A Hundred Years After Keynes

1

Maynard Keynes is best known for advancing the theory of effective demand and laying the foundations of modern macroeconomics. But in his earlier years he served as a bureaucrat in the British India Office before WWI, and then represented the Treasury at the Versailles negotiations at the war’s end — Keynes’s insight came directly out of his lived experience of the violent shocks to Europe’s economic order before and after the war. The Economic Consequences of the Peace, written in 1919 in the wake of Versailles, sounds like a critique of the harsh terms imposed on Germany and the slow recovery they would cause. At a deeper level, it is an anatomy of capitalism’s underlying growth logic and its fragility. From 1921 on, Keynes managed the King’s College Cambridge endowment and effectively became a forerunner of long-horizon equity investing — a transition that can be read as a direct outgrowth of his evolving economic thought.

Among the lesser-recognized wonders of the twentieth century, Philip Morris — the maker of Marlboro — should probably be on the list: roughly 17.0% annualized over its 1925–2003 run, ending in something like 250,000× total return (15.4% annualized from 1925 to today, ~2,000,000×). Set against the Netherlands — capitalism’s birthplace — during the heyday of the VOC across 1500–1700, with annualized growth of just 0.26% and cumulative growth of 1.68×, the contrast sounds like fiction. (Double-entry bookkeeping, paper money, marine insurance, public debt, partnership companies, savings banks, and early central banking can all be traced back to Renaissance Florence and Venice — but Italy was unable to sustain growth: under hard population constraints, total scale stalled around 1500 and then declined as Atlantic trade displaced the Mediterranean.) Philip Morris is not the exception. The S&P 500’s 10.2% annualized, ~18,000× cumulative return is already an astronomical figure utterly unimaginable in pre-modern times.

2

Keynes’s opening claim is this: the sustained compounding growth produced over the roughly 50 years between the maturing of the Industrial Revolution around 1870 and the end of WWI was not inevitable. It was the result of hundreds and even thousands of years of institutional evolution in Europe stretching back to the Middle Ages.

Across the long stretch of history outside the Industrial Revolution, the Malthusian trap — capping a region’s economic scale at some steady state through famine, plague, and war — was the norm. The Taiping Rebellion around 1850, which caused roughly 50 million direct or indirect deaths (10–15% of late-Qing China’s population), is the most recent and most vivid example.

But just as someone today, swiping their phone in a luxury apartment in New York or Shanghai — ordering goods from anywhere in the world, or carrying nothing but a credit card or Alipay across borders without ever touching local cash — the bourgeoisie sipping their morning tea in bed in London a hundred years ago took it equally for granted that the world’s goods were within easy reach, the world’s labor was at their service, and global travel and business faced no barriers. The slightest inconvenience was treated as something annoying and avoidable.

Until the governments they elected and the media they controlled dragged them into a war that lasted four years, with mass casualties to witness. The hyperinflation that followed wiped out their monetary holdings — and that finally destroyed their confidence.

And the unwritten social contract that capitalism actually relied on to keep functioning — the propertyless accept inequality in exchange for security and social mobility; the propertied hold most of society’s wealth but, like Puritans, voluntarily restrain consumption and reinvest, growing total social wealth and ultimately raising aggregate welfare — quietly fractured under the conspicuous consumption of the Belle Époque and the devastating shock of the First World War. Hyperinflation enriches some unexpectedly and dispossesses others without their noticing — debtor-creditor-contract relations themselves dissolve, the sense of dispossession converts into political pressure, and the Russian October Revolution of 1917 follows. And the mechanism is not confined to 1919 — over the past thirty years, asset-price inflation has done the same: multiplying the wealth of the asset-owning few while those whose savings sit in cash are dispossessed in relative terms.

3

In a market economy, economic activity is largely organized through firms. To quantify the vitality of the corporate sector as a whole, one direct synthetic indicator is its total profit.

