A monopoly model, read off twenty-four companies' filings

By market

A company's markup depends on the market it sells into, so this section groups companies by market and reads the whole market at once: how concentrated it is (the HHI, or Herfindahl-Hirschman Index: each firm's share of the market in percent, squared, then added up) and how large the markup is across it (each firm's Lerner index times its share, added up). Only US mobile wireless is here so far; more markets are added one at a time.

In the textbook Cournot model, where firms pick quantities, the share-weighted Lerner index equals HHI divided by the market's demand elasticity (HHI taken as a fraction, so 3,000 is 0.30), so across markets the points on this chart should rise as HHI rises.

• US mobile wireless

  Year-end 2023, measured in postpaid and prepaid mobile wireless connections at the four largest publicly traded carriers, as the FCC's table gives them from their annual reports (each firm's count in the table below is that same FCC table's; a company's own entry further down this page may count its subscribers another way, as AT&T's does, and its markup does not depend on either count). Market total 375.3 million: FCC, 2024 Communications Marketplace Report (FCC 24-136). Table: Fig. II.B.3, Reported Postpaid and Prepaid Connections for Publicly Traded Facilities-Based Mobile Wireless Service Providers, YE 2023 column: Top 4 Providers Total 375,348 thousand (page 49). The table's note: the figure includes only the providers it lists.

  Firm	Connections	Share	Lerner
  AT&T106.3 million connections	106.3 million	28.3%	0.31
  Verizon144.8 million connections	144.8 million	38.6%	left out
  T-Mobile119.7 million connections	119.7 million	31.9%	0.18

  HHI (a lower bound)

  at least 3,307

  Share-weighted Lerner

  0.15

  Named firms hold 98.8% of the market total. Because firms are missing, 3,307 is a lower bound: the HHI over every firm the total counts is at least that. By the 2023 Merger Guidelines that is at least highly concentrated (above 1,800 is highly concentrated, above 1,000 concentrated); by the 2010 guidelines, at least highly concentrated (highly concentrated above 2,500, concentrated from 1,500).

  Left out of the weighted Lerner: Verizon (no figures for fiscal 2023). The scored firms hold 60.2% of the market total, so the weighted Lerner covers only that share; their shares are used as they are, not scaled up to 100%.

  Not in the table: UScellular (4.6 million connections at year-end 2023), the fourth carrier in the FCC's table and in its total. Smaller carriers and cable resellers are not in the FCC's table, so they are outside this count.

  • AT&T connections: FCC, 2024 Communications Marketplace Report (FCC 24-136). Table: Fig. II.B.3, Reported Postpaid and Prepaid Connections for Publicly Traded Facilities-Based Mobile Wireless Service Providers, YE 2023 column: AT&T 106,340 thousand (page 49). The table's note: AT&T also reported about 135 million connected device and wholesale connections, which the table leaves out because the other carriers do not report them.
  • Verizon connections: FCC, 2024 Communications Marketplace Report (FCC 24-136). Table: Fig. II.B.3, Reported Postpaid and Prepaid Connections for Publicly Traded Facilities-Based Mobile Wireless Service Providers, YE 2023 column: Verizon 144,751 thousand (page 49)
  • T-Mobile connections: FCC, 2024 Communications Marketplace Report (FCC 24-136). Table: Fig. II.B.3, Reported Postpaid and Prepaid Connections for Publicly Traded Facilities-Based Mobile Wireless Service Providers, YE 2023 column: T-Mobile 119,700 thousand (page 49); the same count as T-Mobile's own fiscal 2023 Form 10-K, Total customers

What these numbers cannot tell you

• Where one market ends and another begins is a judgment: this one counts the postpaid and prepaid connections of the four largest publicly traded carriers, as their annual reports give them; the FCC's table leaves out AT&T's connected-device and wholesale connections, about 135 million, because the other carriers do not report theirs.
• The market total is the FCC table's Top 4 total, so firms outside those four, such as smaller carriers and resellers, are not in it: every share printed is an upper bound on that firm's real share of all US connections, and the HHI is the four carriers' HHI, not one over every US connection.
• Cost of revenues, or here cost of operating, is not marginal cost; it holds fixed cost, so each firm's Lerner index reads low.
• Where the named firms' shares add to less than 100 percent, a firm the total counts is not named, so the printed HHI is a lower bound on the HHI over every firm in the total: that figure is at least as high.
• The firm counts and the market total come from one FCC table, the current edition's, counted the same way. A count of every US connection, such as CTIA's estimate, is measured another way and is not used here.

Every company is still shown below on its own, whether or not it sits in a market yet.

Careers as markets: the wage markup over opportunity cost

The firm side of this page measures a markup of price over marginal cost. A job has a price too, the wage, and a cost of supplying it: the years of unpaid training, the debt and the hours it takes to get there. This section asks how much of each profession's pay premium the cost of training explains, and how much is left over. The left-over part is the most that a licence or a limited number of training places could be worth. It is not a measure of that, because the same left-over holds skill, demand and luck.

Dollars a year. Benchmark wage $50,980, the median for all workers in May 2025: Bureau of Labor Statistics, Occupational Outlook Handbook: Electricians (Pay). Quote: "In May 2025, the median annual wage for all workers was $50,980." The right-hand columns are modeled, not published by anyone; the method is below.

