1 Introduction

As greenhouse gases increase in the atmosphere, there are growing concerns about their negative impacts on the environment, climate change and global warming. The condensation of various greenhouse gases in the atmosphere led to many problems for the life on the earth, such as various diseases, hot weather, off-season floods, storms, forest fires, and so on. One of main contributors to GHG emissions, particularly CO2 emissions, is the transport sector (Solaymani 2019). In 2020, the transport contributes to 22.4% of the world's total CO2 emissions that generated by electricity and heat (43%) (IEA 2022).

To reduce the level of CO2 emissions, developed and many developing countries have implemented various adaptation policies and regulations since the early 1970s when global oil prices increased significantly. These policies and regulations include carbon and fuel taxes, emissions trading schemes, taxes on other gases and so on. The use of renewable energy commodities such as solar, wind, hydroelectricity, geothermal and others can also significantly reduce the level of greenhouse gas emissions.

Another category of renewable energy is bioenergy, which can be derived from biomass, such as biodiesel energy obtained from food crops, vegetable oils, and sugar crops. Palm oil as a biofuel, which can also be used for cooking oil, is used to produce bioethanol for use in transport fuel for diesel engines.Footnote 1 This bioenergy is used with fossil diesel in the transport sector as a type of fossil fuel. It has been proven that the use of renewable energy can reduce CO2 emissions in the atmosphere (Chen et al. 2019; Razmjoo et al. 2021).

The use of bioenergy with CO2 capture and storage technologies is one of the proposed policies to reduce CO2 emissions to meet the goals of the Paris Agreement (Mendiara et al. 2018). The predictions made by researchers and international agencies show the importance of achieving new energies for energy security and countries’ efforts to achieve a high share of renewable energy consumption in total energy consumption for sustainable development. For example, the International Energy Agency predicted that the share of various renewable energies in electricity production is expected to achieve 40–70% by 2050, compared to the current value, which is less than 10% (IEA 2021). It also predicted that bioenergy consumption in 2040 will increase by up to 10% of total energy consumption (IEA 2021).

Moreover, the share of modern bioenergy increases over time and will be more than doubled by 2030 (IEA 2021). It also predicted a very rapid increase in renewable energy consumption at an average annual growth rate of 2.3% during 2015–2040 (EIA 2017). On the other hand, biofuel production affects food production and supply because of the limitation of land supply, which affects the production and prices of agricultural commodities.

Munoz (2008) showed that transferring land and forests, which are already used for agricultural commodities, to biofuels reduces the availability of crops and livestock products and extremely affects social sustainability and food security in the future. These factors eventually increase feedstock and food prices (Zhang et al. 2010). Since there is no clear vision of the future of agricultural production, the potential of bioenergy for reducing climate change is unclear (Popp et al. 2014). Some studies have suggested that bioenergy production could improve food production and the economy of rural areas (Matemilola et al. 2019).

Bioenergy from biomass contributes significantly to the global energy system and can help in carbon reduction not only in the transport sector but also in manufacturing. Palm oil biomass possesses the remarkable ability to generate energy sources for the transportation industry that are not only cleaner and more cost-effective, but also abundantly available for future use (Kaniapan et al. 2021). The use of bioenergy is not the only way to reduce CO2 emissions from the transport and industry; rather, investment in a basket of CO2 removal technologies is the best option (Butnar et al. 2020).

Large-scale development of bioenergy can help to reduce global warming significantly (Baik et al. 2018; Creutzig et al. 2015). Gelfand et al. (2020) empirically showed that the direct impact of the use of bioenergy products, like ethanol, in the transport sector on CO2 emissions reductions is similar to that of using electric vehicles. Using sustainable forest biomass as an energy source (for heat, electricity, or transportation fuels) offers a promising solution to decrease our dependency on fossil fuels and reduce CO2 emissions in the short run (Cowie et al. 2021). Therefore, the studies above have clearly demonstrated the significant role that bioenergy plays in reducing CO2 emissions in the transportation sector. This highlights the crucial need for further research and exploration of bioenergy's potential in other regions.

Rapid economic growth and the increased need for transportation leads to greater demand for energy in Malaysia. The transport sector is the second largest contributor to CO2 emissions following electricity and heat production, of which 97.4% comes from road transport (IEA 2020). To reduce the level of these emissions, the government has implemented some policies and economic regulations.

