Why Budgets Need More Than One Forecast
A budget is a plan expressed in financial and operational terms. A forecast is an estimate of what may happen over a stated period, given specified assumptions. Both can be useful for decisions, but neither should hide uncertainty behind a single number. A budget may show the expected annual expense while a plausible high-cost scenario implies a need for additional funding, staff, or capacity.
Risk-informed budgeting makes the assumptions visible. It identifies which assumptions drive the result, distinguishes an expected outcome from outcomes that could require a different response, and shows how the plan changes when assumptions change. This does not mean treating every imaginable event as equally likely. It means using a structured set of relevant assumptions and being clear about their limits.
For a discrete set of mutually exclusive, collectively exhaustive scenarios \(j=1,\ldots,n\), let \(p_j\) be the probability of scenario \(j\), and let \(Z_j\) be the budget result in that scenario. The scenario probabilities must satisfy \(p_j\geq 0\) and \(\sum_{j=1}^{n}p_j=1\). The expected result is:
The same calculation applies to costs, revenues, demand, or resource use, provided the scenario values represent the same quantity and period. A scenario set is only as useful as its assumptions: probabilities, definitions, and scope must be stated, and important omitted effects should not be silently treated as zero.
Set Up the Forecast Before Calculating
A practical forecast begins by fixing the decision and time period. For example, a budget may concern the next financial year and estimate total claim cost, while a capacity forecast may concern completed services in that same year. Then identify the assumptions that can change the result: inflation, volume, unit cost, staffing, capacity, or implementation timing.
Separate ordinary variation already represented in the central estimate from uncertainties that could materially alter the plan. A small change in a low-impact assumption may not justify separate scenarios. An assumption that could change funding needs, service capacity, or the planned operating result usually deserves explicit treatment.
Specify whether the model estimates annual cost, operating result, demand, or another quantity, and use consistent units.
Record the central assumptions and the calculation that turns them into the planned budget.
Choose plausible scenarios that change an important driver, and state the probability or other basis used for each.
Apply the same model consistently, then compare the resulting amounts and resource needs.
Show the range of scenario outcomes as well as the weighted average; an expected value alone may conceal a material shortfall.
For the examples below, all amounts are hypothetical one-year amounts in dollars. No discounting or life-contingent payments are involved, so interest and mortality assumptions do not apply. Each example specifies its discrete scenario distribution and probabilities.
Expected Costs Under Alternative Assumptions
When a cost changes in direct proportion to an uncertain rate, the scenario method is straightforward: calculate the cost under each rate, then weight those costs by their probabilities. This separates the calculation from the decision about how much funding to plan for.
Worked Example: Forecasting Claims Inflation
A hypothetical portfolio has forecast claim cost of $900,000 before next year's inflation. Assume the one-year inflation rate is 3% with probability 0.60 or 7% with probability 0.40. The scenarios are mutually exclusive and exhaustive. Calculate the expected claim cost and assess a proposed budget of $940,000.
Method and assumptions. Use the two-point discrete distribution for the inflation rate: \(P(I=0.03)=0.60\) and \(P(I=0.07)=0.40\). Assume the same $900,000 base cost applies in either scenario and that inflation scales the entire cost by \(1+I\).
Step 1: Calculate each scenario cost. At 3% inflation, the cost is \(900{,}000(1.03)=\$927{,}000\). At 7% inflation, it is \(900{,}000(1.07)=\$963{,}000\).
Step 2: Weight the scenario costs. The expected claim cost is \(0.60(\$927{,}000)+0.40(\$963{,}000)=\$556{,}200+\$385{,}200=\$941{,}400\).
Step 3: Compare with the proposed budget. The expected cost exceeds the $940,000 budget by \(\$941{,}400-\$940{,}000=\$1{,}400\). The high-inflation scenario exceeds that budget by \(\$963{,}000-\$940{,}000=\$23{,}000\).
Check. The expected inflation rate is \(0.60(3\%)+0.40(7\%)=4.6\%\). Applying it to the base gives \(\$900{,}000(1.046)=\$941{,}400\), matching the scenario-weighted result. This check works here because the cost is linear in the inflation rate. The budget comparison is not a recommendation about how much funding to approve; it shows the expected shortfall and the scenario-specific shortfall under the stated assumptions. The common mistake is to weight the inflation rates incorrectly or stop at the expected rate without checking the resulting dollar cost.