How the modeled columns are built

• Pay premium = the profession's median annual wage minus the benchmark wage below. The benchmark is the median for all workers, standing in for what the person would earn if they skipped the training. It is a rough stand-in: it is not the wage of a school leaver, which this section did not pull.
• Compensating differential, modeled = (unpaid training years x the benchmark wage + direct costs of training) / the working years below. It is the yearly pay that would make the training break even with no premium at all, with no interest and no discounting. Ignoring interest makes this part too small, so the left-over is too large. One thing pushes the other way: the doctor's medical-school debt includes living costs that were borrowed, and those living costs overlap the forgone earnings already counted for the same years, so the doctor's compensating differential may be too large and the left-over too small. No correction is estimated for either effect.
• Left over, modeled = pay premium minus the compensating differential. For a licensed profession this is a ceiling on the licensing rent: the most that limiting supply could be worth, with skill and demand not separated out. For an unlicensed profession it is not licensing rent at all.
• A cost the section could not source is listed as a gap on that profession's card and is left out of the sum. A sum with a gap in it is marked a floor, and the left-over beside it is marked a ceiling. Nothing is estimated to fill a gap.
• Hours are not in the sum for any profession. The Bureau of Labor Statistics page for each one says in words whether long or irregular hours are common, quoted on the card; no hours figure by profession was pulled, so no hours cost is counted. A profession whose Bureau of Labor Statistics hours text reports long, irregular or over-40-hour weeks is treated as a floor, as for a gap: its compensating differential is shown as "at least" and its left-over as "at most".
• The Compensating and Left over columns are modeled: they are arithmetic on the sourced figures above, with the assumptions listed here, and not figures any agency publishes.
• "At least" and "at most" apply to the unrounded values; each figure is printed rounded to the dollar.
• Working years after training: 40, the same for every profession. This is an assumption of the model, not a sourced figure.
• Years in school or in unpaid training are valued at the benchmark wage. Paid training years (a paid apprenticeship, a paid residency) count as zero forgone earnings. Where that training pay is not sourced (the electrician's apprentice pay), this is a stated gap on that profession's card.

Each profession, with its sources

• Doctor (physicians and surgeons)

  Median pay, May 2025

  $275,930

  Pay premium over the benchmark

  $224,950

  Compensating differential (modeled), a floor

  at least $16,183

  Left over (modeled), a ceiling

  at most $208,767

  Bureau of Labor Statistics, Occupational Outlook Handbook: Physicians and Surgeons. Pay: median annual pay $275,930, May 2025; 862,800 jobs in 2025.

  Training

  Four years for a bachelor's degree and four for a medical degree, unpaid. Then a residency of 3 to 9 years depending on the specialty, which is paid. Unpaid years counted: 8.

  Bureau of Labor Statistics, Occupational Outlook Handbook: Physicians and Surgeons. How to Become a Physician or Surgeon: a medical degree takes an additional 4 years; residency typically lasts 3 to 9 years depending on the specialty.

  • Bachelor's degree, 4 years of tuition and required fees at a public 4-year school (4 x $9,872): $39,488. National Center for Education Statistics, Digest of Education Statistics 2024, Table 330.10. Public 4-year institutions, 2023-24, tuition and required fees $9,872 (in-district rates for public schools); the average across all students, not this profession's.
  • Medical education debt, the median among graduates who borrowed: $200,000. AAMC, Medical Student Education: Debt, Costs, and Loan Repayment Fact Card for the Class of 2024 (October 2024). Education Debt Breakdown: medical education debt, 67% of graduates, median $200,000. The debt includes living costs borrowed, and 33% of graduates borrowed nothing for medical school, so it is not the typical outlay of every doctor.

  Counted: forgone earnings $407,840 plus direct costs $239,488, spread over 40 working years.

  What limits supply

  Every state requires a licence: an accredited medical degree, a residency in the specialty and national exams. Training places are limited: in the 2025 Main Residency Match, 52,498 applicants registered for 43,237 positions offered, and 93.5% of U.S. MD seniors matched to a first-year position. Most U.S. MD seniors do match, so these figures alone do not show that places bind them; they bind other applicants.

  National Resident Matching Program, press release "NRMP Releases 2025 Main Residency Match Results and Data Report" (May 29, 2025). Positions offered 43,237; registered applicants 52,498; U.S. MD senior PGY-1 match rate 93.5 percent. The licensing steps are from the Bureau of Labor Statistics page above.

  Hours (not counted in the sum)

  "Most physicians and surgeons work full time." Some work over 40 hours a week, and many work long shifts with irregular or overnight hours or on-call duty.

  Bureau of Labor Statistics, Occupational Outlook Handbook: Physicians and Surgeons. Work Environment: work schedules.

  The residency years are treated as paid at no cost to the doctor: the Association of American Medical Colleges reports a median first-year resident stipend of $65,100 (2024, preliminary), above the benchmark wage. The hours of residency are not counted. The $200,000 debt includes borrowed living costs, which overlap the 8 years of forgone earnings counted for the same years, so this compensating differential may be too high and the left-over too low - the opposite direction to the interest left out.

  AAMC, Medical Student Education: Debt, Costs, and Loan Repayment Fact Card for the Class of 2024 (October 2024). Last page: "2024 first post-MD year median stipend: $65,100 (preliminary data)".

• Lawyer

  Median pay, May 2025

  $159,670

  Pay premium over the benchmark

  $108,690

  Compensating differential (modeled), a floor

  at least $9,909

  Left over (modeled), a ceiling

  at most $98,781

  Bureau of Labor Statistics, Occupational Outlook Handbook: Lawyers. Pay: median annual pay $159,670, May 2025; 863,700 jobs in 2025.

  Training

  Usually 7 years of full-time study after high school: about four for a bachelor's degree and three for a law degree. No paid training stage. Unpaid years counted: 7.

  Bureau of Labor Statistics, Occupational Outlook Handbook: Lawyers. How to Become a Lawyer: "Becoming a lawyer usually takes 7 years of full-time study after high school".