In this country, palm biodiesel, as a blending program, has been used in the transport sector since June 2011. The government in Malaysia had planned to increase the share of biodiesel composition in current diesel combinations to 7%, 10%, and finally 15% by 2020 to decrease dependence on petro-diesel (Hossain et al. 2020; Wahab 2015). This blending mix was delayed until mid-2022 due to the world oil depreciation caused by COVID-19 pandemic.

Therefore, this study investigates the impact of biodiesel consumption on transport CO2 emissions in Malaysia as one of the largest producers of biodiesel production source, i.e. palm oil, and one of the palm biodiesel producers and consumers in the world. This study employed time series data from 1990 to 2019. This specific sample was selected because, in order to estimate a robust econometric model, time series of at least 30 years of data is required. It also included the duration that Malaysia produced biodiesel from palm oil. It used the traditional ARDL model, which we called the general ARDL model, and a new autoregressive distributed lag (ARDL) model, recognized as dynamic ARDL simulations. The latter method was used to predict the impact of the expected share of biodiesel in transportation fuel in 2020 (i.e., 10% = B10 program) on CO2 emissions from the road transport sector.

The present study contributes to the literature in the following ways. It filled the gap in the literature on the impact of biodiesel consumption on transport CO2 emissions in Malaysia. In addition, it provides an economic-based investigation and discussion on the impacts of biodiesel consumption in the transport sector as one of the major contributors to CO2 emissions. It also predicts the impact of the 10% increase in biodiesel consumption on transport CO2 emissions which has not been investigated in the literature. Moreover, the study investigates how increasing biodiesel consumption affected transport CO2 emissions over time, and if increasing consumption was effective in reducing transport CO2 emissions.

Below is how the rest of the study is managed. The next section looks at the literature on the impact of biodiesel consumption on the economy. Section 3 explains the ARDL methodology and data. A discussion on the results is presented in section 3.1. Section 3.2 concludes by summarizing the results and drawing some policy recommendations.

2 Literature review

Transport emissions increase as a result of economic growth, urbanization and the energy consumption in the road sector (Ahmed et al. 2020; Habib et al. 2021). Reducing energy intensity shows energy efficiency improvement that can reduce CO2 emissions in this sector (Solaymani 2020). Some studies showed that transport renewable energy consumption can reduce total CO2 emissions in the atmosphere (Dai et al. 2023). Not only the use of renewable energy can reduce CO2 emissions but energy efficiency and energy related policies also have meaningful impacts (Sun et al. 2021; Solaymani 2013). The consumption of renewable energy in European countries contributes to a substantial 12 percent reduction in CO2 emissions from transportation (Amin et al. 2020; Godil et al. 2021). Using electric vehicles could effectively reduce carbon emissions from private transportation, while also foster a beneficial synergy with the growing utilization of renewable energy sources (Bellocchi et al. 2018).

Studies on the importance of biofuels show that compared to conventional fossil fuels they reduce greenhouse gas emissions, but they are energy-intensive (Ližbetin et al. 2018). Bildirici (2017) used the auto-regressive distributed lag (ARDL) model which showed that biofuel consumption and economic growth had positive impacts on CO2 emissions in the United States. Based on an experimental investigation, Cassiers et al. (2020) found that the biodiesel concentration in European urban bus fuel increases nitrogen oxide (NOx) emissions and decreases the particle number (PN) and slightly decreases carbon monoxide (CO) emissions.

Using fully modified ordinary least squares, Xu et al. (2020) showed that the impact of biofuel se on carbon emission is negative. Qiao et al. (2016) used a panel data regression model showing that biofuel production contributes to economic growth in China. Similarly, by applying two panel data models for EU countries Simionescu et al. (2017) found that biodiesel consumption in transport positively affects economic growth and there is a two-way relation between them. The results of other panel data studies showed that biofuel consumption rises GDP and reduces the level of pollution (Al-Mulali 2015).

