Measuring firm efficiency with Data Envelopment Analysis (DEA)
A plain-language introduction to DEA, one of the methods I use to compare firm performance and investment attractiveness, and what it can and cannot tell us.
Part of my work at the Digital Empowerment Lab is using data analytics and DEA models to assess firm performance and investment attractiveness. DEA comes up often when I explain my research, so here is an introduction written for readers without an operations research background.
What DEA is
Data Envelopment Analysis (DEA) is a non-parametric method, based on linear programming, for comparing the relative efficiency of a set of similar units. In DEA these units are called decision-making units (DMUs). A DMU can be a firm, a bank branch, a hospital, a school, or any unit that turns inputs into outputs.
The appeal of DEA is that it handles multiple inputs and multiple outputs at the same time, without needing a predefined formula for how they relate. Inputs might be employees, assets or costs; outputs might be revenue, profit or the number of customers served.
How it works, intuitively
DEA looks at all the units together and builds a "best-practice frontier" from the ones that turn inputs into outputs most efficiently. Units on the frontier receive an efficiency score of 1. Units below it receive a lower score, which shows how far they are from the best performers in the same group.
Two points are important:
- The scores are relative. A unit is efficient compared with the others in the sample, not against some absolute standard.
- Because it is non-parametric, DEA does not assume a particular production function. The data shape the frontier.
CCR and BCC
The two classic models are:
- CCR, named after Charnes, Cooper and Rhodes (1978), which assumes constant returns to scale: doubling inputs should double outputs.
- BCC, named after Banker, Charnes and Cooper (1984), which allows variable returns to scale, so that small and large units are not judged by exactly the same yardstick.
Comparing the two helps separate inefficiency that comes from how a unit is managed from inefficiency that comes from its size.
Why I find it useful
My research asks how technology and digital transformation relate to firm performance. DEA fits that question well because firm performance has many dimensions, and because comparing firms against their peers is often more meaningful than comparing them against an abstract ideal. It can also be combined with other analytical steps, for example to explore which factors are associated with higher or lower efficiency.
Its limits
DEA is not magic. The results depend heavily on which inputs and outputs are chosen and on which units are included. Outliers and data errors can move the frontier. And an efficiency score describes where a unit stands; it does not explain why. Choosing variables carefully and being honest about these limits is part of the work.
Coming next
I will write more about how I apply DEA once the related studies are published. If you work with DEA, or have a dataset where it might help, I am always happy to talk.
Photos: chris810 on Pixabay, used under the Pixabay Content License. These are not my photographs.
Related
3 min readResearch Areas
Views: 1Digital transformation, innovation and firm competitiveness
Digitization, digitalization and digital transformation are often used as if they meant the same thing. Here is how I separate them, and the questions about competitiveness that interest me.
3 min readResearch Areas
Views: 1Digital servitization: why products are turning into services
One of my core research interests: how firms move from selling products to selling outcomes, and what digital technologies change in that shift.