A funder reviewing an investment appraisal typically looks for a single headline number: Is the net present value positive? Does the internal rate of return clear the hurdle rate? While that metric matters, in isolation it reveals far less than it appears to. This note sets out what each analytical discipline can and cannot answer, why switching values are the most actionable page in any appraisal, and which three common pitfalls most frequently render a sound-looking analysis misleading.
Three questions, three tools
Economic cost-benefit analysis (CBA) asks whether society as a whole is better off with the project than without it. It accounts for all affected stakeholders, not merely the executing agency. It values project inputs and outputs at their economic opportunity cost: using shadow prices for labour, conversion factors for goods traded in distorted markets, and stripping out taxes and subsidies because they represent transfers between groups rather than net resource consumption. Future flows are discounted at a social discount rate. Its primary outputs are the economic net present value (ENPV), the economic internal rate of return (EIRR), and the benefit-cost ratio (BCR).
Financial analysis asks whether the entity responsible for the project can actually fund it, year by year, in cash. It relies on prevailing market prices (including taxes), tracks only the implementing entity's own cash receipts and outlays, and focuses on cash-flow timing: when upfront grant funding is exhausted, when recurrent liabilities commence, and when financing gaps emerge. A project can be economically outstanding yet financially unviable, or vice versa.
Cost-effectiveness analysis (CEA) identifies which delivery option achieves a predetermined target at the lowest unit cost: cost per child vaccinated, per kilometre of road maintained in passable condition, or per compliance inspection completed. It is the appropriate tool when outcomes cannot be credibly monetised or when the operational target is non-negotiable. However, it cannot determine whether achieving that target represents good value in the first place. Answering that question requires either a monetised valuation of benefits or an explicit political choice.
| Analytical Dimension | Economic CBA | Financial Analysis | Cost-Effectiveness Analysis |
|---|---|---|---|
| Core Question | Is society as a whole better off? | Can the implementing entity pay, year by year? | Which option achieves the target at lowest unit cost? |
| Whose Costs & Benefits | All affected members of society | The implementing agency | The specific programme |
| Pricing Basis | Economic (shadow prices, excluding transfers) | Market prices (including taxes and tariffs) | Prevailing financial prices |
| Primary Outputs | ENPV, EIRR, Benefit-Cost Ratio | Cash-flow balances, funding gaps, financial NPV | Cost per unit of outcome |
| Blind Spot | Cannot identify who funds recurrent costs | Cannot determine if the project is worth doing | Cannot determine if the outcome is worth its cost |
The two worked examples on this site demonstrate the sharp divergence between the first two columns:
- In the rural road maintenance appraisal, the economic NPV is comfortably positive, yet the highway authority must secure an additional US$1.2 million per year on average from its constrained recurrent maintenance budget.
- In the digitisation investment scenarios, the apparent benefit of faster contractor payments collapsed once the agency's own financing cost of early disbursement was deducted: the payment speed-up was primarily a cash transfer from the public purse to private contractors, causing the baseline scenario to fail the economic test. The remaining gains consisted of saved staff time that did not reduce payroll expenditure, while annual software licensing costs opened an ongoing financing deficit.
A funder reviewing only the economic column would overlook the recurrent fiscal hazard in the road programme and the fact that the digital business case rested largely on an internal cash transfer.
Why switching values beat point estimates
An NPV reported as US$5.79 million conveys a deceptive impression of precision. That figure merely represents the mechanical product of a dozen underlying assumptions, several of them speculative. A Monte Carlo simulation offers a more candid representation of risk; however, its headline output—the probability of a negative NPV—remains contingent on user-selected distribution shapes and correlation matrices that are typically assumed away for convenience.
By contrast, a switching value asks an operational question that an executive can act upon: how far would this key parameter have to deteriorate before the investment ceases to be viable? HM Treasury’s Green Book defines switching values as the point to which an input assumption would have to change for an option to forfeit its value-for-money justification (Green Book, para. 2.27). Three characteristics make switching values superior to static point estimates:
- They transform abstract uncertainty into an audit checklist. If an unpaved road rehabilitation project ceases to be viable once traffic falls from 120 to 90 vehicles per day, the funder's priority is not further econometric modeling—it is commissioning an on-site physical traffic count. If an administrative automation project breaks even only if invoice clearance speeds up by 29 days, management must empirically verify the root causes of current payment delays before approving the grant.