Using quantile regressions, Sharif et al. (2021) suggested that biofuel energy consumption reduces CO2 emissions in the United States of America (USA). Similarly, based on USA data and using wavelet coherence analyses, Bilgili et al. (2016) found that biomass usage reduces CO2 emissions over the long run. Likewise, based on the results of the ARDL technique, Kim et al. (2020) argued that if biomass consumption increases by 1%, CO2 emissions decrease by 0.65% in the long run in the USA. Schlamadinger and Marland (1996) also proved that forest and bioenergy strategies reduce CO2 emissions into the atmosphere. Using several economic indicators, Tan et al. (2019) showed that the application of biodiesel in the power sector in Malaysia can reduce CO2 emissions. Szulczyk et al. (2021) also showed that palm biodiesel is not cost-competitive with diesel and shows modest reductions in carbon emissions.

Evidence also showed that using some biofuels like wood and its substitution for coal increases CO2 emissions into the atmosphere (Sterman et al. 2018). Murta et al. (2021) argued that sugarcane and palm oil are the two most efficient crops with high energy densities that can reduce greenhouse gas emissions. Fulton et al. (2015) in their study emphasized the widespread use of biofuels to develop sustainable and low-carbon biofuels.

Based on the literature provided, it is apparent that the majority of studies relied on econometric techniques which fall short in accurately forecasting the effects of increased biodiesel consumption on transport-related CO2 emissions. There is a notable absence of studies that examine the impact of changes in biodiesel consumption on transportation CO2 emissions over time, as well as the actual effectiveness of increasing biodiesel consumption in reducing such emissions. Furthermore, there are few studies on the impact of biofuels on transport CO2 emissions and of course on the prediction of the impact of bio-biofuels. While many studies have delved into the world of science and engineering, there is a noticeable lack of attention given to economic-based analyses and discussions surrounding biodiesel consumption in the transport sector. Therefore, this study tries to expand the limited literature and investigate the impact of biodiesel on transport CO2 emissions to fill the above-mentioned gaps.

3 Methodology and data

3.1 Model construction and data

Belucio et al. (2022) evaluated the impacts of different fuels on CO2 emissions in Europe and introduced CO2 emissions as a function of natural gas consumption, renewable energy consumption, oil consumption, and GDP. Mahmood (2022) also introduced CO2 emissions as a function of GDP, natural gas consumption, and oil consumption for the Gulf Cooperation Council (GCC) countries. Furthermore, many studies analyzed the impacts of oil consumption, GDP, and other variables on CO2 emissions in developing countries (Alam et al. 2015; Khan et al. 2019). Shaari et l. (2020), using a panel ARDL model, also set up the CO2 emissions as a function of GPG, consumption of oil, and natural gas consumption. Other studies applied such modeling to the transport sector. For example, Alkhathlan and Javid (2015) specified CO2 emissions in Saudi Arabia’s transport as a function of its oil consumption.

In this study, we introduce CO2 emissions from road transport as a function of fossil fuels and biodiesel as the major contributors to CO2 emissions. This study investigates the link between road transport CO2 emissions and GDP from transport using the Granger causality test on the basis of the vector autoregressive (VAR) model. Therefore, the initial functional form of the main model of this study as a function of diesel, motor petrol,Footnote 2 and biodiesel, is given as follows:

$$CO_2 = f (motor\,petrol, diesel, biodiesel)$$
(1)

This study uses time series data that covers a period from 1990 to 2019. Figure 1 shows trends for all data used in the study in their logarithmic form.

Fig. 1
figure 1

Trends of the variables of the study

Full size image

3.2 Methodology

The natural logarithmic form of Eq. (1) was used in this study to eliminate the heteroscedasticity issue. The ARDL bounds testing method is based on the ordinary least squares estimation of an unrestricted error correction model to obtain co-integration between time series variables. The bounds testing equation is as follows:

$$\Delta {LCO}_{2t}={z}_{0}+{z}_{1}{L(C{O}_{2})}_{t-i}+{z}_{2}{L(motor\,petrol)}_{t-i}+{z}_{3}{L(diesel)}_{t-i}+{z}_{4}{L(biodiesel)}_{t-i}+\sum\nolimits_{i=1}^{r}{\theta }_{1}\Delta {L(C{O}_{2})}_{t-i}+\sum\nolimits_{i=1}^{r}{\theta }_{2}\Delta {L(motor\,petrol)}_{t-i}+\sum\nolimits_{i=1}^{r}{\theta }_{3}\mathrm{\Delta L}{(diesel)}_{t-i}+\sum\nolimits_{i=1}^{r}{\theta }_{4}\mathrm{\Delta L}{(biodiesel)}_{t-i}+{\varepsilon }_{t}$$
(2)

where Δ is the first-order difference, and t is the trend; ρ1-ρ4 and θ1-θ4 show the long-run and the short-run coefficients, respectively, and ɛ is the error term. r indicates the optimal lag length based on the Schwartz information criterion.

