From spending to outcomes: cost-effectiveness in resource-constrained agencies

Cost-effectiveness analysis boils down to a single ratio:

Cost per outcome = Total program cost / Number of outcomes achieved

The trick is defining the outcome precisely and identically across programs. Vague outcomes ("improved wellbeing") cannot be compared. Countable ones ("people housed and still housed after 12 months") can.

Let's run our two homelessness programs. These figures are illustrative, chosen to show the method.

Program A: Emergency shelter

• Budget: $2,000,000
• People served: 200
• Outcome definition: person is stably housed 12 months later
• Observed result: 30 of 200 reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → stable housing

Cost per outcome = $2,000,000 / 30 = $66,667 per person stably housed

Program B: Housing First

• Budget: $2,000,000
• People served: 120 (fewer, because apartments cost more than shelter beds)
• Outcome definition: same, stably housed 12 months later
• Observed result: 84 of 120 reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → stable housing

Cost per outcome = $2,000,000 / 84 = $23,810 per person stably housed

Program A serves more people and looks busier. But Program B produces the target outcome at roughly one third the cost per success.

If the appropriation is truly capped, Program B delivers about 54 more stably housed people for the same $2 million.

This mirrors real findings. Multiple studies have suggested Housing First approaches can be cost-effective because stable housing reduces expensive emergency room visits, jail stays, and crisis shelter use. The exact savings vary by city and are debated, so treat any single number as an estimate, not a law of nature.

Watch the denominator: not all outcomes are equal

Cost per outcome is only honest if the outcomes are genuinely comparable. Three traps:

1. Cream-skimming. If Program B only accepts easy-to-house clients and Program A takes everyone including people with severe untreated illness, B's numbers look better for reasons that have nothing to do with program quality. Always ask who each program serves.

2. Short time horizons. "Housed after 30 days" is easy. "Housed after 12 months" is the outcome that matters. Cheap short-term wins can hide expensive long-term failures.

3. Ignoring avoided costs. A program that costs more up front may reduce spending elsewhere in government (hospitals, courts, jails). Those are avoided costs, and a fair analysis counts them. A narrow agency budget view can make the smarter program look worse.

Cost-effectiveness vs cost-benefit: know the difference

These two terms get mixed up constantly. Define them once and you will never be confused.

Cost-effectiveness analysis (CEA): compares cost per unit of a single outcome measured in natural units (people housed, cases closed, students graduated). You do NOT convert the outcome to dollars. Best when you have one clear goal and want the cheapest path to it.

Cost-benefit analysis (CBA): converts everything, costs and benefits, into dollars, then compares. It answers "is this worth doing at all?" but requires putting a dollar value on outcomes, which can be controversial (what is a year of housing stability worth in dollars?).

For an agency choosing between two programs with the same goal, CEA is usually cleaner and less politically loaded. For deciding whether to launch a program at all, CBA has its place.

The federal government's own guidance, OMB Circular A-94, lays out how agencies should approach benefit-cost and cost-effectiveness analysis. It is dry but authoritative and free.

Building the analysis: a simple worksheet

You do not need special software. A clean table forces discipline. Here is the minimum structure.

The two bottom rows are the honesty check. If you cannot fill them in confidently, your comparison is fragile. Note them as open questions rather than pretending they do not exist.

Making the decision defensible

A capped appropriation means saying no to something. That decision will be challenged, sometimes publicly. Protect it three ways.

Document the outcome definition first, before you see results. Defining "success" after the data arrives invites bias and looks like cherry-picking.

Report cost per outcome with its assumptions. Never hand a decision-maker a single number with no context. "$23,810 per person housed, assuming both programs serve comparable populations and measuring at 12 months" is defensible. "$23,810" alone is not.

Name what you did not measure. Program A's shelter may provide value the ratio misses: immediate safety on a freezing night, a stabilizing entry point. Cost-effectiveness analysis is a powerful lens, not the whole picture. Acknowledge the gaps and you gain credibility.

When cost per outcome can mislead

The ratio is seductive because it produces one clean number. Guard against three failure modes.

Equity blindness. The cheapest-per-outcome program may achieve its numbers by serving the least difficult clients. If your mandate is to reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → the hardest cases, a higher cost per outcome can be exactly right. Efficiency and equity are different goals, and finance staff should present both, not quietly collapse them into one.

Gaming the metric. Whatever you measure, people will optimize for it. If cost per "person housed at 12 months" drives funding, some programs may push out clients who look shaky at month 11. Build in spot checks.

False precision. Outcome counts in social programs carry real uncertainty. Presenting "$23,810" to the dollar implies a confidence you do not have. Round it, or present a range, and say so.

Bringing it back to the budget meeting

Return to the director choosing between two $2 million programs. With cost-per-outcome analysis in hand, the conversation shifts from "which program is more popular" to "which program converts our capped appropriation into the most measurable public value, and who might we leave behind."

That is the entire discipline: same money, more outcome, stated assumptions, named tradeoffs.

Key Takeaways

• Judge outcomes, not outputs. Meals served and shelter nights provided are activities. People stably housed at 12 months is the result that matters.
• Cost per outcome = total cost / outcomes achieved. One ratio, but only honest if the outcome is defined identically across programs and before you see results.
• Check the denominator for traps: cream-skimming, short time horizons, and ignored avoided costs can all distort the comparison.
• Use CEA to compare programs with the same goal; use CBA to decide whether a program is worth doing at all. Do not confuse the two.
• Efficiency is not the only value. The cheapest-per-outcome option may skip the hardest-to-serve. Present equity alongside cost, and name what your ratio does not capture.