an oil price index. This is because oil price might not affect the consumer goods and services directly but its impacts on PPI and IPI could be more direct as oil and energy are the main sources used in production. The closer relationship between oil price and PPI, and IPI can be observed in Figures 2 and 3.

Figure 1.

Plots of sectoral of CPI Malaysia.

Figure 2.

Plots of sectoral of IPI Malaysia.

Figure 3.

Plots of sectoral PPI Malaysia.

Figure 2 shows the plots of three IPI sectors of Malaysia. It is observed that the PPI of the electricity and manufacturing sectors is more reactive to oil price movement than the mining sector. The disruptions were around the 2000s. Both electricity and manufacturing are energy-intensive sectors. Therefore, these sectors are sensitive to oil price changes. Figure 3 are the plots of the sectoral PPI of Malaysia. PPI sectors (chemicals/ CHEMI, fuel mineral/ FUEL, crude food/ CRUDE, and animals & vegetable oils (ANI)) experienced sharp decreases coincided with big drops in the oil price (mid-1980, late-2000, and mid-2010). These sectors are highly linked to oil and energy and are more affected by oil price changes as they are energy-intensive sectors.

4 Data

The study is focused on Malaysia. The dependent variables include consumer price index (CPI), industrial production index (IPI), and producer price index (PPI) across sectors. These price indices data are collected from the Department of Statistics Malaysia. There are 9 sectors of CPI, 3 sectors of IPI, and 9 sectors of PPI. CPI is measured on base year 2000 = 100, PPI base year 2010 = 100, and IPI base year 2005 = 100. The data are collected in annual frequency from 1980 to 2019 as summarized in Table 1.

Table 1.
Data description.

Data Description

Sectoral price inflation (%) – dependent variable

CPI

DLCPI_FOOD Food

DLCPI_BEVE Beverages & tobaccos

DLCPI_CLOTH Clothing & footwear

DLCPI_FUEL Gross rent, fuel & power

DLCPI_FUR Furniture, furnishings & household equipment & operation

DLCPI_HEA Medical care & health expenses

DLCPI_TRANS Transport & communications

DLCPI_REC Recreation, entertainment education & cultural services

DLCPI_MISC Miscellaneous goods & services

IPI

DLIPI_ELEC Electricity sector

DLIPI_MANU Manufactured sector

DLIPI_MINING Mining sector

PPI

DLPPI_ANI Animal & vegetable oils & fats

DLPPI_BEVE Beverages & tobacco

DLPPI_CHEMI Chemicals

DLPPI_CRUDE Crude material – inedible

DLPPI_FOOD Food

DLPPI_FUEL Mineral, fuel & lubricants

DLPPI_MANU Manufactured goods

DLPPI_MISC Miscellaneous manufactured articles

DLPPI_TRANS Machinery & transport equipment

Explanatory factors

DLGDP GDP growth (%)

DLRER Changes in real effective exchange rate (%)

I Real interest rate (%)

M Broad money (% of GDP)

DLOIL Oil price changes (%)

DLOIL+ Oil price increases (%)

DLOIL- Oil price decreases (%)

Data	Description
	Sectoral price inflation (%) – dependent variable
	CPI
DLCPI_FOOD	Food
DLCPI_BEVE	Beverages & tobaccos
DLCPI_CLOTH	Clothing & footwear
DLCPI_FUEL	Gross rent, fuel & power
DLCPI_FUR	Furniture, furnishings & household equipment & operation
DLCPI_HEA	Medical care & health expenses
DLCPI_TRANS	Transport & communications
DLCPI_REC	Recreation, entertainment education & cultural services
DLCPI_MISC	Miscellaneous goods & services
	IPI
DLIPI_ELEC	Electricity sector
DLIPI_MANU	Manufactured sector
DLIPI_MINING	Mining sector
	PPI
DLPPI_ANI	Animal & vegetable oils & fats
DLPPI_BEVE	Beverages & tobacco
DLPPI_CHEMI	Chemicals
DLPPI_CRUDE	Crude material – inedible
DLPPI_FOOD	Food
DLPPI_FUEL	Mineral, fuel & lubricants
DLPPI_MANU	Manufactured goods
DLPPI_MISC	Miscellaneous manufactured articles
DLPPI_TRANS	Machinery & transport equipment
	Explanatory factors
DLGDP	GDP growth (%)
DLRER	Changes in real effective exchange rate (%)
I	Real interest rate (%)
M	Broad money (% of GDP)
DLOIL	Oil price changes (%)
DLOIL+	Oil price increases (%)
DLOIL-	Oil price decreases (%)

The oil price index is based on the West Texas Intermediate (WTI) price, Dollars per barrel. Interest rate, broad money, real effective exchange rate, CPI, and GDP are collected from the World Bank while the oil price data is collected from the Federal Reserve Bank of St Louis. All data are transformed into natural log form (except interest rate and broad money) and the rate of change or inflation rate is calculated as the log difference of the series (CPI, PPI, IPI, OIL, GDP, RER).

