Real wages and wage distributions
14 min read
Lesson objectives
- Deflate nominal pay and compare mean, median and hourly earnings with explicit coverage.
Please sign in to use all the tools
Once you sign in, the reading tools are yours: larger or smaller text, typeface, line spacing, focus mode, highlighting and bold, and notes saved to your account and gathered in My study.
A worker's monthly pay rises from 1,000 to 1,080 currency units. Is the worker better off? The nominal increase alone cannot tell us: prices and hours may also change. Nominal earnings are the money amount for a period. Real earnings express their purchasing power relative to a defined price basket. In a simple example with a base-year price index of 100, divide nominal earnings by the index and multiply by 100 to express earnings in base-year purchasing power. This is a comparison measure, not an amount the employer will pay. Use a consistent price index, earnings concept and period.
Even with stable prices, an 'average wage' can mislead. Suppose five workers earn hourly amounts of 10, 11, 12, 13 and 54. Their arithmetic mean is 100÷5 = 20, but the median, the middle ordered value, is 12. One unusually high wage lifts the mean much more than it describes a typical worker in this group. That does not make the median superior for every question. Total payroll and mean earnings matter for an employer's spending; the median is useful for locating the centre without being pulled strongly by an extreme value. Hours matter too: a monthly 2,400 for 160 hours means 15 per hour, whereas the same 2,400 for 200 hours means 12 per hour.
ILOSTAT's metadata show why two numbers should not be compared merely because both are called 'wages'. They may be monthly or hourly, gross or net, and an establishment survey may cover registered workers in larger firms while a household survey reaches more workers but relies on recall. Workers' earnings also differ from employers' labour cost, which can include contributions that are not employees' earnings. Before drawing a conclusion, write down the unit, period, survey source, coverage and price base. Nor should a statutory minimum from one jurisdiction be treated as a universal rule: its coverage and design differ.
Two sheets arrive before a pay review. One says earnings increased; the other says workers can buy less. Both can describe the same situation. First name your question: the cash amount, purchasing power, or earnings per hour? Each requires different inputs. Separate these questions before ranking workers or years. We will reconcile a payroll record with hours and prices, then examine changes in the people covered. The figures below are fictional teaching data in abstract currency units. Keep the original amount beside each calculated value so a colleague can spot a transcription error before debating its meaning. When a result looks surprising, check the unit first, the arithmetic next, and the economic explanation afterwards. This sequence avoids a lengthy argument about a change produced solely by inconsistent definitions.
Start with payroll components. Suppose basic pay is 2,000, a travel allowance is 200, and a bonus is 300. Their cash total is 2,500. If you compare gross earnings, hold the included components constant across periods. Take-home pay requires the actual deductions. An employer contribution to a protection fund increases labour cost but may fall outside your employee-earnings definition. Read the source definition before adding it. This matters when employer costs rise because of a contribution while the employee's cash receipts stay unchanged. Keep gross earnings, net earnings and labour cost on separate lines, with their components listed. Ask whether a bonus is regular or relates to a longer period. Comparing an annual-bonus month with an ordinary month may reveal payment timing rather than a change in typical pay.
A price index has a base year that sets its scale and a basket that defines its coverage. Rescaling the same series preserves growth rates. If two index values are 100 and 120, expressing them as 50 and 60 leaves their ratio at 1.2. Use the ratio between price indices when calculating real-earnings growth. Changing the basket or coverage changes the comparison itself and requires a separate explanation. A household facing a particularly large housing-cost increase can experience something different from the basket average. The general index describes its defined average; household experience requires expenditure information. State the index name, period and base year. When both annual and monthly indices are available, choose a period consistent with the earnings measure. Matching periods is part of correct calculation, especially when bonuses shift between months or prices move rapidly.
When combining hourly earnings, decide whether you want the average across people or earnings per paid hour. One person works ten hours at 20 and another works forty hours at 10. Their equally weighted average hourly rate is 15. Total earnings, however, are 600 over fifty hours, giving 12 per hour. The second calculation gives greater weight to the person with more hours. Put the weighting rule beside the result so readers can reproduce it. Survey weights representing population counts may add another layer; distinguish these from hours weights because they serve different purposes. Check whether the source reports paid hours or hours actually worked. Paid leave may fit a paid-earnings question, whereas work-intensity analysis needs a different description. Let the question determine the denominator, then keep that definition stable across periods.
A grouped-data median requires ordering people by pay and weighting each category by its headcount. Suppose thirty workers earn 10 and ten earn 30. Averaging the two listed rates gives 20, but ignores that most workers belong to the lower category. In the ordered set of forty people, both central observations lie among those earning 10, so the median is 10. Mean earnings are 15. With broad pay bands, the middle observation may fall inside a band whose boundaries alone are known. State that limitation or explain an estimation method. A displayed number can suggest more precision than its inputs permit. Keep category counts so an auditor can locate the middle, and distinguish an estimated band midpoint from a median calculated from individual records.
Real files contain blank cells and zero values. Establish their meaning before averaging. A zero might mean no earnings were paid in the reference period, or it might code nonresponse. Treating a missing-data code as actual pay artificially lowers the mean. Follow the source documentation and retain the count of excluded observations. Suppose nine of ten records report earnings of 1,000 and the tenth is unanswered. The respondent mean is 1,000, covering nine people. Entering zero for the tenth produces 900 by adding information the person never supplied. Excluding missing values still leaves possible differences between respondents and nonrespondents. Record response coverage and treatment rules, then describe the result as applying to those observed. This protects calculation and interpretation while allowing revision when the missing record arrives.
Lesson summary
Deflate nominal pay and compare mean, median and hourly earnings with explicit coverage.