Structural Decomposition Techniques: Sense and Sensitivity

Abstract

Structural decomposition techniques are widely used to break down the growth in some variable into the changes in its determinants. In this paper, we discuss the problems caused by the existence of a multitude of equivalent decomposition forms which are used to measure the contribution of a specific determinant. Although it is well known that structural decompositions are not unique, the extent of the problem and its consequences seem to have been largely neglected. In an empirical analysis for The Netherlands between 1986 and 1992, results are calculated for 24 equivalent decomposition forms. The outcomes exhibit a large degree of variability across the different forms. We also examine the two approaches that have been used predominantly in the literature. The average of the two so-called polar decompositions appears to be remarkably close to the average of the full set of 24 decompositions. The approximate decomposition with mid-point weights appears to be almost exact. Although this last alternative might seem a solution to the problem of the marked sensitivity, in fact, it only conceals the problem.Decomposition techniques, input-output framework, sensitivity analysis,

Comparing the Gtap-Mrio and Wiod Databases for Carbon Footprint Analysis

“…We will follow the simplified method proposed (Dietzenbacher and Los, 1998), based on the average of the two polar decompositions. The starting point is the difference in the emissions embedded in the final demand of region r calculated using the two different databases:…”

China's Growing CO2 EmissionsA Race between Increasing Consumption and Efficiency Gains

“…By considering the change in each variable over time the SDA of (1) can be expressed as where the first term represents the contribution to change in emissions, ∆f, due to aggregated changes in the emission intensities (efficiency), F; the second term represents changes in the production structure, L; the third term represents changes in the consumption structure, ys; and the fourth term represents changes in the consumption volume (GDP), y v. When performing the SDA it is possible to compare terms relative to the start or end of each time-period, and this leads to a uniqueness problem. To account for this issue we take the average of all the first-order decompositions (28)(29)(30). These issues are detailed in the Supporting Information.…”

Material and Energy Dependence of Services and Its Implications for Climate Change

“…: (i) material goods (including energy) absorbed by material goods, (ii) material goods absorbed by services, (iii) services absorbed by material goods, and (iv) services absorbed by services. Then, by originally applying input-output partitioning techniques (see Miyazawa (26) for the seminal contribution; Sonis et al (27,28) for its application to regional studies; Fritz et al (29) for its application to environmental studies) and Dietzenbacher's full structural decomposition (30,31) techniques, the indirect domestic CO 2 emissions associated with household consumption can be additively decomposed into df d for the on-site emissions of the industries concerned and p j kl for supply chain emissions, as expressed in eq 2, in which k, l ) 1 are the group of commodities for material goods and k, l ) 2 are the group for services.…”