Why Can Productivity Outpace Pay? Deflators and Labor's Share

Labor productivity measures real output per hour worked. Compensation per hour measures labor costs, including wages and benefits, per hour. To compare compensation across time, statisticians adjust nominal dollars for a price index. A gap appears when the resulting productivity and real-compensation growth rates differ over the chosen period.

The word *real* does not name one universal calculation. CPI-adjusted compensation asks how the amount compares with consumer prices and purchasing power. Compensation adjusted by an output-price deflator asks how it compares with the prices of what a sector produces. Both can be valid, but they answer different questions.

The U.S. Bureau of Labor Statistics (BLS) described this distinction in an industry-level analysis covering 1987–2015. The paper divides the measured gap into a price-deflator component and a labor-share component. That framework helps explain the arithmetic; its historical sample is not a current estimate for every worker, sector, or country. See the BLS analysis of the productivity–compensation gap.

The Consumer Price Index (CPI) follows prices consumers pay for a defined basket. It is useful when the question is whether compensation gained purchasing power over consumer goods and services. A household's own spending differs from the average basket, so CPI-adjusted average compensation is still not an individual's personal cost-of-living measure.

An output-price deflator follows prices for goods or services produced by a sector or industry. It is useful when comparing the value of production with compensation paid to produce it. BLS methods distinguish consumer, value-added-output, and sectoral-output deflators. The BLS calculation guide defines real hourly compensation, output-price deflators, and labor share separately.

The BEA GDP price index is a broad measure of prices for U.S.-produced goods and services. It includes exports and excludes imports, while a consumer price index describes consumer purchases. A GDP index is not automatically the exact industry-level output deflator needed for every productivity comparison. The BEA GDP price index definition explains its production boundary.

Use a hypothetical value-added sector and match output, compensation, and hours. At the starting date, real output is 100 units per hour, the output-price index and CPI are both 100, nominal value added is $100 per hour, and hourly labor compensation is $60. Labor's share of value added is therefore 60%. All dollar amounts in this hypothetical example are U.S. dollars (USD).

At the later date, output per hour rises 5% to 105 units. The output-price index rises 1% to 101, so nominal value added per hour is $105 × 1.01 = $106.05. Suppose labor's share falls by 1% proportionally, from 60% to 59.4%. Compensation per hour is then about $106.05 × 0.594 = $62.99. The CPI rises 3% to 103, so compensation in starting-date dollars is about $62.99 ÷ 1.03 = $61.16, a 1.93% increase from $60.

With CPI adjustment, productivity rises 5% while real compensation rises 1.93%, a simple growth-rate gap of about 3.07 percentage points. With the output-price deflator, compensation is $106.05 × 0.594 ÷ 1.01 ≈ $62.37 in starting-date dollars, up 3.95%; the productivity–compensation gap narrows to about 1.05 percentage points. These are invented inputs to show how the choice of deflator changes a comparison. They are not BLS or BEA observations.

To isolate the effect of the price measure, hold labor share at 60% in a second hypothetical. Hourly compensation is $106.05 × 0.60 = $63.63. Deflated by the output-price index, it is $63.63 ÷ 1.01 = $63.00, up 5% and matching productivity growth. CPI-adjusted compensation is $63.63 ÷ 1.03 ≈ $61.78, up about 2.96%, leaving a CPI-based gap of about 2.04 percentage points. This counterfactual separates the price-index effect; it is not an observed wage or price path.

The CPI-based and output-price comparisons differ by about 2.02 percentage points in this example. The remaining output-deflated gap is close to the proportional decline in labor share. Small discrepancies arise because ordinary percentage-growth differences do not add exactly; a log-growth decomposition gives an exact accounting identity under compatible definitions.

For a value-added production measure, labor share is labor compensation divided by nominal value-added output. It summarizes how much of the measured value accrues to labor under that accounting definition. If productivity rises while labor compensation grows more slowly than nominal output, labor share can decline and output-deflated compensation can lag productivity.

