1.  

Starting from January 1, 2001, Taiwan has imposed a Petroleum Fund Fee (PFF) on the crude oil.  The rate is NT$ 303 (equivalent to U.S. $8.66) per kiloliter.  And the revenue of PFF is estimated as 8.47 billion New Taiwan Dollar (or U.S. $242 million in 2002), which will be used solely for the following purposes: (1) government strategical stock-piling, (2) subsidizing the oil supply facilities in remote areas (3) promoting the oil & gas exploration and development, (4) Energy R&D and (5) others which related with the affairs of oil supply stabilization.

The purpose of this paper is to evaluate the effect of imposing Petroleum Fund Fee on the growth and price of oil sector and its direct and indirect impact on energy demand, CO₂ emission and the economy as a whole by 2001. 

For this evaluation, a Dynamic General Equilibrium Model of Taiwan (DGEMT) is employed.  DGEMT model, which is an extension of the Jorgenson-Liang (1985), consists of the following four sub-models: (1) producer's model, (2) consumer's model, (3) DGBAS Macro-economic model and (4) ITRI MARKAL engineering model.

     In addition to introduction, the following sections are: 2. Theoretic Model—Dynamic Generalized Equilibrium Model, 3. The Simulation Methodology and Procedure, 4. Simulation Result and 5. Conclusion.

2. Theoretic Model—Dynamic Generalized Equilibrium Model

        The dynamic generalized equilibrium model consists of the following four submodels: (1) producer’s model; (2) consumer’s model; (3) DGBAS, macro-economic model and (4) ITRI’s MARKAL engineering energy model.

2.1        Producer’s Model

The producer’s model decomposes Taiwan's economy into twenty-nine sectors, namely, eight main sectors (including agriculture, mining, manufacturing, construction, public utility, transportation, service and industry (mining, manufacturing, construction and public utility)), seventeen manufacturing sectors (including food, beverage & tobacco, textiles, clothes & wearing apparel, leather & leather products, wood & bamboo products, furniture products, paper & printing, chemical & plastic, rubber products, non-metallic mineral, basic metal, metal products, machinery & equipment, elect. mach. & electronics, transport equipment and miscellaneous) and four energy sectors (including coal mining, oil refinery, natural gas and electricity).

        We assume that the sectoral cost function is of the translog form with homothetic weak separability of energy and material inputs.  The model actually consists of four submodels (for each sector): an aggregate submodel, an energy submodel, a non-energy intermediate input submodel, and an oil product submodel.  The aggregate submodel includes one output price equation and five equations relating to the cost shares of capital, labor, energy, non-energy intermediate inputs and the rate of technological change.  The energy submodel has one price (energy price) equation and four share equations explaining the cost shares of coal, oil products, natural gas, and electricity.  The non-energy intermediate submodel is composed of one price (material price) equation and five equations for the cost shares of agricultural intermediate inputs, industrial intermediate inputs, transportation's intermediate inputs, service intermediate inputs, and imported intermediate inputs.  Similarly the oil product submodel has one price (oil price) equation and four share equations explaining the cost shares of gasoline, diesel, fuel oil and other oil products.  Diagram 1 presents the tier structure of submodels in producer's model.  With the sole exception of the oil submodel, the explanatory variables consist of input prices and time as an index for the level of technology.  As for the oil submodel, the explanatory variable consists of input prices only. 

   Taking the aggregate input submodel as an example, the output price(P) equation is:

            (1)

where denotes capital, labor, energy and intermediate input respectively.  T denotes time as an index for the level of technology.

The input cost share equations are: [1]

 subscripts                     (2)

and the rate of technical change (-R_T) is:

                                          (3)

        The basic approach of the model, which is modified from the Hudson-Jorgenson (1974) model, is an integration of econometric modeling and input-output analysis.  However, to reflect the dramatic changes in both industrial structure and energy consumption patterns of Taiwan's economy, a time trend is included in the energy and material submodels.  This innovation makes this Jorgenson-Liang(1985) model significantly different from most of the work of Jorgenson and his associates', which are studies based on highly developed economies, such as the United States, Japan and West Germany.  This kind of model will be also useful for the case studies of the other Newly Industrialized Countries (NICs).

        Liang (1987), Jorgenson and Liang (1985) and Liang (1999) contain detailed descriptions of this theoretical model, estimation method, data compilation and the results of coefficients estimated.  It is noted that Liang (1999) is a revised model of Jogenson-Liang (1985) by updating time series data of producer’s model from 1961-1981 to 1961-1993, and to combine the consumer’s model (Liang (1983)) the macro-economic model of the Directorate General of Budget, Accounting and Statistics, Executive Yuan (Ho-Lin-Wang (2001)) and the MARKAL Engineering Model of the Industrial Technology Research Institute (Young (1996)).