What’s interesting is that, just like income-side GDP, total profit is also a flow concept, and there is an accounting identity for it (derived in 1942 by the Polish economist Michał Kalecki from the national income identity; Keynes himself expressed a similar but less rigorous view in The General Theory):

Profits = I + (G − T) + NX + C_capitalist − S_worker

Where the identity comes from:

• Starting from the income side of GDP: Y = Wages (to labor) + Profits (to investors)
• The expenditure side decomposes into consumption, investment, trade, and government spending: Y = C + I + G + NX
• Consumption splits into capitalist consumption and worker consumption: C = C_capitalist + C_worker
• The identity uses the condition that any wages workers don’t spend, they save: W = C_worker + S_worker
• Folding the tax term T into government spending gives (G − T)

Like a law of physics, we start from the basic fact that “income equals expenditure — if anyone earns money, someone must have spent money,” so the identity must be true. But also like a law of physics, the identity tells us something not at all obvious: the sources of aggregate profit in an economy add up to just four positive contributors, plus one negative drag.

The conclusion: even though profit looks like the thing every participant in the economy most wants to forecast and finds hardest to grasp, if we hold investment, trade, government spending, and the savings rate fixed at the macro level, short-run aggregate profit is uniquely determined.

The identity holds in any economy, not just market economies. In a planned economy, profit is administratively set to zero, prices equal calculated average cost, and high investment must be offset by a combination of high government surplus (T > G, so the (G − T) term is negative) and high worker savings — exactly the situation in pre-1978 reform China. Or take Japan after the bubble burst in the 1990s: investment slowed sharply but corporate profits did not fall into a serious long-run decline, only stagnation, because the Japanese government subsidized the corporate sector for years through deficits (QE combined with fiscal expansion).

The modern undergraduate textbook answer is investment (I) — more capital input on the supply side, physical assets like plants and machinery — and technology (TFP) — R&D and technical innovation.

But Keynes’s 1936 General Theory of Employment, Interest and Money offered a completely different answer — "in the long run we are all dead," and in the short run, demand is the lifeblood of all market economic activity. When workers, out of insecurity, prefer saving to consuming each marginal dollar earned, and investors choose savings over investment, the economy slips into severe depression — even in the 1930s, when the second industrial revolution was producing rapidly compounding outputs, when automobiles, electrical equipment, and chemical industries reached unprecedented levels. The Faustian human appetite for knowledge does not change with the economic system, from Socrates to Fourier. Abel, who built the modern foundations of algebra, did not stop his research because nineteenth-century Norway could not support a living through mathematics. But conversely, the modern consumer society of endless wanting — of always wanting more and better — is the most fundamental source of economic growth. When demand exists, profit’s lure gives people inexhaustible drive to apply the knowledge and technology that exists or is about to be created, and to manufacture supply matched to that demand. From the opening of new spice trade routes to the race against the physical limits of the transistor under Moore’s Law — through repeated amplifications of human desire, civilization has reached a wholly new chapter.

The economic historian Robert Allen takes the argument one level deeper: even the direction of technological invention is determined by demand-side relative factor prices. Britain in the 18th century was first to invent the steam engine and textile machinery not because the British were smarter than the French, Indians, or Egyptians — it was because only Britain simultaneously had high wages and cheap coal, which made machinery that substituted capital and energy for labor profitable to invent. The same technologies were not worth inventing in low-wage economies, nor even worth importing. Demand does not only determine the speed of growth; it shapes the form of growth itself — technological progress is "induced" by demand.

4

The statement above isn’t logically airtight, of course: Sumer had Gilgamesh wanting eternal life, ancient Greece had Icarus wanting to fly, but it wasn’t until the Wright Brothers that humanity actually got the chance to soar.

The main institutional developments are:

• Property rights (whatever I produce, what I get keeps belonging to me)
• Money (I can produce for future consumption)
• Credit (others can trade their future production for my current production)
• The corporate form (the productive organization, as a legal entity, persists regardless of any natural person’s life or death)
• Double-entry bookkeeping (producers and owners can be separated — the precondition for large-scale organizations)

These have become almost foundational laws in any market economy.

