Attraction Market-Share Model
Also known as: Market Share Theorem Models, MCI Model, Multiplicative Competitive Interaction Model, Differential-Effects Attraction Model
The attraction market-share model expresses each brand's market share as its own 'attraction' divided by the total attraction of all brands competing in the market, guaranteeing shares that are non-negative and sum to one by construction. Its theoretical foundation is the 1975 'market share theorem' of Bell, Keeney, and Little, who proved that the familiar share-equals-effort-over-total-effort relationship follows from three mild axioms about attraction. Cooper and Nakanishi turned this into a practical empirical technology, the Multiplicative Competitive Interaction (MCI) model and its exponential MNL variant, in which attraction is built from the marketing mix — price, distribution, advertising, promotion. A log-centering transformation converts the inherently nonlinear share equation into a linear regression that can be estimated by ordinary or generalized least squares. The fitted parameters yield managerially crucial own- and cross-elasticities of share that respect the logically-consistent zero-sum nature of competition. The approach is a cornerstone of competitive market-share analysis and a close cousin of brand-choice logit models.
Key highlights
- Guarantees logically consistent predictions: shares are always non-negative and sum to one, unlike linear share regressions.
- Rests on an axiomatic foundation — the Bell, Keeney, and Little market share theorem — rather than ad hoc functional choices.
- The log-centering transformation linearizes an inherently nonlinear competitive system, allowing simple least-squares estimation.
- Yields closed-form own- and cross-elasticities of share that capture the zero-sum nature of brand competition for managerial planning.
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
Use an attraction market-share model when you have brand-level shares for a well-defined, roughly closed category and want to explain and forecast how the marketing mix drives competitive share in a logically consistent way. It is the natural choice when management thinks in terms of share rather than absolute volume, when you need own- and cross-elasticities that respect the constraint that shares sum to one, and when you have syndicated or scanner tracking data with price, distribution, advertising, and promotion for all major competitors. The model is less appropriate when category sales themselves expand or contract substantially in response to marketing (an attraction model holds the pie fixed and should be paired with a category-volume model), when the relevant decision is individual purchase rather than aggregate share (use a brand-choice logit on panel data instead), or when many brands have zero shares in some periods, which the log transform cannot accommodate without adjustment.
Strengths & limitations
- Guarantees logically consistent predictions: shares are always non-negative and sum to one, unlike linear share regressions.
- Rests on an axiomatic foundation — the Bell, Keeney, and Little market share theorem — rather than ad hoc functional choices.
- The log-centering transformation linearizes an inherently nonlinear competitive system, allowing simple least-squares estimation.
- Yields closed-form own- and cross-elasticities of share that capture the zero-sum nature of brand competition for managerial planning.
- Holds total category volume fixed, so it cannot by itself capture marketing's effect on primary (category) demand and must be combined with a volume model.
- Zero or near-zero shares break the logarithmic transformation and require imputation or aggregation that can distort estimates.
- Aggregate share data confound heterogeneous consumer behavior, so parameters reflect a market average that may mask important segment differences.
- Specifying which rivals' variables enter a brand's attraction (cross-effects) quickly explodes the parameter count and risks multicollinearity in tracking data.
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 exactly does the 'market share theorem' guarantee?
Bell, Keeney, and Little proved that if attraction is non-negative, if brands with equal attraction receive equal shares, and if a fixed increase in any brand's attraction affects every other brand's share through the same mechanism, then market share must take the form of each brand's attraction divided by the sum of all attractions. The practical consequence is logical consistency: any model built this way automatically produces shares between zero and one that sum to exactly 100%. This is what separates attraction models from linear share regressions, which can predict negative or above-100% shares that make no economic sense.
How do the MCI and MNL attraction specifications differ?
Both define share as attraction over total attraction; they differ in how the marketing mix builds attraction. The Multiplicative Competitive Interaction (MCI) model makes attraction a product of marketing-mix variables each raised to a power, so a percentage change in a variable produces a constant proportional change in attraction. The multinomial-logit (MNL) form makes attraction the exponential of a linear index of the mix variables, the same algebra as a brand-choice logit. After the log-centering transformation, MCI is linear in the logged variables while MNL is linear in the raw variables; the better choice is an empirical question about whether marketing effects compound multiplicatively or operate through an exponentiated linear utility.
Why can't I just run a normal regression of share on the marketing mix?
A linear regression of share on marketing variables ignores the two constraints that define a share: it can predict shares below zero or above one, and the predictions across brands will not sum to one. That makes forecasts logically incoherent and elasticities misleading because they do not respect the zero-sum nature of competition. The attraction model encodes these constraints structurally, and the log-centering transformation lets you still estimate it with familiar least squares. So you get the convenience of linear estimation without sacrificing the logical consistency that aggregate share analysis demands.
Sources
- 1.Bell, D. E., Keeney, R. L., & Little, J. D. C. (1975). A Market Share Theorem. Journal of Marketing Research, 12(2), 136-141.
- 2.Cooper, L. G., & Nakanishi, M. (1988). Market-Share Analysis: Evaluating Competitive Marketing Effectiveness. Kluwer Academic Publishers.ISBN 9780898382785
You have read it. What now?
Cite this page
ScholarGate. (2026, June 23). Attraction Market-Share Model. ScholarGate. https://scholargate.app/marketing/attraction-market-share-model