Three-steps are required to implement the ARDL bounds test method. The first step involves determining whether there is a long-run co-integration relationship between the variables of the model. The long-run link between the variables is checked based on the Wald coefficient test or the F test. In this study, a joint F test for the null hypothesis, i.e., no existence of cointegration, (H0: z1 = z2 = z3 = z4 = 0) against the alternative hypothesis (H1: z1 ≠ z2 ≠ z3 ≠ z4 ≠ 0) are used for Eq. (2). At this stage, according to the conventional significance levels (1, 5, and 10%), the estimated F-statistic was compared with the case of the critical bound values ​​provided in the Pesaran et al.'s (2001) study. If the value of the estimated F-statistic exceeds the higher limit of the critical value, the null hypothesis (i.e., co-integration does not take place) cannot be accepted. If the value of the estimated F-statistic is under the lower limit of the critical value, then the null hypothesis cannot be rejected. However, the decision will not be definitive if the F-statistic falls between the higher and lower limits. The second step contained estimating the coefficients of the long-run relationships and quantifying them. Since the model has a log–log form, these coefficients are elasticities. The second step is implemented only if a long-run relationship between the model’s variables exists in the first step. Finally, in the last step, short-term elasticities were estimated from the coefficients of the first difference of the variables in the ARDL model.

The ARDL model makes it possible to obtain a dynamic error correction model. This method provides direct and clear long- and short-run impacts of regressors with a specific lag length. It can also estimate the positive and negative impacts of one of the explanatory variables on the responding variable and analyze it graphically with other things held constant. The error correction model of the dynamic ARDL simulations model, which is obtained from the ARDL methodology, can be formulated as follows:

$$\Delta {CO}_{2t}={\alpha }_{0}+{\theta }_{0}\Delta {L(C{O}_{2})}_{t-1}+{\theta }_{1}\Delta {L(motor\,petrol)}_{t}+{\rho }_{1}{L(motor\,petrol)}_{t-1}+{\theta }_{2}\Delta L{(diesel)}_{t}+{\rho }_{2}{L(diesel)}_{t-1}+{\theta }_{3}\Delta L{(biodiesel)}_{t}+{\rho }_{3}{L(biodiesel)}_{t-1}+{\varepsilon }_{t}$$
(3)

In Eq. 3, ρi explains the dynamic short-run coefficients, while θi shows the long-run coefficients.

4 Results and discussion

Before estimating the main results of the model we needed to perform several pre-estimation tests. The first primary test was to check the stationary of the data with the unit root test. To avoid spurious regression, model variables need to be stationary and do not represent random walks. Otherwise, it is necessary to use the difference of variables that is usually stationary. If the variables used in the estimation of the model represent random walks, while there may be no logical relationship between the independent and dependent variables, the obtained coefficient of determination can be very high and cause the researcher to make incorrect inferences about the degree of relationship between the variables. In the unit root test, the null hypothesis is based on the existence of a single root and the opposite hypothesis is the stationary of at least one panel member.

After selecting the model variables and checking their stationarity, it is necessary to find an optimal lag length for all the model variables. The vector autoregressive (VAR) was used to determine the optimal lag length. In this methodology, it is essential to find an appropriate lag length and the optimal lag must not be too large (leading to a loss in degrees of freedom) or too small (causing the model’s specification error) (Enders 2014). Different criteria were used to determine the optimal lag length, as reported in Table 2. In each of these criteria, the optimal lag was selected in such a way that the degree of freedom was not lost, and the residual terms of the equations had no autocorrelation issues.

One of the characteristics of most time-series data is their stationary over time. Since this study used time series data, the possibility of spurious regression was not far from the expectation. Therefore, in the first step, the unit root test was performed. The results of the Augmented Dickey-Fuller (ADF) test proposed by Dickey and Fuller (1979) and the Phillips-Perron test proposed by Phillips and Perron (1988) are presented in Table 1.. The Phillips-Perron test is reported to support the results of the ADF test. Based on the results obtained, most variables were non-stationary at their level and became stationary at the first difference. The exception here is the coefficient of CO2 emissions, which was stationary at its level. We can conclude that the considered variables were integrated at the first degree and in this research, we could use the ARDL approach.