That statement describes an accounting relationship, not a complete causal explanation. Labor share can change with industry mix, business organization, relative prices, the treatment of self-employment, and other components of production income. A falling aggregate share does not tell us by itself which force changed or how income shifted among individual households. The BLS labor-share study discusses measurement choices and why differently scoped series should not be treated as interchangeable.

Do not assume every published labor-share measure uses the same denominator. BLS uses value-added output for some measures and a different output and cost boundary for some detailed industries. The ratio must match the output concept and sector being analyzed. A labor share from one statistical series cannot simply be substituted into a formula built for another.

For a compatible value-added measure, let P be real value-added output per hour, deflated by Pₒ; R be hourly compensation adjusted by CPI; CPI be the consumer price index; Pₒ be the output-price deflator; and s be labor compensation divided by nominal value-added output. The ratio of productivity to CPI-adjusted compensation is the CPI-to-output-price ratio divided by labor share. Taking changes in natural logs gives:

Δln(P) − Δln(R) = Δln(CPI) − Δln(Pₒ) − Δln(s)

In plain language, the CPI-versus-output-price change contributes one part; the change in labor share contributes another. When labor share falls, Δln(s) is negative, so subtracting it adds to the measured productivity–compensation gap. Log changes are close to ordinary percentage changes for small rates, but the exact identity uses matched log changes and consistent boundaries.

This formula is an accounting decomposition, not a model that predicts wages or identifies causes. It also depends on using compatible series: the same period, hours, workers, and value-added production boundary. A comparison using sectoral output, a different labor group, or mismatched time periods needs definitions suited to those measures.

BLS hourly compensation is an aggregate measure. Depending on the series, it can include wages, salaries, employer benefits, and compensation imputed for proprietors. The mix of workers, hours, industries, and benefits can change over time. Average compensation may therefore move differently from the pay of a particular person, occupation, or median worker.

Productivity is also an aggregate ratio of output to hours, not an individual performance score. It reflects capital, technology, materials, capacity use, management, worker composition, and other inputs alongside labor. If output per hour rises, that does not establish that each employee became more productive by the same amount or that a specific pay policy caused the change.

For worker purchasing power, use a real-pay measure whose population and deflator match the question. For production costs, compare compensation with output using a compatible producer-side measure. For distribution questions, examine labor share and other income measures with their own scope. These views complement one another; none answers all three questions alone.

Before quoting a gap, record the output measure, compensation measure, price index, labor-hour concept, worker population, sector, and period. Check whether growth is quarterly, annualized, year over year, or cumulative over several years. Annualizing a quarterly rate describes a compounded pace under an assumption; it is not the same as an observed one-year change.

Index levels also need context. A price or productivity index rebased to 100 changes its reference point, not the underlying economic event. Compare growth rates across indexes only after aligning periods and transformations. BLS methods may revise output, hours, and compensation estimates as source data are updated, so cite the release or vintage when the precise estimate matters.

Finally, distinguish a broad U.S. GDP price index from an industry-specific output deflator and from the CPI. The GDP index covers domestic production, including exports and excluding imports. The CPI concerns consumer purchases. Neither should be used as a universal replacement for every country or industry measure.

A productivity–compensation gap can describe how aggregate output per hour and a chosen real-compensation series changed relative to each other. Separating the price-deflator and labor-share components helps readers see why CPI-based and producer-price comparisons can differ.

The gap alone does not prove that productivity gains were withheld from every worker, that a specific technology caused pay to lag, or that consumers faced a particular inflation rate. The result depends on price indexes, statistical boundaries, aggregation, and dates. BLS's 1987–2015 industry analysis is useful for understanding the decomposition, not for asserting a current universal trend.

To explore related measures, see unit labor costs and inflation, nominal versus real wages, and CPI, PCE, and the GDP deflator.