And the conditions Keynes described — taken for granted a hundred years ago, but by no means inevitable — are:

• A stable, open financial-and-trade system underwritten by hegemony. If a country’s people are forced to buy only domestically made goods, even at the same income level they will visibly find the things they can buy shrinking.
• A currency with stable value. Under hyperinflation, if creditors cannot predict the real purchasing power of future repayments, the most basic borrowing-and-lending activity in the economy becomes speculation.
• Workers willing to enter a social contract with investors. Because of wealth-concentration effects, workers always make up the absolute majority under any system; if they are no longer willing to keep this contract, the political pressure for redistribution will become very strong.
• An abundant entrepreneurial spirit. If a society no longer believes in value creation and pays more attention to the inheritance of cross-generational wealth and power, the economy will lose its vitality.

Looking at the U.S. over the past decade, almost all four of these conditions are deteriorating: from Trump’s mercantilism, to the inflation pressure that has lingered after the pandemic, to the redistribution pressure brought on by conspicuous-consumption culture (such as California’s billionaire wealth tax), to estate lawyers in Wyoming who can’t keep up with their caseload — dynasty trusts are quietly rebuilding the hereditary model that the U.S. was supposed to reject from the Enlightenment onward.

And yet, at least in major cities like New York and San Francisco, the overall economy still looks vibrant. If we trace the composition of total profit, we find that under the triple influence of 1990s financial liberalization, the new economy, and globalization:

• Investment demand has fallen. The share of corporate profit going to physical capex has slid from 80% in the 1990s down to 50%; meanwhile, the share going to share buybacks has surged from below 20% to above 50% — the leftover cash didn’t go to new factories, it went back to shareholders. Today, America’s most profitable companies all use their abundant cash flow to buy back their own stock. But deindustrialization is not only because globalization shifted capex to other countries — industry concentration (which weakens competitive investment), asset-light models like financial services and software (which require less investment to begin with), and an aging population all stack as overlapping factors.
• Personal savings rates have fallen. The rise in asset prices for stocks and real estate makes people no longer feel the need to save against risk.
• The drag on profit from trade deficits had already halved by the 2008 global financial crisis.

Decompose those three structural shifts through the Kalecki identity, allocating each component’s contribution per dollar of after-tax corporate profit, averaged by decade:

Across three decades: the physical investment supporting each dollar of profit fell from $2.98 to $1.56 (nearly halved), while the contribution from government deficits jumped from $0.30 to $0.72 (more than doubled). This is the accounting picture of the three structural shifts above — investment-driven growth replaced by deficit-driven support.

But this "per-dollar" view risks being misread as "investment is shrinking" — it isn’t. The same identity laid out in absolute amounts ($B/year):

Over thirty years, profit grew 6.7×, but the components grew at very different speeds. Government deficit grew 16× — 2.4 times faster than profit itself. That is what "investment-driven growth replaced by deficit-driven support" actually looks like in dollars. Investment did grow 3.5×, but couldn’t keep up with the expanding profit pool — which is why the per-dollar investment ratio fell from $2.98 to $1.56.