Why Average Inputs Can Misstate the Forecast
Averaging the inputs before calculating an outcome is not always equivalent to calculating each scenario outcome and then averaging. The two methods agree for a linear calculation, but not generally when the model includes a cap, threshold, or other nonlinear feature. A capacity limit is a common example: completed services cannot exceed either demand or available capacity.
The reason is that the cap acts on each possible value, not on the average demand. If demand is below capacity in one scenario and above it in another, the averaging operation and the cap do not commute.
Worked Example: Forecasting a Capacity-Constrained Service
A hypothetical service receives $180 per completed service, incurs $110 variable cost per completed service, and has annual fixed expense of $300,000. Demand is 4,000 services with probability 0.50 or 5,000 services with probability 0.50. Capacity is 4,500 services. Calculate expected operating result. Assume no other revenue or cost varies by scenario.
Method and assumptions. Demand has a two-point discrete distribution: \(P(D=4{,}000)=0.50\) and \(P(D=5{,}000)=0.50\). The number completed is \(\min(D,4{,}500)\). Revenue and variable cost apply only to completed services; fixed expense is incurred in either scenario.
Step 1: Find the contribution per completed service. The contribution is \(\$180-\$110=\$70\).
Step 2: Calculate completed volume and result by scenario. At demand of 4,000, completed volume is \(\min(4{,}000,4{,}500)=4{,}000\), so the result is \(4{,}000(\$70)-\$300{,}000=\$280{,}000-\$300{,}000=-\$20{,}000\). At demand of 5,000, completed volume is \(\min(5{,}000,4{,}500)=4{,}500\), so the result is \(4{,}500(\$70)-\$300{,}000=\$315{,}000-\$300{,}000=\$15{,}000\).
Step 3: Weight the scenario results. Expected operating result is \(0.50(-\$20{,}000)+0.50(\$15{,}000)=-\$10{,}000+\$7{,}500=-\$2{,}500\).
Why the shortcut fails. Expected demand is \(0.50(4{,}000)+0.50(5{,}000)=4{,}500\). Substituting that average into the capacity model gives \(4{,}500(\$70)-\$300{,}000=\$15{,}000\), which is not the expected result. Expected completed volume is instead \(0.50(4{,}000)+0.50(4{,}500)=4{,}250\), producing \(4{,}250(\$70)-\$300{,}000=\$297{,}500-\$300{,}000=-\$2{,}500\). The weighted scenario result and this expected-volume calculation agree. The shortcut overstates the expected result by $17,500 because it applies the capacity limit to average demand rather than to each scenario.
Compare Budget Choices Under the Same Scenarios
A proposed change to a plan should be evaluated against the same underlying scenarios as the current plan. Otherwise, a difference in expected results may reflect changed assumptions rather than the actual effect of the decision. Where a decision changes capacity or fixed cost, calculate both plans scenario by scenario.
Worked Example: Comparing an Expansion Budget
A hypothetical service earns $150 per completed unit, has variable cost of $90 per completed unit, and annual fixed expense of $200,000. Demand is 3,000 units with probability 0.40 or 4,500 units with probability 0.60. Current capacity is 3,500 units. An expansion raises capacity to 5,000 units and adds $30,000 of fixed expense. Compare expected operating results with and without expansion.
Method and assumptions. Use the same two-point discrete demand distribution for both plans: \(P(D=3{,}000)=0.40\) and \(P(D=4{,}500)=0.60\). Completed units equal the lesser of demand and capacity. The expansion's additional fixed expense is incurred in both demand scenarios.
Step 1: Calculate contribution. Contribution per completed unit is \(\$150-\$90=\$60\).
Step 2: Calculate the result without expansion. At demand of 3,000, all 3,000 units can be completed, giving \(3{,}000(\$60)-\$200{,}000=\$180{,}000-\$200{,}000=-\$20{,}000\). At demand of 4,500, only 3,500 units can be completed, giving \(3{,}500(\$60)-\$200{,}000=\$210{,}000-\$200{,}000=\$10{,}000\). The expected result is \(0.40(-\$20{,}000)+0.60(\$10{,}000)=-\$8{,}000+\$6{,}000=-\$2{,}000\).
Step 3: Calculate the result with expansion. Fixed expense becomes \(\$200{,}000+\$30{,}000=\$230{,}000\). At demand of 3,000, the result is \(3{,}000(\$60)-\$230{,}000=\$180{,}000-\$230{,}000=-\$50{,}000\). At demand of 4,500, all 4,500 units can be completed, giving \(4{,}500(\$60)-\$230{,}000=\$270{,}000-\$230{,}000=\$40{,}000\). The expected result is \(0.40(-\$50{,}000)+0.60(\$40{,}000)=-\$20{,}000+\$24{,}000=\$4{,}000\).
