Table of Contents

Cash management models turn a familiar treasury problem into a decision rule: how much cash should a business keep available, and when should it transfer money between cash and an earning asset? Large cash balances reduce the need for frequent transfers but give up interest. Small balances earn more elsewhere but can create transfer costs and liquidity pressure.

The Baumol model answers this question for steady, predictable cash use. The Miller-Orr model supplies a different rule for uncertain cash flows: let the balance move within a band, then restore it to a return point when a boundary is reached. Both are useful benchmarks whose assumptions matter as much as their formulas.

What Do Cash Management Models Decide?

The decision concerns the allocation between spendable cash and an alternative earning asset, rather than how much revenue or profit the business should generate. A transfer might redeem an investment to fund a bank account or move surplus bank cash into an approved investment. The basic models simplify those choices to two assets with a fixed cost each time money is moved.^[1]

The relevant opportunity cost is the difference between the return available on the alternative asset and the return earned on cash, with comparable risk and availability. It is not automatically the company’s weighted average cost of capital or the inflation rate. Inflation can influence the incentive to economize on money holdings, but the costs of inflation are a broader question.

These models sit within working capital management. A cash conversion cycle describes when operating money is tied up. A cash-management model asks how to manage an available balance given the resulting cash flows, investment alternatives and transfer costs.

The Baumol Cash Management Model

The traditional Baumol model treats cash like an inventory that is consumed at a steady, known rate. The firm periodically converts part of an earning asset into a cash lot, spends that lot down, and replenishes when the discretionary cash balance reaches zero. Large lots mean fewer conversions but more average idle cash.^[2]

The basic setup assumes predictable net cash use, a constant opportunity-cost rate, a fixed replenishment fee and no material delay in obtaining cash. It also assumes funding is available and sets aside price risk, taxes, proportional trading costs and unexpected shortages. Net cash use is the cash drain to be funded after relevant operating receipts; it is not automatically sales or total accounting expenses.

The Baumol Formula and Its Units

Choose a planning period first, such as one year. Let T be net cash use during that period, F the fixed cost per replenishment, C the cash amount transferred each time, and r the opportunity-cost coefficient for the same period, expressed as a decimal.

• T / C: modeled replenishments per period.
• C / 2: average discretionary cash when it falls steadily from C to zero.
• F × T / C: total replenishment expense.
• r × C / 2: the period’s forgone return under the model’s linear holding-cost convention.

Supplier payments and payroll consume the cash, but their principal amounts are not added to this objective: those operating payments occur regardless of the lot size. The objective compares the incremental costs of managing the balance.

With annual T and annual r, the answer is a dollar lot, not dollars per year. F times T has dollar-squared units, and the square root returns dollars. Monthly cash use must instead be paired with a matching monthly opportunity-cost coefficient. The inventory logic resembles the economic order quantity model, with cash replenishment replacing an inventory order.^[2]

Baumol Example: Choosing the Transfer Size

Consider a hypothetical business with $1,200,000 of steady net cash use per model year. Each replenishment costs $60. Assume a 4.00% annual opportunity-cost coefficient, applied linearly to the average cash balance for that year. This is an illustrative cost assumption, not a current market yield. The base case has no required reserve, and the earning asset is assumed available for immediate conversion.

Now hold annual cash use, the fee and opportunity cost fixed, and vary only the transfer lot. The chart compares total annual relevant cost with its two components. Numeric horizontal spacing reflects the actual difference between lot sizes; this is a cost schedule, not cash over time.

Cash lots balance two costs (hypothetical)

Cash transfer lot (USD)	Total cost (USD/year)	Transfer cost (USD/year)	Holding cost (USD/year)
20000	4000	3600	400
30000	3000	2400	600
40000	2600	1800	800
60000	2400	1200	1200
80000	2500	900	1600
120000	3000	600	2400

Hypothetical Baumol schedule: annual use $1.2 million, $60 per replenishment and 4% linear annual holding cost. Lines connect calculated sample lots.

At a $20,000 lot, frequent transfers dominate; at $120,000, holding cost dominates. The $60,000 optimum balances both at $1,200. A nearby $80,000 lot costs $2,500, illustrating that a precise mathematical optimum can sit within a relatively shallow cost curve.

The result also passes a calculus check: the first derivative is −FT / C² + r / 2, which equals zero at $60,000. The second derivative, 2FT / C³, is positive for C > 0. Thus the interior solution minimizes this model’s cost function, rather than merely beating the few lots plotted.

Twenty equal cycles imply an average interval of 1 / 20 year, or 18.25 days on a 365-day illustration. An actual payment calendar may require different dates or rounded lots. If a separate fixed reserve must always remain in the account, total average cash becomes reserve + C / 2 under an otherwise unchanged additive model; other constraints can require reoptimization.

The Miller-Orr Model: Control Limits and a Return Point

When receipts and payments vary, cash need not follow a steady decline. The simple zero-drift Miller-Orr model instead assumes unpredictable net operating cash changes with a stable variance, serially independent increments and zero expected change. Transfers between the cash account and earning asset are excluded when measuring those operating changes.^[1]

The policy uses three cash levels:

• L, the lower limit: a minimum balance chosen outside the basic optimization.
• Z, the return point: the balance restored by a transfer when a boundary is reached.
• H, the upper limit: the level that triggers a transfer of excess cash into the earning asset.

