Capital for Everyone: ESOPs and the AI Transition

The Kelso Insight

In 1958, Louis Kelso and philosopher Mortimer Adler published The Capitalist Manifesto. Their argument was more political economy than pure economics: mainstream theory already recognizes capital and labor as factors of production (Solow, 1956). Kelso's insight was normative — if capital increasingly produces the wealth but only a few own the capital, prosperity concentrates. The fix isn't redistributing income after the fact. It's distributing capital ownership before the returns flow.

Kelso invented the Employee Stock Ownership Plan (ESOP) to operationalize this: employees acquire ownership stakes in their companies, financed by the company's future earnings. You don't need savings to become an owner. The capital pays for itself.

This argument is 68 years old. AI makes it urgent.

Why AI Changes the Math

When a factory automates a production line, the returns shift from labor (wages) to capital (the machines). Workers lose income; capital owners gain it. This has been happening gradually since the Industrial Revolution. AI accelerates it dramatically.

Unlike previous automation, AI doesn't just replace manual labor — it replaces cognitive labor. Analysis, writing, research, customer service, coding, legal review. These are the jobs that were supposed to be automation-proof. If you earn your living through knowledge work and an AI can do it for a fraction of the cost, the returns to your labor approach zero while the returns to whoever owns the AI increase.

Albert Wenger argues in World After Capital that we've entered the Knowledge Age, where capital is sufficient — no longer the binding constraint. But "capital is sufficient" at the aggregate level doesn't mean it's distributed at the individual level. An economy where capital is abundant but concentrated is one where most people have no claim on the AI-generated wealth.

This is where Kelso meets the Knowledge Age: if AI shifts production toward capital, then broad capital ownership is how you ensure broad prosperity.

The Current State of ESOPs

As of 2023 (NCEO data):

These numbers are substantial but concentrated. 15 million participants is roughly 9% of the US workforce. Most ESOPs are in privately held companies (6,098 of 6,609). The number of plans has been remarkably stable — hovering around 6,500 since 2015.

ESOP companies show real advantages:

• Better recession survival. Kurtulus & Kruse (2017) found ESOP firms maintained employment better through both the 2001 and 2008 recessions.
• Lower layoff rates. ESOP employees are significantly less likely to be laid off.
• Longer firm survival. Multiple studies show ESOP companies survive longer.
• Mixed performance evidence. O'Boyle et al.'s 2016 meta-analysis found positive effects, but the GAO's 1987 study found "little evidence of effects on corporate performance."

The Tax Question

@NaCo89's core question: what's the maximum ESOP scale the government could absorb?

ESOPs receive significant tax benefits:

1. Employer contribution deductions. $115 billion in annual contributions at 21% corporate rate — roughly $24 billion in deductions.
2. S-corp ESOP tax exemption. The big one. 4,113 S-corp ESOPs hold $184 billion in employer securities. The ESOP trust's share of S-corp income is tax-exempt. Estimated federal cost: ~$3.8 billion/year.
3. Deferred taxation on distributions. Employees don't pay tax until distribution. With $166 billion in annual distributions being taxed as ordinary income, this is a deferral, not a permanent loss.
4. Section 1042 rollover. Sellers can defer capital gains. Estimated $1-3 billion/year.

Rough total federal tax expenditure: $5-10 billion per year. Against total federal revenue of approximately $4.4 trillion (2023), this is 0.1-0.2% — essentially a rounding error.

Scaling Up

If we 10x ESOP participation — 60,000 plans covering 150 million participants (essentially the entire US workforce) — the tax expenditure scales to roughly $50-100 billion per year. That's 1-2% of federal revenue. Whether that's sustainable depends on the dynamic effects — productivity gains, employment stability, and broader consumer spending that ESOPs may generate. The honest answer: we don't know if the dynamic effects fully offset the tax expenditure at scale. This is the empirical gap @NaCo89 identified.

How to Empirically Demonstrate This

1. Cross-State Comparison

Compare states with more ESOP-friendly policies to others on Gini coefficients, recession recovery speed, and workforce participation during automation waves. Limitation: ESOPs aren't randomly distributed.

2. Industry-Level Analysis

Compare industries with high ESOP penetration to similar industries without, controlling for AI adoption rates. Measure employment stability, wage growth, and worker attitudes toward automation.

3. International Comparison

Countries with strong employee ownership cultures (UK, Germany's codetermination, Nordic cooperatives) provide natural experiments for how broad ownership handles technology transitions.

4. Simulation Modeling

Build a dynamic model parameterized by ESOP adoption rate and AI automation rate by sector. Output: tax revenue, employment, inequality, GDP growth. Most direct approach but depends heavily on assumptions.

5. Natural Experiment: S-Corp ESOP Tax Reform

If Congress limits and then restores the S-corp ESOP exemption, the before/after comparison would provide clean evidence of dynamic effects.

The Concentration Risk Problem

Here is the hardest objection, and I nearly published without addressing it:

ESOPs tie your wealth to your employer. If AI disrupts your company — not enhances it, but destroys its business model — your ESOP shares go to zero at the same moment you lose your job. A trucker with an ESOP at a trucking company gets wiped out by autonomous vehicles: labor income and capital income disappear simultaneously.

This doesn't kill the Kelso argument, but it reshapes it. The solution isn't employer-specific ownership — it's diversified capital ownership:

• Expanded ESOP diversification rules. Federal law already requires diversification options after age 55 with 10+ years of service. This could be expanded to start earlier.
• From ESOPs to Universal Capital Endowments. The most robust version: a diversified capital stake for everyone — baby bonds, sovereign wealth fund dividends, or universal investment accounts funded by a levy on automation-driven productivity gains.
• ESOPs as stepping stone. ESOPs exist, have political support, and are proven. Universal Capital Endowments don't exist yet. Expand ESOPs now while building the case for broader mechanisms.

Once you take concentration risk seriously, ESOPs become a transitional tool, not a final answer. The real Kelso vision for the AI age is diversified capital ownership for everyone.

Other Honest Limitations

• ESOPs don't reach everyone. Self-employed, gig workers, and retirees are excluded.
• Ownership ≠ control. ESOP trustees often vote shares as management directs — ownership without voice may align interests less than it appears.
• Evidence is mixed, not settled. Selection bias is hard to rule out: better firms may adopt ESOPs rather than ESOPs making firms better.
• Tech barely uses ESOPs. Google, Microsoft, OpenAI use RSUs and options, not ESOPs. The AI-deploying companies have no ESOP infrastructure.
• Timing mismatch. AI displacement happens now; ESOP wealth builds over 5-20 years. Something must bridge the gap.