Table 1 Unit root test results
Full size table

To apply the bounds testing method, in the first step, we need to identify the optimal lag length using the unrestricted VAR model based on the Schwartz-Bayesian criterion. Table 2 presents the outcomes of the lag length selection and indicates that the optimal lag length was 1. After determining the optimal lag length, the bounds test was used to estimate the presence of a long-run link between CO2 emissions and the explanatory variables of the model.

Table 2 Lag selection criteria
Full size table

The bounds test results indicate that, in the linear symmetric ARDL model, the F-statistic calculated from the critical values ​​of the upper bound was even greater at a significant level of 1% and, therefore, the presence of a long-run link between the variables of the model with 99% confidence was accepted (Table 3).

Table 3 The results of the bounds test for cointegration
Full size table

The outcomes of the Granger test in Table 4 show that the GDP from transport was a cause of palm biodiesel consumption in the transport sector and the bilateral relationship did not occur. This is because the share of biodiesel production and consumption in the economy is very low. This result suggests that higher economic growth is necessary for Malaysia to increase biodiesel consumption in transport to improve environmental protection. Evidence argued that the contribution of biodiesel consumption to improving economic development depends on the factors such as more investment in equipment and plant, more income taxes, value-added to feedstock, and more rural manufacturing jobs (Demirbas 2007; Streimikiene et al. 2019; Solaymani 2022). Furthermore, diesel consumption, motor petrol consumption, and CO2 emissions are also the causes of palm biodiesel consumption in the transport sector. For Malaysia, biodiesel production does not seem to be a major problem because the palm oil industry, as the major source of biodiesel, is the largest contributor to agricultural value-added and the agricultural policy is oriented towards more palm oil production, which is consistent with the EU policy. The key challenge is to increase the use of biodiesel in the transport sector. The result of the causal link is not similar to the one achieved for economic growth and energy consumption in Malaysia (Li and Solaymani 2021) and also for Iran (Solaymani 2021). Moreover, there exists a causal relationship between CO2 emissions and diesel or motor petrol consumption and between diesel consumption and motor petrol consumption and transport CO2 emissions during the study period.

Table 4 Chi-square statistics of the VAR-Granger causality wald test (1987-2019)
Full size table

The values of the F statistic, FCO2 = f(CO2 | diesel, motor petrol, biodiesel) = 10.227, which are reported in Table 4, demonstrate that there exists a long-run cointegration relationship among the variables at the 1% level of significance, because its value exceeded the critical upper limit of the value proposed by Pesaran et al. (2001) (I(0) = 5.070, I(1) = 6.768). Therefore, we can estimate the ARDL model and find out the long- and short-run results.

Table 5 reports the long and short-run results of the study based on the ARDL model. The motor petrol variable had a positive coefficient that was statistically significant at the 1% level. It shows that motor petrol had a positive impact on carbon emissions from the road transport sector in Malaysia. In other words, it demonstrates that if motor petrol consumption increases by 1%, the level of CO2 emissions in road transport increases by 0.55%. The diesel energy consumption also has a positive coefficient that is statistically significant at the 1% level. It shows that diesel fuel consumption increases carbon emissions from the road transport sector (positive impact). This means that when diesel fuel consumption increases by 1%, the level of CO2 emissions in road transport increases by 0.45%. These results support previous findings from the study conducted by Koossalapeerom et al. (2019). They argued that gasoline motorcycles emit more than twice as much CO2 emissions as electric motorcycles. Finally, the outcomes indicate that the coefficient of palm biodiesel was negative and statistically significant at the 1% level. It demonstrates that palm biodiesel can reduce carbon emissions caused by the road transport sector. It shows that as biodiesel consumption in road transport increases by 1%, the level of carbon emissions declines by 0.001%, which is very low, because the current share of biodiesel in transport fuel is very low (5%), which cannot have a high impact on the high level of carbon emissions caused by the road transport sector in Malaysia. The magnitude of this coefficient may increase by increasing the share of palm biodiesel in transport fuel. These results confirm the findings of the study conducted by Mourad and Mahmoud (2019) who suggested that the use of biofuels reduces the level of CO2 emissions.

