When evaluating the performance of an organization, industry, or region, a simple ratio of outputs to inputs often obscures more than it reveals. If a manufacturing plant produces twice as much output per dollar as its competitor, is it because its managers are exceptionally skilled, because the plant operates at a massive scale that unlocks cost efficiencies, or because it has adopted a brand-new proprietary technology?
To answer these questions, researchers use efficiency decomposition. By applying non-parametric frontier techniques like Data Envelopment Analysis (DEA), this methodology breaks down aggregate productivity metrics into distinct, actionable components. Instead of treating efficiency as a single black box, decomposition isolates the separate impacts of managerial competence, operational scale, and macro-level technological progress.
The Geometric Intuition of the Frontier
To understand why decomposition works, imagine plotting a system's input (such as labor or capital) on the horizontal axis and its output on the vertical axis. If you draw a curve that tightly envelopes the top-performing units in the dataset, you create an empirical production frontier. This frontier represents the absolute best-practice performance given the current state of technology.
Output
^
| * (Frontier Unit)
| / \
| / \ * (Frontier Unit)
| / \ /
| / o <--- Inefficient Unit (Distance to frontier = Inefficiency)
| /
+-----------------------------------> Input
Any unit sitting directly on this curve is fully efficient. Any unit sitting below the curve is inefficient, and its vertical distance to the frontier represents its overall Technical Efficiency (TE).
However, a simple distance metric doesn't explain why a unit is falling short. Efficiency decomposition solves this by breaking down that distance into two distinct dimensions: internal execution and physical size.
Static Decomposition: Management vs. Scale
As demonstrated by Yongli Zhang et al. in Evaluation of Green Industry Innovation Efficiency Based on Three-Stage DEA Model: A Case Study of Chinese Information Technology Industry, utilizing a Variable Returns to Scale (VRS) framework allows researchers to mathematically separate overall technical efficiency into a multiplicative identity:
$$\text{Technical Efficiency (TE)} = \text{Pure Technical Efficiency (PTE)} \times \text{Scale Efficiency (SE)}$$
This mathematical separation isolates two completely different operational realities:
- Pure Technical Efficiency (PTE): This measures how well a unit utilizes its inputs under a localized, size-adjusted frontier. It serves as a direct proxy for internal management quality, operational workflows, and resource coordination.
- Scale Efficiency (SE): This measures whether the unit is operating at an optimal physical size. It determines if a firm is suffering from diminishing returns because it is too large, or failing to capture economies of scale because it is too small.
By calculating these components separately, researchers can diagnose the exact structural bottlenecks of an industry. For instance, in an analysis of regional land utilization, Xin Luo et al. showed in The Differential Analysis on Intensive Use Efficiency of Construction Land that while most provinces operated at a near-optimal scale ($SE \approx 1$), their primary bottleneck was low pure technical efficiency. This meant that underperformance was driven by poor internal resource management rather than the physical size of the land developments.
Conversely, a study by Qian Bai on healthcare delivery in Do China rural traditional Chinese medicine hospitals provide efficient healthcare to the people? Empirical study from 2013 to 2018 using data envelopment analysis applied this exact decomposition to reveal that low hospital efficiency stemmed from internal resource redundancies and management practices rather than an inappropriate scale of operations.
When practitioners can isolate these factors, they avoid misallocating capital. As noted by Honglian Wang et al. in Research of the Improved Eco-Efficiency Evaluation Model of Iron and Steel Industry Based on PCA and Fan Feng et al. in Research on the Knowledge Exchange Efficiency of Academic Journals in the Field of Journalism and Publishing Based on the Three-Stage DEA Model, a deficit in PTE requires internal managerial restructuring, whereas a deficit in SE requires scaling the physical operation up or down to hit the sweet spot of the production curve. This distinction is further supported by Wei Sun in Transformation efficiency of resource-based cities in the Yellow River Basin and its influencing factors, who utilized variable returns to scale modeling to show how massive capital injections can fail if they merely induce scale inefficiencies without improving underlying technical processes.
Dynamic Decomposition: Catch-Up vs. Frontier Shifts
Static snapshots only show where a unit stands relative to a fixed frontier at a single point in time. To understand long-term growth, researchers track changes over time using index-based frameworks like the Malmquist Productivity Index (MPI) or the Luenberger Productivity Index.
As explained by Longfei Xue et al. in Productivity Decomposition of Chinese Energy Generation Enterprises: an Index Calculation Approach Embedded in the Data Envelopment Analysis Framework, traditional models often separate static snapshots from dynamic tracking, but index-based decomposition unifies them. When Total Factor Productivity (TFP) changes from year $t$ to year $t+1$, the index decomposes this shift into two distinct dynamic forces:
Output
| New Frontier (Period t+1) -> Technological Progress (Frontier Shift)
| /---/
| / /
| / o (Unit at t+1)
| / /
| / o (Unit at t) -> Efficiency Change (Catch-up Effect)
| /___/
| Old Frontier (Period t)
+-----------------------------------> Input
- The Catch-Up Effect (Technical Efficiency Change): This measures whether an individual unit is successfully crawling closer to the frontier over time. It captures how effectively a firm is adopting existing best practices and optimizing internal resource allocation.
- The Frontier Shift (Technological Change): This measures the movement of the frontier itself. If the entire curve shifts upward, it means industry-wide innovation, new technologies, or structural breakthroughs have raised the ceiling of what is possible for all players, regardless of their individual management.
This distinction is crucial for diagnosing industry-wide stagnation or growth. For example, Longfei Xue et al. discovered a striking paradox in the energy sector: while enterprises successfully improved their internal management and scale (a positive catch-up effect), a lack of industry-wide technological innovation caused the frontier to stagnate, leading to a net annual decline in overall TFP.
