English Русский (автоперевод)

Sources and Observations: How This Research Works

Why I’m Writing This First

Before showing you what I found, I want to show you how I looked — and why that matters.

Most economic commentary you encounter doesn’t tell you where the numbers come from. A pundit says “wages are up” without specifying: whose wages, measured how, adjusted for what, compared against what baseline? A politician says “the economy is growing” without noting that GDP growth and typical household prosperity have had an increasingly weak relationship for the past fifty years.

The methodology matters as much as the findings. This piece documents the approach. The three pieces that follow it show what that approach uncovered.

The critics of mainstream economic measurement Problem: What Are the Real Numbers?

My starting point was an idea I encountered in the work of heterodox economists: that modern economies produce two sets of numbers — official statistics designed for political palatability, and the underlying ratios that describe what’s actually happening in households, labor markets, and production.

This analytical tradition has sometimes called its approach finding the “hidden numbers” — but a more precise description is shifting the analytical lens from aggregate to distributional metrics. The official GDP growth figure, for example, is real — but it aggregates across a distribution so skewed that the aggregate conceals more than it reveals. When the top 1% of earners captures 20% of national income, a GDP growth rate of 2.5% means something very different for the top and bottom halves of the distribution. The aggregate is not false; it is simply the wrong unit of measurement for most questions about typical household experience.

The distributional metric that critics of mainstream economic measurement want you to find is: who captured that 2.5% growth, and through what mechanisms? When you reframe the question this way — substituting distributional measures for aggregate ones — the picture changes dramatically.

This framing shaped the entire research project. I wasn’t asking “is the economy growing?” I was asking: “who is capturing what it produces, through what mechanisms, and what are they doing with it?” Those are different questions, and they require different data.

Data Source Discipline: The Three Rules

Anyone who has done serious empirical research knows the temptation to “clean” your data — to smooth over conflicting figures, pick the number that fits, quietly correct the outlier. I set explicit rules against this.

Rule 1: Every Data Point Has a Named Source

No figures appear in this research without a citation: which dataset, which organization published it, what year it covers, and which specific table or series it came from. Where FRED data is used, I include the series ID (e.g., TDSP for Household Debt Service as % of Disposable Personal Income). Where academic papers are cited, I include the author, institution, and publication.

This sounds obvious. In practice, a surprising amount of economic commentary doesn’t do it — numbers float free of their origins, get repeated, and eventually become “common knowledge” whose provenance no one can trace. I’m trying not to contribute to that.

Rule 2: Source Conflicts Are Data, Not Errors

Different sources often give different numbers for the same phenomenon. When the Tax Foundation says the US labor share of national income is ~69% and the BLS/Brookings measure shows ~57%, that is not an error to be corrected — it is evidence of a methodological dispute that itself tells you something important.

The Tax Foundation figure includes “imputed income” — statistical estimates of non-cash economic activity like owner-occupied housing. The BLS figure excludes this. Which one better describes what workers actually take home? That question is not purely technical; it has political and distributional implications. Rather than silently picking one number, this research documents both, explains the methodological difference, and notes what each version implies.

Rule 3: Never Silently Correct

If data from 1958 shows something surprising — a number that seems inconsistent with later data, or that contradicts a popular narrative — I don’t “correct” it. I document it, note the apparent inconsistency, and look for explanations. Sometimes the surprise is a data quality issue. More often, the surprise is real, and understanding why the number looks wrong is part of understanding how the economy works.

This discipline is especially important when working with historical data, where series definitions change, methodologies get revised, and political pressures on statistical agencies are not hypothetical.

This rule-based, source-tracking approach prioritizes auditability over narrative smoothness—each data point traceable to origin, each conflict preserved rather than resolved.

The Thread-Pulling Approach

The research structure was determined by following connections rather than testing a predetermined thesis. I call this “thread-pulling” — you start with one observable anomaly, pull the thread, and follow it wherever it leads.

The First Thread: Real Wages vs. Household Debt

The starting anomaly: official statistics show the US economy roughly doubling in real terms since 1958, but most people don’t feel twice as wealthy as their grandparents. Why?

The most obvious explanation is that the gains weren’t evenly distributed — but the scale of the maldistribution turned out to be far larger than popular discourse suggests. The most direct test was the primary source of household income: wages.

What the thread revealed: Real wages for production and nonsupervisory workers — the majority of the American workforce — peaked in January 1973 and were essentially flat for the following 45 years. At the same time, household debt exploded from roughly 25% of GDP in the late 1950s to a peak of 96% of GDP in Q3 2007 (Federal Reserve Z.1 data, Table D.3). The standard of living that Americans maintained over those decades was financed not by rising wages but by rising debt. A hidden decline in per-worker living standards becomes visible when you account for the shift from one-income to two-income households: it now takes two incomes to achieve what one income provided in 1973. This two-earner adjustment represents a hidden 50% decline in per-worker living standards not captured in individual wage statistics.