All the sectoral CPI, IP, and PPI indices are transformed into natural log differences to proxy for domestic price inflation (treated as dependent variables). There are six explanatory variables (DLGDP, DLRER, I, M, DLOIL +, DLOIL-). DLGDP or GDP growth is included following the economic theory (for instance, Keynesian theory and open economy model) in modeling the Phillips curve or price changes. There is a continuous debate on the relationship. In general, economists accept the view of the New Keynesian hypothesis that there is a short-run trade-off between inflation and output (i.e. high output is associated with high inflation) and weak or no relationship in the long run. But there is no clear distinction between short- versus long-run trade-offs although the debate is still going on. Sim⁵⁴ provided a review of the literature. DLRER, I, and M are included as explanatory variables as they are monetary policy tools. Since price stability is one of the main policy targets highlighted by policymakers, these variables might have some impacts on inflation. The relationship between these three variables with inflation is explained in economic theory in which the change in the exchange rate, policy rate, and money supply might influence the price stability as the value of the domestic currency might change by adjusting these variables. According to the classical economic theory, inflation is a monetary phenomenon (Friedman.⁵⁵⁾ The Quantity Theory of Money postulated a positive relationship between money supply and price level is given other factors stay constant. On the other hand, the opposed theory (the Keynes theory) believed that aggregate demand (through fiscal policy and output) is the main determinant of economic fluctuation and inflation. It is expected that money (M) will have a positive relationship with inflation, while the policy rate (I), will have a negative relationship with inflation.

According to Dornbusch,⁵⁶ exchange rate fluctuations can have a vital impact to affect prices in emerging economies. In general, appreciation of the domestic currency is associated with a lower price level. However, the net effect might depend on how the economic variables react to exchange rate changes. The exchange rate can affect inflation through different channels such as through imported input cost channel, real balance channel, relative price, and wage.

In order to examine the asymmetric effect of oil price changes, we decompose the oil series into oil increases and decreases series, following the approach by Ibrahim.⁵

LOIL + = \sum_{j = 1}^{t} {(DLOIL +)}_{j} = \sum_{j = 1}^{t} max (DLOI L_{j}, 0)

(1a)

LOIL - = \sum_{j = 1}^{t} {(DLOIL -)}_{j} = \sum_{j = 1}^{t} min (DLOI L_{j}, 0)

(1b)

DLOIL + = (LOIL +)_{t} - (LOIL +)_{t - 1}

(2a)

DLOIL - = (LOIL -)_{t} - (LOIL -)_{t - 1}

(2b)

5 Methodology

In this study, the Markov-switching regression is applied. For a robustness comparison, the study will perform the MS-AR and MS-DR models for both time-variant and time-invariant specifications. Three models are considered. Model I is the model with no lag term, with a mean regime-dependent term. Model II is the model as applied in Hamilton³¹ to capture the business cycle fluctuation with AR term while Model III is the model with dynamic effect to consider for gradual adjustment in economic variables. Model IV is as in Kim, Nelson, and Startz⁴⁴ to handle the high volatility/ heteroscedastic data, with mean zero and the variance is regime-dependent. These models although are very simple, can explain the real economic condition (for example business cycle fluctuation) very well.

Model I: no lag term

y_{t} = μ_{s_{t}} + ε_{t}

Model II: AR(1) model

y_{t} = μ_{s_{t}} + ϕ_{1} (y_{t - 1} - μ_{s_{t - 1}}) + ε_{t}

Model III: DR(1) model

y_{t} = μ_{s_{t}} + ϕ_{1} y_{t - 1} + ε_{t}

Model IV: heteroskedastic model

\begin{aligned} y_{t} = ε_{t} \\ ε_{t} \sim N (0, σ_{s_{t}}^{2}) \end{aligned}

s_{t} = j, (for j = 1, 2)

is the latent indicator to the state of regime at time t, takes the value either 1 or 2, indicating the state of Regime 1 or 2. Regime 1 is the period with downward trending and is interpreted as low inflation regime while regime 2 is indicates to the upward trending regime and implies to high inflation equation. The error term

ε_{t}

is independent and identically standard normal distributed;

μ_{s_{t}}

is the mean of regime-switching indicator,

s_{t}

. Model I, II, and III are mean regime-dependent models where

μ_{s_{t}}

changes across regimes. For instance, Model II shows that

y_{t}

is evolved following the realization of

(y_{t - 1} - μ_{s_{t - 1}})

which is mean regime-dependent and is called as Markov-switching mean model (Krolzig)⁵⁷ while Model III is also a mean regime-dependent model but

y_{t}

is gradually adjusted by the lagged term (dynamic effect). Model IV is a variance regime-dependent model in which

y_{t}

is regime-dependence on its error term (heteroscedastic model).