  • Bachelor's degree, 4 years of tuition and required fees at a public 4-year school (4 x $9,872): $39,488. National Center for Education Statistics, Digest of Education Statistics 2024, Table 330.10. Public 4-year institutions, 2023-24, tuition and required fees $9,872 (in-district rates for public schools); the average across all students, not this profession's.

  Gap: Law-school tuition or debt. No figure was read from a primary source: the search turned up debt medians in secondary reports and one press release whose text could not be read, and none is used here.

  Counted: forgone earnings $356,860 plus direct costs $39,488, spread over 40 working years.

  What limits supply

  A state licence, which usually requires passing a bar examination. There is no cap on training places comparable to the residency Match, and no bar-passage figure was pulled, so how much the licence limits supply is a gap.

  Bureau of Labor Statistics, Occupational Outlook Handbook: Lawyers. How to Become a Lawyer: licensing.

  Hours (not counted in the sum)

  "Most lawyers work full time, and some work more than 40 hours per week."

  Bureau of Labor Statistics, Occupational Outlook Handbook: Lawyers. Work Environment: work schedules.

• Software developer

  Median pay, May 2025

  $135,980

  Pay premium over the benchmark

  $85,000

  Compensating differential (modeled)

  $6,085

  Left over (modeled)

  $78,915

  Bureau of Labor Statistics, Occupational Outlook Handbook: Software Developers, Quality Assurance Analysts, and Testers. Pay: median annual pay for software developers $135,980, May 2025 (the page's headline figure, $134,040, includes quality assurance analysts and testers).

  Training

  A bachelor's degree, typically in computer and information technology or a related field. Some employers prefer a master's degree; that is not counted. Unpaid years counted: 4.

  Bureau of Labor Statistics, Occupational Outlook Handbook: Software Developers, Quality Assurance Analysts, and Testers. How to Become a Software Developer: education.

  • Bachelor's degree, 4 years of tuition and required fees at a public 4-year school (4 x $9,872): $39,488. National Center for Education Statistics, Digest of Education Statistics 2024, Table 330.10. Public 4-year institutions, 2023-24, tuition and required fees $9,872 (in-district rates for public schools); the average across all students, not this profession's.

  Counted: forgone earnings $203,920 plus direct costs $39,488, spread over 40 working years.

  What limits supply

  None. The How to Become section lists no licence or certification requirement, so there is no licensing rent to find: the whole left-over is something else, such as skill and demand.

  Bureau of Labor Statistics, Occupational Outlook Handbook: Software Developers, Quality Assurance Analysts, and Testers. How to Become a Software Developer: the section has no licence or certification requirement.

  Hours (not counted in the sum)

  "Most software developers, quality assurance analysts, and testers work full time."

  Bureau of Labor Statistics, Occupational Outlook Handbook: Software Developers, Quality Assurance Analysts, and Testers. Work Environment: work schedules.

• Electrician

  Median pay, May 2025

  $63,190

  Pay premium over the benchmark

  $12,210

  Compensating differential (modeled), a floor

  at least $0

  Left over (modeled), a ceiling

  at most $12,210

  Bureau of Labor Statistics, Occupational Outlook Handbook: Electricians. Pay: median annual pay $63,190, May 2025; about 821,000 jobs in 2025.

  Training

  A high school diploma, then a 4- or 5-year apprenticeship, with about 2,000 paid on-the-job hours a year plus technical instruction. The apprentice is paid, so no earnings are counted as forgone; what the apprentice is paid is not pulled. Unpaid years counted: 0.

  Bureau of Labor Statistics, Occupational Outlook Handbook: Electricians. How to Become an Electrician: apprenticeships.

  Gap: The cost of the technical instruction in the apprenticeship, and the apprentice's pay. The Bureau of Labor Statistics page says the on-the-job hours are paid but gives neither figure.

  Counted: forgone earnings $0 plus direct costs $0, spread over 40 working years.

  What limits supply

  "Most states require electricians to pass a test and be licensed." Requirements vary by state. How far this limits supply is a gap: no count of places or pass rate was pulled.

  Bureau of Labor Statistics, Occupational Outlook Handbook: Electricians. How to Become an Electrician: licenses.

  Hours (not counted in the sum)

  "Almost all electricians work full time. Work schedules may include evenings and weekends." Overtime is common.

  Bureau of Labor Statistics, Occupational Outlook Handbook: Electricians. Work Environment: work schedules.

  Plumbers are the other trade the Bureau of Labor Statistics describes this way (median $63,800, May 2025, a 4- to 5-year paid apprenticeship). They are not scored separately here.

  Bureau of Labor Statistics, Occupational Outlook Handbook: Plumbers, Pipefitters, and Steamfitters. Pay: median annual pay $63,800, May 2025; How to Become a Plumber, Pipefitter, or Steamfitter: a 4- to 5-year paid apprenticeship.

• High school teacher

  Median pay, May 2025

  $72,040

  Pay premium over the benchmark

  $21,060

  Compensating differential (modeled), a floor

  at least $6,085

  Left over (modeled), a ceiling

  at most $14,975

  Bureau of Labor Statistics, Occupational Outlook Handbook: High School Teachers. Pay: median annual wage $72,040, May 2025; 1,087,500 jobs in 2025.

  Training

  At least a bachelor's degree, with a supervised student-teaching placement as part of the preparation program. Unpaid years counted: 4.

  Bureau of Labor Statistics, Occupational Outlook Handbook: High School Teachers. How to Become a High School Teacher: education and student teaching.

