- Email: [email protected]
Fuel price determination in transportation sector using predicted energy and transport demand
ARTICLE IN PRESS
Energy Policy 34 (2006) 3078–3086 www.elsevier.com/locate/enpol
Fuel price determination in transportation sector using predicted energy and transport demand Soner Haldenbilen Department of Civil Engineering, Engineering Faculty, Pamukkale University, Denizli 20017, Turkey Available online 14 July 2005
Abstract This study determines fuel price based on estimated sectoral energy and transport demand using pumping prices. Three approaches are first used for estimating energy and transportation demand based on linear time series, polynomial time series and genetic algorithm based (GATEDE and GATDETR), as multi-parameter, models. Then, future fuel prices and marginal costs of the energy consumption are obtained. Transport demand-based energy efficiency methods are also developed. The fuel prices (FP) are analyzed under two scenarios: Linear and exponential price scenarios. Results showed that if the FP increases linearly, the marginal cost will slightly decreases from current trend, but will increases if demand increases exponentially. Results also showed that the demand-based pricing policy would help to develop a new pricing policy for fuel use in order to control fast growing demand on this sector. The exponential price increase would also help to locate financial sources to create environmentally friendly transportation systems. r 2005 Elsevier Ltd. All rights reserved. Keywords: Energy demand; Transport demand; Fuel price
1. Introduction Energy plays a vital importance in daily life and it may be considered as one of the lifelines for the humanity. It affects the living standards in the community. In addition to the energy importance, the transportation is also vital importance for community. But, energy, especially fuel as being a scarce sources and non-renewable, needs to carefully be planned. Over consumption of energy sources may lead to decrease scarce sources and also pollutes our environment. Therefore, the planning of the energy and transport demand needs to be investigated to control growing demand to meet future needs of humanity to achieve sustainable objectives. In order to lessen the environmental affect of energy consumption, the Kyoto protocol is singed some countries in 1990.
Tel.: +90 258 2134030; fax: +90 258 2125548.
E-mail address: [email protected]. 0301-4215/$ - see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.enpol.2005.06.007
Total energy consumption increased 500% and the petroleum and natural gas consumption increased about 900%, while the world population growth increased 200% between 1945 and 1985 (Haldenbilen, 2003). Thus, there is a huge demand to energy and transport. Transportation is one of the biggest sectors to use non-renewable energy sources as fuel. The energy consumption in this sector covers both private and commercial vehicles. Per capita sectoral energy consumption in European Union (EU) was about 0.4 tone equivalent petrol (TEP) in 1970 and was increased to a level of 0.77 TEP in 1995 (Banister et al., 2000). The energy consumption in this sector was increased from 0.1 to 0.2 TEP between 1970 and 2000 in Turkey as well. While there is a decrease on energy consumption per vehicle-km, there is an increase on demand per capita (Haldenbilen, 2003). The reason for is that the improvement of the vehicle technology will cause to reduce the energy consumption. In 1970, the energy consumption on rural roads was 0.52 kilogram equivalent petrol (KEP)/veh-km and decreased to a level of
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3079
as a fuel price. For this aim, the transport and energy demand is forecasted first, and then the fuel price is analyzed under two different pricing trends. The relationship between the fuel price and transport demand (TD), and also energy demand (ED) are evaluated based on marginal cost (MC).
0.28 KEP/veh-km in 2000 (Haldenbilen and Ceylan, 2005a, b). The efficient use of energy is generally justified with pricing policy and technological improvements, but the demand for gasoline is inelastic with price changes according to State Planning Organization (SPO, 2001). The report indicated that the changes on fuel price did not affect the growing demand for transportation for the period of 1990–2000. The fuel tax increased from 56% to 72% in same period and reached to a level of EU average. The high tax rate and the price on fuel will subsidize the national budget and they are not used for managing and planning energy sources. The price for operating private vehicles are relatively decreased when it is compared with public transport, for instance, the increase on operating public transport in terms of fuel cost increased by about 53% while on private transport it decreased by about 38% including insurance, service, road tax costs (Banister et al., 2000). Designing efficient and cost effective energy and transport systems that meet environmental conditions is one of the foremost challenges that engineers face. In the world with scarce natural resources and large energy demands, modeling becomes increasingly important to understand the mechanisms that degrade energy and resources and to develop systematic approaches for improving systems and, thus, also reducing the impact on the environment. Moreover, fuel price and marginal cost analysis help to identify the components where inefficiencies occur. Improvements should be done to these components to increase their efficiency. However, estimation of fuel price and marginal cost analysis of energy and transport demand is quite new. Thus, this study investigates relations between the fuel price (FP), energy and transport demand. Also demand-based marginal cost analysis is proposed. The main objective of this study is to investigate the cost of the energy consumption and transportation cost
2. FP The FP in Turkey, as one of a developing country, has been increasing and the tax on fuel is also increasing. The FPs is given in Fig. 1 for the period of 1990–2004. The effect of economic crisis, which hit Turkey two times, can clearly be seen for which the prices decreased due to the big variations on the value of money, but the general trend followed the trends of the developed world. The fuel price per liter was about 0.50$/lt in 1990 and it increased to a level of 1.8$/lt. As figure indicates there is an increasing trend in dollar basis. The prices are given in the middle of month for each year, where the National Statistics (NS, 2005) records data on midterm basis. The tax that is taken from fuels is very high about 75%, which supports the National Budget (NB) about 10% in Turkey. Tax rates, which are taken from motorized vehicles, are given in Table 1 as a proportion of the NB, where FT is the fuel tax, MOT is the motor vehicle tax, and the PT is the purchase tax (Ministry of Finance, MF, 2003). An average of 12% of the NB comes from road transport taxes. However, the revenues collected from road transport are not fully used in transportation sector although there is a growing demand. General Directorate of Turkish Highways (GDTH) budget has been decreasing when it is compared with the NB even though the road tax revenues are increasing to a level of 11% in 2000 (GDTH, 2004). The highest share is in 1960 that can be the political reasons and one of the critical stages in that
2.00 1.60
$/lt
1.20 0.80 0.40
Years
Fig. 1. Gasoline pumping prices (NS, 2005).