5.1 Innovations

There are several innovations made on the original Markov-switching regression in Model I, II, and III which include:

Some explanatory/ control variables (x) are included and specified as non-regime switched variables

The MS model is added with the asymmetric terms to capture the oil price increases and decreases through data decomposition approach

Consider different specifications (Case I, II, and III) for time-variant and time-invariant analysis

In terms of innovations (1) and (2), the explanations can refer to the innovative model:

Model I*:

y_{t} = μ_{s_{t}} + β_{i} x_{i t} + α_{1} (DLOIL +)_{t} + α_{2} (DLOIL -)_{t} + ε_{s_{t}}

Model II*:

y_{t} = μ_{s_{t}} + β_{i} x_{i t} + α_{1} (DLOIL +)_{t} + α_{2} (DLOIL -)_{t} + ϕ (y_{t - 1} - μ_{s_{t - 1}}) + ε_{t}

Model III*:

y_{t} = μ_{s_{t}} + ϕ y_{t - 1} + β_{i} x_{i t} + α_{1} (DLOIL +)_{t} + α_{2} (DLOIL -)_{t} + ε_{t}

Model IV: heteroskedastic model

\begin{aligned} y_{t} = ϕ y_{t - 1} + β_{i} x_{i t} + α_{1} (DLOIL +)_{t} + α_{2} (DLOIL -)_{t} + ε_{t} \\ ε_{t} \sim N (0, σ_{s_{t}}^{2}) \end{aligned}

The dependent variable

y_{t}

is represented by the sectoral CPI, IPI, and PPI inflation price series while the explanatory variables are added to Model I, II and III respectively. All variables and parameters are as in Model I, II, and III except the newly added one.

x_{i t}

is the explanatory or control variables,

β_{i}

is the coefficient capturing the effects of explanatory variables, where i indicates to different factors. The details of variables can refer to Table 1.

(DLOIL +)_{t}

and

(DLOIL -)_{t}

are the oil price inflation increases and decreases series obtained using data decomposition as shown in equation (1a, 1b) and (2a, 2b).

α_{1}

and

α_{2}

captures the asymmetric effects of oil price increases and decreases respectively. The inclusion of DLOIL+ and DLOIL- has changed the linear standard model to the nonlinear asymmetric form. This new model can capture the asymmetric effects of oil price changes and the interpretation of results under different inflation regimes.

In terms of innovation (3), the MS models allow a change in mean or a change in variance or both. However, the original AR and DR models are based on a change in mean alone while the heteroscedasticity model is with a change in variance. The earlier studies assumed a time-invariant in the transition probability function. For better comparisons, we perform models with different specifications.

Case (A): change in mean only –Model I*, II*, III*

Case (B): change in variance only- Model IV*

Case (C): change in both mean and variance - Model I*, II*, III*

Model II* and II* are estimated applicable to Case (A) and (C) as they are mean regime-dependent models (must have a change in mean) while Model I* is applicable for Case (B) (with zero mean and a change in variance) using the CPI, IPI, and PPI sectoral inflation data. A comparison of results is made between time-variant and time-invariant in transition probability function.

6 Results

Before conducting the regression, all variables are tested for their stationarity via unit-root tests (Phillips-Perron (PP) and Augmented Dicky-Fuller (ADF)). All variables are stationary as reported by both PP and ADF tests, i.e. reject the null hypothesis of unit-root (not stationary). After confirming the stationarity of all variables, we proceed with the Markov-switching estimation.

The analysis involves several steps. First, three models are estimated under different cases based on time-invariant transition probability. Model II* and III* are estimated under Case I and III while Model I* is estimated under Case II. Then Case I and III will select the best model between Model II* and III respectively. Each case will select the best model. Second, three cases with their best models will be compared to identify the final best model. Third, the final best model will be estimated using time-variant transition probability and results will be compared between time-variant and time-invariant specifications. The performances of models are evaluated based on the forecast error indicators of MAE, RMSE, and bias proportion. The model with the lowest error indicators will be selected as the best model. The best model for each sector's price inflation is summarized in Table 2.

Table 2.
Best model.

Dependent variable Best model Dependent variable Best model

DLCPI_BEVE III(Case A) DLPPI_ANI I(Case A)

DLCPI_CLOTH III(Case A) DLPPI_BEVE I (Case C))

DLCIP_FOOD III* (Case A) DLPPI_CHEMI I* (Case C)

DLCPI_FUEL III* (Case A) DLPPI_CRUDE III(Case C)

DLCPI_FUR III (Case A) DLPPI_FOOD I(Case A)

DLCPI_HEA III (Case A) DLPPI_FUEL I* (Case A)

DLCPI_MISC III* (Case A) DLPPI_MANU III* (Case A)

DLCPI_REC III* (Case A) DLPPI_MISC I* (Case A)

DLCPI_TRANS III(Case C) DLPPI_TRANS I (Case C)

DLIPI_E I(Case C)

