---
title: "Internal capital markets: when a group allocates well"
description: "Head office can back the right division, yet one model says weak ones still win too much capital. Internal capital markets: Stein, Scharfstein and BCG."
url: "https://staging.khoi.co.uk/en/blog/thi-truong-von-noi-bo-tap-doan"
lang: en-GB
author: "Lê Quốc Khôi"
date: 2026-11-02
updated: 2026-11-02
topics: ["Corporate finance"]
tags: ["Internal capital market", "BCG matrix", "Capital allocation", "Agency problem", "Diversification"]
---

# Internal capital markets: does a group allocate its own cash well?

## Key takeaways

- An internal capital market is a group allocating cash among its divisions in place of outside markets. Stein (1997): head office has control rights that let it shift funds between projects ("winner-picking"), which a bank lender lacks, so it can create value even when it cannot ease the group's overall credit constraint; the mechanism may work better when head office oversees a small, focused set of projects (Stein, 1997).
- In Scharfstein and Stein's (2000) model, division managers can rent-seek, and head office pays them partly through favoured capital budgets rather than only in cash, so strong divisions may subsidise weak ones (Scharfstein and Stein, 2000).
- The BCG matrix says what head office should do: a cash cow's excess cash should not be poured back into it, and a "pet" is evidence of failure (Henderson, 1970). Setting each division's share of capital beside its share of revenue is my way of framing questions, not a test the authors ran.
- Both papers are theoretical models, and I read the working-paper versions, not the journal versions.

In May 1994 Kodak announced it was putting its Sterling Winthrop drug unit and two other businesses up for sale, together about 23% of revenues. In a Wall Street Journal quotation reproduced by Stein, chairman George Fisher explained that Kodak had been passing up too many small investments because it was short of cash, and that keeping both health and imaging would have shortchanged both (Stein, 1997). A CFO holding the budget for a four-division group faces the same question: is the money flowing to the division that deserves it, or to the one that lobbies best? An [internal capital market](https://staging.khoi.co.uk/en/glossary#internal-capital-market) can do either.

## How does an internal capital market differ from a bank loan?

It differs in control. Stein argues that head office holds residual control rights over the projects it finances and a bank lender does not, so head office has both the authority and the incentive to shift funds between projects (Stein, 1997). In his account a bank can create value only by easing credit constraints on the projects it lends to, while head office can create value merely by dividing the same pot of money better (Stein, 1997).

Stein cites Lamont's study of the 1986 oil price fall: the major oil companies cut investment even in non-oil divisions, most strikingly in petrochemicals, while independent petrochemical firms raised it (Stein, 1997). Stein notes this might be efficient reallocation or simply head office subsidising a pet project (Stein, 1997). I have not read Lamont's paper.

## When can focus help head office pick winners?

In some cases: when head office divides a fixed sum among a few divisions and its misjudgments tend to run in the same direction across them. Head office is then mostly ranking projects against each other, so errors that move together do no harm. Stein's example: with two toy divisions, head office may be too optimistic about Christmas demand and still pick the better toy maker; with a toy division beside a biotechnology division, misjudging toy demand would misallocate the money (Stein, 1997). He also presents this as a trade-off: if project outcomes are highly correlated and can be predicted with little error, focus makes an internal capital market less valuable (Stein, 1997).

A hypothetical example, with numbers I made up to show the same logic. A group has 10 billion dong to give to one division. Head office is 10 percentage points too optimistic about the coffee market.

| Division | Return head office believes | True return |
|:--|--:|--:|
| Coffee roasting A | 18% | 8% |
| Coffee roasting B | 12% | 2% |
| Property C | 9% | 9% |

Between A and B the error cancels: A still beats B. Between A and C, the same error makes A look better than C (18% against 9%) when in truth A is slightly worse (8% against 9%). The more alike the divisions, the more head office's errors tend to move together; that is the logic behind Stein's leaning towards focused strategies under some conditions (Stein, 1997).