Expand 1990-2024 year-by-year detail

Year	Profit ($B)	Investment ($B)	Govt Deficit ($B)	Net Exports ($B)	−Personal Saving ($B)
1990	$284	$993	+$221	−$78	−$361
1991	$308	$944	+$269	−$29	−$401
1992	$335	$1,013	+$290	−$35	−$450
1993	$357	$1,107	+$255	−$65	−$393
1994	$438	$1,256	+$203	−$92	−$357
1995	$501	$1,318	+$164	−$90	−$379
1996	$541	$1,432	+$107	−$96	−$370
1997	$592	$1,596	+$22	−$102	−$372
1998	$533	$1,737	−$69	−$163	−$425
1999	$570	$1,887	−$126	−$260	−$316
2000	$553	$2,038	−$236	−$381	−$317
2001	$547	$1,935	−$128	−$377	−$363
2002	$646	$1,930	+$158	−$440	−$452
2003	$777	$2,027	+$378	−$522	−$440
2004	$1,010	$2,281	+$413	−$634	−$417
2005	$1,326	$2,535	+$318	−$740	−$209
2006	$1,464	$2,701	+$248	−$786	−$276
2007	$1,422	$2,673	+$161	−$736	−$263
2008	$1,183	$2,478	+$459	−$741	−$452
2009	$1,302	$1,930	+$1,413	−$419	−$625
2010	$1,606	$2,166	+$1,294	−$532	−$671
2011	$1,589	$2,333	+$1,300	−$580	−$776
2012	$1,881	$2,622	+$1,077	−$552	−$976
2013	$1,858	$2,838	+$680	−$478	−$616
2014	$1,947	$3,074	+$485	−$509	−$712
2015	$1,841	$3,288	+$442	−$524	−$791
2016	$1,864	$3,278	+$585	−$503	−$746
2017	$1,998	$3,468	+$666	−$543	−$842
2018	$2,020	$3,725	+$779	−$593	−$997
2019	$2,078	$3,894	+$984	−$577	−$1,178
2020	$2,212	$3,763	+$3,096	−$619	−$2,662
2021	$2,898	$4,246	+$2,774	−$849	−$2,177
2022	$3,007	$4,844	+$1,374	−$938	−$633
2023	$3,242	$5,024	+$1,688	−$786	−$1,159
2024	$3,499	$5,259	+$1,815	−$898	−$1,193

The deeper question: why did profit grow 6.7×? The number factors into three independent multiplicative layers:

6.66 ≈ 1.95 (real GDP growth) × 1.69 (general price-level rise) × 2.02 (profit-share-of-GDP rise)

So the corporate sector’s profit expansion came from three quite different sources (with log contribution shares of roughly 35% / 28% / 37%):

• Real economic activity (~35%) — productivity, labor force growth (including immigration), capital accumulation, technological change. This is a supply-side question driven by total factor productivity, population, and capital inputs, and not the focus of this essay.
• General price-level rise (~28%) — cumulative inflation over 30 years, mostly determined by monetary and fiscal policy.
• Profit’s share of GDP (~37%) — declining labor share (wages lagging productivity), industry concentration (monopoly rents), globalized cost reductions, the 2017 TCJA corporate tax cut (federal rate from 35% to 21%), and the migration of the profit pool from heavy industry to high-margin asset-light models in tech, finance, and services.

That is, even at a 30-year horizon, the distribution-driven profit contribution (~37%) already slightly exceeds the supply-side real-growth factor (~35%) — Keynes’s effective demand theory is not merely a short-run diagnosis; it applies in the long run too. For the profit pool to expand 6.7× in 30 years — twice as fast as GDP itself — the right-hand side of the Kalecki identity needed sustained demand-side releases. Over those decades the U.S. was running four such "savings releases" in parallel: (1) household saving rate fell from ~8% to ~5%, with part of the foregone precautionary saving substituted by the wealth effect from rising asset prices; (2) the federal deficit expanded from ~$134B/year (1990s avg) to ~$2,149B/year (2020-24 avg) — by a wide margin the single largest source; (3) sustained foreign capital inflows financed the persistent U.S. trade deficit; (4) the corporate sector cut physical investment and ramped up share buybacks, redirecting retained cash from productive capital into financial capital. The largest releaser was the government deficit, not the household sector.

And in the short term, during both the 2008 financial crisis and the pandemic, the U.S. government stabilized sharp drops in investment and consumption with massive deficits.

Where exactly did those deficits go? Take 2024 as an example: federal expenditures of $6.75 trillion, revenue of $4.92 trillion, deficit of $1.83 trillion. Of which:

• Social Security — $1.47 trillion
• Medicare (health insurance for those 65 and older) — $0.92 trillion
• Medicaid (low-income / disability health insurance, federal portion) — $0.62 trillion
• Interest on federal debt — $0.88 trillion
• Defense — $0.85 trillion
• Income security — $0.70 trillion

By comparison, total AI capex from U.S. hyperscalers across 2024–2027 (projected) comes to only $1.41 trillion. Put differently, just one year of U.S. healthcare spending alone (Medicare $0.92T + federal Medicaid $0.62T = $1.54T) already exceeds the entire hyperscaler industry’s four-year AI spending total. The aggregate profit pool is still growing, but the bulk of the deficit is flowing into expenditures that do not directly generate future growth (healthcare, Social Security, debt interest) — and U.S. healthcare in particular costs 2–3× per unit of what other developed countries pay. These categories may be justifiable on welfare grounds, but they do not expand productive capacity or raise productivity, so the fiscal model that supports profits is itself unsustainable.