DLIPI_M III(Case C)

DLIPI_MI I*(Case C)

Dependent variable	Best model	Dependent variable	Best model
DLCPI_BEVE	III*(Case A)	DLPPI_ANI	I*(Case A)
DLCPI_CLOTH	III*(Case A)	DLPPI_BEVE	I* (Case C))
DLCIP_FOOD	III* (Case A)	DLPPI_CHEMI	I* (Case C)
DLCPI_FUEL	III* (Case A)	DLPPI_CRUDE	III*(Case C)
DLCPI_FUR	III* (Case A)	DLPPI_FOOD	I*(Case A)
DLCPI_HEA	III* (Case A)	DLPPI_FUEL	I* (Case A)
DLCPI_MISC	III* (Case A)	DLPPI_MANU	III* (Case A)
DLCPI_REC	III* (Case A)	DLPPI_MISC	I* (Case A)
DLCPI_TRANS	III*(Case C)	DLPPI_TRANS	I* (Case C)
DLIPI_E	I*(Case C)
DLIPI_M	III*(Case C)
DLIPI_MI	I*(Case C)

6.1 Results of MS regression

Table 3(a, b, and c) summarize the results of the best MS model (time-invariant and time-variant) for the sectoral CPI, IPI, and PPI inflation respectively. Referring to the CPI inflation, the recreation sector shows the lowest value in the coefficient of mean regime-dependent µ in both regimes, indicating that the inflation rate is relatively lower compared to other sectors. On the other hand, beverages show the highest value in µ in Regime II. CPI inflation exhibits a gradual adjustment with the dynamic effect, especially in the food, clothing, and fuel sectors. These sectors show a larger value in the dynamic effect coefficient (DR(-1)), with slower adjustments in prices. The consumer prices of these sectors are stable and rigid as the main products (such as rice, sugar, oil, and petrol) are under price controls. The LOGSIGMA is the standard deviation in log form, a larger negative of LOGSIGMA indicates to smaller variance or lower fluctuation.

The oil price has a limited or small effect on CPI as its impact is not directly at the consumer level but directly on the production level. According to Eilyn et al.⁵⁸ changes in commodity prices may affect domestic inflation either directly or indirectly. The direct effect is through the pass-through into final goods and services. The indirect effect is through the change of cost pressures which may influence the pricing decision of final goods. In other words, the indirect effect of commodity prices is through changes on production cost The change of commodity prices may pass through into producer price at the first level/ immediately then from the production level pass-through into final consumer prices as a second level/ indirect effect. From here, the relationship between commodity price and producer price (PPI/ IPI) is direct or very high as compared to the commodity-CPI relationship which is indirect. For instance, Long and Liang²⁷ found that the impact of oil price on PPI is much larger than that on CPI. Our results also report the same condition. In terms of IPI inflation, oil price increases have a larger impact in electricity and manufacturing sectors than its decreases but the effect is not significant in the mining sector. Both electricity and manufacturing sectors are more oil-intensive in production and yet receive direct and larger impact compared to CPI inflation which is indirect. Higher oil price increases lead to higher IPI inflation in the manufacturing sector as the cost of production increases. But the effect in the electricity sector is reversed, possibly due to the price control policy and subsidy imposed by the government.

In all three price indices, the effect of oil price changes is asymmetric, i.e. the effect of oil price increases is different from that of decreases. The impact also varies across sectors. For instance, oil price changes have no significant effect on less oil-intensive sectors (beverages, food, miscellaneous), and price-control & subsidized sectors (crude inedible, and transport). On the other hand, the PPI inflation in oil-intensive sectors like fuel, chemicals, and manufacturing sectors tend to increase in response to oil price increases and/ or decreases. This is because crude oil is used intensively in production. Changes in the oil price might cause fluctuation in production cost. The PPI of the animal & vegetable oils and fats sector (ANI) sector seems to react differently. Oil price increases trigger PPI in ANI to drop while oil price decreases induce a higher PPI of ANI. This seems that animal and vegetable oils and fats can substitute for crude oil and energy for certain usages. Vegetable oils can be an alternative to fuel diesel engine (Corsini et al.).⁵⁹

The real exchange rate has the largest determination effect on CPI with the most impacted sectors being food and beverages. In general, the exchange rate may affect price stability in several ways and appreciation of the domestic currency is associated with lower prices. First, if the price of imported substitute goods effect dominated, then appreciation of domestic currency implies higher domestic price relative to foreign good price, hence there is a shift of demand to foreign goods. Hence, export drops, causing deficits in trade balance and current account balance and this leads to lower prices. Next, appreciation of domestic currency also implies cheaper imported inputs and lower production costs and wages, hence lower consumer prices. However, exchange rate changes might also affect the expectation of traders and speculators on future pricing due to uncertainties of trading and foreign exchange. Hence, the net effect may depend on which effect dominates the outcome.