This thread led immediately to a question: if wages weren’t growing, why did productivity keep rising? Someone was capturing those productivity gains. Which led to the second thread.

The Second Thread: Where Did the Productivity Go?

Between 1948 and 1979, productivity and worker compensation grew together — productivity +112.5%, worker pay +90.2%. Then something changed. Between 1979 and 2023, productivity grew another 85.1% while typical worker compensation grew only 13.2%. Roughly 72 percentage points of productivity growth went somewhere else.

What the thread revealed: The missing productivity appears to have split roughly as follows (approximate ranges with significant overlap and measurement uncertainty): ~25–30% to expanded corporate profit margins, ~15–20% to top earner and executive compensation, ~20–25% to capital income (dividends, interest, capital gains), and ~10–15% to the expanding financial sector. These estimates are not precise — the categories overlap, methodologies differ across sources, and no single authoritative decomposition exists. The specific percentages should be read as order-of-magnitude indicators, not exact measurements. What is more robustly established is the directional evidence: across four independent measurement vectors (labor share of national income, corporate profit share of GDP, top 1% income share, and financial sector share of GDP), all shifted simultaneously in the same direction, starting between 1979 and 1982. The directional consistency across independent measures is the stronger finding; the specific allocation percentages are illustrative approximations.

That convergence pointed at something structural, not cyclical. Which led to the third thread.

The Third Thread: What Is the Financial Sector Actually Doing?

The financial sector in 1950 was roughly 2.5% of GDP (FRED series VAPGDPFI; Federal Reserve historical financial accounts). Today it is 7–8% (finance and insurance alone) or 20%+ (including real estate). It earns 25–30% of all corporate profits. But what does it actually produce that justifies a profit share three to four times larger than its share of economic output?

What the thread revealed: research on financial intermediation costs shows that the cost of financial intermediation has remained essentially flat for a century despite massive technological change. In competitive markets, productivity gains translate to lower costs for consumers. The evidence does not support the efficiency hypothesis - the financial sector appears to extract rents rather than create proportional value. research on financial sector efficiency at NYU Stern documented that the unit cost of financial intermediation — the cost of moving a dollar from saver to borrower — has not declined despite massive technological advances over 70 years. In a normally competitive industry, technology would drive costs down. That costs stayed flat while the sector’s profit share tripled suggests the financial sector isn’t competing on efficiency. It appears to extract rents.

One plausible model consistent with the data suggests a self-reinforcing cycle: stagnant wages may push households toward borrowing; the financial sector profits from this borrowing; those profits correlate historically with lobbying that expanded lending regulation; which in turn expanded lending capacity. This feedback loop represents a theoretical framework explaining the correlations, not a proven causal chain. The evidence suggests the financial sector is not merely a passive intermediary, but an active participant whose growth is structurally linked to the household debt it helps create.

The Key Data Sources

In brief: three main datasets tracked worker wages over decades; two tracked household debt; two tracked financial sector size. The detailed source list follows for those who want to verify the numbers.

FRED (Federal Reserve Economic Data) — St. Louis Fed

The primary quantitative database for this research. Covers macroeconomic time series going back decades, with consistent methodology documentation. Key series used:

Limitation: Most FRED series don’t go back to 1958 in continuous comparable form. Pre-1979 real wage data requires combining series with different coverage periods — specifically, splicing CES0500000013 (available from 1964) with supplementary BLS historical data for 1958–1964, then deflating using CPI-U series CPIAUCSL. This stitching methodology involves judgment calls (splice point selection, deflation method) that affect results. Data-stitching code for this analysis is not currently published, but the series IDs and splice point (January 1964) are documented here to allow independent replication. The underlying raw series are all publicly accessible via FRED and the BLS website.

BLS (Bureau of Labor Statistics)

Source for wage and labor force data. The 2017 Monthly Labor Review article on estimating the US labor share is a particularly useful document that acknowledges the methodological disputes openly.

EPI (Economic Policy Institute)

The most comprehensive and rigorously documented source for productivity-pay gap analysis. Their ongoing Productivity–Pay Gap tracker synthesizes BLS and BEA data. EPI has an identifiable center-left orientation, which is documented and should be factored in — but their underlying data sources are the same government datasets everyone uses.

OECD Data

Used for international comparisons. The OECD database provides standardized cross-country labor statistics that allow comparison of US trends against other developed economies.

