Cash management and cash flow forecasting

Holding cash has an opportunity cost: money in the vault earns nothing while it could be generating returns elsewhere. Cash management balances the liquidity of cash against the profitability forgone by holding it, and the rule is exact — hold cash only until the marginal value of liquidity equals the interest lost. Beyond that point, excess cash is idle capital.

In practice the job is two things at once: optimising the amount of cash available to meet obligations, and maximising the interest earned on surplus funds not needed immediately. The time value of money shapes both. Banks pay more on longer-notice accounts — the longer the money is tied up, the better the rate — and charge more on overdrafts than on term loans, because flexibility is what you are paying for. Treasury switches money between accounts to minimise aggregate cost, netting the interest differentials against the transaction costs of moving the funds.

The Baumol model finds the desirable cash balance when cash needs are known with certainty. It treats cash like inventory: withdrawing from the investment account costs a fixed fee per transaction, while holding cash forgoes interest — and the optimum balances the two, exactly as EOQ balances ordering against holding costs.

Its assumptions are the model's boundaries: disbursements are known and occur uniformly through the period, the opportunity cost is constant, and the transaction cost is constant regardless of the amount moved.

The lecture's example: ABC Ltd disburses ZMW 300,000 a year, the bank pays 10% on money held on deposit, and each withdrawal costs ZMW 20. C = √(2 × 20 × 300,000 ÷ 0.10) = √120,000,000 ≈ ZMW 11,000.

At the optimum the two costs nearly equalise — ZMW 540 of fees against ZMW 550 of interest forgone — which is the signature of the model working. Any higher balance holds too much idle cash; any lower balance churns too many withdrawals.

Miller and Orr relax the assumption Baumol cannot live without: certainty. Their model is stochastic, built for the real world where daily cash flows wander unpredictably. Instead of one optimum balance it sets control limits — a lower limit at which you sell marketable securities to raise cash, an upper limit at which you buy securities to shed cash, and a return point the balance is brought back to after either transaction.

The lecture's example: a lower limit of ZMW 1,000, a daily interest rate of 0.025%, daily cash-flow standard deviation of ZMW 500 (variance 250,000), and ZMW 20 per securities transaction. Z = 3 × ∛(3 × 20 × 250,000 ÷ (4 × 0.00025)) = 3 × ∛15,000,000,000 ÷ … ≈ ZMW 7,400.

Read as instructions: if cash falls to ZMW 1,000, sell securities to bring the balance back to ZMW 3,467. If it rises to ZMW 8,400, buy securities down to the same return point. Between the limits, do nothing — the model's insight is that constant tinkering is itself a cost.

Forecasting is what turns cash management from reaction into planning: estimate future inflows and outflows, generate a pro-forma cash position for the period, then decide in advance how to cover the deficits (borrowing, asset sales) and where to place the surpluses. It serves liquidity management first, but also control — an unexplained gap between forecast and actual is how mistimed payments, collection delays and fraud surface — plus capital budgeting, FX exposure planning and regulatory compliance.

The workhorse is the receipts and disbursements method: schedule expected collections and other inflows, schedule payments — purchases, payroll, taxes, interest, dividends, capex — and net them against the minimum cash balance required. Distribution forecasts refine single large events, using the historical pattern of how (say) a big collection actually arrives over several days. Statistical methods project the routine flows.

The lecture's numbers: five days of collections of ZMW 110,000, 120,000, 115,000, 122,000 and 126,000 give a moving-average forecast of ZMW 118,600 for day six. Day six actually brings ZMW 124,000 — an error of ZMW 5,400. Exponential smoothing with α = 0.40 then forecasts day seven as 0.40 × 124,000 + 0.60 × 118,600 = ZMW 120,760: the forecast leans towards the newest evidence without surrendering to it. Where a cash flow tracks another variable — sales volume, production units — regression (Y = a + bX) forecasts it from the relationship instead.

Surpluses arise from seasonality, uneven project timing, unexpected recoveries. Before placing one, ask four questions: how long can the cash be tied up, how much is available, what return can be earned, and what does early withdrawal cost?

On the borrowing side, the default tool is the bank overdraft: a short-term advance on the current account up to an agreed limit, with interest charged on the daily balance actually drawn. Its strengths are flexibility, light documentation and paying only for what you use; its weaknesses are that it is repayable on demand, secured, and floats with base rates — the healthier the borrower, the better the rate. The forecast is what keeps the overdraft cheap: shortfalls seen in advance can be funded on better terms than shortfalls discovered on the day.

A multi-site organisation holds many accounts across regions and units, and scattered money is hard to use: centralised payments are awkward to fund, surpluses too small to invest well, the true cash position invisible. Cash concentration pools it. There are two families of technique.

Physical sweeping moves the cash. A zero balance account (ZBA) arrangement sweeps every subsidiary account to zero into the concentration account at each day's end, transferring funds back when a subsidiary goes overdrawn. Constant balancing keeps a pre-set minimum in each account and sweeps only the excess. Trigger balances sweep only when an account passes a threshold — fewer transfers, at the cost of more cash sitting decentralised between sweeps.

Notional pooling moves nothing: the bank calculates interest on the combined credit and debit balances of all the accounts while each subsidiary keeps daily control of its own money. Interest is earned on the net pool, intercompany loans and transfer fees never arise, and global pooling can even offset multicurrency balances without FX transactions. The catch is regulatory — notional pooling is not permitted in some countries, including parts of Africa.