An illustrative example: a reunion in a church basement, and two cousins put starting salaries on a paper napkin, one from the late 1970s and one from last spring. The older number looks small. The newer number looks large. Neither number is a standard of living. A fair comparison across decades has to say what is being held constant. Prices are the first adjustment people remember. Taxes, benefits, and hours are the three that quietly decide the result after the price adjustment is done.
Prices first, and only prices so far
The sticker wage is nominal. To ask what basket it matched, scale it by a published price index, using one series at both ends. For a national U.S. consumption comparison, the usual public tool is the all-items CPI-U, U.S. city average, not seasonally adjusted, which is what the BLS inflation calculator uses. The arithmetic is a ratio of two index levels. It is explained without a sample index value in purchasing power versus nominal dollars, and the ways a form can smuggle in a different series are in how inflation calculators actually work.
Stop at that step and you have answered a limited question: how the sticker wage compares after repricing a national urban consumption basket. You have not answered how much of the paycheck survived taxes, what the employer added in benefits, or how many hours the salary bought. BLS is plain that the CPI excludes income and Social Security taxes, because those are not the price of a consumer good, and that it excludes investment items such as real estate. A salary comparison that stops at CPI-U has ignored the tax line on purpose. Say so, or do the next steps.
Which CPI belongs in a legal formula is a separate choice. Social Security's COLA follows the CPI-W from third quarter to third quarter, on the rule the Social Security Administration describes for years after 1983. Federal income-tax parameters follow the all-items CPI-U. A family argument about "what people earned" is not either of those formulas. Labeling your napkin CPI-U when you meant the benefit formula, or the reverse, imports a population you did not intend. The CPI-W covers about 30 percent of the population. The CPI-U covers more than 90 percent. The BLS FAQ states both shares.
Taxes are a second ledger
A before-tax salary and an after-tax salary are different objects. The CPI will not convert one into the other. Payroll taxes, federal income tax, state income tax, and credits all sit between the offer letter and the deposit. They have changed across decades for reasons that are legislative, not statistical. Two sticker wages that look equivalent after a CPI scaling can leave different deposits once the tax code of each year is applied.
This page will not invent a tax bill for a sample household. The fair method is to use the actual tax rules of each year, or a published historical series you can cite, and to keep the price adjustment and the tax adjustment in separate columns. If you cannot do the tax column, the honest sentence is "CPI-adjusted, before tax," not "what the job was worth."
A further trap is comparing one person's salary with a household's income. A single earner in one decade and a two-earner household in another are not a pay raise. They are a change in how many people worked. Keep the unit fixed: one job, or one household, and say which.
Benefits are not in the hourly cash average
Employer health coverage, retirement contributions, and paid leave changed the composition of pay across the late twentieth century and after. Cash wages can look stagnant in a period when a growing share of compensation arrived as a premium paid to an insurer. The reverse can happen when a plan is cut. The CPI medical index prices out-of-pocket medical spending. It does not record the premium an employer paid on a worker's behalf. That premium shows up more naturally in compensation statistics than in the CPI.
BLS maintains the distinction on purpose. Current Employment Statistics average hourly earnings exclude benefits and the employer's payroll taxes. The Employment Cost Index includes the cost of benefits and holds a fixed basket of jobs, so a shift in who is employed does not look like a raise. A decades-long comparison that uses only average hourly earnings is a cash comparison. A comparison that uses the ECI is an employer-cost comparison. Wage growth that outruns prices separates those two publications for a single business cycle. Across decades the same split applies, with the added warning that benefit designs are not a single price.
If you are comparing two specific jobs rather than national averages, ask for the benefits summary in each year: what the plan covered, who paid the premium, and whether retirement was a pension, a contribution, or nothing. A salary line that rose in CPI-adjusted dollars while a pension disappeared is not an unambiguous gain. You do not need a fake dollar value for the pension to refuse the unambiguous claim.
Hours and the shape of the job
A salary is often quoted per year. Living standards are closer to pay per hour, and to pay per hour after commuting. An annual salary that rose while the job absorbed evening work can be a decline in hourly pay. CES publishes average weekly hours beside average hourly earnings because the two margins move separately. Your own comparison needs the same pair: dollars and hours.
Overtime, unpaid extra time, and weeks worked per year belong in the hours column. A school-year contract and a twelve-month contract are not the same denominator. A union scale that specifies a weekly hour count and a professional salary that does not are not the same kind of offer. Write the hours down before you scale the dollars. If the older generation cannot remember the hours, the comparison is unfinished, not "roughly fine."
Occupation and credentials are part of the job, too. A starting wage for a teenager and a starting wage for a licensed trade are different products. National averages mix them. The ECI's fixed basket is one way statistical agencies refuse that mix. A family story can refuse it by naming the occupation at both ends.
A sheet that keeps the adjustments apart
| Column | What goes in it | What it is not |
|---|---|---|
| Sticker pay | The offer, with the year and the period (hour, week, year) | A standard of living |
| Price scaling | One named index, same series at both dates | A tax adjustment |
| Hours | Hours required in a normal week, and weeks in the year | Something the CPI measures |
| Benefits | What the employer paid for or promised, in words and in any documented dollars | Out-of-pocket medical CPI |
| Taxes | Withholding and income tax under that year's rules, if you can document them | A guess |
| Result | A sentence that lists every column you actually filled | "People were richer then" |
Leave a cell blank rather than filling it with a round number you hope is close. Blank cells keep the sentence honest. The result line should name the blank cells. "CPI-U adjusted, before tax, benefits unknown, hours assumed equal" is a responsible finding. It is weaker than a reunion speech, and it is stronger than a false one.
Housing needs its own warning inside the price column. Shelter in the CPI is rent and owners' equivalent rent, not the price of buying a house. Two salaries can look aligned after CPI-U while the deposits required to buy a house in those two years tell a different story. That second story uses a house-price index such as the FHFA series, and it should stay in its own column. The reason the columns diverge is house prices, rent, and the same number trap.
Geography and household size
A national index erases place. A salary in a coastal metro and a salary in a small city can be identical in CPI-U dollars and unequal in rent. BLS publishes some metropolitan price indexes, and they are the closer tool when the question is local prices rather than a national basket. Even then, a metro CPI is an average for that area, not your lease.
Household size changes what a salary has to cover. The same real wage supports a different life for one adult and for an adult with children. Per-person comparisons divide by people. They do not divide by the CPI. If the cousins at the reunion are comparing one income against a household that included a second earner and two children, they are comparing unlike units. Say the unit out loud before the index gets the blame.
The glossary entries for purchasing power, cost of living, and inflation are worth reading beside the sheet, because "cost of living" in casual speech includes taxes, housing deposits, and family size that the CPI deliberately leaves in other statistics. BLS itself says the CPI is not a complete cost-of-living index.
Limit of the comparison
A careful salary comparison is a set of labeled columns, not a verdict on a generation. Tax rules, benefit plans, and local prices have to be documented for the years you care about; this article does not supply those documents or a personal calculation. It is general education, not tax advice and not a claim about what anyone should have been paid.
Which column on your own napkin is still blank: prices, taxes, benefits, or hours?