 

2.2        The Consumer’s Model

Following Jorgenson-Slesnick(1983), we assume that the k^th household allocates its expenditures in accordance with the translog indirect utility function. Under exact aggregation condition, the vector of aggregate expenditure shares can be written in the form:

           (4)

     Under exact aggregation, systems of individual expenditure shares for consuming units with identical demographic characteristics can be recovered in one and only one way from the system of aggregate expenditure shares.[2]

    Equation (4) implies that the vector of the expenditure shares of the household sector (private consumption) are determined by commodity prices (P), expenditure structure (ΣM_KlnM_K ⁄M) and the joint distribution of household expenditure and the attributes (ΣM_KlnA_K ⁄M), where  and  denote the k^th household's expenditure and attributes respectively.   is a vector of ones.  We divide private consumption into five categories:

(1) Food: Expenditures on food, beverages and tobacco.

(2) Clothing: Expenditures on clothing, apparel.

(3) Housing: Expenditures on rent and non-energy utilities, furniture, furnishing and household equipment, household operations and services.

(4) Energy: Expenditures on fuel and electricity including fuel for vehicles.

(5) Recreation, Transportation and Miscellaneous: Expenditures on recreation, amusement and education, medical and health care, transportation and miscellaneous consumption expenditures.

And hence the vector of expenditure share (S) in fact consists of the above five types of expenditure share.  The following demographic characteristics are employed as attributes of households:

1.          Family size: 1, 2, 3, 4, 5, 6, 7, 8 or more.

2.          Occupation: Nonfarmer and farmer.

3.          Number of persons employed: 1, 2, 3 or more.

For detailed description of the model, please refer to Liang (1983).[3]  The consumer’s model links with the producer’s model through output prices by sector; while it links with the DGBAS Macro economic model via total private consumption.  (See next section)

 

2.3        DGBAS Macro-Economic Model[4]

The macro-economic model of Directorate General of Budget Accounting and Statistics (DGBAS) is a Keynsian model which consists of 159 equations.  We retrieve the following projection data from the Macro-Economic model as initial values in the baseline projection: (1) GDP growth rate, (2) wage, (3) interest rate, (4) private consumption, (5) CPI, (6) WPI, (7) Investment, (8) Government expenditure, and (9) Exports.  CPI and WPI are affected by output prices by sector.  The GDP, wage, interest rate and private consumption are functions of CPI or WPI in the DGBAS macro-economic model. Thus, there are feed-back relationships between DGBAS macro-economic model and the producers’ model if sectoral output prices change due to energy tax implementation.

The total supply is composed of the intermediate demands from industries and the final demands of private consumption (C), investment (I), government expenditures (G), and net export (X) minus import (M).  Markets are cleared by the prices of domestically produced commodities by sector (P_i).[5] 

  ,        i, j=1….29     (5)

 

2.4        ITRI MARKAL Engineering Energy Model[6]

By employing linear programming method, the ITRI MARKAL engineering model combines the information of growth of industries, energy supply and energy technologies to achieve the best energy mix.  This model is developed by the Institute of Energy and Resources, Industrial Technology Research Institute (ITRI).

Because information for future energy technology development is well considered in the model, we use the aggregate of the energy demand by types projected by the ITRI MARKAL engineering model to control the total energy demand projected from the producer’s and consumer’s models.

 

3. The Simulation Methodology and Procedure

        The simulation framework of the model is presented in Diagram 1.

Base case projection

To assess the effect of PFF we must first determine the future path of the Taiwanese economy in the absence of the tax.  We call such a scenario a base case.  The base case projection is conducted by the following steps:

(1) We insert the values of capital services price (P_K), wage (P_L) and price of import intermediate input (P_m) projected by the DGBAS macro-economic model into the producer's model. Thus, we obtain the prices and factor cost shares in 29 sectors in 2001.

(2) By employing 1996 input-output table, we then convert the 29 sectoral output prices into prices of 5 consumer's goods by 2001. Inserting the prices of 5 consumer's good together with the private consumption as projected by the macro-economic model into the consumer's model, we got the shares of 5 consumer's goods in total private consumption.

(3) The demand for types of energy by sector, taking oil as an example, is derived by multiplying the oil coefficient (O/Q) with the total output (Q) by sector.  The oil coefficient (O/Q) can be calculated by the following equation :

                                                                                    ( 6)

where  S_E : Energy shares in total cost,

       S_O : Oil share in energy cost,

       P : Output price,

       P_O : The price of oil products,

  And  S_E, S_O, P, and P_O are endogenously determined in the model.

The projected growth rate of sectoral output by 2001 is derived by: (i) the GDP growth rate obtained from the Macro-Economic model, (ii) the industrial structure projection provided by this study, and (iii) employing the sectoral value- added shares in total output which are endogenously determined from simulation of this model.

(4) The demand for energy in household sector (E_H) is derived by

.                                                 (7)

Here, ,  and  denote, respectively, the energy expenditure share in private consumption, energy price and private consumption.  Both  and  are determined endogenously from the consumer's model, while  (private consumption) comes from the projection of DGBAS Macroeconomic model.

(5) The demand for types of energy are then converted into CO₂ emission by employing the conversion factor projected by MARKAL engineering model, such as: coal (3.53 ton CO₂/KLOE)[7], oil products (2.89 ton CO₂/KLOE), and natural gas (2.09 ton CO₂/KLOE).  This finishes the whole process for baseline projection.

