cost sharing

Cost sharing is the problem of dividing a joint cost among parties who benefit unequally from the arrangement. Unlike simple Allocation - where you distribute a resource across uses - cost sharing asks who pays when the thing being paid for serves everyone, but not equally.

The core tension comes from Game Theory: every party has an Outside Option (going it alone). If the cost-sharing arrangement charges someone more than their standalone cost, they walk away. A valid cost-sharing scheme must satisfy two constraints:

1. 1)No party pays more than standalone - otherwise they leave and the arrangement collapses
2. 2)Total allocation equals the actual joint cost - no surplus left unassigned, no shortfall to cover

Within those bounds, many possible splits exist. The question is which one is principled rather than just negotiated.

Cost sharing hits your P&L in three places:

Internal overhead allocation. Shared infrastructure - platforms, support teams, office space - generates overhead that must be charged back to Cost Centers. If you split a $500K analytics platform equally across three Cost Centers when one barely uses it, you distort which lines of the business look profitable. Bad allocation leads to bad capital discipline decisions downstream.

Post-merger integration. When PE portfolio companies merge Operations, the savings are real but the split determines whether integration succeeds. A warehouse consolidation that saves $1.1M per year sounds great - until the acquired company's leadership feels they're subsidizing the acquirer and quietly sabotages the Turnaround.

Consortiums and shared infrastructure. Building something together - a shared logistics network, a co-developed product - means agreeing upfront who pays what. Get the split wrong and the consortium collapses at the first unexpected cost increase, destroying the surplus that made collaboration worthwhile.

The common thread: an unfair split doesn't just feel bad. It creates incentives to defect, which destroys Value Creation for everyone.

What doesn't work:

Equal split ignores that parties have different standalone costs and different usage levels. If your Outside Option is $100K but your equal share is $167K, you walk away.

Usage-proportional split is better but still arbitrary. Why should cost scale linearly with usage when infrastructure has Fixed vs Variable Costs? And who defines "usage" - queries, storage, headcount?

The principled approach: Shapley value

The Shapley value answers the question: if parties joined the arrangement one at a time in a random order, what's the average marginal contribution each party makes to the total cost?

The intuition: imagine building the shared arrangement by adding one party at a time. The first party in bears their full standalone cost. The second party adds only their incremental cost on top. The Shapley value computes this for every possible ordering and averages the result.

For two parties, Shapley always splits the surplus equally. The cost allocation may be asymmetric (the bigger party pays more in absolute terms), but both sides save the same dollar amount. This makes it a natural starting point in mergers.

For three or more parties, the allocation reflects each party's true marginal contribution to costs - averaged across all orderings - which naturally accounts for asymmetry without requiring any arbitrary weighting.

Why Shapley - the uniqueness result

Shapley (1953) proved this is the only allocation method satisfying four axioms simultaneously:

1. 1)Efficiency - every party's share adds up to the total cost. Nothing left over, no shortfall.
2. 2)Symmetry - parties who add the same marginal contribution in every possible subgroup pay identical amounts.
3. 3)Null player - a party whose presence adds zero marginal cost in every subgroup pays nothing.
4. 4)Additivity - if you combine two separate cost problems, the allocations add up. No one can game the result by splitting or bundling the arrangement.

This is a mathematical result, not a negotiation outcome. No other method satisfies all four.

A separate practical property: in arrangements where combining parties always produces savings (the typical case - sharing a warehouse is cheaper than two warehouses), the Shapley allocation also guarantees no party pays more than their standalone cost. This isn't one of the four axioms, but it's why the method works in Operations - nobody has an incentive to walk away.

Where the cost estimates come from

The Shapley calculation requires knowing what each possible subgroup of parties would pay on its own. For two parties, that's three numbers (A alone, B alone, A+B together). For three parties, it's seven (three standalone, three pairwise, one for all three). These estimates come from the same place any Budget estimate comes from: vendor quotes, capacity models, historical Operating Statement data from similar arrangements, or engineering estimates of shared infrastructure costs. The Shapley arithmetic is straightforward - the hard work is getting credible cost estimates for each subgroup. In mergers and consortium design, this estimation is part of the M&A due diligence process.

Scalability

With N parties, there are N! (N factorial) orderings to consider. For 3 parties, that's 6 orderings - easy by hand. For 5 parties, it's 120 - tedious but manageable in a spreadsheet. For 8 or more parties sharing infrastructure (common in a Holding Company with many Cost Centers), you'd face 40,000+ orderings. Standard approximation methods handle this: sample a few thousand random orderings and average the results. The estimates converge quickly. Shapley is exact and fast for small-N problems (2-5 parties), and approximate but still principled for large-N.

Use cost sharing when:

• •Merging Operations across PE portfolio companies
• •Allocating overhead to Cost Centers that share infrastructure
• •Entering a consortium
• •Any shared resource where at least one party has a credible Outside Option

Decision rule: If walking away is a credible threat for at least one party, you need a formal cost-sharing framework. Without one, the party with the best Outside Option has all the Bargaining power, and the arrangement is one disagreement away from collapse.

If no one can walk away (say, internal Cost Centers that can't spin up their own infrastructure), you still want principled allocation - but the urgency shifts. Now the risk isn't defection; it's Goodhart's Law. Charge a Cost Center too much for shared services and they'll restructure to minimize the internal charge rather than maximize Value Creation.

Don't overcomplicate it when:

• •One party has overwhelming Bargaining power (it's just a negotiation, not a fairness problem)
• •The shared cost is small relative to Revenue (use any reasonable split and move on)
• •The arrangement is temporary with a short Time Horizon (the transaction cost of computing Shapley may exceed the benefit)

Cost sharing extends Allocation from distributing resources to funding shared ones, and applies Game Theory - Outside Options, incentives, stability analysis - to cooperative settings where the question is whether every party stays in. The Shapley value bridges both: marginal contribution from Allocation, stability from Game Theory. Downstream: mergers and M&A due diligence, consortium design, and Efficient Allocation of overhead across any Holding Company sharing Operations across PE portfolio companies.