Stage 2 — Analysis, Memo, Prompt Log

Objective

Explain the verdicts by mechanism, follow the money to the asset that was capturing it, show what entry did to both — and write the recommendation to somebody who has to act on it.

Learn — the four things the analysis has to establish

The verdicts, by mechanism

Not a restatement of the table. The part-timer is hit hardest because the car's costs do not scale: a full payment and full insurance sitting on half the revenue. The office worker's positive economic profit is the same statement in reverse — and it is the reason not everyone quits the office. The cab driver wins at the driver level and hands $36,000 a year to somebody who never drives, which is the hook into the second half.

Net-to-net discipline

Why the implicit cost is $52,200 and not the $57,600 gross, and why the part-timer's alternative is halved. If your Stage 1 figures matched the check figures you have already applied this; here you articulate it, because a reader who cannot follow the reasoning cannot trust the verdicts.

Where the money went

The medallion is a factor in fixed supply — 13,587 and no more — so the lease is economic rent, and an asset earning rent indefinitely is worth rent ÷ required return. $36,000 ÷ 3.5% = $1,028,571 against an observed peak above $1M. $18,000 ÷ 6.0% = $300,000 against roughly $335,000 observed. One division, both eras, within about 10% each time. The arithmetic is the story — an analysis that narrates the collapse without it has described the weather.

Why the rent collapsed, and what the moat was actually protecting

App vehicles went from about 40,000 to more than 120,000. Fares fell, so the lease fell; the stream got riskier, so the required return rose. Both moves shrink the same fraction, which is why 70–85% of the value evaporated without anyone revoking a license. And the cap blocked entry into yellow cabs, never into rides — rent survives only as long as the moat surrounds the market rather than the product.

Do — write the analysis

Create analysis/economic-profit-analysis.md. Open by reproducing your Stage 1 hypothesis unedited, followed by an honest verdict: where it landed, and what you misjudged. Wrong but well-reasoned scores as well as right.

Then the four things above from your own model's numbers, plus the supply-and-demand section: which curve shifted and in which direction — supply, right, massively; demand also right, but the supply shock dominates — and what happened to price, quantity, and whose surplus. The Econ Policy Lab draws those surplus areas if the geometry is not obvious yet.

Use your own sensitivity run. The 30 → 22 days-per-month result is the gig-economy fragility point in a single number: every driver verdict goes deep negative one input away from the base case.

Export at least two figures into analysis/figures/ and reference each in the text. Natural candidates: accounting against economic profit across the four workers; capitalized value against observed price in both eras; the sensitivity result. A figure earns its place by carrying evidence the text uses — here a decorative one costs the argument the space it should have served.

Do — close the arc

One paragraph connecting this case to the one before it. The medallion's rent behind a moat is the seed patent in miniature: a legal barrier creates rent, the rent capitalizes into an asset price, and entry that bypasses the barrier destroys both. Draw the logic — naming the other case is not the same as connecting it.

That link is what makes three separate engagements one argument: competition, then market power, then where the profits actually went.

Do — write the memo and update the log

The memo goes in docs/decisions/economic-profit-memo.md — half a page to somebody who has to act. A regulator deciding whether to cap app vehicles, a lender deciding what a medallion is worth as collateral, or a driver deciding whether to lease one; pick your reader and write to them. Recommendation, the reasoning that drives it, the judgment call where the evidence ran out, and what would change your answer. The memo template has the structure.

Then update prompt-log.md at the repository root with a dated section covering both stages, and close with a reflection of 300 words or fewer naming one AI error you caught and how you verified it. Economic versus normal profit is a reliable place for a model to stumble, and it stumbles confidently.

Finally, update capabilities/economic-profit/README.md so its "exercised in:" line points at the analysis and the memo as well as the brief.

Deliverable

• Hypothesis reproduced unedited at the top, with an honest verdict against the model
• The four verdicts explained by mechanism — the part-timer's non-scaling car costs named explicitly
• Net-to-net discipline articulated: why $52,200 and not $57,600, and why the part-timer's alternative is halved
• The medallion story told with the capitalization arithmetic, both eras
• Both capitalized values tied to their observed prices, and the ~10% agreement noted
• The collapse explained by both moves — lease down and required return up — plus the honest note that lending inflated the peak
• The moat point landed: the cap blocked entry into yellow cabs, never into rides
• Supply and demand: correct curve, correct direction, price, quantity, and whose surplus
• Cross-case link drawn on the moat → rent → entry logic, not name-dropped
• The 30 → 22 sensitivity used as the fragility argument, with your own numbers
• At least two figures in analysis/figures/, each referenced and analytically load-bearing
• Figures render on the GitHub page
• Memo written: recommendation, reasoning, the judgment call, what would change your answer
• prompt-log.md updated across both stages, with a reflection of ≤300 words
• The reflection names a concrete AI error caught, or the checks that cleared it
• capabilities/economic-profit/README.md "exercised in:" line updated
• At least two descriptive commits for this stage

Check yourself