  • Bachelor's degree, 4 years of tuition and required fees at a public 4-year school (4 x $9,872): $39,488. National Center for Education Statistics, Digest of Education Statistics 2024, Table 330.10. Public 4-year institutions, 2023-24, tuition and required fees $9,872 (in-district rates for public schools); the average across all students, not this profession's.

  Gap: Licence exam fees, and the cost of a master's degree in the states that require one after hiring.

  Counted: forgone earnings $203,920 plus direct costs $39,488, spread over 40 working years.

  What limits supply

  "All states require teachers in public schools to be licensed or certified in the specific grade level they will teach." Private schools usually do not require it. How far this limits supply is a gap: no count was pulled.

  Bureau of Labor Statistics, Occupational Outlook Handbook: High School Teachers. How to Become a High School Teacher: licenses.

  Hours (not counted in the sum)

  Teachers generally work during school hours, often also grade and prepare in the evenings and on weekends, and most do not teach during the summer.

  Bureau of Labor Statistics, Occupational Outlook Handbook: High School Teachers. Work Environment: work schedules.

  Hours are not treated as long here (long_hours is false), so they do not make this sum a floor: the Bureau of Labor Statistics describes work during school hours, and while it mentions grading and preparing in the evenings and on weekends, it gives no over-40-hour figure.

• Farmer

  Median pay, May 2025

  $89,900

  Pay premium over the benchmark

  $38,920

  Compensating differential (modeled)

  not modeled

  Left over (modeled)

  not modeled

  Bureau of Labor Statistics, Occupational Outlook Handbook: Farmers, Ranchers, and Other Agricultural Managers. Pay: median annual pay $89,900, May 2025; 788,700 jobs in 2025. The occupation title covers other agricultural managers as well as farmers, so it is not an owner-operator's income.

  Training

  A high school diploma or equivalent. No licence is required. Unpaid years counted: 0.

  Bureau of Labor Statistics, Occupational Outlook Handbook: Farmers, Ranchers, and Other Agricultural Managers. How to Become a Farmer, Rancher, or Other Agricultural Manager: education; licensing.

  Gap: The real cost of becoming a farmer is the land and equipment, and the risk of owning them. No source for that was pulled, and the Bureau of Labor Statistics occupation covers other agricultural managers as well as farmers, so no defensible source was found for an owner-operator's income either. The section shows the median and stops: it does not model a compensating differential or a left-over for farmers.

  What limits supply

  None. There is no licence, so there is no licensing rent: a premium here would come from something else.

  Bureau of Labor Statistics, Occupational Outlook Handbook: Farmers, Ranchers, and Other Agricultural Managers. How to Become a Farmer, Rancher, or Other Agricultural Manager: licenses.

  Hours (not counted in the sum)

  "Most farmers, ranchers, and other agricultural managers work full time, and many work more than 40 hours per week."

  Bureau of Labor Statistics, Occupational Outlook Handbook: Farmers, Ranchers, and Other Agricultural Managers. Work Environment: work schedules.

Boom premiums: what a booming industry pays above the economy, and for how long

A job in a booming industry can pay more than the same job elsewhere. This section measures that premium year by year for two booms, the information and tech sector and oil and gas extraction, asks how fast it fades, and adds up what an entrant would have collected from it. The question behind the careers section above is the same one run through time: how much of a pay gap is a lasting wall and how much is a spike that competition takes back.

How the lines are built

• Premium(t) = the mean annual pay for an occupation inside the boom industry in year t, minus the mean annual pay for the same occupation across all industries, as a percent of the all-industry pay. Pay is the Bureau of Labor Statistics' Occupational Employment and Wage Statistics (OEWS) national mean annual wage for May of the year. The occupations are software developers, accountants and auditors, and general and operations managers: jobs that exist in many industries, so the same job can be compared.
• Two kinds of line, kept apart. The same-job lines (OEWS) hold the occupation fixed. The industry-average lines (Quarterly Census of Employment and Wages, QCEW, private sector, whole-year average pay) compare everyone in an industry with everyone in the private sector, so they also carry the mix of jobs the industry has, and QCEW wages include bonuses and stock options. They are larger, and they are the gross industry gap, not a same-job premium. They are used because the OEWS industry files this page pulled start in May 2003 and so cannot see the dot-com spike.
• Every premium here is gross. Part of any premium can be a compensating differential, pay for a cost or a risk of working there: a boomtown's cost of living, the chance the boom ends in layoffs. This section pulled no cost-of-living or layoff-risk figure, so it computes no net premium. Netting those out could only lower the numbers shown.
• OEWS smooths. Each estimate combines six half-year survey panels collected over three years, and older wages are aged forward (by the Employment Cost Index through May 2020, by a regression model from May 2021). A spike therefore shows up later and flatter than it happened. QCEW is a count of all covered jobs for the year, not a sample.
• Baseline = the mean premium over the five years before the boom's start year. Peak = the highest premium inside the peak window. Half-life = the years from the peak to the first later year when the premium is at or under half-way between the baseline and the peak. A line whose peak is under 5 percentage points above its baseline gets no half-life, because it has not risen enough to decay. The start year and the peak window are this page's choices, written in the data file beside each line, and moving them moves the readouts.
• Entrant math. An entrant takes a job in the boom industry in the boom's start year plus a break-in lag (0, 3 or 10 years, the time to get hired into the premium). Total extra earnings = the sum, over every later year with data up to the last, of that year's industry pay minus the all-industry pay for the same job, in nominal dollars: no inflation adjustment, no discounting. It counts the average worker in the cell, not a particular entrant, and is gross. A year with no cell is left out of the sum and listed as a year left out, never estimated; the sum says nothing about which way the missing years would have moved it. Each year's dollars compare the same job inside and outside the industry in that year, but a sum whose years span a change in the job's definition adds years of one job to years of a wider one, so the sum names the change.
• A point is the source's own figure, and a cell with a standard error over 5% is flagged as imprecise. The mean wage is used, not the median, because the mean is published for every cell and the median is hidden above a pay ceiling in some. The OEWS files were read from the Internet Archive's copies of the Bureau of Labor Statistics files, because the Bureau's site refuses scripted downloads; the links cite the Bureau's own addresses.