Table 5 ARDL(1,1,1,1) regression results: dependent variable carbon emissions
Full size table

In the short run, the use of motor petrol and diesel fuel increases CO2 emissions from the road transport sector (positive impact). Results show that if the consumption of motor petrol or diesel fuel increased by 1%, the level of CO2 emissions increased by 0.53% and 0.45%, respectively. T is noteworthy that the long-run coefficients of motor petrol and diesel are greater than their short-run coefficients. The coefficient of biodiesel was positive and statistically significant at the level of 1%. It implies that a 1% increase in biodiesel consumption in road transport increases the level of CO2 emissions in this sector by 0.004%. This may occur due to the increase in the replacement of natural forests with oil palm trees to produce more palm oil. This means that the oil palm plantation needs logging and burning of a large number of natural forests that release large quantities of carbon dioxide. Wan Mohd Jaafar et al. (2020) showed that reducing natural forests and oil palm plantations in both Sarawak and Sabah in Malaysia led to an increase in total CO2 emissions by 0.01663 Gt CO2-c/yr.

The error correction term (ECT) coefficient shows that in each period, a small percent of the short-term CO2 emissions imbalances was adjusted to achieve long-term equilibrium. This means that it takes several periods for CO2 emissions to resume their long-term trend. The coefficient of the error correction term in this model was 0.57, which shows that, in each short-term period, 57% of the imbalance (or deviation from the long-term trend) in CO2 emissions is adjusted and approaches its long-term trend.

The diagnostic outcomes of the estimated ARDL model are reported in Table 5. These tests show that the autocorrelation and heteroscedasticity problems did not occur in the estimated model, and it did not have normality and parameter stability problems. We also checked the normal distribution of residuals of the estimated ARDL(1,1,1,1) model, the results of which are reported in Figs. 2 and 3. These figures confirm that the residuals are normally distributed. Finally, to check parameter stability we estimated potential structural breaks by employing the cumulative sum test. Evidence from Fig. 4 confirms that the estimated coefficients were stable over time and that the estimated test statistic occurred within the 95% confidence interval.

Fig. 2
figure 2

Plot for standardized normal probability

Full size image
Fig. 3
figure 3

Relationship between residuals and normal distribution quantiles

Full size image
Fig. 4
figure 4

Parameter stability test (cumulative sum test) using ordinary least square (OLS) CUSUM plot

Full size image

The dynamic ARDL model, as a novel methodology, introduced by Jordan and Philips (2018) can estimate, stimulate and plot forecasts of the effect of counterfactual change in an independent variable on the dependent variable, while all other explanatory variables remain constant. To use the dynamic ARDL model, the data series for the model estimation must be integrated of order one and co-integrated, where the time series variables of this current study met this requirement. This study used 5000 simulations in the dynamic ARDL error correction model. The outcomes of the dynamic ARDL simulations methodology are reported in Table 6.

Table 6 Estimates of dynamic ARDL simulates
Full size table

Apparently, the error correction term with a negative and statistically significant coefficient (-0.57) shows that if a disequilibrium occurs in the model, it adjusts over the short run with a speed of 57% because variables move backward to achieve a long-run stable relationship. This value is the same as the estimated value of the error correction term in the general ARDL model. The coefficient of biodiesel was negative in the long run, which shows that if biodiesel consumption increases by 1%, the level of transport CO2 emissions initially declines by 0.001%. This is in line with the results of the studies conducted by Bilgili (2012) and Sarkodie et al. (2019). But, the short-run coefficient of this variable was positive and statistically significant at the 1% level. As discussed above, this may happen because of deforestation and the replacement of natural forests that are associated with logging and forest fires. The study also found a positive and statistically significant short-run link between road transport CO2 emissions and motor petrol and diesel consumption and a significant and positive link between CO2 emissions from road transport and the consumption of both motor petrol and diesel in the long run. These outcomes are consistent with those of other studies, such as Bekun et al.'s (2019) study about South Africa and Sarkodie et al.'s (2019) study about Australia.

To estimate the effect of increasing the share of palm biodiesel in fuel transport on CO2 emissions, we measured the counterfactual shocks by applying the dynamic ARDL simulations. To do this, we introduced a 10% increase in the share of palm biodiesel in the transport fuel portfolio and the estimated period for de-carbonization. Figure 5 plots the results of simulated shock and indicates that a 10% increase in the share of predicted biodiesel consumption may positively affect road transport emissions in the first period but becomes negative and decreases over time in the long run. But its negative impact on CO2 emissions from road transport was very low as reported in Table 6. Thus, decarbonizing the transport sector will have no long-run impact on sustained CO2 emissions.

