The final demand vector, Y, can be treated as exogenous to the system; for example, the level of total production research use can be determined by the final demand (2):Y=BX.(2)Input-output model can be applied to calculate each sector’s indirect energy consumption regardless of the length and complexity of their production processes by using the energy input-output table (Wu and Chen, 1990; Peet, 1993). In energy input-output tables, energy sectors should be represented both in monetary and energy terms for computing the direct energy consumption coefficient matrix [14]. Assume that in input-output tables the economy can be categorized into n sectors, which includes k energy sectors and n-k nonenergy sectors.

Hence we can write an equation representing the way in which energy sectors distribute their products to energy sectors, nonenergy sectors, and final demands in physic units:Ak,1+Ak,2+?+Ak,k+Ak,k+1+?Ak,n+fk=xk.(3)Using energy input-output tables, the direct energy intensity and total energy intensity of each economic sector can be calculated. Direct energy intensity of one sector is calculated as the ratio of direct energy consumption (in physical terms) to total inputs (in monetary terms). Total energy intensities are calculated by multiplying direct energy intensity matrix with the Leontief inverse matrix of the corresponding energy input-output table. Embodied energy use in infrastructure investment can be calculated by multiplying total energy intensities with infrastructure investment:ei=��i=1kEj,1Xi,(4)etotal=e(I?A)?1,(5)EII=etotalYII=e(I?A)?1YII.

(6)ei is the direct energy intensity of sector i, e is the direct energy intensity matrix, and etotal is the total energy intensity matrix. EII is the embodied energy in infrastructure investment, and YII is the infrastructure investment.2.2. Structural Decomposition AnalysisBased on (6), the change of embodied energy use in infrastructure investment is driven by several factors, such as growth in infrastructure investment, energy efficiency AV-951 improvement, and industrial structure changes. Aiming at identifying the driving factors for changes in embodied energy use in infrastructure investment overtime, we applied input-output structural decomposition analysis on (6).