In the food and beverages sector, the sign is negative, where appreciation in the domestic currency relative to other traders leads to higher flow out with more spending and lower demand on domestic goods/ export, hence lower CPI. The opposite relationship is observed in the beverages sector due to the market uncertainties and expectations, leading to higher demand and prices on beverages. On the other hand, the main determinant to IPI inflation is GDP growth where increases in GDP growth stimulate higher IPI inflation. In terms of the main determinants of PPI inflation, the real exchange rate is the main factor contributing to lower PPI inflation in ANI, chemicals, crude, and manufacturing sectors. CPI inflation or the aggregate supply is the main factor that causes higher PPI inflation in the sector of food, chemicals, food, fuel, and miscellaneous.

As claimed by many studies, emerging markets are vulnerable to exchange rate fluctuation and they are always termed to exhibit the behavior of ‘fear of floating’, i.e. the policymaker in emerging markets tend to intervene in the foreign exchange rate to reduce the volatility of exchange rate, leading the currency to be not freely floating as claimed. The reason that emerging markets are highly impacted by exchange rate volatility is due to the economic structure such as being very open in trades but financially not stable or undeveloped. Hence, exchange rate movements might affect the trades and balance of payment, as well as the cost of production and budget balance. Besides, most of the trades and international borrowings are denominated in dollar currency. The change in the exchange rate might affect financial stability and liability. Malaysia is a small economy but highly open in trades. Besides, according to Bank Negara,⁶⁰ Malaysia has adopted the managed floating system since 2005, which allows for government intervention to monitor the currency, but is subject to market conditions and economic fundamentals when the currency is too volatile.

Apart from the main influences from the real exchange rate, other tools of monetary policy (real interest rate and money supply) are also influential. The tightening of monetary policy of lifting real interest rate and the loosen policy of increasing money supply leads to lower and higher domestic price inflation respectively in majority sectors but the influences are small. This result holds for CPI, IPI, and PPI inflation. To sum up, monetary policy is effective in controlling domestic price inflation with the most impactful tool through real exchange rates. Fiscal policy (petrol and necessity goods subsidies) and price control action may also affect price stability. Although oil price changes induced domestic price fluctuation at producer level (IPI and PPI) with larger impact, the effect pass-through to consumer level (CPI) is limited or partial and does not lead to higher CPI inflation due to the price control scheme and subsidy allocated by the government. However, the cease of the fiscal policy/ subsidies might cause the tendency of domestic price inflation as shown in the results in the next section below.

6.2 Results of transition probability and expected duration

Next, the trend and behavior of CPI inflation can be studied from the transition probability and expected duration of inflation. The results are summarized in Table 4(a, b, and c) for CPI, IPI, and PPI respectively. Transition probabilities P11 and P22 indicate the transition of probability to remain in Regime 1 (low inflation) and Regime 2 (high inflation) respectively while P21 and P12 are the transitions of the probabilities between both regimes. The results show that all sectors exhibit a high probability in p11 and P21, indicating high probabilities to remain or transit to Regime 1 (low inflation). All these imply that domestic price inflations tend to remain in the low inflation regime. The exceptions are found in the CPI recreation and transportation sectors and the PPI animals and vegetables oils & fats (ANI), and fuel (FUEL) sectors. The ANI sector is very reactive to real exchange rate and oil price changes while the fuel sector shows fluctuations in prices probably due to the cease of the subsidy on sugar, cooking gas, and petrol since 15 July 2010 as a reformation plan to improve the balance sheet (Malaysiakini).⁶¹ Therefore, prices in these two sectors should be monitored so that prices remain stable.

6.3 Policy implication and recommendation

Overall, the results have evident on the asymmetric effect of oil price changes on domestic prices inflation with larger impacts felt in oil-intensive sectors. Apart from the oil intensity factor, the influence of policy also matters. For instance, the sectors that are listed under price control products and with subsidies show lower price volatility and are less affected by oil price changes. However, the change in the policy action and decision might cause price fluctuations and inflationary shocks. For instance, the cuts in the subsidies for fuel, sugar, and cooking gas since 15 July 2010, the replacement of fuel subsidies with a managed float fuel price mechanism in May 2018 and the re-implementation of Sales and Service Tax (SST) to replace the former Goods and Services Tax (GST) in 2018 caused price uncertainty and fluctuation as the market needs time to adapt and adjust to the new policy and operating system. The price control and subsidy schemes are among good options to monitor the stability of prices, in addition to taking care of the life quality and welfare of the public especially the lower-income group. However, such schemes might be affected by budget constraints. Hence the implementation of these schemes should follow by a proper operating/ monitor plan and backups to ensure the effectiveness of the policy. Such price control and subsidy schemes should co-exist with the effective monetary policy to enhance price stability as results show that monetary policy action through policy rate and money supply adjustments are effective to control price stability. All these are the shorter-term policy action in which the policy decision could be revised and updated by considering the change in the economic situation and budget constraint.