What it rests on

• Krueger and Summers (1988), "Efficiency Wages and the Inter-Industry Wage Structure", Econometrica 56(2): 259-293. They compared wages for equally skilled workers across industries and found that the gaps stay large after controls for ability and working conditions, which they read as noncompetitive rents in high-wage industries. That is why this section treats an industry's pay gap as something that can outlast a boom, and then checks whether it does.
• Kline, Petkova, Williams and Zidar (2019), "Who Profits from Patents? Rent-Sharing at Innovative Firms", Quarterly Journal of Economics 134(3): 1343-1404. Using patent grants as a shock to a firm's profit, they found workers capture roughly 30 cents of each dollar of extra surplus as higher earnings, about twice that for workers already at the firm. That is the mechanism by which a boom in profit can become a premium in pay, and why the premium goes to those already inside before it reaches an entrant.

Left out, and why

• The dot-com spike in the same-job lines. The OEWS industry files this page pulled begin in May 2003; the earlier survey used a different industry system and was not pulled. The 1998-2000 spike appears only in the industry-average lines.
• Software developers in oil and gas extraction. The cell holds 30 to 630 workers in the years it exists and is missing in 2005, 2006 and 2011, so it cannot carry a line. Accountants and general and operations managers, whose cells hold thousands, stand in.
• Accountants and general and operations managers in computer systems design. The cells exist and are not drawn, to keep the page readable.
• Railroads. They are a mention and not a series: no occupation-by-industry wage source reaching back to the 19th-century railroad booms was pulled.
• A net-of-compensating-differential premium, as set out in the method.
• Standard errors for May 2025, which come from the Bureau's data interface and were not pulled; the 2025 point of each same-job line says 'SE not pulled'.

Railroads are named only as an earlier boom that this question also fits. No occupation-by-industry wage source reaching back that far was pulled, so there is no railroad series and no railroad number.

Netflix US price changes: each hike as a dated step

This section lists each US plan price change the cited reports record, from October 2017 to March 2026, as a dated step, with the plan names and the old and new monthly price the cited report gives, and draws the same steps as a line per plan. It is descriptive: it records what the price was and when it moved, nothing more.

• Illustrative only. The prices are the ones the linked reports state; the section draws no conclusion from them.
• It makes no claim that any price change caused anything, and it sets no subscriber or revenue figure beside a price, so the steps are not read against either.
• Each step is dated by the report that carries it, which is when the change was announced or noticed. New and existing customers were often moved on different days; the step's note says so where the report does.
• A plan a report does not price is left as a gap on the line, never filled in.

Monthly price by plan

• Premium
• Standard
• Standard with ads
• Basic

Gap, Standard with ads: The reports for the earlier steps do not price this plan, so its line starts at the first report here that states its price (October 2023). No earlier price is estimated.

Gap, Basic: The January 2025 and March 2026 reports do not state a Basic price, so its line stops at October 2023. The October 2023 report says Basic was no longer offered to new US subscribers. No later price is estimated.

Every step, with its source

1. Oct 5, 2017

   • Basic: $7.99, unchanged
   • Standard: $9.99 to $10.99 a month (+$1.00)
   • Premium: $11.99 to $13.99 a month (+$2.00)

   The report says the new prices applied to new customers right away and to existing customers from November.

   Sources for Oct 5, 2017: 9to5Google, "Netflix price jump", Oct 5, 2017.

2. Jan 15, 2019

   • Basic: $7.99 to $8.99 a month (+$1.00)
   • Standard: $10.99 to $12.99 a month (+$2.00)
   • Premium: $13.99 to $15.99 a month (+$2.00)

   Sources for Jan 15, 2019: Light Reading, "Netflix Raises Prices as Content Spending, OTT Competition Heat Up" (prices to the cent; names the plans by stream count), Jan 16, 2019; Fortune, "Netflix price increase 2019" (whole-dollar prices: $8 to $9, $11 to $13, $14 to $16), Jan 15, 2019.

3. Oct 29, 2020

   • Basic: $8.99, unchanged
   • Standard: $12.99 to $13.99 a month (+$1.00)
   • Premium: $15.99 to $17.99 a month (+$2.00)

   The report says new customers paid the new prices at once and existing customers within about two months.

   Sources for Oct 29, 2020: TechCrunch, "Netflix is raising prices on most of its plans in the US", Oct 29, 2020.

4. Jan 14, 2022

   • Basic: $8.99 to $9.99 a month (+$1.00)
   • Standard: $13.99 to $15.49 a month (+$1.50)
   • Premium: $17.99 to $19.99 a month (+$2.00)

   Sources for Jan 14, 2022: TechCrunch, "Netflix raised its prices again", Jan 14, 2022.

5. Oct 18, 2023

   • Basic: $9.99 to $11.99 a month (+$2.00)
   • Standard: $15.49, unchanged
   • Standard with ads: $6.99, unchanged
   • Premium: $19.99 to $22.99 a month (+$3.00)

   The report says Basic was no longer offered to new US subscribers.

   Sources for Oct 18, 2023: MacRumors, "Netflix Raising U.S. Prices Yet Again, Premium Plan Now Costs $23/Month", Oct 18, 2023.