Fig. 5
figure 5

Plot for the counterfactual shock in predicted biodiesel energy consumption

Full size image

Table 7 reports the estimated results for pointwise derivatives obtained from the KRLS model. The estimated results show that all variables in the model were statistically significant at the 1% level and the projecting power of the model was 0.998. This means that the independent variables explain 99.8% of changes in road transport CO2 emissions. Table 7 also reports the estimates of heterogeneous marginal impacts by applying derivatives of independent variables, which are presented as the 25th, 50th, and 75th percentiles. It can be seen that heterogeneous marginal effects do not exist across the model’s variables. Therefore, it confirms the results of the pointwise derivatives. We can observe that the mean pointwise marginal impacts of diesel consumption, motor petrol consumption, and biodiesel consumption were 0.42%, 0.32%, and 0.01%, respectively. This emphasizes the importance of diesel, motor petrol, and biodiesel in sustaining transport CO2 emissions in Malaysia. A question that arises here is how the phasing out of biodiesel influences future carbon emissions in the road transport sector. Furthermore, we estimated the long-run variation in biodiesel consumption and how it influences road transport CO2 emissions and vice versa. To perform this, this study plotted the pointwise derivative of biodiesel consumption in contrast to road transport CO2 emissions to capture the varying marginal effects.

Table 7 Results for the pointwise derivatives with KRLS
Full size table

Figure 6 shows that biodiesel energy consumption did not increase significantly as CO2 emissions increased. We can see that palm biodiesel consumption in the transport sector was not significant (or constant), when CO2 emissions increased. However, increasing biodiesel consumption thereafter decreased CO2 emissions and turned significantly positive when biodiesel consumption declines till it was below the threshold of marginal returns. Thus, biodiesel energy consumption increases marginal returns as CO2 emissions increase. This indicates that palm biodiesel technology or its share in total energy consumption is not sufficient over time as CO2 emissions increase.

Fig. 6
figure 6

Representation of pointwise marginal effect of biodiesel energy

Full size image

5 Conclusions

This study measured the impact of palm biodiesel on CO2 emissions from road transport in Malaysia during 1990–2019. It used three econometric methods, the Granger causality test based on the VAR model, the traditional ARDL model, which we called the general ARDL model, and the dynamic ARDL simulations model.

The causality analysis results show the existence of a causal relationship between all variables in the model (transport real GDP, diesel consumption, motor petrol consumption, and CO2 emissions) and palm biodiesel consumption during the study period. A long-run association also exists between transport carbon emissions and biodiesel consumption, diesel consumption, and motor petrol consumption. The results also show that a 1% increase in motor petrol consumption is associated with 0.55% and 0.32% increase in general and predict carbon emissions from road transport in the long-run, respectively, while, in the short-run, it leads to equal increases of 0.53% in general and predicted carbon emissions from road transport. Moreover, a 1% increase in diesel consumption is related to an increase of 0.45% and 0.26% in general and predicted transport carbon emissions in the long-run, respectively, whereas, in the short run, it leads to an increase of 0.45% and 0.45% in general and predicted transport CO2 emissions, respectively. However, a 1% increase in palm biodiesel consumption is associated with equal decreases of 0.001% in general and predicted carbon emissions from road transport in the long run, whereas, in the short run, it causes equal increases of 0.004% in general and predicted transport CO2 emissions. This is because of the low-level use of biodiesel and other renewable energy commodities in the transport sector.

The Kernel results show that if Malaysia continues the current share of palm biodiesel consumption in transport fuel or uses the current technology for the production of biodiesel it will not be able to reduce the level of CO2 emissions in the future.

These findings suggest that the use of biodiesel in transport fuel does not have a high effect on CO2 emissions in Malaysia. To facilitate this impact, the govenment needs to stimulate the use of clean energy products by implementing effective policies such as low fuel prices; more support for producing other bioenergy products, such as ethanol, for light and private transport, isare necessary; it also needs to invest more in studies on the research and development of technologies in palm biodiesel production to achieve sustainable development goals; research and innovation on affordable technologies for palm biodiesel conversion are necessary.