Merit increase vs cost of living raise: which one to give, and what each one costs
Give a merit increase when you are paying for performance in the current job. Give a cost of living raise when you are protecting everyone's purchasing power regardless of performance. For most US companies under 200 people the right answer is one merit pool sized with inflation in mind, not two programs, because a true across the board adjustment costs about as much as the entire merit budget and buys no differentiation at all. Reserve a real COLA for the narrow cases where it is genuinely the right instrument: a minimum wage floor moving, a multi-year contract, or a relocation between pay markets.
The two get confused because they arrive in the same envelope on the same day and both look like a percentage. They are answering completely different questions. A merit increase asks "how well did this person do the job", and the honest answer varies by person, which is the entire point of having a pool. A cost of living raise asks "what did prices do", and that answer is identical for everyone in the building. Any attempt to make one number serve both purposes ends up doing neither convincingly.
The cost difference is the whole argument
Take a 40 person company with a $3.6 million base payroll. A 3.4 percent across the board cost of living adjustment costs $122,400 a year, plus roughly $9,364 in employer FICA, and every single employee receives exactly 3.4 percent. Nobody is recognized, nobody is corrected, and the money is gone permanently because a raise is a base increase, not a one time payment.
Now spend the same $122,400 as a merit pool. That is still an average of 3.4 percent, but you can pay 5 or 6 percent to the people you cannot afford to lose, 3.5 to the solid majority, and 1 or 2 percent to those coasting. The average employee experience is identical. The retention effect is not remotely identical, because the people most likely to leave are the ones who received a number that told them something.
That is the practical case against running both. Two programs means either double the money or a merit pool so thin that the differentiation is invisible. A 1 percent spread between your best and worst performer is not a message, it is a rounding error, and employees correctly read it as a company that does not actually distinguish.
The decision, in one table
| Question | Merit increase | Cost of living raise |
|---|---|---|
| What it pays for | Individual performance in the current role | Inflation, applied to everyone equally |
| Who decides the amount | The manager, inside a pool and a matrix | Nobody. It is an index, usually CPI |
| Distribution | Uneven, on purpose | Flat, by definition |
| Cost on a $3.6M payroll at 3.4% | $122,400, allocated | $122,400, untargeted |
| Retention effect | Concentrated where turnover hurts most | Spread evenly, including on leavers |
| Fixes underpayment against market | Only by accident | No |
| Best used when | You have real performance data and a band | A floor moved, or a contract requires it |
The row worth staring at is the sixth. Neither instrument reliably fixes someone who is underpaid relative to the market, because both are percentages of an existing salary. If somebody is at $60,000 in a job that now pays $74,000, a 3.4 percent adjustment moves them to $62,040 and leaves them just as underpaid in relative terms. That is a third category, a market adjustment, and it needs its own budget line and its own conversation. Running it through the merit pool quietly punishes everyone else.
What the federal data actually says
You do not need a paid survey subscription to size this. The Bureau of Labor Statistics publishes both halves for free on a fixed schedule, and they are the numbers a defensible budget memo should cite.
The Employment Cost Index for June 2026 reported that wages and salaries for private industry workers rose 3.1 percent over the year, and 3.2 percent for civilian workers. The ECI is the right benchmark because it holds the mix of jobs fixed between periods, so it measures what happened to pay for the same work rather than what happened when hiring shifted toward different occupations.
The Consumer Price Index for July 2026 reported CPI-U up 3.4 percent over 12 months, after 3.5 percent for the year ending June. Put those two together and you get the sentence BLS printed itself in the same ECI release: inflation-adjusted, or constant dollar, private industry wages and salaries decreased 0.4 percent over the year.
That single line should shape how you talk about this year's increases. The average American private sector worker received a raise and still lost ground. If your pool is landing near 3 percent, you are not being stingy by market standards, and you are also not protecting anyone from inflation. Both things are true, and saying only the first one is how a compensation announcement loses the room. Our salary increase calculator shows the real-terms figure next to the headline percentage for exactly this reason.
When a cost of living raise really is the right call
There are four situations where a genuine across the board adjustment beats a merit pool, and they are narrower than the phrase gets used.
A wage floor moved under you. When a state or city minimum wage increases, employees at the bottom of the scale must move, and if you move only them you compress the people just above. That compression is the actual cost of a minimum wage increase for most employers, and it usually forces a small flat adjustment across the lower grades whether you planned one or not.
A contract or agreement requires it. Collective bargaining agreements and some executive contracts specify an indexed annual increase. That is not a discretionary decision and does not come out of the merit pool.