It may not be long before we, too, are jolted into realizing that — just as the sustained growth of those 50 years from 1870 to 1920 was not eternal — what we know as “normal” is not inevitable either.

---

Appendix: Empirical validation

The essay’s main numerical claims were independently audited by the OpenCausality project. Its agentic pipeline translates the essay’s causal story into a DAG (us_profit_deficit_claims) and assigns each claim a specific identification strategy: accounting identity (Kalecki), exogenous-shock identification (Romer-Romer tax local projections), event study (TCJA, ACA, Ramey defense spending), sectoral-balance cross-check, etc. Each estimate is explicitly labeled by what it identifies — separating "deficits correlate with profits" from "deficits cause profits."

Cross-validation sources: (a) the same core data the essay uses (FRED quarterly NIPA series) was re-pulled via OpenCausality’s direct BEA API access (rules out FRED-conversion or caching artifacts); (b) new data not used in the essay was introduced for cross-checks: BEA NIPA sectoral-balance table (private / government / foreign three-way accounts), Romer-Romer 2010 exogenous tax-shock replication archive (1985-2007 quarterly), Ramey 2016 defense-spending news series (1985-2013 quarterly), Medicare / ACA policy event windows. Each layer is an independent falsifiable check. Key findings:

1. Romer-Romer exogenous tax-shock local projections: a 1pp-of-GDP exogenous tax cut raises after-tax profits by ~14.3 units 8 quarters out (p=0.018); the effective tax rate falls 1.8pp (p=0.022). This confirms that tax-policy channels (e.g., 2017 TCJA) mechanically lift after-tax profits without requiring new productive capacity. The TCJA alone cut the effective rate from 16.1% to 12.7% and lifted profits by ~$80B (3.9% of base).
2. Necessary qualifications: the Kalecki identity is an accounting identity, not an identified causal effect; the residual is large because the proxy omits capitalist consumption and sectoral-boundary terms; Ramey’s defense-spending news shocks show no clean causal effect on profits, meaning the simple "deficits cause profits" story is unsupported — the mechanisms supporting profits are specific policy channels (tax cuts, transfers, healthcare), not deficit volume per se.
3. Cyclically-adjusted deficit diagnostic regression and counterfactual: on the ex-crisis subsample (N=31, excluding 2009, 2020, 2021), regressing the profit share on a model-implied cyclically-adjusted deficit (constructed from a CBO-potential-GDP output gap) yields a coefficient of 0.188 (p=0.205, R²=0.787, weak) without the Kalecki residual as a control, and 0.502 (p≈0.000, R²=0.929, significant) once the Kalecki residual is included as a control; the profit-growth specification remains weak (coefficient 0.790, p=0.476). In other words, once the accounting-identity residual absorbs cyclical and mechanism noise, the association between cyclically-adjusted deficit and profit share strengthens substantially — but this is still an associational coefficient, not an identified causal one (endogeneity, since the deficit itself responds to the cycle and to profits, and policy-channel multiplicity, since tax cuts / transfers / healthcare are different mechanisms than "deficit volume," remain unresolved). A counterfactual built on this associational coefficient — where the deficit had been capped near the 1990s non-crisis mean or the 2010-2016 non-crisis mean — implies annual profit reductions of roughly $290–360B over 2010–2024, $727–843B over 2020–2024, and $650–783B in 2024 alone. The OpenCausality team explicitly labels these as associational counterfactuals, not causal predictions — intended to quantify the strength of the association, not to claim "if the deficit had been lower, profits would necessarily have fallen by this amount."
4. Accounting-level replication: Kalecki components numerically replicate the essay’s table; the three-way profit decomposition (profit ×6.66 ≈ real GDP ×1.92 × price ×1.77 × profit-share ×1.99) is independently validated — the three-axis log contribution shares (real / price / share) are about 35% / 30% / 36%, matching the essay’s 35% / 28% / 37% across all three axes; and BEA NIPA sectoral balances cross-validate the government-borrowing/profit ratio rising from 0.66 (1990s) to 0.78 (2020-24) (private net lending + government net lending + current account three-way balance).