- They illuminate the real width of safety margins. The rural road appraisal remains economically justified at a 12% discount rate and even under a zero valuation of travel time, rendering debate over those parameters immaterial. Conversely, it cannot withstand a capital cost overrun exceeding 33%. As the World Bank’s guidance note on social discounting observes, a project that remains viable across high discount rates is structurally robust, whereas one that clears 6% but fails at 6.5% demands urgent scrutiny of its fundamental rationale (World Bank, 2016).
- They expose discrete, non-linear dynamics. Some variables do not alter NPV smoothly. In the gravel road model, a slower pavement deterioration rate defers the scheduled reconstruction by discrete calendar years, causing the NPV curve to jump abruptly. Point estimates mask these step-function behaviors; a switching-value sweep brings them immediately to light.
The EIRR is itself a switching value applied to the social discount rate. It is far more informative when read as a resilience threshold than as an actual rate of return the project will deliver.
Three failure modes
1. Optimism bias
Project appraisals systematically underestimate capital costs and implementation schedules while overestimating operational benefits. The UK Green Book identifies this as a documented institutional tendency and instructs analysts to adjust costs upward and benefits downward before running the appraisal model, rather than as a cosmetic afterthought (para. 2.25).
A credible appraisal should:
- Apply an explicit uplift contingency to baseline capital works;
- Explicitly model a delivery-delay scenario;
- Apply an adoption-attrition factor to benefits that depend on behavioural change (such as staff software utilization or the productive redeployment of saved hours);
- Present net returns both before and after these adjustments. An appraisal base case featuring zero contingency allowances and instant execution should be treated as uncalibrated.
2. Double counting
Economic benefits are easily counted twice under different labels. On an upgraded rural corridor, reduced vehicle operating costs are passed down to farmers via reduced freight rates: counting both the freight cost reduction and the subsequent increase in local farm-gate crop prices tallies the identical economic surplus twice. Similarly, staff time saved through workflow automation is valued once at the shadow wage rate; it must not be added again as incremental administrative output.
Accelerated contractor disbursements represent either a liquidity benefit to the contractor or, if competitive bidding forces the firm to pass the benefit through, lower contract prices for the procuring agency—never both. Furthermore, taxes and duties are transfer payments that must be excised from economic accounting, while induced traffic must be valued strictly according to the rule-of-a-half rather than at the full unit cost saving.
The audit test is straightforward: for every benefit stream, name the specific recipient and the underlying market transaction, and verify that no other line item rewards the same party for the same underlying efficiency gain.
3. Ignoring recurrent liabilities
Many capital investments are justified by socio-economic benefits that materialize only if ongoing maintenance expenditures are sustained long after commissioning: blading for gravel roads, cloud hosting and security patching for software platforms, and consumables and staffing for social clinics.
While an economic CBA incorporates these lifecycle outlays to generate a mathematically sound NPV, political approvals focus predominantly on upfront capital allocations. Recurrent operations fall on domestic line-ministry budgets years after the development funder has exited. This disconnect drives the destructive cycle of build, neglect, and reconstruct, forfeiting the modeled economic returns. The antidote is straightforward: place the multi-year recurrent liability schedule directly alongside the headline NPV, require binding institutional funding commitments, and treat an unfunded recurrent budget line as an active project risk with its own failure scenario.
What to ask for
When an appraisal arrives for executive review, five specific requests separate an actionable decision tool from a purely decorative exercise:
- Demand side-by-side economic and financial balance sheets, paired with a dedicated schedule of recurrent liabilities and named funding commitments.
- Require a switching-value schedule covering the five or six most critical assumptions, with explicit notes explaining how each parameter can be empirically verified in the field.
- Inspect explicit optimism-bias adjustments, reviewing both the unadjusted baseline and the stress-tested figures.
- Enforce an itemized beneficiary mapping for every benefit stream to identify and eliminate double counting and non-economic transfers.
- Request the live computational model itself, enabling independent reviewers to adjust key input assumptions and observe the impact on project returns directly.
The worked models across this site are built to fulfill these five criteria. While they rely on synthetic baselines for open reproducibility, their analytical architecture applies directly to the evaluation of real-world capital programmes.