Academic Papers — Key Sources

Piketty & Saez (2022) — “Striking It Richer”: The foundational long-run US top income share database. Shows top 1% share rising from ~9.6% (1979) to ~20%+ (recent decades).

research on financial sector efficiency, NYU Stern — “Has the U.S. Finance Industry Become Less Efficient?” The key paper documenting that financial intermediation costs haven’t fallen despite technology — the core empirical basis for the rent extraction argument.

Levy Economics Institute, Working Paper 525 — “Financialization: What It Is and Why It Matters.” Provides the historical financial sector debt data (1.35% of GDP in 1946 → 109.5% in 2009).

ScienceDirect — “Local financialization, household debt, and the great recession.” Geographic evidence that financial sector size causes household debt growth (not the reverse).

Tandfonline/Post Keynesian Economics — “Household debt, financialization, and macroeconomic performance, 1951–2009.” Documents the 1982 structural break after which household debt became measurably negative for output.

Federal Reserve Z.1 (Flow of Funds)

The authoritative source for household balance sheet data. Table L.101 provides historical household asset and liability data back to the 1950s — the only source with continuous coverage for the 1958–1980 period that other databases miss.

News Reports (With Source Tracking)

Journalistic sources — Al Jazeera, Pew Research fact-checks, Congressional Research Service summaries — are used as reference points but flagged as secondary sources. When a news report cites data, I traced the data to its primary source and cited the primary source. News reports are used for qualitative narrative context.

The Three Completed Threads: Brief Summary

Thread 1: US Real Wages vs. Household Debt (1958–Present)

Finding: Confirmed. American living standards since 1973 have been maintained by debt accumulation, not real wage growth. Real wages for the typical worker peaked in January 1973 and were essentially flat for the following 45 years. Household debt rose from ~25% of GDP (late 1950s) to a peak of 96% of GDP in Q3 2007 (Federal Reserve Z.1 data, Table D.3). A hidden decline in per-worker living standards becomes visible when you account for the shift from one-income to two-income households: it now takes two incomes to achieve what one income provided in 1973. The two-earner adjustment makes this concrete: 1958 households required one breadwinner; 2024 households typically require two — yet real per-capita work-hour income has barely changed.

Key hidden number: The wage stagnation figures are themselves understated when measured against what households actually need. The official CPI measures an average basket of consumer goods—but the costs of essential life stability have risen far faster.

Housing, healthcare, and higher education have inflated three to six times faster than the official CPI over this same period. This is not a peripheral observation: these are the goods that define a stable life in modern America. Workers are not merely failing to get ahead in general purchasing power—they are falling behind specifically on shelter, health, and education. The stagnation in wages, combined with this essential-cost inflation, is the core mechanism connecting flat paychecks to real household hardship.

Source thread: US Real Wages and Debt Analysis

Thread 2: Productivity Gains Distribution (1979–2019)

Finding: Confirmed with high confidence. Workers received approximately 15% of productivity gains since 1979. The other ~72 percentage points were captured by corporate profits, top earners, capital income, and the financial sector. This is not explained by any single cause — the four independent measurement vectors (labor share, corporate profits, top income share, finance share) all shifted simultaneously, which points to structural institutional change rather than natural market dynamics.

Key hidden number: The official “labor share of national income” figure of ~69% includes imputed housing income and other non-cash items. The BLS measure that tracks what workers actually receive shows ~57% — a 12-percentage-point difference with significant distributional implications.

Source thread: Productivity Gains Distribution

Thread 3: Financial Sector Financialization (1945–Present)

Finding: Confirmed. The US financial sector earns 25–30% of all corporate profits while contributing 7–8% of GDP value — a 3–4× multiplier that has persisted for decades. This is structural rent extraction, not a market efficiency signal. While this rent-extraction interpretation is supported by research on financial intermediation costs, it remains a contested view in economics, with alternative explanations focusing on risk management and liquidity provision services. The financial sector grew by creating household debt, profiting from it, and then using those profits to lobby for further deregulation enabling more debt creation. The 1982 structural break (Garn–St. Germain Act) is the documented inflection point. Specifically, the Garn-St. Germain Depository Institutions Act deregulated savings and loans, removed interest rate ceilings, and dramatically expanded what financial institutions could do with depositor funds - accelerating financial sector growth and the debt-creation capacity of the banking system.

Key hidden number: The number of FDIC-insured banks fell from 10,222 (1999) to 5,002 (2020) (FDIC Historical Statistics on Banking, Table CB01), with only 48 new charters issued in the entire 2010–2020 decade. This is a hallmark of oligopoly — not a competitive market that should be pricing intermediation at near-zero as technology would predict.

Source thread: Financial Sector Financialization

Alternative Explanations Considered

Before presenting what the threads suggest together, it is worth noting the most commonly cited alternative explanations and why the data, as assembled here, does not appear to support them as primary drivers.

