SupportsData Saves America calculation
Pay rises over time because output per hour rises.
U.S. nonfarm business sector, BLS Labor Productivity and Costs via FRED, downloaded 2026-11-10, latest quarter 2026 Q3 (preliminary).
| Period | Output per hour | Pay, consumer prices | Pay, output prices |
|---|
| Per year, 1948–2026 | 2.14%/yr | 1.48%/yr | 1.97%/yr |
|---|
| Per year, 1973–2026 | 1.83%/yr | 0.93%/yr | 1.52%/yr |
|---|
| Per year, 2000–2026 | 2.05%/yr | 0.87%/yr | 1.35%/yr |
|---|
| Total, 1948–2026 | +428% | +218% | +363% |
|---|
How strong: official data; the size of the gap depends on which measures you choose
BLS Labor Productivity and Costs via FRED (OPHNFB, COMPRNFB, HCOMPBS, IPDNBS); Data Saves America calculation
Supports
Most of the apparent productivity-pay gap comes from how it's measured.
U.S. nonfarm business productivity roughly doubled from 1970 to 2006. Total compensation per hour (wages plus benefits) grew at about the same rate when deflated by the same price index as output. Employee compensation was 66% of national income in 1970 and 64% in 2006.
How strong: single short paper by a prominent economist; widely cited; its method (using the output deflator, counting total compensation) is the main point of dispute
Feldstein, Martin (2008). 'Did Wages Reflect Growth in Productivity?' Journal of Policy Modeling 30(4): 591-594. NBER Working Paper 13953, doi:10.3386/w13953
Supports
Productivity growth still lifts typical pay.
U.S., 1973-2016: each 1 percentage point of extra productivity growth went with 0.7 to 1 percentage point of extra growth in median and average compensation, and 0.4 to 0.7 pp for production/nonsupervisory workers. The authors conclude that other forces, not slow productivity, have been holding typical pay down. Productivity growth has pushed it up.
How strong: well-regarded study by economists. The link is strong but not always one-for-one: the paper reports a range of coefficients, not one exact elasticity
Stansbury, Anna M. and Lawrence H. Summers (2017). 'Productivity and Pay: Is the Link Broken?' NBER Working Paper 24165, doi:10.3386/w24165 (presented at the Peterson Institute conference, Nov. 2017)
Mixed
The measured gap splits roughly evenly between price-index choice and a falling labor share.
Across 183 industries, 1987-2015: output prices rose more slowly than consumer prices in 79% of industries (median gap -0.6 pp/yr), and labor's share fell in 77% (median -0.6%/yr). At the median, the two effects were equal in size. Price-index differences mattered most in high-tech industries; the labor-share drop mattered most in manufacturing.
How strong: official BLS analysis
Brill, M., C. Holman, C. Morris, R. Raichoudhary and N. Yosif (2017). 'Understanding the labor productivity and compensation gap.' Beyond the Numbers 6(6), U.S. Bureau of Labor Statistics, June 2017
Counter-evidence
Typical workers' pay has fallen far behind productivity.
EPI finds that typical workers' pay has grown far more slowly than productivity since 1979: in its 2021 update, net productivity rose 59.7% from 1979 to 2019 while a typical worker's compensation rose 15.8%. Official BLS data tell a similar story when pay is measured in consumer prices: since 1973, output per hour grew 1.83% a year and CPI-adjusted hourly compensation 0.93% a year.
How strong: Think-tank study. The figures are accurate for the measures EPI chose, but they compare different groups and price indexes (total-economy productivity vs. one group of workers' pay deflated by consumer prices).
Economic Policy Institute, 'The Productivity-Pay Gap' (2021 update, data through 2019); BLS Labor Productivity and Costs
Supports
Long-run labor productivity growth comes from capital, worker skills and efficiency (TFP).
U.S. private nonfarm business, 1987-2025: labor productivity grew 2.0%/yr. Of that, about 0.9 pp came from more capital per hour, 0.3 pp from a more skilled and experienced workforce, and 0.9 pp from TFP (components sum to 2.1 because of rounding). In 2025 labor productivity rose 2.2%, with TFP up 0.8%. TFP averaged 1.0%/yr in 2019-25, versus 0.6%/yr in 2007-19.
How strong: official data; growth-accounting method is mainstream consensus
U.S. Bureau of Labor Statistics, 'Total Factor Productivity - 2025' news release USDL-26-0504, March 19, 2026, Tables A-B
SupportsData Saves America calculation
Countries where an hour of work produces more have much higher typical incomes.