6. Jan 21, 2025

   • Standard with ads: $6.99 to $7.99 a month (+$1.00)
   • Standard: $15.49 to $17.99 a month (+$2.50)
   • Premium: $22.99 to $24.99 a month (+$2.00)

   Sources for Jan 21, 2025: 9to5Mac, "Netflix price increase, all three plans", Jan 21, 2025.

7. Mar 26, 2026

   • Standard with ads: $7.99 to $8.99 a month (+$1.00)
   • Standard: $17.99 to $19.99 a month (+$2.00)
   • Premium: $24.99 to $26.99 a month (+$2.00)

   FOX 2 Detroit says new customers paid the new prices at once and existing customers on their next bill.

   Sources for Mar 26, 2026: TheWrap, "Netflix price increase 2026" (states the rises of $1, $2 and $2), Mar 26, 2026; FOX 2 Detroit, "Netflix raising prices again on all plans" (states old and new prices), Mar 27, 2026.

Pricing power over time

Each panel is one company, drawn only in the years its filings disclose a unit. Every panel shares the same years and the same scales. A higher L, and an e nearer 1, is more pricing power.

• Netflix

  Fits

  Fiscal 2015 to 2024. The whole company: streaming memberships (revenue through 2023 also holds the small DVD business).

  Lerner index, L

  Implied elasticity, e (log scale)

  From fiscal 2025 Netflix stopped reporting a membership count, so the line ends at 2024.

  Form 10-K, fiscal 2025

  The 10-K reports revenue and cost of revenues but no paid membership count

• Tesla

  Fits

  Fiscal 2017 to 2025. Automotive sales only: energy storage, services, leasing and regulatory credits are left out.

  Lerner index, L

  Implied elasticity, e (log scale)

• Rivian

  Fits

  Fiscal 2022 to 2025. Automotive only: software and services are left out.

  Lerner index, L

  Implied elasticity, e (log scale)

  No markup read in fiscal 2022, 2023, 2024, 2025: the cost scored, listed under the sources below, was at or above revenue, so the model does not apply. Shown as rings on the L chart's zero line, and left blank on the e chart, where the bottom line is the most pricing power.

• General Motors

  Fits

  Fiscal 2001 to 2025. Automotive only, the lending arm left out: automotive net sales and revenue against automotive cost of sales.

  Lerner index, L

  Implied elasticity, e (log scale)

  General Motors Corporation went through Chapter 11 in June 2009 and its assets were sold to a new company, General Motors Company, on July 10, 2009, so figures before and after are two different companies' books and 2009 is split between them.

  Form 10-K, fiscal 2009

  Item 1, Chapter 11 Proceedings and the 363 Sale; Consolidated Statements of Operations show Predecessor Jan 1-Jul 9, 2009 and Successor Jul 10-Dec 31, 2009

• Harley-Davidson

  Partial fit

  Fiscal 2000 to 2025. Motorcycles and Related Products only, the lending arm left out: its revenue against its cost of goods sold. Revenue includes parts, accessories and apparel.

  Lerner index, L

  Implied elasticity, e (log scale)

  From 2022 LiveWire electric motorcycles became their own segment, so units, revenue and cost are the HDMC segment only (LiveWire shipments, a few hundred a year, are no longer in the count).

  Form 10-K, fiscal 2022

  Note 'Segment Information': reportable segments HDMC, LiveWire and HDFS; Item 7 HDMC segment results

• Disney

  Partial fit

  Fiscal 2022 to 2025. Entertainment Direct-to-Consumer (Disney+ and Hulu) only.

  Lerner index, L

  Implied elasticity, e (log scale)

  No markup read in fiscal 2022, 2023: the cost scored, listed under the sources below, was at or above revenue, so the model does not apply. Shown as rings on the L chart's zero line, and left blank on the e chart, where the bottom line is the most pricing power.

• Southwest Airlines

  Partial fit

  Fiscal 1995 to 2025. The whole company: total operating revenues against total operating expenses, every cost of running the airline.

  Lerner index, L

  Implied elasticity, e (log scale)

  No markup read in fiscal 2020: the cost scored, listed under the sources below, was at or above revenue, so the model does not apply. Shown as rings on the L chart's zero line, and left blank on the e chart, where the bottom line is the most pricing power.

• UPS

  Partial fit

  Fiscal 2014 to 2025. U.S. Domestic Package only: segment revenue against segment revenue minus segment operating profit.

  Lerner index, L

  Implied elasticity, e (log scale)

  From fiscal 2024 UPS moved its U.S. air cargo business into this segment, so revenue and cost from 2024 include air cargo that carries no package count; earlier years were not recast, so the line is not bridged.

  Form 10-K, fiscal 2024

  Item 7, Overview: U.S. Air Cargo presented within the U.S. Domestic Package segment as of the fourth quarter of 2024

• AT&T

  Partial fit

  Fiscal 2007 to 2023. Domestic wireless only: its operating revenues against those revenues minus its operating income, as printed under three names over the years - the Wireless segment (2007-2014), the AT&T Mobility supplemental table (2015-2017) and the Mobility business unit of the Communications segment (2018-2023).

  Lerner index, L

  Implied elasticity, e (log scale)

  From 2024 AT&T stopped counting connected devices (tablets, monitors, car systems) as Mobility subscribers, so the count fell by more than half and the line ends at 2023.

  Form 10-K, fiscal 2024

  Item 7, Mobility Subscribers table, footnote 1 to Total Mobility Subscribers

• Verizon

  Partial fit

  Fiscal 2005 to 2018. The Wireless segment only: its operating revenues against those revenues minus its operating income.