Somebody moved between pay markets. A person relocating from a lower cost metro to a higher one is a geographic differential question, not an inflation question, even though it usually gets called a cost of living adjustment. The right number comes from the band for that job in the new location.
You skipped a year. After a pay freeze, a flat catch-up adjustment restores the structure before you resume differentiating. Trying to fix two years of drift through a normal merit cycle produces percentages nobody can explain.
Where the money comes from
The pool has to be funded before it can be allocated, and in a company under 200 people there are only three honest sources: revenue growth, headcount you decided not to add, or operating costs you cut. The first is not in your control by budget season and the second has its own consequences, which leaves the third, and for most small US companies the largest controllable recurring non-payroll line is software. Getting a clear read on what the SaaS and cloud stack actually costs each month is often the fastest way to find a percentage point of payroll without touching a single role, because that spend accumulates through renewals nobody re-examines.
The other half of funding is arithmetic people skip: a raise is permanent and it compounds. Employer FICA adds 7.65 percent while pay stays under the $184,500 Social Security wage base for 2026, so a $122,400 pool really costs about $131,764 in year one. And because next year's percentage is computed on the new, higher base, the pool grows every year even if the percentage never does. Budget the pool as a permanent step in run rate, not as a line item that resets.
How to actually decide, in four steps
One, set the pool before you set anyone's number. Pick a total percentage of base payroll you can sustain next year too, using the ECI at 3.1 percent and CPI at 3.4 percent as your reference points rather than a number somebody remembers from a conference.
Two, pull the market adjustments out first. Identify anyone sitting materially below the range for their job, fund those corrections from a separate line, and only then allocate what is left as merit. Skipping this step is the single most common reason a merit cycle feels unfair to everyone involved.
Three, allocate with a matrix, not a gut feel. Cross the performance rating with position in the band so that two people with the same rating get different percentages when one is already paid above midpoint. A merit matrix is the mechanism, and compa ratio is the input that makes it work.
Four, check the after picture across the team, not per person. Individually reasonable raises can still produce pay compression, particularly where a recent hire came in near the top of the band. Look at the whole grade after allocation and before the letters go out, while it is still cheap to fix.
Frequently asked questions
What is the difference between a merit increase and a cost of living raise?
A merit increase rewards individual performance and is distributed unevenly, so strong performers get more. A cost of living raise protects purchasing power against inflation and goes to everyone at the same rate regardless of performance. One is a reward you allocate; the other is a floor you apply across the whole payroll.
How much is a cost of living raise in 2026?
The break even figure is the 12 month change in the Consumer Price Index, which BLS reported at 3.4 percent for the 12 months ending July 2026. A cost of living adjustment below that number still leaves real pay lower than it was, because prices moved further than the salary did.
What is the average merit increase?
The best free US benchmark is the BLS Employment Cost Index. Wages and salaries for private industry workers rose 3.1 percent over the 12 months ending June 2026. BLS also reported that constant dollar private wages and salaries fell 0.4 percent over that year, so the average raise did not keep pace with prices.
Is a 3 percent raise good in 2026?
It is slightly below break even. With CPI-U at 3.4 percent for the 12 months ending July 2026, a 3 percent raise leaves real purchasing power about 0.4 percent lower than the year before. It is close to the private sector average of 3.1 percent, which itself failed to keep pace with prices.
Can you give both a merit increase and a cost of living raise?
Yes, and larger employers often do, with a small across the board adjustment plus a differentiated merit pool on top. It only works if the two are budgeted separately and communicated separately. Blended into one number, employees read the whole increase as a performance verdict and the inflation protection goes unnoticed.
Do employers have to give cost of living raises?
No US federal law requires any employer to give a cost of living raise or any raise at all. The obligations that do bind are different: state and local minimum wage increases, the salary level test for exempt classification, and any raise promised in a written contract or collective bargaining agreement.
What to do this quarter
Write the pool down as one number and one percentage before any manager sees a spreadsheet. Separate the market corrections onto their own line so they stop competing with performance. Then check that your midpoints are current, because every calculation in this article depends on knowing what each job actually pays now, and a band built two years ago will quietly turn a generous merit cycle into a year of standing still. Public federal wage data from the BLS OEWS program is enough to refresh most of them without buying a survey.
Sources: US Bureau of Labor Statistics, Employment Cost Index news release, June 2026 (USDL-26-1270); US Bureau of Labor Statistics, Consumer Price Index news release, July 2026 (USDL-26-1378); Social Security Administration, Contribution and Benefit Base, 2026. Figures are the published national numbers and are not a substitute for your own market data or for advice from employment counsel on contractual or bargained increases.