However, in the longer-term, the authority should have a long-run economic objective and economic planning, such as investment in green technology, regional trade collaboration, and technology transfer agreement with trade partners/ neighbor countries. The government should initiate the development of green technology and its usage as a long-term economic planning. This does not only help to economic development and investment, but also in the attempt in seeking of alternatives and the shift from oil products to renewable energy. This helps to reduce the oil dependency. Setting a long-term goal and initiative for economic development is important to improve the economic structure/ financial system so that the economy is less impacted by external shocks such as oil price and exchange rate shocks.

7 Conclusion

This paper seeks to examine and capture the asymmetric effects of oil price increases and decreases in affecting the domestic price inflation (CPI, IPI, and PPI) in Malaysia. This paper utilizes the Markov-switching regression in studying the dynamic behavior of CPI, IPI, and PPI inflations in low versus high inflation regimes. To further explore the asymmetric effects of oil price changes, we decompose the oil data into oil price increases and decreases series and incorporate them in the Markov-switching models. Our results reveal asymmetric effects of oil price changes on CPI, IPI, and PPI inflation. The effects differ across sectors with more significant impacts found in oil-intensive or related sectors and larger influences on IPI and PPI relative to CPI. However, the oil price is not the main factor but other factors such as aggregate supply, aggregate demand, and real exchange rate are the main determinants that greatly influence the domestic price changes. Apart from this, the monetary policy is found to be effective to monitor domestic prices inflation across sectors.

The study has identified a few sectors to tend to stay at the high inflation rate such as CPI transportation, PPI animals & vegetable oils, and PPI fuel. Hence, the policymaker plays a crucial role in maintaining price stability. An effective policy can reduce the impacts of external shocks such as oil price fluctuations as well as create a low-inflationary environment. The policymaker should monitor the price stability of these sectors through respective policy actions and remedies. Finally, the findings also provide policy evaluation for continuous improvement. Since the implementation of monetary policy tools (real interest rate, real interest rate, and money supply) and the fiscal tool through subsidies are effective to control domestic prices inflation, the monetary policy should co-exist with the fiscal policy to manage the price stability continuously. Perhaps the price regulations could also help to enhance healthier business and policy operations.

Footnotes

Declaration of conflicting interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Research University Grant Universiti Sains Malaysia (grant number 1001 / PMATHS / 8016115).

ORCID iD

Siok Kun Sek

Appendix

Table 4(c):

Transition probability- PPI inflation

Transition prob.	Time-invariant			Time-variant
Transition prob.	DLPPI_ANI	DLPPI_BEVE	DLPPI_CHEMI	DLPPI_ANI	DLPPI_BEVE	DLPPI_CHEMI
P11	0.9264	0.7999	0.6532	0.9674	0.6374	0.7408
P22	0.9715	0.1723	0.6777	0.0000	0.2912	0.4068
P21	0.0285	0.8277	0.3223	1.0000	0.7088	0.5932
P12	0.0736	0.2001	0.3468	0.0326	0.3626	0.2592
	DLPPI_CRUDE	DLPPI_FOOD	DLPPI_FUEL	DLPPI_CRUDE	DLPPI_FOOD	DLPPI_FUEL
P11	0.9093	0.7689	0.0000	0.9188	0.7695	0.0906
P22	0.3883	0.4890	0.9685	0.3973	0.4587	0.8247
P21	0.6117	0.5110	0.0315	0.6027	0.5413	0.1753
P12	0.0907	0.2311	1.0000	0.0812	0.2305	0.9094
	DLPPI_MANU	DLPPI_MISC	DLPPI_TRANS	DLPPI_MANU	DLPPI_MISC	DLPPI_TRANS
P11	0.0000	0.3938	0.8765	0.0000	0.4373	0.9466
P22	0.4536	0.0000	0.0049	0.5907	0.0000	0.2980
P21	0.5464	1.0000	0.9951	0.4093	1.0000	0.7020
P12	1.0000	0.6062	0.1235	1.0000	0.5627	0.0534

References

Salisu

Isah

Oyewole

, et al. Modelling oil price-inflation nexus: the role of asymmetries. Energy 2017; 125: 97–106.

Hamilton

. Understanding crude oil prices. Energy Journal 2009; 30: 179–206.

Valcarcel

Wohar

. Changes in the oil price-inflation pass-through. J Econ Bus 2013; 68: 24–42.

Herrera

Lagalo

Wada

. Asymmetries in the response of economic activity to oil price increases and decreases? J Int Money Finance 2015; 50: 108–133.

Ibrahim

. Oil and food prices in Malaysia: a nonlinear ARDL analysis. Agricultural and Food Economics 2015; 3: 1–14.

Karantininis

Katrakylidis

Persson

. Price transmission in the Swedish pork chain: Asymmetric non linear ARDL. In EAAE 2011 Congress, Change and Uncertainty Challenges for Agriculture, Food and Natural Resources. European Association of Agricultural Economists Zurich, Switzerland, 30 August – 2 September, 2011. https://econpapers.repec.org/paper/agseaae11/114772.htm.