Technology and skill-biased technical change: The standard mainstream explanation for the wage-productivity gap is that technology increased relative demand for high-skilled workers. This plausibly contributes. If technology alone explained the gap, we would expect a gradual shift across all sectors—but the data shows a sharp break concentrated in the 1979–1983 period, coinciding with specific policy and structural changes rather than a gradual technology adoption curve. Moreover, other technology-adopting economies (Germany, Japan) adopted similar technologies over the same decades yet show far less severe distributional divergence, which the technology hypothesis alone cannot explain.

Globalization and trade competition: Import competition from low-wage economies plausibly depresses wages in tradeable-goods sectors. However, the productivity-pay gap is also observed in non-tradeable service sectors (healthcare, education, retail), which are largely insulated from trade competition. Healthcare workers, educators, and retail employees—in sectors with minimal import competition—showed the same wage stagnation pattern as manufacturing workers directly exposed to global competition. This parallel stagnation across tradeable and non-tradeable sectors limits globalization’s explanatory scope.

Demographic shifts: Baby Boomer cohort dynamics, increased female labor force participation, and immigration can affect average wages. However, the median wage for full-time male workers — a more demographically stable cohort — also shows the same stagnation pattern, suggesting demographics are not the primary driver.

Voluntary household preference for debt: Households may have simply chosen to borrow more for home ownership, investment, or consumption smoothing. The timing evidence is less consistent with this: debt acceleration correlates tightly with real wage stagnation beginning in the mid-1970s, not with an autonomous prior shift in household preferences. Between 1974 and 1985, the US household savings rate fell from approximately 10% to 6% (Bureau of Economic Analysis, NIPA Table 2.1)—a pattern consistent with households compensating for lost purchasing power rather than voluntarily choosing greater debt exposure.

These alternatives are not dismissed — they likely contribute in partial ways. While this analysis found stronger evidence for the distributional/debt/financialization framework, the competing explanations remain active areas of economic debate and likely contributed alongside the mechanisms described here. The argument here is that structural redistribution and financial sector dynamics provide a more complete account of the simultaneous, cross-metric, cross-sector pattern in the data. Readers who weigh these alternatives differently may reach different interpretations, which is why the underlying data sources are fully documented.

What These Threads Converge To

The three threads are not separate phenomena. They are one integrated system:

  1. Productivity gains are systematically redirected from labor to capital (since ~1979)
  2. Household debt appears to compensate for wage stagnation to maintain consumption (and prevent demand collapse)
  3. The financial sector grows by creating, intermediating, and profiting from that debt
  4. The US dollar’s reserve currency status enables the entire system by removing balance-of-payments constraints that would otherwise limit debt expansion (This mechanism will be examined in detail in the Conclusions article of this series.)

The three threads map onto a self-reinforcing cycle: Stagnant Wages → Rising Household Debt → Financial Sector Growth → Political Influence → Deregulation → Enhanced Debt Creation Capacity → Continued Wage Suppression. The loop explains both the persistence of the pattern and the difficulty of reversing it.

One might argue that households chose debt voluntarily - for home ownership, education, or investment. But the timing is telling: debt acceleration begins precisely when real wages stall in the mid-1970s, not before. The composition also shifted: an increasing share of household debt in this period reflects spending on necessities - healthcare, housing, and basics - as wages failed to keep pace with costs.

A natural question arises: if wages were flat for fifty years, where did consumer demand come from to drive corporate profits? The data offers a clear answer: debt. Household debt expanded from 25% of GDP in the late 1950s to over 90% at its 2009 peak. This debt-demand relationship is the hidden connector between all three threads.

This model — described in detail in the research synthesis — is internally non-contradictory, consistent with the empirical data from all three threads, and connects structurally to what some economists call the Crisis of Capital Effectiveness (CCE)—the point where debt-financed growth cannot substitute indefinitely for genuine wage-based demand—and to analyses of dollar reserve currency dynamics.

The next three articles will work through the evidence and implications of each thread in depth, with the data presented accessibly for non-economists while maintaining full source attribution throughout.

A Note on What This Research Is Not

This is not a polemic. The data is what the data is. The findings point in a consistent direction — toward structural redistribution away from labor and toward capital since roughly 1979 — but that direction was not chosen in advance. The methodology was chosen in advance: follow the thread wherever it leads, keep the sources, document the conflicts.

If you want to dispute the findings, the sources are all listed. Pull the threads yourself. That’s exactly what this approach invites. For readers interested in how AI approaches complex empirical tasks, the methodology described here—data stitching, conflict preservation, thread-following without predetermined conclusions—illustrates one path toward systematic analysis without prior expertise.

Next: Part 2 — The Wage-Debt Substitution: How America Maintained Living Standards Without Wage Growth