## Why does the model say a weak division can end up with more capital than it should?

Because in the model division managers can spend their time lobbying head office instead of working. Scharfstein and Stein start from a claim many observers make: that internal capital markets tend to be "socialist", with strong divisions subsidising weak ones. They argue this cannot be explained simply by agency problems at CEO level, since it is hard to believe a CEO gets more private benefit from weak divisions, so they add division managers able to engage in wasteful rent-seeking (Scharfstein and Stein, 2000).

The mechanism runs as follows. A division manager can do productive work or engage in wasteful rent-seeking, and to keep them working head office must pay them. The weaker the division, the lower the manager's opportunity cost of lobbying, so the threat is more credible and the payment larger. Head office is only an agent of outside investors, and it finds it less costly to pay partly by tilting capital spending towards the divisions demanding most, even though investors would prefer payment mainly in cash; when the distortion would become too large, the remainder is paid in cash. The result: the division with the least valuable assets in place gets more than its efficient share of the capital budget (Scharfstein and Stein, 2000).

The authors state their own limits. The result holds under certain conditions, not always. If the strong divisions are run by managers who are better at rent-seeking, the effect can reverse and the strong division gets even more than its share. The model also assumes head office is the only party with authority to allocate capital (Scharfstein and Stein, 2000).

## How can the BCG matrix catch misallocation?

The matrix implies what head office should do; setting it beside the actual budget shows what head office is doing. The [BCG growth–share matrix](https://staging.khoi.co.uk/en/framework/bcg-matrix) rests on the idea that high-share, slow-growth businesses throw off excess cash that should not be reinvested in themselves, while high-growth businesses need cash to grow, and question marks must also buy share (Henderson, 1970). Henderson states bluntly that "pets" are not necessary and are evidence of failure: either the business never won a leadership position during the growth phase, or the group failed to get out and cut its losses (Henderson, 1970).

A hypothetical example: the group name and every number are mine. The group has four divisions and 50 billion dong of capital spending this year.

| Division | BCG cell | Revenue | Share of revenue | Capital spend | Share of capital | Capital/revenue |
|:--|:--|--:|--:|--:|--:|--:|
| Coffee roasting | Cash cow | 200bn | 40% | 5bn | 10% | 2.5% |
| City delivery | Star | 80bn | 16% | 15bn | 30% | 18.8% |
| Restaurant chain | Question mark | 100bn | 20% | 10bn | 20% | 10.0% |
| Contract garments | Pet | 120bn | 24% | 20bn | 40% | 16.7% |

The last row is the one to question. The garment division sits in the "pet" cell yet takes 40% of capital, the most in the group, while producing 24% of revenue; the cash cow receives only 10%.

Scharfstein and Stein propose a test across many conglomerates: whether divisions with the least valuable assets in place (those in low-Q or low-profit industries) systematically invest more than comparable independent firms in the same industry (Scharfstein and Stein, 2000). Comparing a single division with independent firms, as below, is my scaled-down version. Suppose independent garment firms invest about 6% of revenue a year (another invented number): the group's garment division is then investing nearly three times as much (16.7% against 6%) with no reason visible in the matrix. "Who is being humoured?" needs an answer in specific investment opportunities, set against the [opportunity cost of capital](https://staging.khoi.co.uk/en/glossary#opportunity-cost-of-capital), not in the standing of whoever asks.

One caution. Scharfstein and Stein's "weak division" is one whose assets in place are of low value, measured by things like Tobin's Q or profitability (Scharfstein and Stein, 2000); the BCG "pet" is related but not the same idea. The comparison above is my reading, not a test the authors ran, and a gap in shares is only a prompt for questions, since high-growth divisions often take a larger share of capital than of revenue.

## When should a group allocate capital itself, and when not?

It suits a group with few related divisions, where head office's misjudgments are likely to run in the same direction across them (that part is my inference from Stein's argument). Those are the conditions under which Stein's paper leans towards the benefits in some cases (Stein, 1997).

