The free-money era subsidized your salary, not just startup valuations

Eco3min Research · Monetary Regimes

For two years, lending to the US government guaranteed a loss. That subsidy didn’t stop at asset prices — it paid wages.

May 2026 Coverage: 2003–2026 281 monthly observations CC BY 4.0

Between 23 March 2020 and 18 April 2022, the 10-year US real Treasury yield — the market’s inflation-adjusted risk-free rate — stayed below zero for 520 consecutive trading sessions, never once closing positive. For roughly two years, lending to the federal government for a decade guaranteed a loss in real terms.

That negative real cost of capital is usually discussed through one channel: it inflated the price of growth equities. The less-discussed channel is that it also lowered the cost of financing companies that lose money for years — which is also how headcount and pay were funded. The link from cheap capital to valuations is mechanical and visible. The link to employment is an inference this page makes explicit rather than asserts.

01What “cheap capital” actually means

The dominant account of the past four years is a valuation story: the end of “free money” repriced richly valued technology stocks, deflated unicorns, and closed the IPO window. That account is correct as far as it goes. Lower interest rates raise the present value of profits expected far in the future, so the assets most sensitive to rates — long-duration growth equities — rose most when capital was cheapest and fell hardest when it was not.

But the price of capital is not only a discount rate applied to a stock; it is also the cost of staying alive while unprofitable. A company that burns cash funds the gap by raising capital, and the cost of that capital is set by the same real rate. When the real cost of money is negative, the burn is cheap to finance and the market rewards growth over profitability. When it turns positive, the same burn becomes a recurring, expensive liability. The relevant measure is therefore the real cost of capital — the nominal rate stripped of inflation — and the cleanest market reading of it is the 10-year inflation-indexed Treasury yield.

02The real cost of capital changed regimes

The 10-year real yield was negative or near zero for most of 2011–2013, then fell decisively below zero through the pandemic. It reached −1.19% on 3 August 2021, the lowest reading in a series that begins in 2003. It then reversed with unusual speed. By 25 October 2023 the same yield had risen to 2.52% — a level it had not reached since 28 November 2008, in the depths of the financial crisis. The move from trough to peak was 3.7 percentage points in about twenty-six months. As of late May 2026 the real yield stands near 2.06%, still far above its 2010s norm; the 2012 annual average was −0.48%.

Drag the chart below. The shaded band is the stretch where lending to the US government locked in a guaranteed real loss.

The chart plots monthly averages; the −1.19% trough and +2.52% peak cited above are daily extremes within those months.

• For 74.3% of trading days in 2020–2022, the real yield was negative. In 2011–2013 it was 51.1%. Before 2009, almost never.
• The reversal to +2.52% was the first time since the 2008 crisis that real safe capital was this expensive.
• The policy rate tracked the same arc, harder: the real federal funds rate hit about −8.3% in March 2022, with inflation at 9.06% against a near-zero funds rate.

03Why a negative real rate makes loss-making growth cheap

Here is the mechanism the headline rests on, made tangible. A negative real discount rate does something specific to the math of valuing a company: it makes a dollar of profit expected ten years from now worth more than a dollar today. That inversion is the entire reason cheap capital favours companies that promise distant profits over companies that earn now.

The simulator below is pure arithmetic — the present-value formula, nothing forecast. Move the real discount rate and watch what happens to the value, today, of a profit expected in future years.

The lesson sits in the gap between the presets. At the 2021 trough, $100 of profit expected in twenty years was worth more than $100 today. At today’s real rate it is worth roughly two-thirds. The same business plan — profits deferred far into the future — was repriced by the discount rate alone, before a single operating number changed.

Now extend the logic from the share price to the company itself. The same cheap capital that made distant profits valuable also made the cash burn on the way there inexpensive to finance. A business that loses money can raise it at a low real cost and run at a loss for longer. And the activity the burn pays for is, to a large degree, people: software companies that are not yet profitable spend most of their capital on salaries. A lower cost of carrying losses is, in practice, a lower cost of carrying payroll above what current revenue supports.

Put plainly: the cheap capital that lifted valuations did not stop at the share price. It financed the operating losses of a cohort of companies, and those losses were substantially wages. When commentators say free money inflated asset prices, the same money was also, through this chain, subsidizing employment and pay in the parts of the economy most dependent on external financing.

