The Human Capital Markets Model.

What the model is

Ed Walker built the model with an AI assistant in October 2026. It starts from published figures for the U.S. economy and changes one thing on purpose: the share of the economy's income paid as wages and salaries. Then it follows the effects through taxes, household debt, spending, ownership and the nuclear workforce.

It is not an economic model of the kind a central bank or university would publish, and it is not investment, tax or legal advice.

The main assumption

The model lowers the wage share of GDP by 2035, starting from roughly 42 percent. The main cases are declines of about:

These declines are assumptions chosen to stress the system, not predictions.

What it starts from

• Gross domestic product and the share of income paid as wages and salaries
• Federal receipts by source, and payroll-funded programs such as Social Security and Medicare
• Household debt and consumer spending
• How stock ownership is distributed across households
• The nuclear workforce

Each input has a published source. Inputs are re-verified before the book is released, and any change is dated below.

What it calculates

• How much income shifts from wages to capital, and how much of that reaches the bottom 90 percent of households given current stock ownership
• The effect on federal income and payroll tax receipts, and on Social Security and Medicare funding
• Household debt relative to wages
• Consumer spending relative to a path without AI
• Knowledge at risk in the nuclear workforce
• Judgment-based scores for different ways of representing human capital, such as verified capability records and human capital tokens
• Three illustrative futures: Drift, Tax the Machines, and Broad Ownership and Accountability

Its limits

• The assumptions were designed by an AI assistant at Ed Walker's direction and are editable. Different reasonable assumptions give different numbers.
• It holds much of the economy fixed while changing the wage share, which real economies do not do.
• It does not model new kinds of work that AI may create, which is the strongest historical counterargument.
• The scores for capability records and tokens are judgments, not measurements.
• The three futures are illustrations, not political positions or predictions.

Updates

Last reviewed October 23, 2026

No changes since the version described in the book. When an input is re-verified or an assumption changes, the date and the change will be listed here.

The full workbook

The full model workbook is not yet available. It is being cleaned up, documented and given terms of use before it is offered.

To hear when it is updated or available, email Ed. This opens your email app; nothing is sent until you choose to send it, and it does not enroll you in a newsletter.