  Lerner index, L

  Implied elasticity, e (log scale)

  In 2019 Verizon replaced its Wireless segment with Consumer and Business segments that mix wireless with Fios and other wireline services, so the wireless line ends at 2018.

  Form 10-K, fiscal 2019, Exhibit 13

  Exhibit 13, MD&A Overview: strategic reorganization into Verizon Consumer Group and Verizon Business Group, effective April 1, 2019

• T-Mobile

  Partial fit

  Fiscal 2013 to 2023. The whole company: total revenues against total revenues minus operating income, as printed in each fiscal year's own Form 10-K. From April 2020 the figures include Sprint.

  Lerner index, L

  Implied elasticity, e (log scale)

  In 2020 T-Mobile merged with Sprint (April 1) and began counting only postpaid and prepaid customers, leaving wholesale customers out. Its own 2020 report restates 2019 at 67.9 million customers against the 86.0 million first printed, so the line does not join 2019 to 2020.

  Form 10-K, fiscal 2020

  Item 6, Selected Financial Data: Total customers (in thousands), 2019 column as restated

• Boeing

  Partial fit

  Fiscal 2003 to 2025. Commercial Airplanes only: segment revenues against segment revenues minus segment earnings from operations.

  Lerner index, L

  Implied elasticity, e (log scale)

  No markup read in fiscal 2009, 2019, 2020, 2021, 2022, 2023, 2024, 2025: the cost scored, listed under the sources below, was at or above revenue, so the model does not apply. Shown as rings on the L chart's zero line, and left blank on the e chart, where the bottom line is the most pricing power.

  In 2017 Boeing moved its commercial aviation services business (spare parts, modifications, support) out of Commercial Airplanes into a new Global Services segment, so segment revenue and cost from 2017 cover airplanes with far less aftermarket business.

  Form 10-K, fiscal 2017

  Note 21, Segment Information: effective July 1, 2017, BGS brings together the Commercial Aviation Services businesses previously included in the BCA segment

• Marriott

  Partial fit

  Fiscal 2018 to 2025. Owned and leased hotels only: 'Owned, leased, and other revenue' against the owned, leased, and other direct expense line (hotel operating costs, excluding depreciation); total revenue includes cost reimbursements. Starts 2018, the first year on the current revenue standard after the 2016 Starwood merger.

  Lerner index, L

  Implied elasticity, e (log scale)

  No markup read in fiscal 2020: the cost scored, listed under the sources below, was at or above revenue, so the model does not apply. Shown as rings on the L chart's zero line, and left blank on the e chart, where the bottom line is the most pricing power.

  From 2025 Marriott moved certain fee-related costs (guarantee expense, credit losses and similar, $129 million in 2024) out of general and administrative into the owned, leased, and other expense line, so 2025 cost is broader than earlier years.

  Form 10-K, fiscal 2025

  Note 1, Basis of Presentation: reclassification from 'General, administrative, and other' to 'Owned, leased, and other expense'

• Hilton

  Partial fit

  Fiscal 2013 to 2025. Ownership segment's consolidated owned and leased hotels only: 'Owned and leased hotels' revenue (called 'Ownership' from 2025) against the matching expense line (hotel operating costs including rent, excluding depreciation).

  Lerner index, L

  Implied elasticity, e (log scale)

  No markup read in fiscal 2020, 2021: the cost scored, listed under the sources below, was at or above revenue, so the model does not apply. Shown as rings on the L chart's zero line, and left blank on the e chart, where the bottom line is the most pricing power.

  On January 3, 2017 Hilton spun off most of its owned hotels into Park Hotels & Resorts (and its timeshare business into Hilton Grand Vacations), so from 2017 the owned and leased portfolio is about a third of its earlier size.

  Form 10-K, fiscal 2017

  Item 1, Business: spin-offs completed January 3, 2017; Consolidated Statements of Operations restate 2016 owned and leased hotels revenue to $1,452 million

• Apple

  Does not fit

  Fiscal 2009 to 2018. Through fiscal 2018 the 10-K counted iPhones sold, so the line reads the whole company's margin in the years a unit was disclosed. Revenue and cost are company-wide, so this is the loosest line here.

  Lerner index, L

  Implied elasticity, e (log scale)

  From fiscal 2019 Apple stopped reporting unit sales, so the line ends at 2018.

  Form 10-K, fiscal 2019

  Item 7: iPhone net sales discussed with no unit count

• FedEx

  Does not fit

  Fiscal 2001 to 2002. FedEx Ground only, the two years its packages can be counted: average daily packages times operating weekdays, both printed. As at UPS, the segment prints no cost of revenues, so the cost is every segment operating expense. Earlier years are left out: from fiscal 2001 a new FedEx Services unit's sales, marketing and IT costs are charged to the segment, and the filing calls some of the segment's earlier expense figures not comparable.

  Lerner index, L

  Implied elasticity, e (log scale)

  From fiscal 2003 the 10-K stops printing operating weekdays, so FedEx Ground's yearly package count cannot be multiplied out and the line ends at 2002.

  Form 10-K, fiscal 2003

  Item 6, Selected Financial Data, Other Operating Data: aircraft fleet and employees only, no operating-weekday rows; MD&A FedEx Ground table prints average daily package volume only

Where the model fits, and where it does not

• Fits. The filing discloses a unit count, the revenue for those units and a cost of revenues for them, so P, Q and MC all come from the filing.
• Partial fit. One segment has units but the company has several, and the filing does not give all three figures for those units. Scored on that segment and labelled.
• Does not fit. No unit is disclosed; or the company has several segments and none has a unit, where software's near-zero marginal cost makes a gross-margin Lerner index meaningless; or it sells a commodity at the world price, so its margin is the price of oil, not pricing power.

