How to Actually Run a Cost-Benefit Analysis Without Losing Your Mind

Cost-benefit analysis in practice is way messier than the textbooks make it look. Boardman's framework is one of the cleaner approaches out there, but even with a solid structure, you're going to hit edge cases that no formula really covers. I want to walk through how this actually works when you're doing it for real, not how it reads on paper.

What Cost Benefit Analysis Boardman Actually Looks Like in Practice

Anthony Boardman, Aidan Vining, and David Weisbach laid out a methodology in their textbook that's become standard in public policy circles. The core idea is straightforward: identify all costs and benefits, assign monetary values where possible, discount future flows to present value, and compare the net result. Simple in theory. Brutal in execution. The discounting piece is where most people first run into trouble. Pick the wrong rate and your entire analysis flips. A 3 percent discount rate versus a 7 percent rate can turn a project with a negative net present value into one that looks wildly profitable, or vice versa. The choice isn't arbitrary even though it feels like it should be. OMB Circular A-94 gives guidance in the US, suggesting 2 to 3 percent for most public projects. Use something higher if the opportunity cost of capital is genuinely higher in your context. You also need to separate transfers from actual resource costs. If a project taxes one group to pay another, that's a transfer, not a real cost or benefit to society. Beginners miss this constantly. I remember going through a municipal infrastructure proposal where the "cost" included payments to a private operator that were really just redistributions from ratepayers to shareholders. Once I flagged it, the net benefit calculation shifted significantly. The hardest part is putting numbers on things that don't have market prices. Valuing a statistical life, estimating time savings, calculating ecosystem services — these aren't guesswork in the sloppy sense, but they require explicit assumptions and sensitivity testing. If you're not doing at least a basic sensitivity analysis on your key valuation parameters, you're not really doing the analysis. You're doing storytelling with equations. Here's what nobody tells you: the biggest source of error isn't in the math, it's in the scope. Getting the boundary right — deciding what counts, what doesn't, and who falls inside or outside the calculation — usually determines the outcome more than any discount rate tweak. I once worked on a transportation project where the initial analysis excluded commute time savings for a neighboring municipality because they weren't "directly" affected. When we corrected the scope, the project went from barely positive to clearly negative. The model hadn't changed. The map had. One practical trick that helps: write out your value framework before you start crunching numbers. List every category of cost and benefit you intend to include, then go through and mark each one as easily quantifiable, hard to quantify, or effectively incalculable. This forces you to confront what you're about to ignore instead of pretending it doesn't exist.

The Step-by-Step Process

Identify the baseline. This is your counterfactual — what happens if you do nothing. Without a clear baseline, every projected benefit is meaningless because you have nothing to compare it to. The baseline should reflect current policy or the most likely alternative, not a utopian no-intervention scenario. Catalog the effects. Go category by category: direct costs, indirect costs, intangible benefits, distributional impacts. Don't skip the indirect ones. A road project might directly create jobs, but it also changes property values, shifts traffic patterns, and alters local business viability. These ripple effects matter. Monetize what you can. Use market prices when available. For non-market items, apply standard valuation methods: willingness-to-accept or willingness-to-pay surveys, hedonic pricing, travel cost models, or revealed preference approaches. Document every assumption. When someone challenges your number six months later, you should be able to point to exactly where it came from. Discount to present value. Apply your chosen discount rate to all future costs and benefits. The standard formula is NPV equals the sum of net benefits in each period divided by one plus the discount rate raised to the power of that period. Spreadsheet software handles this without issue. Calculate net present value and the benefit-cost ratio. NPV greater than zero means the project passes the basic efficiency test. The BCR gives you a ratio that some decision-makers find more intuitive. Both tell you something different. Report both. Run sensitivity analysis. Change your key assumptions one at a time and see what breaks. If your conclusion flips on a reasonable variation in a single parameter, your result is fragile. That's useful information in itself.

Common Mistakes I See Repeatedly

Double-counting benefits. This is the most common error. A health improvement might show up as reduced medical costs, increased productivity, and higher quality of life. These are three expressions of the same underlying benefit. Count it once, not three times. Ignoring sunk costs. Past expenditures don't factor into a forward-looking analysis. If you've already spent money on a project that's clearly not working, that money is gone. The decision is whether to spend more, not whether to recover what you already lost. Using nominal instead of real values. Make sure your price projections and your discount rate are in the same terms. Mixing them is an easy way to introduce systematic bias. Overstating precision. Your analysis will produce a neat number like $14.7 million in net benefits. That level of precision is false. Round it. Two significant figures is generous at this stage. Failing to account for risk and uncertainty. Expected value calculations assume you know the probability distribution. You rarely do. Monte Carlo simulation or at minimum a range of scenarios gets you closer to reality than a single point estimate.

When This Approach Fails Completely

Cost-benefit analysis Boardman style breaks down when the stakes involve fundamental rights or irreversible environmental damage. You can put a dollar figure on almost anything given enough effort, but some things shouldn't be priced. Air quality standards, species extinction, intergenerational equity — these domains require qualitative judgment alongside quantitative analysis, not in place of it. It also struggles with distributional questions. A project might have positive net benefits overall while concentrating most of the costs on a small, politically vulnerable population. The CBA says proceed. Justice might say stop. Having the analysis doesn't resolve that tension. It just makes it explicit. If your decision involves deep uncertainty where probabilities are unknown rather than merely variable, expected value calculations become misleading. This is the difference between risk and Knightian uncertainty. CBA handles risk fine. Uncertainty is a different problem entirely, and you should flag it honestly instead of pretending the numbers have more authority than they do.

What I Wish I'd Known Earlier

The quality of your analysis depends more on your inputs than your methods. Garbage in, garbage out isn't a catchy phrase here, it's a daily reality. Spend more time on data collection and assumption justification than on the calculation itself. A clean spreadsheet built on shaky foundations is worse than a messy one built on solid ones because it looks more authoritative while being less reliable. Stakeholder input during the scoping phase prevents later surprises. I learned this the hard way on a water quality project where we'd completed a full analysis before realizing a community group considered aesthetic value a primary benefit we'd completely omitted. Adding it required going back and redoing half the work. Twenty minutes of conversation upfront would have saved two days of rework. Finally, document your process thoroughly. Future analysts — or critics — should be able to trace every number back to its source. Transparency isn't just good practice, it's your best defense when someone inevitably challenges your results.