Conversely, in agricultural sectors, single-author work by Cong Wang in An Evaluation of Chinese Rapeseed Production Efficiency Based on Three-Stage DEA and Malmquist Index demonstrated that broad technological progress—rather than internal managerial catch-up—was the primary engine driving overall productivity gains.
By separating the catch-up effect from frontier shifts, researchers can evaluate whether policy interventions are working as intended. For instance:
- Wenkai Sun et al. in Government Subsidies and Innovation Performance of SMEs used a non-parametric index to show that while R&D subsidies successfully push the technological frontier outward by fostering high-quality patents, they can sometimes inadvertently lower internal resource management efficiency (the catch-up component).
- Minghui Liu et al. in The Impacts of Market Segmentation on Thermal Power Generation Efficiency demonstrated that regional trade barriers stifle efficiency by disrupting the "innovation effect," preventing high-efficiency plants from shifting the industry frontier forward.
- Yiming Yuan et al. in How does industrial intelligence affect capacity utilization?—Analysis based on green development perspective isolated competing drivers to show how the positive productivity gains of robotics can be completely drowned out if firms undergo excessive, uncoordinated scale expansions that outpace market demand.
This granular tracking is widely applied across diverse sectors—including energy finance by Zeng Sheng et al. in Research on High Quality Evaluation and Influencing Factors of China Energy Finance: Evidence From A-Share New Energy Companies, public health infrastructure by Chongyi Wang et al. in Efficiency and productivity of county-level CDCs in Qingdao municipal, China: a retrospective study based on 6 years of panel data, and public expenditure tracking by Y. F. Yang in Research Design on the Efficiency of Hubei Provincial Governments Healthcare Expenditure Based on Applied Mathematical Concepts—to determine whether performance changes are driven by internal operational adjustments or macro-level shifts.
Isolating Environmental Noise and Undesirable Outputs
In the real world, efficiency metrics can easily be distorted by external factors. A firm might appear highly efficient simply because it operates in a favorable economic environment, or it might look inefficient because it is heavily investing in reducing pollution—an outcome that traditional economic models penalize as an uncompensated "input cost."
To resolve this, modern decomposition frameworks incorporate advanced adjustments to handle environmental noise and "undesirable outputs" (such as carbon emissions or waste).
Three-Stage Adjustments
To isolate true operational capability from external luck, researchers like Yongli Zhang et al. and Fan Feng et al. utilize a three-stage DEA model. In the second stage of this framework, Stochastic Frontier Analysis (SFA) is used to calculate how much of a unit's input redundancy is caused by the external environment and random noise. The raw inputs are then mathematically adjusted to create a level playing field, ensuring that the final decomposed PTE and SE metrics reflect pure operational execution rather than environmental advantages.
Environmental Productivity Indices
When environmental sustainability is a core objective, traditional productivity indices fail because they treat pollution as a standard cost rather than an undesirable byproduct. To fix this, the Malmquist-Luenberger (ML) index and the Directional Distance Function (DDF) are deployed.
As shown by 岳立 et al. in 环境约束下的中国典型城市土地利用效率—基于DDF-Global Malmquist-Luenberger指数方法的分析, these models credit units for simultaneously maximizing desirable outputs (like GDP or crop yield) while minimizing undesirable outputs (like carbon emissions).
By integrating these constraints directly into the frontier definition, the resulting decomposition yields a true picture of "green" productivity growth. For instance:
- Fei-Fei Ye et al. in Carbon emissions management efficiency evaluation based on indicator information integration and DEA-Malmquist index combined indicator integration with the DEA-Malmquist index to dynamically isolate corporate carbon management efficiency from baseline production scaling.
- Jie Sheng et al. in Measurement, Dynamic Evolution, and Influencing Factors of Total Factor Productivity in Japan’s Beef Cattle Industry utilized the ML index to break down agricultural TFP into pure technical efficiency, scale efficiency, and technological progress under strict environmental constraints, revealing a systemic W-shaped fluctuation in industry performance.
- Xin Song et al. in Digital linkage and innovation promotion: An empirical study on the development of green economy in China and Belarus under the background of the belt and road initiative leveraged the Luenberger Productivity Index to treat carbon emissions as an undesirable output, allowing them to prove that long-term green economic growth depends on high-tech frontier shifts rather than simple infrastructure-heavy scale expansion.
- L.L. SUN et al. in THE IMPACT OF CLIMATE RISK ON THE OPERATIONAL EFFICIENCY OF FISHERY ENTERPRISES: EVIDENCE FROM CHINA used dynamic productivity decomposition to demonstrate how external climate risks suppress operational efficiency, showing that digital transformation can act as a buffer to break the vicious cycle where management inefficiency leads to higher carbon intensity.
Ultimately, whether applied to financial institutions by Akber Aman Shah et al. in Are Sustainable Banks Efficient and Productive? A Data Envelopment Analysis and the Malmquist Productivity Index Analysis or to regional infrastructure networks by Xiaoshu Cao et al. in Measuring the Efficiency of Regional Integrated Transport in China: a Data Envelopment Analysis, efficiency decomposition transforms a blunt, single-number performance metric into a multi-dimensional diagnostic map. It explains not just whether a system is succeeding, but precisely how its management, scale, and technological environment are interacting to drive that success.
Key Foundational Work
The Differential Analysis on Intensive Use Efficiency of Construction Land (2012) — contributed 1 concept elements to this direction.
Chinese companies distress prediction: an application of data envelopment analysis (2014) — contributed 1 concept elements to this direction.
Research of the Improved Eco-Efficiency Evaluation Model of Iron and Steel Industry Based on PCA (2014) — contributed 1 concept elements to this direction.