LMDI Decomposition
Also known as: Logarithmic Mean Divisia Index, LMDI-I Additive Decomposition, LMDI-II Multiplicative Decomposition, Logaritmik Ortalama Divisia İndeksi
Log-Mean Divisia Index (LMDI) Decomposition is a quantitative technique for attributing changes in an aggregate indicator — most commonly energy consumption or CO₂ emissions — to its underlying driving factors, such as activity level, structural mix, and intensity. Introduced in its definitive practical form by B. W. Ang in 2005, LMDI builds on Divisia index theory and uses the logarithmic mean as a weighting function to achieve a mathematically perfect, residual-free decomposition.
Key highlights
- Produces a perfect, zero-residual decomposition in both additive and multiplicative forms.
- Handles multi-period and multi-region analyses consistently without path-dependency issues.
- Results are straightforward to interpret and communicate to policy audiences.
- Widely adopted standard in energy and environmental accounting, enabling direct cross-study comparisons.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
LMDI is appropriate when a researcher needs to attribute a measurable aggregate change — typically in energy use, carbon emissions, or material throughput — to a set of multiplicative drivers across homogeneous sub-groups. The method assumes that the aggregate can be expressed as a product identity and that all sub-group values are strictly positive (zero values require interpolation or small-constant adjustments). It does not model causal mechanisms; it is a descriptive accounting tool. When causal inference is needed, structural equation or regression models are more suitable. For decomposing welfare or productivity aggregates, the related Shapley-Owen approach may be preferred.
Strengths & limitations
- Produces a perfect, zero-residual decomposition in both additive and multiplicative forms.
- Handles multi-period and multi-region analyses consistently without path-dependency issues.
- Results are straightforward to interpret and communicate to policy audiences.
- Widely adopted standard in energy and environmental accounting, enabling direct cross-study comparisons.
- Cannot handle zero-value sub-group observations without a workaround such as small-constant substitution.
- Descriptive by nature — it quantifies contributions but does not identify causal mechanisms or counterfactuals.
- The decomposition identity must be specified a priori; omitting a relevant driver conflates its effect with included factors.
- Sensitive to the choice of aggregation level and sector classification, which can influence the magnitude of individual effects.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
What is the difference between LMDI-I and LMDI-II?
LMDI-I uses an additive framework in which each factor's contribution is expressed as an absolute change in the same units as the aggregate (e.g., Mtoe or MtCO₂). LMDI-II uses a multiplicative framework in which contributions appear as dimensionless index ratios. Both yield zero residuals; the choice depends on whether absolute or relative attribution is more meaningful for the application.
How should zero-value observations be handled?
The logarithmic mean is undefined when either argument is zero. The widely accepted practice is to replace zero values with a very small positive constant (e.g., 0.001) before computation, or to aggregate sectors so that no sub-group has a zero value. Ang (2005) discusses this limitation and recommends the small-constant approach when zero cells are unavoidable.
Is LMDI suitable for causal analysis of emissions drivers?
No. LMDI is an accounting identity decomposition, not a causal model. It quantifies how much each factor arithmetically contributed to an observed change, but it cannot establish that any factor caused the change. For causal questions — such as whether a carbon tax reduced emissions — regression-based or counterfactual methods are required alongside or instead of LMDI.
Sources
- 1.Ang, B. W. (2005). The LMDI approach to decomposition analysis: a practical guide. Energy Policy, 33(7), 867–871.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 2). LMDI Decomposition. ScholarGate. https://scholargate.app/sustainability/lmdi-decomposition