Method · Revision 1.3 · 21 September 2026

How we score dependency

Every item on every map, and every report we write for someone who asks, gets one of four ratings for how far it depends on the people who own companies voting their values. This is the whole of how that rating is reached.

A rating answers one question: how much of this outcome is in the hands of the people who own the companies?

We answer it twice, once for what companies do and once for what they spend to stop the outcome, and take whichever answer is higher. Owners reach both through the same vote, voting alongside a civil society organization they choose.

The two questions

Answered separately, on the evidence, before any rating is written down.

Question 1

Corporate conduct

How far does the outcome depend on companies changing what they make, how they make it, whom they employ, what they sell, and what they emit or discharge?

The bold line is the step. The grey line under it is one example of an item that landed there, not a test to apply.

  • C-decides
    Conduct decides it.Changing what companies do is necessary, and close to sufficient on its own.Forced labor in a supply chain falls sharply when the buyer changes who it buys from.
  • C-required
    Conduct must change, and so must others.Companies have to move, and governments, households or others have to move too.Wages above the legal floor are set inside companies; the floor itself is set by statute.
  • C-marginal
    Conduct helps at the margin.It matters, and it is not on the critical path.Commercial cash bail is a real corporate practice touching the presumption of innocence, and the reform runs through state legislatures.
  • C-none
    Conduct is not a lever here.Nothing a company makes, sells or employs bears on the outcome.Nationality is conferred and revoked by states, and no company issues one.

Question 2

Corporate political influence

How far does the outcome depend on companies stopping or reversing their lobbying, campaign giving and regulatory influence against it?

Again, the grey line is one example, not a test.

  • P-blocks
    Their political spending blocks it.A primary obstacle. The outcome will not arrive while it continues.A bill that passed the House twice and has never reached a Senate vote, kept there by a named, documented lobbying campaign.
  • P-slows
    Their political spending slows it.It narrows or delays progress, and is not the main obstacle.Industry lobbied to weaken a law; the law passed in strengthened form anyway.
  • P-none
    Their political spending is not a factor.We looked for a fight and there is not one.Multiple searches for a lobbying dispute on this subject found none.

Each answer counts on its own, and the higher one sets the rating

AnswersRatingWhat it means for a reader
C-decides, or P-blocksPivotalWhat owners control decides this. Either corporate conduct is close to sufficient on its own, or corporate political spending is the thing standing in the way.
C-requiredNecessaryOwners remove an obstacle nothing else removes, and governments, households or others must also act.
C-marginal, or P-slowsHelpfulOwners help at the margin. Other levers carry most of it.
C-none and P-noneIndependentThis moves without owners. It belongs to governments, courts, aid or households.

Two rules

What is not a lever

Donations, sponsorships, grants and in-kind giving; cost reductions, efficiency and operational improvements; and the bare fact that a company supplies an input to the people doing the work do not count as corporate levers on any topic, since any company can offer them on any issue. A lever is a change in what a company makes, how it makes it, or what it sells, or in its political influence, that a source ranks as material to the obstacle. Where an obstacle's only corporate connection is one of the three excluded kinds, it is rated Independent. A stated absence (“no lobbying was found”) is recorded, not argued from.

Only a public company counts

Only publicly traded companies, or trade associations funded by them, count toward a rating, because only a vote reaches them. Where an obstacle's corporate actors are privately held firms, mutuals, nonprofits or trade associations without listed members, no vote reaches them and the obstacle is rated Independent. We never rate on what a company could do if it were publicly traded.

The filter

A care report is written in one pass by an analyst that searches the web, holds every page it reads as a document the moment it reads it, and names the held document beside every factual sentence it writes. Then our own code, and no model, does three things to the finished draft. Every marked sentence is checked against the document it names: the figures and the names in the sentence must be found in that document, or the sentence is removed and listed under “What was removed, and why” with its text, as is any factual sentence with no marker. Every entry in the advocacy record must point at a source we hold, and which source depends on what is claimed: a filing, a co-filing or a vote result at the SEC filing or a row of our own ballot table, because only those have a filing behind them; and an engagement, a dialogue, a letter, a public statement or a campaign at the page it was read on, with no filing asked of it. The rest are removed and listed, and no address is ever substituted. That no filing is on record may be said only after EDGAR and our ballot table were searched; that no engagement was found may be said only of the documents the report read. And every company the pass names is resolved and marked public, private, mutual, nonprofit or trade association; a rationale whose only reachable actor is not public, or whose lever is a donation or an efficiency, is flagged, and the report is held as a draft for a person to read rather than published. The filter changes no rating, moves no obstacle and writes no sentence. Each version we replace is kept rather than deleted, with the reason beside it.

Name something you care aboutSee the dependency maps

Revision 1.3, 21 September 2026. Claude was used to conduct the analysis. The work is human led, and human review is not yet applied to every item.