15.01.2005
15.01.2004
15.01.2003
15.01.2002
15.01.2001
15.01.2000
15.01.1999
15.01.1998
15.01.1997
15.01.1996
15.01.1995
15.01.1994
15.01.1993
15.01.1992
15.01.1991
15.01.1990
0.00
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3080
Table 1 The share of MOT, PT and FT to the NB (%)
FT MOT PT
Table 2 Observed data
1994
1995
1996
1997
1998
1999
2000
Years
Avarage fuel price
6.7 0.55 0.95
7.4 0.5 1.0
11.3 0.6 1.3
11.1 0.6 1.3
9.2 0.6 1.1
12 0.7 1.1
9.76 0.6 1.2
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
0.63 0.73 0.74 0.65 0.55 0.65 0.69 0.75 0.74 0.85 0.94 0.82 0.99 1.22 1.39
year. Since then, the investment on road system starts to decrease. The budget share of GDTH is 1.8% in 2002. The Turkish Railway (TR, 2003) budget is even lower than the GDTH for the same period. For example, the budget of GDTH is 2.5 times higher than TR budget (2003) in 1980. Investment on infrastructure and demand management on transport is the major parameters to be considered in order to meet growing demand in transport sector. In order to support other sectors and economic development, the taxes on fuel must be kept at least in current level. Although the FT tax rate is about a level of EU average, it is still not enough to control growing demand. 2.1. Expected values of fuel price Many parameters affect the price of fuel such as price of crude oil, inflation rate, value of foreign exchange rate, etc. But, this study uses only the recorded fuel prices as a time series to estimate future values, where other related parameters are not taken into account due to difficulties of their measurement, which may not affect overall objectives of this study. Data are given in Table 2. Two forms of the fitted mathematical model using Table 2 to the current data are obtained and given in expressions (1) and (2), where the first one is the linear fitted trend line (linear price approach, LPA) and the second one is the exponential form of the fitted trend line (exponential price approach, EPA). The LPA model is obtained as Y ¼ 0:042X þ 0:487;
2
R ¼ 0:67,
(1)
where Y is the pumping price as a unit of $/lt and X is a time series for which 1990 ¼ 1, 1991 ¼ 2,y,2004 ¼ 15,y. The EPA model is obtained as Y ¼ 0:547e0:047X ;
R2 ¼ 0:70.
(2)
Although the correlation coefficients are not high in (1) and (2) when they are compared with expressions in (3)–(8), they may roughly be used for estimating the fuel prices. Estimation of the fuel price is a complex task due to inclusion of many economic and political parameters such as foreign department, inflation rate, etc. It would be possible to improve the correlation coefficients in the LPA and the EPA model using the many measurable
independent variables but this study only uses a two form of the very simplistic time series model, which may not affect the main objective of this study. Figs. 2a and b show that the observed data and the error of estimated values for the LPA and the EPA, respectively. Fig. 3 shows the projected fuel price for two models. The FP will be about 1.79$/liter under LPA and 2.34$/ liter under EPA in 2020. Forecasted FP will need to be analyzed with respect to energy and transport demand in order to find changing in marginal cost for using energy and movement. To this purpose, we need to project both energy and transport demands.
3. Energy and transport demand 3.1. Energy demand Energy production and consumption as a million tons of oil equivalent (MTOE) is given in Table 3, which are taken from the Minister of the Energy and Natural Resources (MENR, 2005). The ratio of petroleum production to consumption is about 16% in 1990, while this ratio decreased to a level of 8%. The biggest share of petroleum consumption is taken by transportation sector. The consumption rate of transport sector to the total petroleum consumption was about 40% within last 10 years. The consumption increased from 4.6 MTOE in 1974 to 12.12 MTOE in 2002 (MENR, 2005). The expected energy demand is obtained with three approaches. One of them is the linear, second is the polynomial time series and the third one is the GATEDE model, which is developed by Haldenbilen and Ceylan (2005a, b) based on Genetic Algorithm (GA)
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3081
1.40 LPA
Avarage Fuel Price
Price $/lt
1.20
1.00
0.80
0.20
0.40 1990
1992
1994
1996
(a)
1998
2000
2002
2004
Years 1.40 Avarage Fuel Price
LPA
Price $/lt
1.20
1.00
0.80
0.20
0.40 1990
1992
1994
(b)
1996 Years
1998
2000
2002
2004
Fig. 2. (a) Average fuel price estimation under the LPA model. (b) Average fuel price estimation under the EPA model. 2.50 LPA
EPA
2.25
$/lt
2.00 1.75 1.50 1.25 1.00 2006
2008
2010
2012
2014
2016
2018
2020
Fig. 3. Projected fuel prices with two models.
approach (see for details on GA based demand modeling, Ceylan and ve Ozturk, 2004; Ozturk and Ceylan, 2005; Ozturk et al., 2005). The linear time series (LTS) model is obtained as y ¼ 0:29x þ 3:811;
R2 ¼ 0:92.