Across 108 countries (latest year available 2017-2023 for each), output per hour worked (Penn World Table, 2021 international $) and median daily income or consumption (World Bank PIP, 2021 PPP $) have a log-log correlation of 0.93 (R² ≈ 0.87). The slope is about 0.97: 10% higher productivity goes with about 10% higher median living standards. Examples (output/hour; median $/day): U.S. $83.5, $69.7; Norway $130.1, $77.8; Germany $82.5, $64.2; Mexico $26.8, $12.5; India $8.1, $5.9; Nigeria $6.2, $3.4.
How strong: the pattern is very robust. As a cross-section correlation it does not prove which causes which
Feenstra, Inklaar & Timmer, Penn World Table (2025 release, via Our World in Data); World Bank Poverty and Inequality Platform (2026, via Our World in Data). Calculation: Data Saves America.
MixedData Saves America calculation
Investment, especially housing, swings the most and moves first. Consumption is the smoothest part of spending.
U.S. 1960Q1-2019Q4 (before COVID), using year-over-year real growth rates from BEA via FRED. Volatility relative to real GDP (std. dev. ratio): consumer spending 0.84, services 0.65, nondurables 0.81, durables 2.78, total private investment 4.30, residential investment 6.10, business fixed investment 2.87. Lead/lag (correlation of each series with GDP growth, at 4 and 2 quarters earlier / same quarter / 2 and 4 quarters later): residential 0.42/0.66/0.61/0.17/-0.17 (LEADS GDP); business fixed investment -0.07/0.33/0.75/0.73/0.38 (LAGS); total consumer spending 0.37/0.72/0.87/0.53/0.19 (moves with GDP, slightly ahead); durables 0.31/0.64/0.77/0.40/0.00.
How strong: the volatility ranking is a textbook stylized fact (consensus). The lead/lag pattern is our own simple calculation (overlapping 4-quarter growth rates). Treat it as descriptive
BEA National Income and Product Accounts via FRED; Data Saves America calculation (compare Stock & Watson, 1999)
Mixed
Housing leads the business cycle.
Residential investment is the best early-warning sign of a recession among the parts of GDP. Eight U.S. postwar recessions were preceded by substantial problems in housing and consumer durables.
How strong: influential; widely cited as a stylized fact, though Leamer's own count is 8 postwar recessions, not all of them
Leamer, Edward E. (2007). 'Housing IS the Business Cycle.' Proceedings, Economic Policy Symposium, Jackson Hole, Federal Reserve Bank of Kansas City, pp. 149-233. NBER Working Paper 13428, doi:10.3386/w13428
Supports
Consumption follows income.
U.S. data fit a world where about half of consumers spend their current income ('rule of thumb') and half spend based on expected long-run income. Expected income changes predict consumption changes, and interest rates do not. When consumption is high relative to income, high income growth tends to follow.
How strong: classic, widely replicated
Campbell, John Y. and N. Gregory Mankiw (1989). 'Consumption, Income, and Interest Rates: Reinterpreting the Time Series Evidence.' NBER Macroeconomics Annual 4. NBER WP 2924, doi:10.3386/w2924
Supports
Growth comes first, and higher saving follows.
Across countries, income growth Granger-causes (predicts) higher saving, but saving does not predict later growth. Households expecting faster income growth save MORE, which standard permanent-income models can't explain. The authors argue habit formation can. Implication: earlier estimates of saving's effect on growth are probably too high.
How strong: influential single study; Granger causality shows which comes first in time, not true cause and effect
Carroll, Christopher D. and David N. Weil (1994). 'Saving and Growth: A Reinterpretation.' Carnegie-Rochester Conference Series on Public Policy 40: 133-192. NBER WP 4470
Supports
Borrowing to consume predicts a hangover: slower growth later.
Panel of 30 countries, 1960-2012. A rise in household debt relative to GDP predicts lower GDP growth and higher unemployment over the following three years (published QJE abstract). In the NBER working-paper version, a one-standard-deviation rise in household debt-to-GDP over three years (6.2 percentage points) is followed by GDP growth 2.1 percentage points lower over the next three years. No such relationship holds for business debt.
How strong: Top-journal, widely cited, robust across countries and decades. It measures all household debt, which is mostly mortgages, and the authors explain the slowdown through weaker demand from indebted households.