Step 4: Compare the plans. The expansion raises expected result by \(\$4{,}000-(-\$2{,}000)=\$6{,}000\). Its effect is not uniformly positive: relative to the current plan, the result falls by $30,000 in the low-demand scenario and rises by $30,000 in the high-demand scenario. The expected improvement is \(0.40(-\$30{,}000)+0.60(\$30{,}000)=-\$12{,}000+\$18{,}000=\$6{,}000\).
Check and trap. The expected results from the scenario calculations are \(-\$2,000\) without expansion and \(\$4,000\) with expansion. A common error is to apply the $30,000 expansion cost only in the high-demand scenario. It is a fixed cost of the plan, so it belongs in both scenarios. These figures compare expected operating results only; they do not incorporate other considerations such as implementation timing or costs not specified in the assumptions.
Presenting the Budget as a Decision Tool
A useful forecast communicates both the central estimate and what could make it wrong. A table can make the distinction concrete:
| Forecast element | What to state | Why it matters |
|---|---|---|
| Baseline | Period, measure, and central assumptions | Defines what the planned amount represents |
| Scenario assumptions | Alternative values and their probability or rationale | Makes uncertainty traceable rather than implicit |
| Scenario outcomes | Cost, result, or resource requirement under each alternative | Shows where the plan may be strained |
| Expected outcome | Probability-weighted result under the stated model | Provides a central summary without implying certainty |
| Model limitations | Material effects excluded or simplified | Prevents the forecast from appearing more precise than it is |
The expected amount is useful for summarizing a modeled set of outcomes, but the range and operational implications can matter just as much. For example, a budget may have a manageable expected expense but include a plausible scenario that requires short-term funding or additional service capacity. A sound presentation states that need plainly rather than implying the average will cover every scenario.
Do not mix a scenario estimate with a certainty. If probabilities are judgments rather than measured frequencies, identify them as assumptions. If a scenario set omits a material driver, do not interpret the expected value as a complete forecast. Where several assumptions change together, use coherent combinations rather than combining individually favorable or unfavorable values in a way that cannot occur under the stated plan.
Common Budgeting Traps
- Using an average input in a nonlinear model. With a capacity cap, threshold, or other nonlinear feature, calculate each scenario first and then average the outcomes.
- Reporting only the expected value. A weighted average can conceal the size of a scenario-specific funding or capacity gap.
- Changing the comparison basis. Compare alternatives under the same scenario assumptions so the difference reflects the decision being assessed.
- Omitting costs in one scenario. A fixed cost incurred under the plan belongs in every scenario in which it applies.
- Using inconsistent units or periods. Annual expenses should not be combined with monthly volumes unless the time basis is reconciled.
- Giving probabilities false precision. Make clear whether scenario probabilities are evidence-based estimates or planning judgments.
Check Your Understanding
For each problem, use the stated assumptions and calculate scenario outcomes before taking any weighted average.
- A hypothetical portfolio has a one-year base claim cost of $1,200,000. Inflation is 2% with probability 0.75 or 6% with probability 0.25. What is expected claim cost? (Answer: $1,236,000.)
- A hypothetical unit receives $240 per completed service, incurs $150 variable cost per service, and has $180,000 fixed expense. Demand is 1,500 services with probability 0.40 or 2,500 with probability 0.60; capacity is 2,000. What is expected operating result? (Answer: -$18,000.)
- A hypothetical operation has contribution of $45 per completed unit and fixed expense of $120,000. Demand is 2,000 units with probability 0.30 or 4,000 units with probability 0.70; capacity is 3,000. Calculate expected completed volume and expected operating result. (Answer: 2,700 completed units; $1,500 expected operating result.)
- A hypothetical expense forecast has a $500,000 base amount. Costs rise by 4% with probability 0.50 or 10% with probability 0.50. What is expected cost, and by how much does it exceed a $530,000 budget? (Answer: $535,000 expected cost; $5,000 above budget.)
- A hypothetical department expects workload of 800 hours with probability 0.70 or 1,200 hours with probability 0.30. It plans capacity for 920 hours. What is expected workload, and how many hours of capacity are insufficient in the high-workload scenario? (Answer: 920 expected hours; 280 hours insufficient in the high-workload scenario.)