(3)
The polynomial time series (PTS) model is obtained as y ¼ 0:006x2 þ 0:118x þ 4:698;
R2 ¼ 0:94,
(4)
where y is the transport energy demand in MTOE and x is a time series for which 1974 ¼ 1, 1975 ¼ 2,y, 2004 ¼ 31,y.
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3082
Table 3 Energy Consumption and production (MENR, 2005) Years
1990 1992 1994 1996 1998 2000 2002 2003
Consumption (Co)
Production (Pr)
Petroleum Pr/Co
Petroleum (Pe) (MTOE)
Total (T) (MTOE)
PE/T (%)
Petroleum (MTOE)
Total (MTOE)
PE/T (%)
22.700 23.660 25.859 29.604 29.022 31.072 29.776 30.669
52.987 56.684 59.127 69.862 74.709 81.251 78.711 83.804
0.43 0.42 0.44 0.42 0.39 0.38 0.38 0.37
3.717 4.281 3.687 3.500 3.224 2.749 2.420 2.375
25.478 26.794 26.511 27.386 29.324 26.855 24.727 23.812
0.15 0.16 0.14 0.13 0.11 0.10 0.10 0.10
0.16 0.18 0.14 0.12 0.11 0.09 0.08 0.08
40 35 Observed
GATEDE
LTS
PTS
Demand (MTOE)
30 25 20 15 10 5
2019
2014
2009
2004
1999
1994
1989
1984
1979
1974
0
Years
Fig. 4. Forecasted energy demand.
The GATEDE model is (see for details on this model, Haldenbilen and Ceylan, 2005a, b) given as GATEDE ¼ 0:079X 3:973 0:251X 0:1 1 2 þ 0:42X 30:875 þ 1:562;
R2 ¼ 0:99,
ð5Þ
where X 1 is the gross domestic product (GDP, $109), X 2 is population (106) and X 3 is the total annual veh-km (109). The R2’s of the models are high; therefore it is included to estimate energy demand. The verification and performance of GA-based models can be obtained in Ceylan and ve Bell, 2004, 2005. The forecasted energy demand is given in Fig. 4 for the period of 1974–2020. It is obtained that per capita energy consumption will be 0.20 TEP with the LTS, 0.27 TEP with the PTS and 0.42 TEP with the GATEDE models.
countries. In Turkey, for example, the mobility was about 2.5 109 passenger-km in 1950 in rural roads and reached to a level of 185 109 passenger-km in 2005. The vehicular movement increases from 109 veh-km/ year in 1974 to 45 109 veh-km/year in 2002 according to General Directorate of Turkish Highways (GDTH, 2004). In the same period, the growth of the socioeconomic and demographic indicators was less than transportation demand (NS, 2005). The biggest portion of goods and passenger transport are made with highway transport by about 95% (GDTH, 2004). Three forms of the TD model is developed and given in the following way. The linear form of the transport demand model (LTD) is y ¼ 1:415x þ 5:049;
R2 ¼ 0:95.
(6)
3.2. Transport demand
The polynomial form of the transport demand model (PTD) is
Transport demand (TD) has rapidly been increasing for last half a century in both developed and developing
y ¼ 0:036x2 þ 0:348x þ 10:56;
R2 ¼ 0:98,
(7)
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3083
160
Transport demand (veh-km/year)
Observed
GATDETR
PTD
LTD
120
80
40
2019
2014
2009
2004
1999
1994
1989
1984
1979
1974
0
Years
Fig. 5. Predicted transport demand.
where y is the TD (veh-km (109) and x is a time series for which 1974 ¼ 1, 1975 ¼ 2,y,2004 ¼ 31,y. The GA form of the model (GATDETR) is (see for details, Haldenbilen and Ceylan, 2005b) given as GATDETR ¼ ð0:086X 1 þ 0:010X 2 þ 0:038X 3 Þ1:777 , R2 ¼ 0:99,
ð8Þ
where X 1 is the population (106), X 2 is the gross domestic product per capita (GDPPC, $102) and X 3 is the total annual vehicle number(106). The forecasted TD can be seen in Fig. 5 for the three models. The verification of the models may be obtained in Haldenbilen and Ceylan (2005a, b). The highest transport demand is obtained by GATDETR model and the lowest one is the LTD. Along with the GATDETR both LTD and PTD model is included in the analyses due to their high correlation coefficients.
4. Marginal cost (MC) estimation The MC of energy and the transport with respect to fuel prices is given in Fig. 6 for the period of 1991–2002. The MCs are in big variations, reached in negative values in some years, but it is in a decreasing trend. This trend will show that the operating cost of vehicles is decreasing, for which sustainable objectives may not be achieved. The estimation of the MC is obtained in the following way: MC T ¼
FPn FPn1 ; TDn TDn1
FPn FPn1 MC E ¼ ; EC n EC n1
n ¼ 1991; 1992; . . . , n ¼ 1991; 1992; . . . ,
(9) (10)
where MC T is the marginal cost of transportation ($/lt/ 109 veh-km), FP is average fuel price ($/lt), TD is transport demand (109 veh-km), MC E is the marginal cost of energy use ($/MTOE) and EC is the energy consumption (MTOE). The MC of 1999 is removed from this study due to insignificance of the data, where the prices were not controlled because of big earthquake in Turkey. The following two scenarios may be helpful to develop a fuel pricing policy using the demand-based analyses. 4.1. Linear price scenario Expected marginal costs of energy and transport are given in Figs. 7a and b. The following results may be drawn with this scenario. 1. If energy and transport demand increases linearly, there is no change on marginal costs, indicated as the LTS and the LTD in Figs. 7a and b. 2. If the energy and transport demand increases, indicated as PTS, GATEDE, PTD and GATDETR in Figs. 7a and b, the marginal costs decreases.