Simulation in Petroleum Fund Fee

(6) Next, we evaluate the impact of PFF.  We convert the prices of oil products from endogenous to exogenous.  The price of oil products is modified by incorporating PFF schedules into the producer’s model and consumer’s model, respectively, to calculate their corresponding output prices, cost shares, demand for types of energy and CO₂ emission by sectors as well as the consumption structure and quantity of consumer’s goods. 

(7) However, the above scenarios do not consider the 'feed back' effect in the changes of capital service price (P_K) and wage (P_L) and output caused by PFF implementation. In fact, the implementation of PFF will affect P_K and P_L and total output by sector as well.  In the DGBAS Macro-Economic Model P_K and P_L are affected by the PFF through the increase in general price level.  Hence we insert the GDP deflator into the function of P_K and P_L to get new P_K and P_L, and in turn new values of output price, cost structure and CO₂ emission by sector.

(8) The impact of PFF on total output by sector are evaluated by the following procedure:

(i)    First of all, we calculate the impact of PFF on sectoral output price and general price level (GDP deflator), and in turn, the new values of final demand such as private consumption, investment, government expenditure, net export and GDP.

(ii) Next, we multiply the private consumption with the private consumption shares of five consumer’s goods, which are then deflated by their respective prices to obtain the new values of five consumer’s goods.

(iii)       We then employ the 1996 Input-Output table to convert the changes in five consumer’s goods to the changes in sectoral final demand (FD).[8]

(iv)We obtain the sectoral total output (Q) by using the following standard input–output equation: ;  Here, D denotes the matrix of domestic product input-output coefficient.

(v) We calculate the energy conservation effect on total output of the four energy sectors and the whole economy. The energy conservation effect is obtained by comparing the demand for four types of energy in the base case with that in the 'PFF' case where PFF is implemented.

(9) Finally, the impact of different PFF on the sectoral output prices, demand for types of energy and CO₂ emissions are compared. 

It is noted that the imposition of PFF will reduce total output and further reduce the demand for energy and CO₂ emission.  Therefore, the total impact of energy taxes on CO₂ emission reduction should also accommodate the effect on output growth.  Briefly speaking, we consider not only ‘substitution effect’ but also the ‘income effect’, both in the consumer’s and producer’s models, on demand for energy and CO₂ emissions.

 

4. Simulation Result

Effect on Output Prices

According to the Energy Commission, the budget of the petroleum fund in the 2001 fiscal year is New Taiwan Dollar 8.47 billion (equivalent to USD 242 million).  In contrast, the revenue of whole petroleum sector is forecasted as 304 billion New Taiwan Dollar.  Consequently, the effect i.e. tax rate is 2.786 (=8.47/304) percent which implies that the prices of oil products will directly increase by 2.786 percent.

However, together with indirect effect, the price of oil refinery sector will increase by 2.99 percent (see Table 1).  Among manufacturing industries, non-metallic mineral products, basic metal and chemical & plastic will suffer relative great impact on price increase.  But none of their price increase exceeds 0.32 percent.  Water, electricity & gas (0.56 percent) and transportation & Communication (0.35 percent) have the highest price increase among the seven one-digital sectors.  For the economy as a whole, the GDP deflator will increase by 0.13 percent in 2001.

Effect on Output Growth

The oil refinery sector will also suffer the greatest decline in output growth, i.e., -2.85 percent, where PFF is imposed.  This is due to the ‘substitution effect’ and ‘income effect’ both in the final demand and in the producer’s sector.  Similarly, the basic metal, non-metallic mineral products and chemical & plastic are among the most affected sectors in the manufacturing sector.  However, decline less than 0.2 percent.  All of their output will decrease by less than 0.2 percent.  And electricity, water & gas and transportation & communication sectors have the greatest decline in output growth among the seven one-digital sectors.  For the economy as a whole, the GDP will reduce by a negligible 0.08 percent.

Effect on Energy Demand and CO₂ Emission

Imposing the PFF tax, the CO₂ emission will decrease 1.46 percent for the economy as a whole in 2001.  The energy demand for oil has the greatest decrease, which is -2.93 percent, followed by electricity, -0.20 percent and natural gas, -0.19 percent.  On contrary, the demand for coal will increase by 0.18 percent owing to the ‘substitution’ effect between coal and oil products.

5. Conclusion

The imposition of PFF will raise a sizable fund, i.e., New Taiwan Dollar 8.47 billion for the government to do the following tasks:  (1) government strategical stock-piling, (2) subsidizing the oil supply facilities in remote areas (3) promoting the oil & gas exploration and development and (4) Energy R&D.  The impact of imposing the PFF will only have a negligible impact both on sectoral prices and output growth.  On the contrary, it will reduce the demand for energy and CO₂ emission by as much as 1.46 percent.  The PFF policy of Taiwan might be useful for other countries to refer.  

（本文發表於91.6.26~91.6.28英國蘇格蘭亞伯丁舉辦第二十五屆國際能源經濟學年會）