Inside the band, the model does not trigger a transfer merely because cash differs from Z. This avoids paying the fixed fee to adjust every small fluctuation. The original paper assumes negligible transfer lead time and a fixed fee that is the same in either direction. It also examines extensions beyond the zero-drift case; the compact formula below is for the simpler case.^[1]

The Cube-Root Distance

Let F be the fixed transfer fee, σ² the variance of net cash changes per model day, and i_d the daily opportunity-cost rate in decimal form. Define d as the distance from the lower limit to the return point:

The exponent 1/3 means cube root. Some presentations call 3d the “spread”; confusing it with d triples the wrong quantity. Also, σ² is variance, not standard deviation. With variance measured per unit time and the opportunity rate per the same unit time, Fσ²/i has dollar-cubed units, giving a dollar distance after the cube root.^[1][3]

Miller-Orr Example With a Daily Rate

For a separate hypothetical business, choose a $20,000 lower cash limit and assume a $60 fee per transfer in either direction. The standard deviation of one-day net operating cash changes is $6,000. Assume cash earns no interest and the earning asset yields 4.00% effective annually, with 365 equal model days per year. The asset’s yield is therefore the opportunity cost of cash. These parameters are invented, not estimates for a real firm.

First convert the rate consistently:

Dividing 4% by 365 would be a simple-rate approximation, not this exact effective-rate conversion. A quoted nominal rate with a specified day-count basis may use a different conversion. The difference between effective and nominal interest rates determines which calculation is appropriate. The preceding Baumol example separately specified a linear annual cost coefficient.

All calculations use the unrounded daily rate and cube root before rounding dollars to cents. The original model minimizes long-run average transfer and holding costs within its assumed policy framework; the $44,703.38 figure is not an independently estimated liquidity guarantee.

How Much Should Be Transferred?

Apply the calculated levels to the same hypothetical business. Assume the balance reaches a boundary exactly and funds can be moved as the model requires. The amounts below are transfer principal; the $60 fee is an additional management expense funded separately from these illustrated account movements.

The upper-bound transfer is twice the lower-bound transfer, before rounding: H − Z = 2d, while Z − L = d. The return point is therefore below the band’s midpoint. Returning to the midpoint, to L, or to H would be a different policy.

The model’s long-run average cash balance is also different from Z: L + 4d / 3 = $52,937.84 here.^[1] That is a theoretical steady-state average, not an average calculated from an observed company cash path.

Real balances can jump past a limit between checks. If observed cash is X below L, restoring Z requires Z − X, rather than automatically transferring d; above H, it requires X − Z. Delays, jumps and funding constraints also challenge the assumptions behind the classical limits. They must be addressed before implementing the rule.

How Cash-Flow Volatility Changes the Band

Return to the Miller-Orr example and hold the $20,000 lower limit, $60 fee, 4% effective annual spread and 365-day convention fixed. Vary only the standard deviation of one-day net cash changes, squaring each value before calculating the limits. This parameter comparison shows how greater uncertainty changes the policy; it is not a forecast, probability distribution or simulated cash path.

Volatility widens the band (hypothetical)

SD of one-day net cash change (USD)	Return point (USD)	Upper limit (USD)	Lower limit (USD)
3000	35562.15	66686.46	20000
6000	44703.38	94110.14	20000
9000	52370.58	117111.75	20000
12000	59214.17	137642.51	20000

Hypothetical Miller-Orr sensitivity; L and all cost/rate inputs fixed. Variance equals squared daily SD; limits rounded to cents after calculation.

Doubling the daily standard deviation from $6,000 to $12,000 quadruples variance. The distance d rises from $24,703.38 to $39,214.17, about 1.5874 times, because it responds to the cube root of variance. The lower limit remains $20,000 because it was supplied as an input.

Higher fixed transfer fees also widen the band; a higher opportunity-cost rate narrows it. These are model comparative statics. A treasury team may separately raise its minimum balance when liquidity risk increases, so a fixed-L sensitivity does not describe every real response to stress.

Baumol vs. Miller-Orr

Both models compare the cost of cash with the cost of transfers, but they organize the decision differently. Neither replaces a forecast of known obligations.