4.2. Exponential price scenario Marginal costs of the energy and transport demand are given in Figs. 8a and b under this scenario. The following results are: 1. If the FP increases exponentially, the marginal costs of energy and transport stay even based on GATEDE and GATDETR models in Figs. 8a and b; and
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3084
0.40 TD
ED
0.20
Marginal cost ($)
0.00 -0.20 -0.40 -0.60 -0.80 -1.00
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
-1.20
Years Fig. 6. The variations on marginal cost of transport energy use.
Marginal costs of energy ($)
0.160
0.120 LT S
PT S
GATE DE
0.080
0.040
2018
2020 2020
2016
2018
(a)
2014
2012
2010
2008
2006
0.000
Years
Marginal cost of transport Demand ($)
0.024
0.018
0.012
0.006 LT D
PT D
GAT DETR
(b)
2016
2014
2012
2010
2008
2006
0.000
Years
Fig. 7. (a) The pricing policy with linear growth of energy demand. (b) The pricing policy with linear growth of transport demand.
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3085
0.400
Marginal cost of energy demand ($)
LTS
PTS
GATEDE
0.300
0.200
0.100
(a)
2020
2018
2016
2014
2012
2010
2008
2006
0.000
Years
Marginal cost of transport demand ($)
0.060 LTD
PTD
GATDETR
0.040
0.020
(b)
2020
2018
2016
2014
2012
2010
2008
2006
0.000
Years
Fig. 8. (a) The pricing policy with exponential growth of energy demand. (b) The pricing policy with exponential growth of transport demand.
2. if the FP increases exponentially, the marginal costs of energy and transport will increase based on LTS, PTS, LTD and PTD in Figs. 8a and b. When the FP increases exponentially, the demand for energy and transport may be under control. But, the elasticity of the FP to the demand is still needs to be investigated to find robust results.
5. Conclusions This study evaluates and analysis the energy and transport demand with respect to fuel prices. For this
purpose, three of the energy demand model is developed as LTS, PTS and GATEDE models. For the transportation energy demand, the LTD, PTD and GATDETR models are developed. The marginal cost of using energy and movement is calculated to find the future trends of fuel pries. The following results can be obtained from this study: 1. Fuel prices will increases about 100% with LPA and 200% with EPA when it is compared with 2005 prices. The increase rate of estimated energy and transport demand is far over the FP in between 2005 and 2020. 2. Planning only the energy and transportation demand with respect to the FP using the historical data will
ARTICLE IN PRESS 3086
S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
not be enough to control demand for which the marginal cost of energy and movement is in decreasing trend. Therefore, new measures and policies should be developed as demand-based pricing policy. 3. Planning of the FP for future should be carried out with multivariable demand analysis such as the GATEDE and GATDETR, which help to develop a dynamic FP determination that lead us to intervene to the marginal costs. 4. The road use charging for private transport may control the growing demand if the MC for the private operators is increased. 5. The acceptable level of increase on marginal cost of energy and movement may be used to effectively plan energy demand and to develop environmentally friendly transportation systems to create financial sources.
References Banister, D., Stead, D.ve., Steen, P., 2000. European Transport Policy and Sustainable Mobility. Spon Press, London, New York. Ceylan, H., ve Bell, M.G.H., 2004. Traffic signal timing optimization based on genetic algorithm approach, including drivers’ routing. Transportation Research 38B (4), 329–342. Ceylan, H., ve Bell, M.G.H., 2005. Genetic algorithm solution for the stochastic equilibrium transportation networks under congestion. Transportation Research-B 39B, 169–185.
Ceylan, H., ve Ozturk, H.K., 2004. Estimating energy demand of Turkey based on economic indicators using genetic algorithm approach. Energy Conversion and Management 45, 2525–2537. GDTH, 2004. General Directorate of Turkish Highways, Annual Statistics of Turkish Highways, http://www.kgm.gov.tr (in Turkish). Haldenbilen, S., 2003. Evaluation of Sustainable transport indicators for Turkey based on genetic algorithm approach. Ph.D. Thesis, Institute of Science and Technology, Pamukkale University, Denizli, Turkey (in Turkish). Haldenbilen, S., Ceylan, H., 2005a. Genetic algorithm approach to estimate transport energy demand in Turkey. Energy Policy 33 (1), 89–98. Haldenbilen, S., Ceylan, H., 2005b. The development of a road tax policy in Turkey, using a genetic algorithm approach for demand estimation. Transportation Research A, in press. MENR, 2005. The Ministry of Energy and Natural Resources, http:// www.enerji.gov.tr (in Turkish). MF, 2003. Ministry of Finance, (Downloadable from. http://www.maliye.gov.tr). NS, 2005. National Statistics, (http://www.die.gov.tr). Ozturk, H.K., Ceylan, H., 2005. Forecasting total and industrial sector electricity demand based on genetic algorithm approach: Turkey case study. International Journal of Energy Research 29, 829–840. Ozturk, H.K., Ceylan, H., Canyurt, O.E., ve Hepbasli, A., 2005. Electricity estimation using genetic algorithm approach: a case study of Turkey. Energy 30 (7), 1003–1012. SPO, 2001. State Planning Organization, (Downloadable from http:// www.dpt.gov.tr). TR, 2003. Turkish Railways, (Downloadable from http:// www.tdcc.gov.tr (in Turkish).