Sek

. Impact of oil price changes on domestic price inflation at disaggregated levels: evidence from linear and nonlinear ARDL modelling. Energy 2017; 130: 204–217.

Choi

Furceri

Loungani

, et al. Oil prices and inflation dynamics: evidence from advanced and developing economies. J Int Money Finance 2018; 82: 71–96.

Ibrahim

Said

. Disaggregated consumer prices and oil price pass-through: evidence from Malaysia. China Agricultural Economic Review 2012; 4: 514–529.

10.

Hansman

McConnachie

Wollersheim

, et al. The impact of oil prices on the air transportation industry. National Center of Excellence for Aviation Operations Research, 2014. https://isr.umd.edu/NEXTOR/pubs/NEXTOR-II-Oil-Impact-3-2014.pdf

11.

Álvarez

Hurtado

Sánchez

, et al. The impact of oil price changes on Spanish and euro area consumer price inflation. Econ Model 2011; 28: 422–431.

12.

Country analysis executive summary: Malaysia. U.S. Energy Information Administration, 25 January 2021.

13.

Crude petroleum in Malaysia. The Observatory of Economic Complexity. https://oec.world/en/profile/bilateral-product/crude-petroleum/reporter/mys.

14.

De Gregorio

Landerrectche

Neilson

. Another pass-through bites the dust? Oil prices and inflation. Central Bank of Chile, 2007. https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.497.8411&rep=rep1&type=pdf.

15.

Chen

. Oil price pass-through into inflation. Energy Economics 2009; 31: 126–133.

16.

Hooker

. Are oil shocks inflationary? Asymmetric and nonlinear specifications versus changes in regime. Journal of Money. Credit Bank 2002; 34: 540–561.

17.

Filis

Chatziantoniou

. Financial and monetary policy responses to oil price shocks: evidence from oil-importing and oil-exporting countries. Review of Quantitative Finance and Accounting 2014; 42: 709–729.

18.

Gelos

Ustyugova

. Inflation responses to commodity price shocks – How and why do countries differ? J Int Money Finance 2017; 72: 28–47.

19.

Bala

Lee

Maijama’a

. Asymmetric pass-through effects of oil price on economic growth in Malaysia. International Journal of Business and Society 2021; 22: 753–764.

20.

Yang

T-Y

Thanh

Yen

C-W

. Oil threshold value between oil price and production. Panoeconomicus 2019; 66: 487–505.

21.

Donayre

Wilmot

. The asymmetric effects of oil price shocks on the Canadian economy. International Journal of Energy Economics and Policy 2016; 6: 167–182.

22.

Ozdemir

Akgul

. Inflationary effects of oil prices and domestic gasoline prices: Markov-switching –VAR analysis. Ege Akademik Bakis Dergisi 2015; 15: 367–378.

23.

Pagliacci

Barraez

. A markov-switcing model of inflation: looking at the future during uncertain times. Análisis Económico 2010; XXV: 25–46.

24.

Behera

Patra

. Measuring trend inflation in India. RBI Working Paper Series No. 15, Reserve Bank of India, 2020.

25.

Chou

Lin

. Oil price shocks and producer prices in Taiwan: an application of non-linear error-correction models. Journal of Chinese Economics and Business Studies 2013; 11: 59–72.

26.

Bala

Chin

. Asymmetric impacts of oil price on inflation: an empirical study of African OPEC member countries. Energies 2018; 11: 1–21.

27.

Long

Liang

. Asymmetric and nonlinear pass-through of global crude oil price to China’s PPI and CPI inflation. Economic Research Ekonomska Istra- zivanja 2018; 31: 240–251.

28.

Babuga

Mohd Naseem

. Asymmetric effect of oil price change on inflation: evidence from Sub Saharan African countries. International Journal of Energy Economics and Policy 2021; 11: 448–458.

29.

Lily

Kogid

Nipo

, et al. Oil price pass-through into inflation revisited in Malaysia: the role of asymmetry. Malaysian Journal of Business and Economics 2019; Special Edition (2): 129–139. https://jurcon.ums.edu.my/ojums/index.php/mjbe/article/view/2123/1569

30.

Xuan

Chin

. Pass-through effect of oil price into consumer price: an empirical study. International Journal of Economics and Management 2015; 9: 143–161.

31.

Razmi

Azali

Chin

, et al. The role of monetary transmission channels in transmitting oil price shocks to prices in ASEAN4 countries during pre- and post-global financial crisis. Energy 2016; 101: 581–591.

32.

Hamilton

. A new approach to the economic analysis of non-stationary time series and the business cycle. Econometrica 1989; 57: 357–384.

33.

Lindgren

. Markov regime models for mixed distributions and switching regressions. Scandinavian Journal of Statistics 1978; 5: 81–91.