It suits less well a group with many unrelated divisions, including a weak one whose head has real influence over head office. This is my reading, combining two mechanisms: in Stein's account, unrelated divisions make head office's errors less correlated, which reduces the benefit of focus (Stein, 1997); in Scharfstein and Stein's, division managers have a reason to lobby where they can rent-seek, and their model does not address whether divisions are related (Scharfstein and Stein, 2000). The options to compare are keeping the division but forcing more transparent capital requests, or spinning it off or selling it. Scharfstein and Stein present divestment as an implication shared by three models, not a result they analyse separately (Scharfstein and Stein, 2000).

One inference of mine from the model, not a recommendation by the authors: if you reward division managers, tie a cash bonus to the division's results and avoid using the size of the capital budget as the reward. Scharfstein and Stein note that output-based pay only weakens, and does not remove, the tendency to rent-seek (Scharfstein and Stein, 2000).

## Going deeper: what these two papers are, and what they have not shown

Stein (1997) appeared in The Journal of Finance, 52(1), pp. 111–133; Scharfstein and Stein (2000) in the same journal, 55(6), pp. 2537–2564 (Stein, 1997; Scharfstein and Stein, 2000). For the first I read the journal abstract and the full text of NBER working paper 5101 (May 1995); for the second I read the full text of NBER working paper 5969 (March 1997). The details on Kodak, Lamont and the toy example come from the working paper, so the published version may differ.

Both are theoretical models: the conclusions hold under the model's assumptions, such as head office acting as an agent of outside investors. Scharfstein and Stein propose data tests and note that results could be masked because talented managers are assigned to strong divisions (Scharfstein and Stein, 2000). The empirical studies of diversified firms' value that the papers cite, such as Lang and Stulz or Berger and Ofek, I have not read, so this article does not use them. The four-division group and the coffee and property example are hypothetical.

## Frequently asked questions

### Is it always wrong for a weak division to receive a lot of capital?

No. Henderson says a high-growth, low-share business (a "question mark") needs very large cash inputs to buy market share, so heavy funding is reasonable if the business can become a leader (Henderson, 1970). In my view, what deserves suspicion is a weak division receiving more than independent rivals in its industry spend, with no investment opportunity to explain it.

### Is it wrong to reward a division manager with the power to spend capital?

In Scharfstein and Stein's model, that is a channel through which the distortion appears: head office pays division managers partly through a larger capital budget, even though outside investors would prefer to pay mainly in cash (Scharfstein and Stein, 2000). The authors say output-based incentive pay only weakens, and does not remove, the rent-seeking tendency (Scharfstein and Stein, 2000). That is a model result, not empirical proof for every group.

### Should a weak division be spun off or sold?

Scharfstein and Stein say three models (theirs and two related ones) share the implication that divesting weak divisions would increase value; they do not analyse divestiture separately (Scharfstein and Stein, 2000). That is a suggestion, not a rule: in my view the decision must also ask whether the weak division supplies inputs or customers to another division.

### Do these two papers contain empirical data?

They are theoretical models. Scharfstein and Stein outline tests that could be run across many conglomerates, such as comparing the investment of divisions with the least valuable assets against independent firms in the same industry, and say plainly that selection effects could mask the result (Scharfstein and Stein, 2000).

## References

- Henderson, B.D. (1970) *The product portfolio*. BCG Perspectives (The Boston Consulting Group). Available at: https://www.bcg.com/publications/1970/strategy-the-product-portfolio.
- Scharfstein, D.S. and Stein, J.C. (2000) ‘The dark side of internal capital markets: divisional rent-seeking and inefficient investment’, *The Journal of Finance*, 55(6), pp. 2537–2564. https://doi.org/10.1111/0022-1082.00299
- Stein, J.C. (1997) ‘Internal capital markets and the competition for corporate resources’, *The Journal of Finance*, 52(1), pp. 111–133. https://doi.org/10.1111/j.1540-6261.1997.tb03810.x

---

Canonical HTML version: https://staging.khoi.co.uk/en/blog/thi-truong-von-noi-bo-tap-doan