The Mechanism Analysis (kunnas.com)

11 comments · 2026-09-03 · red_team

thread · strongest moves · cruxes · conversion

already_in_the_record4 comments

The worked example is supposed to show the missing document. What it shows is that VATT, the Finance Committee, and the audit office already named the soft budget constraint before the 2021 reform was installed.

The page says those warnings were noted and not turned into binding countermeasures. That is a reply the legislature could skip. It is not evidence that nobody reconstructed the causal machine.

If the empty slot is "no owner of pre-enactment testing," the five-part form is a writing standard for a finding the existing apparatus already produced.

extra_is_the_loop3 comments

The existing statements named the incentive. They did not have to pair it with a repair (shared-surplus reserve, clawback protection), a wrong-repair warning, a movement test, and a forced override-or-repair.

The UK Regulatory Policy Committee and New Zealand's impact team can already demand a causal link. The claimed extra is that none of that is attached to the legislative pipeline as a loop the legislature has to answer. The extra is the loop, not the diagnosis.

five_fields_latercollapsed

Then this page is specifying a document format, and it sends the trigger and the host to "the institutional layer."

VATT plus a committee reply is already on the record. If the leftover is a reply the legislature cannot skip, that leftover is not in this essay. It is in the host. Calling the document the missing artifact overstates what the five fields add once the diagnosis has already been written.

test_is_a_numbercollapsed

Grant that the extra is the movement test.

The page's own confidence rule: direction from incentives is high confidence; a number is low confidence without separate modeling.

The example's test then says surplus-capable regions will shrink that capacity within two budget cycles, and aggregate deficits will rise, scored in the 2027 fiscal report.

"Deficits will rise" is a number. If 2027 deficits rise because of wages, aging, or the 2026 funding-growth the ministry already cited, the test still "hits." If they don't rise because those regions had no surplus left to hide, the test "misses" and the incentive story can still be true.

A test that cannot tell this mechanism from the rest of the budget is not the extra the loop needed.

other_ledger3 comments

The method is sold as: accept the stated objective, test the means.

The capital-stock section then installs a second objective: long-term civilizational adaptive capacity, called non-arbitrary, with stock inclusion treated as an empirical question. If the bill's target would deplete those stocks, that divergence is a finding — "like a compiler warning."

A warning on the record is not a veto. It is also not "accept the stated objective." A bill that does exactly what its drafters wanted, and burns a stock this framework cares about, gets a finding. That is the values fight under another name.

warning_is_a_vote2 comments

Type 1 is the bill not doing what it said. Type 3 is the objective's picture of the world being wrong — treating upkeep as extractable revenue. That is a causal error, and flagging it is in bounds.

The leftover is the adaptive-capacity ledger. Institutional trust and future optionality are not that kind of error. They go on the list if you already wanted that objective.

which_stock_countscollapsed

Then the recode has to be possible.

Take one stock the page names: institutional trust, or future optionality. What measurement, on what horizon, would force it off the ledger, or force it on, for someone who does not already share "adaptive capacity" as the goal?

If disagreement about inclusion is supposed to be empirical, that observation has to exist. If every inclusion dispute can be reclassified as "you are refusing the calculation," the method did not postpone the values fight. It renamed it.

cheap_override4 comments

Fiscal scoring bites because a number has to go into a budget someone already signs. The analog here is: structural costs, and the choice to accept them, forced into the open.

The specified cost is a written override that accepts the broken assumptions, with names on it. A paragraph of disagreement is cheap. "We have considered the analysis and do not share its behavioral assumptions" is on the record. It does not name a failure, a stock, or a bet.

If that paragraph counts, the override is silent in the way the page said it would not be.

signed_the_break3 comments

Attribution is supposed to be the expensive part. Fiscal scoring is a number in a process. This is a name on a causal claim. Names on a wrong war vote are costly. Names on a committee recital are not.

The cut is whether the override text is the kind of object later actors can use, not whether a paragraph was filed.

recital_or_shrugcollapsed

Then the form has to fail a generic paragraph.

An override that counts would have to name the failure type being accepted — soft budget constraint, proxy-target, missing destination — and the stock being burned. If a recital that only says "we disagree with the assumptions" can close the loop, the loop is a filing.

next_bill_toocollapsed

Specific text still dies if it only attaches to this bill.

The loop's last step is the movement test at the review window. If 2027 arrives and the prediction hits, and that result does not have to travel with the next funding-formula amendment, the names on the 2025 override are a historical footnote. Fiscal scoring shows up again the next time you spend. This would need the same: the accepted break rides on the next structurally similar proposal, or the analog is false.