Energy Policy 34 (2006) 3078–3086 www.elsevier.com/locate/enpol
Fuel price determination in transportation sector using predicted energy and transport demand Soner Haldenbilen Department of Civil Engineering, Engineering Faculty, Pamukkale University, Denizli 20017, Turkey Available online 14 July 2005
Abstract This study determines fuel price based on estimated sectoral energy and transport demand using pumping prices. Three approaches are first used for estimating energy and transportation demand based on linear time series, polynomial time series and genetic algorithm based (GATEDE and GATDETR), as multi-parameter, models. Then, future fuel prices and marginal costs of the energy consumption are obtained. Transport demand-based energy efficiency methods are also developed. The fuel prices (FP) are analyzed under two scenarios: Linear and exponential price scenarios. Results showed that if the FP increases linearly, the marginal cost will slightly decreases from current trend, but will increases if demand increases exponentially. Results also showed that the demand-based pricing policy would help to develop a new pricing policy for fuel use in order to control fast growing demand on this sector. The exponential price increase would also help to locate financial sources to create environmentally friendly transportation systems. r 2005 Elsevier Ltd. All rights reserved. Keywords: Energy demand; Transport demand; Fuel price
1. Introduction Energy plays a vital importance in daily life and it may be considered as one of the lifelines for the humanity. It affects the living standards in the community. In addition to the energy importance, the transportation is also vital importance for community. But, energy, especially fuel as being a scarce sources and non-renewable, needs to carefully be planned. Over consumption of energy sources may lead to decrease scarce sources and also pollutes our environment. Therefore, the planning of the energy and transport demand needs to be investigated to control growing demand to meet future needs of humanity to achieve sustainable objectives. In order to lessen the environmental affect of energy consumption, the Kyoto protocol is singed some countries in 1990.
Tel.: +90 258 2134030; fax: +90 258 2125548.
E-mail address: [email protected]. 0301-4215/$ - see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.enpol.2005.06.007
Total energy consumption increased 500% and the petroleum and natural gas consumption increased about 900%, while the world population growth increased 200% between 1945 and 1985 (Haldenbilen, 2003). Thus, there is a huge demand to energy and transport. Transportation is one of the biggest sectors to use non-renewable energy sources as fuel. The energy consumption in this sector covers both private and commercial vehicles. Per capita sectoral energy consumption in European Union (EU) was about 0.4 tone equivalent petrol (TEP) in 1970 and was increased to a level of 0.77 TEP in 1995 (Banister et al., 2000). The energy consumption in this sector was increased from 0.1 to 0.2 TEP between 1970 and 2000 in Turkey as well. While there is a decrease on energy consumption per vehicle-km, there is an increase on demand per capita (Haldenbilen, 2003). The reason for is that the improvement of the vehicle technology will cause to reduce the energy consumption. In 1970, the energy consumption on rural roads was 0.52 kilogram equivalent petrol (KEP)/veh-km and decreased to a level of
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3079
as a fuel price. For this aim, the transport and energy demand is forecasted first, and then the fuel price is analyzed under two different pricing trends. The relationship between the fuel price and transport demand (TD), and also energy demand (ED) are evaluated based on marginal cost (MC).
0.28 KEP/veh-km in 2000 (Haldenbilen and Ceylan, 2005a, b). The efficient use of energy is generally justified with pricing policy and technological improvements, but the demand for gasoline is inelastic with price changes according to State Planning Organization (SPO, 2001). The report indicated that the changes on fuel price did not affect the growing demand for transportation for the period of 1990–2000. The fuel tax increased from 56% to 72% in same period and reached to a level of EU average. The high tax rate and the price on fuel will subsidize the national budget and they are not used for managing and planning energy sources. The price for operating private vehicles are relatively decreased when it is compared with public transport, for instance, the increase on operating public transport in terms of fuel cost increased by about 53% while on private transport it decreased by about 38% including insurance, service, road tax costs (Banister et al., 2000). Designing efficient and cost effective energy and transport systems that meet environmental conditions is one of the foremost challenges that engineers face. In the world with scarce natural resources and large energy demands, modeling becomes increasingly important to understand the mechanisms that degrade energy and resources and to develop systematic approaches for improving systems and, thus, also reducing the impact on the environment. Moreover, fuel price and marginal cost analysis help to identify the components where inefficiencies occur. Improvements should be done to these components to increase their efficiency. However, estimation of fuel price and marginal cost analysis of energy and transport demand is quite new. Thus, this study investigates relations between the fuel price (FP), energy and transport demand. Also demand-based marginal cost analysis is proposed. The main objective of this study is to investigate the cost of the energy consumption and transportation cost
2. FP The FP in Turkey, as one of a developing country, has been increasing and the tax on fuel is also increasing. The FPs is given in Fig. 1 for the period of 1990–2004. The effect of economic crisis, which hit Turkey two times, can clearly be seen for which the prices decreased due to the big variations on the value of money, but the general trend followed the trends of the developed world. The fuel price per liter was about 0.50$/lt in 1990 and it increased to a level of 1.8$/lt. As figure indicates there is an increasing trend in dollar basis. The prices are given in the middle of month for each year, where the National Statistics (NS, 2005) records data on midterm basis. The tax that is taken from fuels is very high about 75%, which supports the National Budget (NB) about 10% in Turkey. Tax rates, which are taken from motorized vehicles, are given in Table 1 as a proportion of the NB, where FT is the fuel tax, MOT is the motor vehicle tax, and the PT is the purchase tax (Ministry of Finance, MF, 2003). An average of 12% of the NB comes from road transport taxes. However, the revenues collected from road transport are not fully used in transportation sector although there is a growing demand. General Directorate of Turkish Highways (GDTH) budget has been decreasing when it is compared with the NB even though the road tax revenues are increasing to a level of 11% in 2000 (GDTH, 2004). The highest share is in 1960 that can be the political reasons and one of the critical stages in that
2.00 1.60
$/lt
1.20 0.80 0.40
Years
Fig. 1. Gasoline pumping prices (NS, 2005).