Scheduled tax payments, payroll runs and seasonal collections contain information that the pure random-flow setup does not exploit. Miller and Orr themselves discuss forecasts and more complex cash rules in the original paper. Treating predictable events as random noise can produce a mathematically tidy answer with little operational value.^[1]

Using the Models in Real Treasury Work

Start with a cash forecast, then determine whether a simple transfer rule adds value. As a concrete practice reference, the Government Finance Officers Association recommends ongoing forecasts, updated assumptions and comparisons of actual cash flows with projections for governments. That is government guidance, rather than a corporate legal mandate, but it illustrates the role of forecasting that a formula alone cannot fill.^[4]

1. Define accessible cash. Separate available balances from restricted, pledged or currency-specific funds. The same consolidated balance may not be spendable by every entity.
2. Forecast known payments and receipts. Link the near-term cash calendar to broader pro forma financial modeling, including payroll, taxes, financing and unusual transactions.
3. Set the minimum independently. Consider contractual compensating balances, operational reserves, payment cutoffs and risk tolerance. The cube-root formula does not decide these requirements.
4. Specify the funding/investment mechanism. Compare actual money market instruments, settlement times, redemption rules, principal risk and counterparty limits. A positive yield is not the only criterion.
5. Check the credit backstop. A line of credit has a limit, conditions, fees, possible collateral requirements and draw timing. It changes funding choices but is not automatically cash on demand.
6. Estimate and challenge inputs. Exclude treasury transfers from operating-flow variance, distinguish predictable patterns, and use sensitivity analysis for uncertain costs or volatility.
7. Monitor outcomes. Compare realized shortages, transfer counts, holding costs and exceptions with the benchmark; revise the policy when conditions change.

For multinational groups, pooling, netting, currency constraints and legal-entity access add another layer. Those organizational questions are covered in international trade finance and cash management. Optimizing one unconstrained account is a different task from making cash available across a group.

Common Mistakes With Cash Management Models

1. Confusing the transfer lot with average cash. Baumol’s C* is the replenishment amount. With steady depletion to zero, average discretionary cash is half that amount.

2. Mixing daily flows with an annual rate. Miller-Orr’s variance and opportunity rate must use the same time basis. Define whether the rate is effective or nominal and whether the data use calendar or operating days.

3. Using standard deviation instead of variance. A $6,000 daily standard deviation becomes 36,000,000 in dollar-squared variance units. Substituting 6,000 directly changes both the dimensions and the result.^[3]

4. Taking the square root in both models. Baumol uses a square root. The Miller-Orr distance uses a cube root.

5. Calling d the full spread. Here d = Z − L. The complete band is H − L = 3d, and the upper transfer is 2d.

6. Treating the lower limit as an output. L is specified before the simple Miller-Orr calculation. A contractual or operational minimum cannot be discarded to make the calculated balance look efficient.

7. Ignoring settlement or funding constraints. A theoretical instruction to sell an asset is not proof that cash will arrive before an obligation is due. A constructed cash path would illustrate rule execution, not establish statistical validity or optimality.

Limitations of Cash Management Models

Unstable cash flows. Seasonal patterns, trends, serial dependence and rare large payments can undermine the steady-use or stationary zero-drift assumptions. Historical variance can also miss future stress. Different processes or costs may require different control rules.^[1]

More complex costs. Proportional trading charges, different fees for investing and raising cash, taxes and changing yields are outside the simplest formulas. Fees do not disappear because transfers are digital. Very low rates or high fees can also favor a boundary solution rather than a useful interior transfer lot.

Operational feasibility. Transfer approval, bank cutoffs, asset settlement, market closures, credit covenants and restricted funds impose constraints absent from a frictionless two-asset model. Minimum balances may need to change over time.

Precision and scope. A return point printed to cents can rest on uncertain estimates. These models address cash-allocation costs; they do not choose the firm’s capital structure, guarantee its solvency or replace management judgment about liquidity risk.

Sources

Sources consulted November 5, 2026. Examples are hypothetical U.S.-dollar educational cases. Historical model papers support the analytical framework, not current rates, settlement availability or a real firm’s optimal balances.

1. Merton H. Miller and Daniel Orr, A Model of the Demand for Money by Firms, Quarterly Journal of Economics 80(3), August 1966, pp. 413–435. University-hosted copy; Baumol setup pp. 413–415, assumptions and control policy pp. 417–420, solution and average balance pp. 422–424, applicability and forecasts pp. 429–432.
2. Kevin D. Hoover, Duke University, Baumol’s (1952) Model of the Transactions Demand for Money, two-page academic handout; assumptions, units, costs and square-root derivation. No publication date displayed; this is an exposition, not the original 1952 article.
3. Association of Chartered Certified Accountants, Financial Management: September/December 2024 Examiner’s Report, Guillermo Co requirement (c), pp. 18–19; variance, matched daily rate and spread/return-point terminology. Its exam rate convention differs from the expressly defined effective-rate example here.
4. Government Finance Officers Association, Using Cash Forecasts for Treasury and Operations Liquidity, board approved September 30, 2019; government forecasting guidance, regular updates and actual-versus-projected comparisons.

Keep reading

More in Accounting

• GAAP Conceptual Framework for Financial Reporting

  The GAAP conceptual framework is the foundation of U.S. financial reporting. Learn about qualitative characteristics, recognition criteria, measurement bases…

  16 min read

• Income Statement & Comprehensive Income: Structure, Components & Reporting

  A complete guide to the income statement and comprehensive income under U.S. GAAP — income statement structure, OCI components, discontinued operations…

  15 min read

• Statement of Cash Flows: Direct Method, Indirect Method & Preparation

  Learn the statement of cash flows indirect method and direct method with a complete worked example. Covers operating, investing, and financing activities…

  16 min read