Full DAG, LP coefficients, event windows, and reproducibility paths at the OpenCausality project repository.

Appendix B: How to read the identity

The accounting identity is a workhorse, but it can mislead a reader who treats it as a behavioral or causal statement. These four notes address the places where the identity’s algebraic structure invites a deeper question.

1. Worker consumption is a leakage, not an engine.

The Kalecki identity Π = I + (G−T) + NX + C_capitalist − S_worker can mislead a reader who notes that worker consumption (C_worker) does not appear on the right-hand side. The cancellation is algebraic, not economic. In the derivation, every wage dollar paid out (W on the income side) and every wage dollar spent (C_worker on the expenditure side) sits on opposite sides of the same equation; the round trip nets to zero. What survives is S_worker, the wages workers earned but did not spend — equivalently, (1 − MPC) × W. Worker consumption has not been removed from the framework; it has been split into a recycling-neutral part (which legitimately disappears) and a leakage part (which stays visible as the −S_worker term).

This matters because the essay’s “demand drives growth” argument and the Kalecki identity address two different dependent variables. Real GDP (Y) is multiplier-driven by MPC: higher worker consumption keeps the multiplier alive and lets autonomous injections amplify into a larger pie. The profit flow (Π) is determined by injections minus leakages, and worker consumption enters only via the −S_worker drag. Worker consumption is necessary — if MPC collapsed to zero, S_worker would consume the entire wage bill and profits would vanish — but it is not sufficient: no amount of MPC = 1 worker consumption produces profits without autonomous injections, because recycled wages net to zero by construction.

The clearest empirical refutation of the strong “worker consumption is the only real demand” reading is Engels-Pause Britain, 1770–1830. Real worker consumption stagnated for sixty years; profits and the capital share rose dramatically over the same window. The offsetting injections — investment doubling as a share of GDP, exports tripling, and Napoleonic-War-era defense spending — fully substituted for the missing worker-consumption growth. The identity holds; the strong reading does not.

2. Investment is a primitive in accounting, a propensity in behavior.

A natural follow-up is whether investment (I) is itself a derived term, parallel to how worker behavior shows up as the saving rate inside −S_worker. The answer has three layers.

Algebraically, no. Investment is a primitive in the expenditure decomposition Y = C + I + G + NX. Unlike C_worker, which gets squeezed into a saving rate by cancellation, I never had a cancellation structure — it stands as an explicit variable on the right-hand side.

Behaviorally, yes. Fazzari, Hubbard, and Petersen (1988) and the subsequent internal-funds literature find a robust elasticity of corporate fixed investment with respect to current cash flow, typically 0.3–0.5. Firms with higher profits invest more — partly because internal financing is cheaper than external, partly because current profits update expectations of future demand. The essay’s per-dollar-of-profit table is exactly an estimate of this propensity over time: physical capex as a share of profits fell from roughly 80% in the 1990s to roughly 50% in 2020–24, while buybacks rose from below 20% to above 50%. That decline is the corporate sector’s marginal propensity to invest out of profits collapsing.

Causally, Kalecki himself rejected the framing. His slogan — “workers spend what they earn, capitalists earn what they spend” — runs the causality from I to Π, not the other way around. Capitalists decide today’s investment based on expected future demand and animal spirits, and those decisions create today’s profits via the identity. A propensity to invest out of current profits is theoretically awkward in his own framework, though a propensity to invest out of expected profits is fine and is how the modern accelerator literature handles it.

The structural point underneath is that workers and capitalists face different decision sets. Workers receive income and choose between spending and saving — a binary disposition. Capitalists allocate profits across at least four channels: physical investment, distributions (buybacks and dividends), executive consumption, and corporate cash retention. Worker behavior collapses to a single MPC; capitalist behavior decomposes into a vector. The 80% → 50% capex/profits shift is one component of that vector moving.

