Skip to content
Back to search results

Herfindahl Diversification Index

Introduced by
Alexis Jacquemin & Charles Berry; Krishna Palepu
Year
1979
Type
Continuous index of corporate diversification with related/unrelated decomposition
Family
Strategic management / corporate strategy measurement
Font size

What does it do?

The Herfindahl diversification index and its entropy cousin turn a firm's spread across businesses into a single continuous number, with the decisive advantage that the entropy form can be cleanly split into related and unrelated diversification.

The Herfindahl-based measure is one minus the sum of squared segment revenue shares; the entropy measure, introduced for diversification by Jacquemin and Berry in 1979, is the share-weighted sum of the logged inverse shares. Jacquemin and Berry's key contribution was showing that total entropy decomposes additively into within-industry-group (related) and between-group (unrelated) components. Palepu's 1985 study applied this entropy decomposition to strategic management, finding that related diversification was associated with superior profit growth and giving the field an objective, replicable alternative to categorical schemes.

The intuition behind the method

Suppose you want one number for how diversified a company is. If it earns all its revenue in a single business, it is undiversified; if its revenue is spread evenly across many businesses, it is highly diversified. The Herfindahl approach captures this by squaring each business's revenue share and summing — a focused firm scores near one, a spread-out firm near zero — and then flipping it so higher means more diversified. The entropy measure does something similar with logarithms but adds a crucial trick: because logs turn products into sums, total diversification splits exactly into the part coming from operating in several closely related industries and the part coming from operating in entirely different industry groups. That decomposition matters because strategy theory says related diversification (sharing capabilities) and unrelated diversification (a financial portfolio) have very different consequences, and the entropy index measures each separately.

History

The Herfindahl-Hirschman index has long measured market concentration, and Caves and others adapted concentration logic to firm diversification, with the Berry index expressing diversification as one minus the Herfindahl of segment shares. Jacquemin and Berry's 1979 Journal of Industrial Economics paper advanced the field decisively by promoting the entropy measure and proving its additive decomposition into related (within-group) and unrelated (between-group) diversification. Krishna Palepu's 1985 Strategic Management Journal study brought this entropy decomposition into strategic management, demonstrating that it predicted profit performance and arguing it combined the objectivity of indices with the conceptual richness of Rumelt's categories. The entropy and Herfindahl measures became standard quantitative tools in corporate-strategy and industrial-organization research on the antecedents and consequences of diversification.

When to use it

Use Herfindahl and entropy diversification indices when you need an objective, continuous, replicable measure of corporate diversification for econometric analysis — for example, as a regressor explaining performance, risk, or value, or as an outcome capturing diversification strategy over time. The entropy decomposition is the right tool whenever the related-versus-unrelated distinction matters and you have segment data that can be grouped into broader industry categories. These measures are less suitable when segment data are too coarse or inconsistent to compute reliable shares, when you need the rich strategic interpretation that judgment-based categorical schemes provide, or when industry grouping is ambiguous enough that the related/unrelated split becomes arbitrary. In practice they are often used together with categorical classifications to combine objectivity with strategic meaning.

Limitations and considerations

  • Accuracy depends heavily on the granularity and consistency of segment reporting, which varies across firms and over time.
  • The related/unrelated split hinges on how industries are grouped, and reasonable people can group them differently.
  • Revenue-based shares ignore operational, financial, and capability linkages that the strategic notion of relatedness implies.
  • As a purely descriptive measure it says nothing about causation, and diversification remains endogenous to firm choices.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Reduces complex diversification to a continuous, objective, replicable number suitable for statistical modeling.
  • The entropy form decomposes additively into related and unrelated diversification, capturing the key strategic distinction.
  • Weights segments by revenue importance rather than merely counting businesses, reflecting the firm's actual portfolio.
  • Computable from standard segment-revenue disclosures, enabling large-sample and longitudinal studies.

Applications

  • Constructing continuous diversification variables for regressions explaining firm performance, risk, or market value.
  • Separating related from unrelated diversification to test which form of corporate scope creates value.
  • Tracking changes in firms' diversification over time, including refocusing and conglomeration waves.
  • Measuring product, geographic, or technological diversification using the same Herfindahl and entropy machinery.
  • Cross-validating judgment-based categorical diversification classifications with an objective continuous index.

How it works

How it works

1

The measure begins with the firm's revenue shares across its business segments: the fraction of total sales coming from each industry in which it operates. These shares, which sum to one, are the raw inputs to every diversification index. Jacquemin and Berry build the entire framework on these shares, so the analysis depends on how finely segments are defined and on consistent industry classification — too coarse a partition understates diversification, while inconsistent coding makes firms incomparable. Getting clean, comparable segment shares is the practical foundation of the whole exercise.