34.

Baum

Petrie

Soules

, et al. A maximization technique occurring in the statistical analysis of probabilistic functions of markov chains. Annals of Mathematical Statistics 1980; 41: 164–171.

35.

Poritz

. Linear predictive hidden markov models and the speech signal. Acoustics, speech and signal processing. IEEE Conference on ICASSP ’82, 1982: 1291–1294. Doi: 10.1109/ICASSP.1982.1171633.

36.

Rabiner

. A tutorial on hidden Markov models and selected applications in speech recognition. Proc IEEE 1989; 77: 257–286.

37.

Hamilton

Raj

. New directions in business cycle research and financial analysis. Empir Econ 2002; 27: 149–162.

38.

Diebold

Lee

Weinbach

. Regime switching with time-varying transition probabilities. In: Hargreaves

(eds) Nonstationary time series analysis and cointegration. Oxford University Press, 1994, pp.283–302.https://www.sas.upenn.edu/~fdiebold/papers/paper69/pa.dlw.pdf

39.

Filardo

. Business-cycle phases and their transitional dynamics. J Bus Econ Stat 1994; 12: 299–308.

40.

Shumway

Stoffer

. Dynamic linear models with switching. Journal of the American Statistical Asociation 1991; 86: 763–769.

41.

Kim

Yoo

. New index of coincident indicators: a multivariate markov switching factor model approach. J Monet Econ 1995; 36: 607–630.

42.

Chauvet

. An econometric characterization of business cycle dynamics with factor structure and regime switches. Int Econ Rev (Philadelphia) 1998; 39: 969–996.

43.

Xaba

Moroke

Rapoo

. Modeling stock market returns of BRICS with a moarkov-switching dynamic regression model. Journal of Economics and Behavioral Studies 2019; 11: 10–22.

44.

Kim

Nelson

Startz

. Testing for mean reversion in heteroskedasticdata based on Gibbs-sampling-augmented randomization. Journal of Empirical Finance 1998; 5: 131–154.

45.

Hamilton

Susmel

. Autoregressive conditional heteroskedasticity and changes in regime. J Econom 1994; 64: 307–333.

46.

Klaassen

. Improving GARCH volatility forecasts with regime-switching GARCH. Empir Econ 2002; 27: 363–394.

47.

Brunetti

Scotti

Mariano

, et al. Markov switching GARCH models of currency turmoil in Southeast Asia. Emerging Markets Review 2008; 9: 104–128.

48.

Mandilaras

Bird

. A markov switching analysis of contagion in the EMS. J Int Money Finance 2010; 29: 1062–1075.

49.

Giordani

Kohn

van Dijk

. A unified approach to nonlinearity, structural change and outliers. J Econom 2007; 137: 112–133.

50.

Clements

Krolzig

H-M

. A comparison of the forecast performance of Markov-switching and threshold autoregressive models of US GNP. Econom J 1998; 1: C47–C75.

51.

Elliott

Siu

Lau

. A hidden markov regime-switching smooth transition model. Studies in Nonlinear Dynamics & Econometrics 2018; 22: 1–21.

52.

Economic Development in 2010, Annual Report 2011. Bank Negara Malaysia, 2011.

53.

Hamid

Rashid

. Cambodia’s electricity sector in the context of regional electricity market integration’. In: Wu

Shi

Kimura

(eds) Energy market integration in east Asia: theories, electricity sector and subsidies, ERIA research project report 2011-17. Jakarta: ERIA, 2012, pp.207–252.

54.

Sim

. A review on output-inflation trade-off based on New Classical and New Keynesian theories. MPRA Paper No. 105767, 12 Feb 2021.

55.

Friedman

. Studies in the quantity theory of money. 1st edn Chicago: University of Chicago Press, 1956.

56.

Dornbusch

. Expectations and exchange rate dynamics. J Polit Econ 1976; 84: 1161–1176.

57.

Krolzig

. Markov-switching vector autoregressions: modelling, statistical inference, and application to business cycle analysis. Springer Verlag, 1997. https://link-springer-com-443.web.bisu.edu.cn/book/10.1007/978-3-642-51684-9

58.

Eilyn

Ramlee

Tan

. Global commodity prices, inflation and monetary policy: A study of selected commodity exporters and importers. Working papers WP8/2017, Bank Negara Malaysia. 2017.

59.

Corsini

Marchegiani

Rispoli

, et al. Vegetable oils as fuels in diesel engine. Engine performance and emissions. Energy Procedia 2015; 81: 942–949.

60.

Bank Negara Malaysia. Malaysia adopts a managed float for the Ringgit exchange rate, 2005. http://www.bnm.gov.my/index.php?ch=en_press&pg=en_press&ac=904&lang=en.

61.

5-in-1 price hike: Fuel, sugar and gas up, Malaysiakini, 15 July 2010. https://www.malaysiakini.com/news/137433