Mian, Atif, Amir Sufi and Emil Verner (2017). 'Household Debt and Business Cycles Worldwide.' Quarterly Journal of Economics 132(4): 1755-1817. doi:10.1093/qje/qjx017. NBER WP 21581
Supports
The more credit fuels a boom, the worse the bust.
More than 200 recessions in 14 advanced economies, 1870-2008. Expansions with bigger credit build-ups are followed by deeper recessions and slower recoveries, both in financial-crisis recessions and in normal ones. Financial-crisis recessions cost more output.
How strong: highly cited, long historical sample
Jordà, Òscar, Moritz Schularick and Alan M. Taylor (2013). 'When Credit Bites Back.' Journal of Money, Credit and Banking 45(s2): 3-28. doi:10.1111/jmcb.12069. NBER WP 17621
Mixed
Output lost to financial crises is often never recovered.
In a panel of rich, emerging, developing and transition economies, output losses from financial crises (and some political crises) are highly persistent, with no significant rebound. The authors also find strong evidence of over-optimistic growth before financial crises.
How strong: influential; the timing of the effects has been debated (Mueller 2012 AER comment)
Cerra, Valerie and Sweta Chaman Saxena (2008). 'Growth Dynamics: The Myth of Economic Recovery.' American Economic Review 98(1): 439-457. doi:10.1257/aer.98.1.439
SupportsData Saves America calculation
'Consumer spending is 70% of GDP, so it drives the economy' confuses an accounting share with a cause.
Q2 2026 (seasonally adjusted annual rates, BEA via FRED): consumer spending (PCE) $21.99T, the average of the April-June monthly figures; GDP $32.56T; Gross Output $56.15T. PCE = 67.5% of GDP but 39.2% of Gross Output. Gross Output is about 1.72x GDP. PCE share of GO was 39.2-39.7% in every quarter of 2025-26. GDP = C + I + G + (X - M) splits up spending on FINAL goods. It is an identity that holds by definition, so it cannot show what causes what. Every dollar of consumer spending is also a dollar of revenue for someone who produced the good, and it is paid for out of income earned in production, or out of borrowing against future income.
How strong: the arithmetic is exact. Which measure is the right lens is contested (see BEA caveat)
BEA GDP, personal consumption expenditures and Gross Output via FRED (PCE, GDP, GOAI); Data Saves America calculation
Counter-evidence
Drops in demand clearly cause recessions in the short run.
Mian & Sufi find that the drop in household spending after the housing bust (falling home equity) accounts for almost 4 million of the jobs lost in 2007-09, 65% of the job losses in their data. The losses were concentrated in local non-traded industries such as restaurants and retail in heavily indebted counties, while job losses in traded industries were spread evenly. The pattern is not explained by industry supply shocks, construction exposure, uncertainty, or credit tightening.
How strong: top-journal, widely accepted
Mian, Atif and Amir Sufi (2014). 'What Explains the 2007-2009 Drop in Employment?' Econometrica 82(6): 2197-2223. doi:10.3982/ECTA10451. NBER WP 17830 ('What Explains High Unemployment? The Aggregate Demand Channel', 2012)
Counter-evidence
Fiscal multipliers. Government spending raises output in the short run, but by roughly one dollar or less per dollar.
Ramey surveys the decade of research after the financial crisis. Average government-spending multipliers mostly fall between 0.6 and 1. Tax-change multipliers mostly fall between -2 and -3: a tax cut of 1% of GDP raises output by about 2-3%. Multipliers can fall outside these ranges in some conditions (e.g., at the zero lower bound or in deep slack, which the paper discusses).
How strong: authoritative survey; ranges are consensus-ish, with real disagreement at the edges
Ramey, Valerie A. (2019). 'Ten Years after the Financial Crisis: What Have We Learned from the Renaissance in Fiscal Research?' Journal of Economic Perspectives 33(2): 89-114. doi:10.1257/jep.33.2.89
Supports
The common middle ground: demand shocks fade, supply shocks last.
Separating U.S. output and unemployment movements into two kinds of shock: demand-type shocks have a hump-shaped effect that peaks after about a year and fades within 2-5 years, while supply-type shocks raise output permanently. The demand component tracks most NBER peaks and troughs. Its share of output fluctuations at a one-year horizon ranges from about 40% to over 95% depending on assumptions.
How strong: classic. The 'demand shocks don't last' part is an ASSUMPTION the method imposes, not something it finds
Blanchard, Olivier J. and Danny Quah (1989). 'The Dynamic Effects of Aggregate Demand and Supply Disturbances.' American Economic Review 79(4): 655-673. NBER WP 2737