Formula
P_j=\dfrac{\text{sales in segment }j}{\text{total sales}},\quad \sum_j P_j=1
2

The Herfindahl-based diversification measure is one minus the sum of squared segment shares (the Berry index), the complement of the familiar Herfindahl concentration index. A firm with all sales in one segment has a sum of squares of one, so its diversification is zero; a firm spread across many equal segments approaches the maximum. This index is intuitive and bounded, weighting larger segments heavily through the squaring, and it provides a simple continuous diversification score, though on its own it does not separate related from unrelated diversification.

Formula
D_{H}=1-\sum_j P_j^2,\qquad D_{H}\in[0,\,1-1/N]
3

The total entropy measure of diversification is the sum over segments of each share multiplied by the natural log of its reciprocal. Like the Herfindahl form it rises with the number of segments and with the evenness of the distribution, reaching zero for a single-segment firm. Jacquemin and Berry favored entropy because of a property the Herfindahl form lacks: the logarithm makes the measure additively decomposable, which is what allows total diversification to be separated into meaningful strategic components rather than remaining a single undifferentiated number.

Formula
DT=\sum_j P_j\,\ln\!\left(\dfrac{1}{P_j}\right)
4

The signature step is decomposing total entropy into related and unrelated diversification. Segments are grouped into broader industry groups; the unrelated component is the entropy across these groups (diversification between unrelated areas), while the related component is the share-weighted entropy within groups (diversification among related businesses). Jacquemin and Berry proved these sum exactly to total diversification. This additive split is the measure's defining advantage: it operationalizes, in objective and replicable numbers, the strategically central distinction between expanding into related businesses and assembling an unrelated portfolio.

Formula
DT=\underbrace{\sum_g P^{g} D^{g}_{w}}_{DR\ (\text{related})}+\underbrace{\sum_g P^{g}\ln\!\big(1/P^{g}\big)}_{DU\ (\text{unrelated})}
5

The related and unrelated diversification components then enter performance models as continuous regressors, typically alongside controls for size, leverage, risk, and industry. Palepu used exactly this approach to test the diversification-performance relationship, regressing profit growth on the entropy-derived related and unrelated measures and finding that firms diversifying along related lines showed significantly better profit growth than those diversifying into unrelated areas. Because the components are continuous and objectively computed, they slot directly into econometric models, complementing or replacing the judgment-based categorical schemes used earlier in the field.

Formula
Y_i=\beta_0+\beta_1 DR_i+\beta_2 DU_i+\boldsymbol{\gamma}'\mathbf{x}_i+\varepsilon_i

Common pitfalls

  • Computing diversification from coarse segment data that hides true diversity, biasing the index downward.
  • Choosing industry groupings arbitrarily so that the related/unrelated decomposition reflects the analyst's coding rather than economic relatedness.
  • Confusing the Herfindahl concentration index with its diversification complement, or mixing definitions across studies.
  • Treating the entropy split as if relatedness were fully captured by industry codes, when codes only proxy the underlying capability linkages.

Frequently asked questions

What is the difference between the Herfindahl and entropy diversification measures?

Both summarize how a firm's revenue is spread across segments. The Herfindahl-based Berry index is one minus the sum of squared shares, weighting larger segments heavily and giving an intuitive bounded number. The entropy measure uses share-weighted logs of inverse shares. Their key practical difference is decomposability: because of the logarithm, total entropy splits additively into related (within-group) and unrelated (between-group) components, which the Herfindahl form does not. Jacquemin and Berry favored entropy precisely for this decomposition, which is essential when the related-versus-unrelated distinction is the object of study.

How is total diversification split into related and unrelated parts?

Segments are grouped into broader industry groups. The unrelated component is the entropy computed across these groups, reflecting diversification between distinct areas of business. The related component is the share-weighted average of the entropy within each group, reflecting diversification among businesses that are economically close. Jacquemin and Berry proved that these two components sum exactly to total entropy diversification. Palepu used this split to show that related diversification, not unrelated, was associated with superior profit growth, making the decomposition central to corporate-strategy research.

Why use a continuous index instead of Rumelt's categories?

Rumelt's categorical scheme is rich in strategic meaning but relies on subjective coding, which threatens replicability. The entropy and Herfindahl indices are computed objectively from segment-revenue data, are continuous (so they fit naturally into regressions), and are easy to replicate across large samples. Palepu argued they combine the simplicity, objectivity, and replicability of the index approach with the essential richness of relatedness through the entropy decomposition. In practice many researchers report both, using the continuous index for estimation and the categories for interpretation.

Related methods

Sources

  1. Jacquemin, A. P., & Berry, C. H. (1979). Entropy measure of diversification and corporate growth. The Journal of Industrial Economics, 27(4), 359-369.

    Open DOI reference
    Copied.
  2. Palepu, K. (1985). Diversification strategy, profit performance and the entropy measure. Strategic Management Journal, 6(3), 239-255.

    Open DOI reference
    Copied.