3. Liquidity preference: the bridge from the identity to the General Theory.

The Kalecki identity is silent on the monetary side — the interest rate does not appear on either side. To see how worker saving, capitalist investment, and the rate of interest interact, the natural bridge is Keynes’s General Theory, which has three relevant building blocks.

The consumption function (Chapters 8–10) posits a “fundamental psychological law” that as income rises, consumption rises by less than the rise in income, so MPC < 1. This generates the multiplier 1 / (1 − MPC) — worker consumption matters via the slope, exactly the role identified in the Kalecki identity, expressed differently.

The inducement to invest (Chapters 11–12) is not driven by current profits but by the Marginal Efficiency of Capital: the discount rate that equates the expected stream of future returns to the current cost of the capital asset. Investment proceeds until MEC equals the interest rate r. MEC is expectations-driven, and Chapter 12 famously argues that uncertainty makes investment governed less by precise calculation than by spontaneous “animal spirits.”

Liquidity preference (Chapters 13–15) is Keynes’s sharpest break from the classical tradition. The interest rate is not determined in the loanable-funds market by S equating I; it is determined in the money market, where the supply of money M (set by the central bank) equates the demand to hold money L(r, Y) for transactions, precaution, and speculation. Because r is a monetary phenomenon, the classical auto-equilibration — workers save more → r falls → investment rises by exactly the amount saving rose — does not hold. Higher saving without an offsetting fall in r becomes a leakage that depresses aggregate demand. This is the paradox of thrift.

The modern application is direct. The 80% → 50% capex-share-of-profits decline reflects two channels Keynes would recognize: a falling MEC (high-margin software and asset-light platforms have fewer high-MEC physical-capex projects than the manufacturing-heavy firms of 1990 did), and rising corporate liquidity preference (firms holding large cash balances and distributing via buybacks rather than reinvesting). The federal deficit expansion from $134 billion to $2,149 billion average is fiscal injection substituting for missing private investment — the GT prescription when monetary policy is at its lower bound. One wrinkle Keynes did not fully anticipate: non-financial corporations themselves becoming major holders of liquid assets. Apple-scale cash hoards are corporate liquidity preference at the firm level — neither investment, nor distribution, nor traditional expense.

4. The intangibles wrinkle.

The categories on the right-hand side of the Kalecki identity are NIPA categories, and NIPA categories carry accounting conventions that can rewrite economic history when revised.

The clearest example is the 2013 BEA revision that added intellectual property products (IPP) to gross fixed investment. Before 2013, R&D was treated as a current expense and disappeared into the intermediate-input wash; after 2013, it is capitalized as investment with depreciation spread over future periods. The revision raised measured US GDP by roughly 3 percent overnight and raised measured corporate profits, because what was previously hitting the income statement as expense now spreads as a depreciation charge over future years.

The intangibles question reaches further than R&D. Corrado, Hulten, and Sichel (2009) argue that brand investment (advertising and customer acquisition that creates durable customer relationships), worker training (which builds human capital), and organizational capital (process improvements and internal know-how) are all economically equivalent to investment in physical plant, even though NIPA expenses them. Their adjustments raise total US business investment by 30 to 50 percent. By their accounting, the modern American firm invests heavily — just in categories that don’t get capitalized.

This matters for reading the essay’s per-dollar-of-profit table. The 80% → 50% decline in physical capex as a share of profits is partly a recategorization story: as the corporate sector shifts from manufacturing toward asset-light models, more of what is functionally “investment” sits in expensed categories that don’t show up in I.

But the conventional adjustment cannot do all the work. The investment-versus-distribution boundary — capex versus buybacks — is not a measurement question. Buybacks unambiguously route corporate cash to shareholders, who then choose between consumption and saving in their personal capacity. The rise from below 20% to above 50% of profits going to buybacks is real: corporations are returning a much larger share of their cash flow to shareholders than they did a generation ago. Some of the 80% → 50% capex/profit decline is recategorization; most of it, particularly the buyback portion, is a real shift from reinvestment to financial distribution.