15.01.2005
15.01.2004
15.01.2003
15.01.2002
15.01.2001
15.01.2000
15.01.1999
15.01.1998
15.01.1997
15.01.1996
15.01.1995
15.01.1994
15.01.1993
15.01.1992
15.01.1991
15.01.1990
0.00
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3080
Table 1 The share of MOT, PT and FT to the NB (%)
FT MOT PT
Table 2 Observed data
1994
1995
1996
1997
1998
1999
2000
Years
Avarage fuel price
6.7 0.55 0.95
7.4 0.5 1.0
11.3 0.6 1.3
11.1 0.6 1.3
9.2 0.6 1.1
12 0.7 1.1
9.76 0.6 1.2
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
0.63 0.73 0.74 0.65 0.55 0.65 0.69 0.75 0.74 0.85 0.94 0.82 0.99 1.22 1.39
year. Since then, the investment on road system starts to decrease. The budget share of GDTH is 1.8% in 2002. The Turkish Railway (TR, 2003) budget is even lower than the GDTH for the same period. For example, the budget of GDTH is 2.5 times higher than TR budget (2003) in 1980. Investment on infrastructure and demand management on transport is the major parameters to be considered in order to meet growing demand in transport sector. In order to support other sectors and economic development, the taxes on fuel must be kept at least in current level. Although the FT tax rate is about a level of EU average, it is still not enough to control growing demand. 2.1. Expected values of fuel price Many parameters affect the price of fuel such as price of crude oil, inflation rate, value of foreign exchange rate, etc. But, this study uses only the recorded fuel prices as a time series to estimate future values, where other related parameters are not taken into account due to difficulties of their measurement, which may not affect overall objectives of this study. Data are given in Table 2. Two forms of the fitted mathematical model using Table 2 to the current data are obtained and given in expressions (1) and (2), where the first one is the linear fitted trend line (linear price approach, LPA) and the second one is the exponential form of the fitted trend line (exponential price approach, EPA). The LPA model is obtained as Y ¼ 0:042X þ 0:487;
2
R ¼ 0:67,
(1)
where Y is the pumping price as a unit of $/lt and X is a time series for which 1990 ¼ 1, 1991 ¼ 2,y,2004 ¼ 15,y. The EPA model is obtained as Y ¼ 0:547e0:047X ;
R2 ¼ 0:70.
(2)
Although the correlation coefficients are not high in (1) and (2) when they are compared with expressions in (3)–(8), they may roughly be used for estimating the fuel prices. Estimation of the fuel price is a complex task due to inclusion of many economic and political parameters such as foreign department, inflation rate, etc. It would be possible to improve the correlation coefficients in the LPA and the EPA model using the many measurable
independent variables but this study only uses a two form of the very simplistic time series model, which may not affect the main objective of this study. Figs. 2a and b show that the observed data and the error of estimated values for the LPA and the EPA, respectively. Fig. 3 shows the projected fuel price for two models. The FP will be about 1.79$/liter under LPA and 2.34$/ liter under EPA in 2020. Forecasted FP will need to be analyzed with respect to energy and transport demand in order to find changing in marginal cost for using energy and movement. To this purpose, we need to project both energy and transport demands.
3. Energy and transport demand 3.1. Energy demand Energy production and consumption as a million tons of oil equivalent (MTOE) is given in Table 3, which are taken from the Minister of the Energy and Natural Resources (MENR, 2005). The ratio of petroleum production to consumption is about 16% in 1990, while this ratio decreased to a level of 8%. The biggest share of petroleum consumption is taken by transportation sector. The consumption rate of transport sector to the total petroleum consumption was about 40% within last 10 years. The consumption increased from 4.6 MTOE in 1974 to 12.12 MTOE in 2002 (MENR, 2005). The expected energy demand is obtained with three approaches. One of them is the linear, second is the polynomial time series and the third one is the GATEDE model, which is developed by Haldenbilen and Ceylan (2005a, b) based on Genetic Algorithm (GA)
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3081
1.40 LPA
Avarage Fuel Price
Price $/lt
1.20
1.00
0.80
0.20
0.40 1990
1992
1994
1996
(a)
1998
2000
2002
2004
Years 1.40 Avarage Fuel Price
LPA
Price $/lt
1.20
1.00
0.80
0.20
0.40 1990
1992
1994
(b)
1996 Years
1998
2000
2002
2004
Fig. 2. (a) Average fuel price estimation under the LPA model. (b) Average fuel price estimation under the EPA model. 2.50 LPA
EPA
2.25
$/lt
2.00 1.75 1.50 1.25 1.00 2006
2008
2010
2012
2014
2016
2018
2020
Fig. 3. Projected fuel prices with two models.
approach (see for details on GA based demand modeling, Ceylan and ve Ozturk, 2004; Ozturk and Ceylan, 2005; Ozturk et al., 2005). The linear time series (LTS) model is obtained as y ¼ 0:29x þ 3:811;
R2 ¼ 0:92.