04What this does not prove

The employment link is the weakest part of the chain by construction, and three qualifications bound it.

It is not a measured pass-through. No public series isolates “jobs financed by cheap capital.” The mechanism above is consistent with the data and with how loss-making firms allocate capital, but it cannot be quantified the way the yield itself can. It belongs in the category of defensible inference, not counted fact.

Cheap capital was one driver among several. The 2020–2022 period also saw a genuine surge in demand for digital services, real network and platform economics in some firms, and measurable productivity gains in others. Disentangling the share attributable to the cost of capital from the share attributable to demand or to genuine value creation is not possible from these series alone.

05What the uneven normalization implies

If the real 10-year yield holds near 2% — the most positive real-rate regime since before the 2008 crisis — the discount applied to distant, uncertain profits is structurally higher than it was through the 2010s, and the cost of carrying operating losses is no longer near zero. Historically, regimes of positive real rates have been associated with wider dispersion between profitable and unprofitable companies and with compression in the multiples of the longest-duration equities.

The unfinished part of the adjustment is where the signal lies. Long real rates have normalized; the rest of the cost of capital has not. The real federal funds rate was back to −0.49% in April 2026, and the high-yield spread, at 2.71% in late May 2026, sits near the tightest of this cycle. The risk premium remains compressed on top of a restrictive real base. Whether the cost of risky capital fully reprices depends less on the risk-free rate, which has already moved, than on the credit spread, which has not.

06Methodology & data

The real cost of capital is proxied by the 10-year inflation-indexed Treasury yield (the market’s real risk-free rate); the policy and credit dimensions by the federal funds rate and the ICE BofA US high-yield option-adjusted spread. The link to employment is treated as an inferential mechanism, not a measured series. All inputs are public, and the counted figures are computed by Eco3min from daily and monthly data.

import pandas as pd

# 10Y real yield (TIPS) straight from the US Treasury — no API key
url = ("https://home.treasury.gov/resource-center/data-chart-center/"
       "interest-rates/daily-treasury-rates.csv/2026/all"
       "?type=daily_treasury_real_yield_curve&field_tdr_date_value=2026&_format=csv")
df = pd.read_csv(url, parse_dates=["Date"])

# Longest unbroken run of sub-zero real yield
neg = df["10 YR"] < 0
run = (neg != neg.shift()).cumsum()
longest = df[neg].groupby(run).size().max()
print(longest)  # 520 trading days (Mar 2020 – Apr 2022)

07Data & reproducibility

The monthly real-yield series used in the chart and simulator is provided in open format, updated when the US Treasury publishes.

License: Creative Commons Attribution 4.0 (CC BY 4.0). Free for research, academic and journalistic use with attribution.

08Questions & answers

09Sources & limitations

• PrimaryUS Treasury — Daily Treasury Real Yield Curve Rates (10Y TIPS), 2003–present.
• PrimaryFederal Reserve — Effective federal funds rate, 1954–present.
• PrimaryBureau of Labor Statistics — CPI-U, used for the real policy rate.
• PrimaryICE BofA — US High Yield option-adjusted spread.

• The employment link is an inference, not a measured series. No public data isolates wages financed by cheap capital; the mechanism is defensible but not quantified.
• Real-policy-rate and CPI figures are scoped to the recent period. The −8.3% real funds rate and 9.06% CPI are not presented as multi-decade records.
• The simulator is a present-value identity, not a valuation. It holds the cash flow fixed and varies the discount rate; it is not a security price or a forecast.
• One real-rate proxy. The 10-year TIPS yield is the risk-free real rate; the cost of risky capital adds a credit spread that has not normalized in step.

10Cite this

No investment advice — informational analysis only. The decomposition presented is descriptive: it reports the real cost of capital and its historical distribution. The present-value simulator is an illustrative computation, not a forecast or a security valuation. Past distributions are not predictive of future levels.

11Related research

• The gap between the Fed’s balance sheet and the market’s liquidity
• A history of 2s10s yield-curve inversions and their lead times
• How monetary transmission reaches corporate earnings and margins