(3)
The polynomial time series (PTS) model is obtained as y ¼ 0:006x2 þ 0:118x þ 4:698;
R2 ¼ 0:94,
(4)
where y is the transport energy demand in MTOE and x is a time series for which 1974 ¼ 1, 1975 ¼ 2,y, 2004 ¼ 31,y.
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3082
Table 3 Energy Consumption and production (MENR, 2005) Years
1990 1992 1994 1996 1998 2000 2002 2003
Consumption (Co)
Production (Pr)
Petroleum Pr/Co
Petroleum (Pe) (MTOE)
Total (T) (MTOE)
PE/T (%)
Petroleum (MTOE)
Total (MTOE)
PE/T (%)
22.700 23.660 25.859 29.604 29.022 31.072 29.776 30.669
52.987 56.684 59.127 69.862 74.709 81.251 78.711 83.804
0.43 0.42 0.44 0.42 0.39 0.38 0.38 0.37
3.717 4.281 3.687 3.500 3.224 2.749 2.420 2.375
25.478 26.794 26.511 27.386 29.324 26.855 24.727 23.812
0.15 0.16 0.14 0.13 0.11 0.10 0.10 0.10
0.16 0.18 0.14 0.12 0.11 0.09 0.08 0.08
40 35 Observed
GATEDE
LTS
PTS
Demand (MTOE)
30 25 20 15 10 5
2019
2014
2009
2004
1999
1994
1989
1984
1979
1974
0
Years
Fig. 4. Forecasted energy demand.
The GATEDE model is (see for details on this model, Haldenbilen and Ceylan, 2005a, b) given as GATEDE ¼ 0:079X 3:973 0:251X 0:1 1 2 þ 0:42X 30:875 þ 1:562;
R2 ¼ 0:99,
ð5Þ
where X 1 is the gross domestic product (GDP, $109), X 2 is population (106) and X 3 is the total annual veh-km (109). The R2’s of the models are high; therefore it is included to estimate energy demand. The verification and performance of GA-based models can be obtained in Ceylan and ve Bell, 2004, 2005. The forecasted energy demand is given in Fig. 4 for the period of 1974–2020. It is obtained that per capita energy consumption will be 0.20 TEP with the LTS, 0.27 TEP with the PTS and 0.42 TEP with the GATEDE models.
countries. In Turkey, for example, the mobility was about 2.5 109 passenger-km in 1950 in rural roads and reached to a level of 185 109 passenger-km in 2005. The vehicular movement increases from 109 veh-km/ year in 1974 to 45 109 veh-km/year in 2002 according to General Directorate of Turkish Highways (GDTH, 2004). In the same period, the growth of the socioeconomic and demographic indicators was less than transportation demand (NS, 2005). The biggest portion of goods and passenger transport are made with highway transport by about 95% (GDTH, 2004). Three forms of the TD model is developed and given in the following way. The linear form of the transport demand model (LTD) is y ¼ 1:415x þ 5:049;
R2 ¼ 0:95.
(6)
3.2. Transport demand
The polynomial form of the transport demand model (PTD) is
Transport demand (TD) has rapidly been increasing for last half a century in both developed and developing
y ¼ 0:036x2 þ 0:348x þ 10:56;
R2 ¼ 0:98,
(7)
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3083
160
Transport demand (veh-km/year)
Observed
GATDETR
PTD
LTD
120
80
40
2019
2014
2009
2004
1999
1994
1989
1984
1979
1974
0
Years
Fig. 5. Predicted transport demand.
where y is the TD (veh-km (109) and x is a time series for which 1974 ¼ 1, 1975 ¼ 2,y,2004 ¼ 31,y. The GA form of the model (GATDETR) is (see for details, Haldenbilen and Ceylan, 2005b) given as GATDETR ¼ ð0:086X 1 þ 0:010X 2 þ 0:038X 3 Þ1:777 , R2 ¼ 0:99,
ð8Þ
where X 1 is the population (106), X 2 is the gross domestic product per capita (GDPPC, $102) and X 3 is the total annual vehicle number(106). The forecasted TD can be seen in Fig. 5 for the three models. The verification of the models may be obtained in Haldenbilen and Ceylan (2005a, b). The highest transport demand is obtained by GATDETR model and the lowest one is the LTD. Along with the GATDETR both LTD and PTD model is included in the analyses due to their high correlation coefficients.
4. Marginal cost (MC) estimation The MC of energy and the transport with respect to fuel prices is given in Fig. 6 for the period of 1991–2002. The MCs are in big variations, reached in negative values in some years, but it is in a decreasing trend. This trend will show that the operating cost of vehicles is decreasing, for which sustainable objectives may not be achieved. The estimation of the MC is obtained in the following way: MC T ¼
FPn FPn1 ; TDn TDn1
FPn FPn1 MC E ¼ ; EC n EC n1
n ¼ 1991; 1992; . . . , n ¼ 1991; 1992; . . . ,
(9) (10)
where MC T is the marginal cost of transportation ($/lt/ 109 veh-km), FP is average fuel price ($/lt), TD is transport demand (109 veh-km), MC E is the marginal cost of energy use ($/MTOE) and EC is the energy consumption (MTOE). The MC of 1999 is removed from this study due to insignificance of the data, where the prices were not controlled because of big earthquake in Turkey. The following two scenarios may be helpful to develop a fuel pricing policy using the demand-based analyses. 4.1. Linear price scenario Expected marginal costs of energy and transport are given in Figs. 7a and b. The following results may be drawn with this scenario. 1. If energy and transport demand increases linearly, there is no change on marginal costs, indicated as the LTS and the LTD in Figs. 7a and b. 2. If the energy and transport demand increases, indicated as PTS, GATEDE, PTD and GATDETR in Figs. 7a and b, the marginal costs decreases.
4.2. Exponential price scenario Marginal costs of the energy and transport demand are given in Figs. 8a and b under this scenario. The following results are: 1. If the FP increases exponentially, the marginal costs of energy and transport stay even based on GATEDE and GATDETR models in Figs. 8a and b; and
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3084
0.40 TD
ED
0.20
Marginal cost ($)
0.00 -0.20 -0.40 -0.60 -0.80 -1.00
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
-1.20
Years Fig. 6. The variations on marginal cost of transport energy use.
Marginal costs of energy ($)
0.160
0.120 LT S
PT S
GATE DE
0.080
0.040
2018
2020 2020
2016
2018
(a)
2014
2012
2010
2008
2006
0.000
Years
Marginal cost of transport Demand ($)
0.024
0.018
0.012
0.006 LT D
PT D
GAT DETR
(b)
2016
2014
2012
2010
2008
2006
0.000
Years
Fig. 7. (a) The pricing policy with linear growth of energy demand. (b) The pricing policy with linear growth of transport demand.
ARTICLE IN PRESS S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
3085
0.400
Marginal cost of energy demand ($)
LTS
PTS
GATEDE
0.300
0.200
0.100
(a)
2020
2018
2016
2014
2012
2010
2008
2006
0.000
Years
Marginal cost of transport demand ($)
0.060 LTD
PTD
GATDETR
0.040
0.020
(b)
2020
2018
2016
2014
2012
2010
2008
2006
0.000
Years
Fig. 8. (a) The pricing policy with exponential growth of energy demand. (b) The pricing policy with exponential growth of transport demand.
2. if the FP increases exponentially, the marginal costs of energy and transport will increase based on LTS, PTS, LTD and PTD in Figs. 8a and b. When the FP increases exponentially, the demand for energy and transport may be under control. But, the elasticity of the FP to the demand is still needs to be investigated to find robust results.
5. Conclusions This study evaluates and analysis the energy and transport demand with respect to fuel prices. For this
purpose, three of the energy demand model is developed as LTS, PTS and GATEDE models. For the transportation energy demand, the LTD, PTD and GATDETR models are developed. The marginal cost of using energy and movement is calculated to find the future trends of fuel pries. The following results can be obtained from this study: 1. Fuel prices will increases about 100% with LPA and 200% with EPA when it is compared with 2005 prices. The increase rate of estimated energy and transport demand is far over the FP in between 2005 and 2020. 2. Planning only the energy and transportation demand with respect to the FP using the historical data will
ARTICLE IN PRESS 3086
S. Haldenbilen / Energy Policy 34 (2006) 3078–3086
not be enough to control demand for which the marginal cost of energy and movement is in decreasing trend. Therefore, new measures and policies should be developed as demand-based pricing policy. 3. Planning of the FP for future should be carried out with multivariable demand analysis such as the GATEDE and GATDETR, which help to develop a dynamic FP determination that lead us to intervene to the marginal costs. 4. The road use charging for private transport may control the growing demand if the MC for the private operators is increased. 5. The acceptable level of increase on marginal cost of energy and movement may be used to effectively plan energy demand and to develop environmentally friendly transportation systems to create financial sources.
References Banister, D., Stead, D.ve., Steen, P., 2000. European Transport Policy and Sustainable Mobility. Spon Press, London, New York. Ceylan, H., ve Bell, M.G.H., 2004. Traffic signal timing optimization based on genetic algorithm approach, including drivers’ routing. Transportation Research 38B (4), 329–342. Ceylan, H., ve Bell, M.G.H., 2005. Genetic algorithm solution for the stochastic equilibrium transportation networks under congestion. Transportation Research-B 39B, 169–185.
Ceylan, H., ve Ozturk, H.K., 2004. Estimating energy demand of Turkey based on economic indicators using genetic algorithm approach. Energy Conversion and Management 45, 2525–2537. GDTH, 2004. General Directorate of Turkish Highways, Annual Statistics of Turkish Highways, http://www.kgm.gov.tr (in Turkish). Haldenbilen, S., 2003. Evaluation of Sustainable transport indicators for Turkey based on genetic algorithm approach. Ph.D. Thesis, Institute of Science and Technology, Pamukkale University, Denizli, Turkey (in Turkish). Haldenbilen, S., Ceylan, H., 2005a. Genetic algorithm approach to estimate transport energy demand in Turkey. Energy Policy 33 (1), 89–98. Haldenbilen, S., Ceylan, H., 2005b. The development of a road tax policy in Turkey, using a genetic algorithm approach for demand estimation. Transportation Research A, in press. MENR, 2005. The Ministry of Energy and Natural Resources, http:// www.enerji.gov.tr (in Turkish). MF, 2003. Ministry of Finance, (Downloadable from. http://www.maliye.gov.tr). NS, 2005. National Statistics, (http://www.die.gov.tr). Ozturk, H.K., Ceylan, H., 2005. Forecasting total and industrial sector electricity demand based on genetic algorithm approach: Turkey case study. International Journal of Energy Research 29, 829–840. Ozturk, H.K., Ceylan, H., Canyurt, O.E., ve Hepbasli, A., 2005. Electricity estimation using genetic algorithm approach: a case study of Turkey. Energy 30 (7), 1003–1012. SPO, 2001. State Planning Organization, (Downloadable from http:// www.dpt.gov.tr). TR, 2003. Turkish Railways, (Downloadable from http:// www.tdcc.gov.tr (in Turkish).