• Netflix

  Fits

  One product sold by the membership. The 10-K gives the count of paid memberships, the revenue they bring in and the cost of revenues behind them.

  Fiscal 2024. The whole company: streaming memberships (revenue through 2023 also holds the small DVD business). Scored revenue $39.0B, 100% of the company's.

  Units, Q

  301.6 million paid memberships

  Price per unit, P

  $129

  MC per unit

  $70

  Lerner index, L

  0.46

  Elasticity, e

  2.2

  Producer surplus

  $18.0B

  Consumer surplus, a band over two demand shapes

  $9.0B to $33.3B

  $9.0B if demand is a straight line, $33.3B if its elasticity is constant. The gap is how much the answer rests on a shape no filing shows.

• Tesla

  Fits

  Cars are counted. The 10-K gives vehicles delivered, the revenue from selling them and the cost of those sales.

  Fiscal 2025. Automotive sales only: energy storage, services, leasing and regulatory credits are left out. Scored revenue $65.8B, 69% of the company's.

  Units, Q

  1.6 million vehicles delivered

  Price per unit, P

  $40,135

  MC per unit

  $34,309

  Lerner index, L

  0.15

  Elasticity, e

  6.9

  Producer surplus

  $9.6B

  Consumer surplus, a band over two demand shapes

  $4.8B to $11.2B

  $4.8B if demand is a straight line, $11.2B if its elasticity is constant. The gap is how much the answer rests on a shape no filing shows.

• Rivian

  Fits

  Cars are counted, as at Tesla. But it has sold every car for less than its cost of revenues, so there is no markup for the model to read.

  Fiscal 2025. Automotive only: software and services are left out. Scored revenue $3.8B, 71% of the company's.

  Units, Q

  42,247 vehicles delivered

  Price per unit, P

  $90,657

  MC per unit

  $100,883

  No markup read - model does not apply.

• General Motors

  Fits

  Cars are counted, as at Tesla: GM prints the vehicles it sells wholesale to its dealers, and the automotive revenue and cost of sales for them. Old GM (to 2008) and new GM (from mid-2009) are two companies' books, kept apart; 2005 to 2010 are blank: old GM stopped printing a wholesale count after 2004, 2009 is split between the two companies, and new GM's first worldwide count is for 2011. From 2018 the cost line also holds Cruise, the self-driving unit.

  Fiscal 2025. Automotive only, the lending arm left out: automotive net sales and revenue against automotive cost of sales. Scored revenue $168.0B, 91% of the company's.

  Units, Q

  3.8 million wholesale vehicle sales

  Price per unit, P

  $44,215

  MC per unit

  $41,887

  Lerner index, L

  0.05

  Elasticity, e

  19.0

  Producer surplus

  $8.8B

  Consumer surplus, a band over two demand shapes

  $4.4B to $9.3B

  $4.4B if demand is a straight line, $9.3B if its elasticity is constant. The gap is how much the answer rests on a shape no filing shows.

• Harley-Davidson

  Partial fit

  Motorcycles are counted: the 10-K prints motorcycles shipped to dealers. But it prints a cost of goods sold only for the whole motorcycle business, which also sells parts and accessories and apparel - roughly a fifth to over a quarter of that business's revenue, with no unit count - so the revenue and cost scored cover more than the bikes counted, and price per motorcycle runs above what a bike sells for. Buell motorcycles, sold through 2011 inside the same revenue and cost, are counted with them (2 to 5% of units).

  Fiscal 2025. Motorcycles and Related Products only, the lending arm left out: its revenue against its cost of goods sold. Revenue includes parts, accessories and apparel. Scored revenue $3.6B, 80% of the company's.

  Units, Q

  124,477 motorcycles shipped

  Price per unit, P

  $28,747

  MC per unit

  $21,785

  Lerner index, L

  0.24

  Elasticity, e

  4.1

  Producer surplus

  $0.9B

  Consumer surplus, a band over two demand shapes

  $0.4B to $1.1B

  $0.4B if demand is a straight line, $1.1B if its elasticity is constant. The gap is how much the answer rests on a shape no filing shows.

• Disney

  Partial fit

  Only the streaming business has units. Parks, TV networks and film have none, and even for streaming the 10-K gives operating income rather than a cost of revenues, so the cost here is every streaming cost.

  Fiscal 2025. Entertainment Direct-to-Consumer (Disney+ and Hulu) only. Scored revenue $24.6B, 26% of the company's.

  Units, Q

  195.7 million paid streaming subscriptions

  Price per unit, P

  $126

  MC per unit

  $119

  Lerner index, L

  0.05

  Elasticity, e

  18.5

  Producer surplus

  $1.3B

  Consumer surplus, a band over two demand shapes

  $0.7B to $1.4B

  $0.7B if demand is a straight line, $1.4B if its elasticity is constant. The gap is how much the answer rests on a shape no filing shows.

• Southwest Airlines

  Partial fit

  Passengers are counted: the 10-K prints revenue passengers carried every year since 1995. But airlines print no cost-of-revenues line, so the cost here is every operating expense, which overstates marginal cost more than a cost of revenues would. 2011 on includes AirTran, bought that May; 2020 and 2021 costs are net of federal payroll support.

  Fiscal 2025. The whole company: total operating revenues against total operating expenses, every cost of running the airline. Scored revenue $28.1B, 100% of the company's.

  Units, Q

  134.1 million revenue passengers carried

  Price per unit, P

  $209

  MC per unit

  $206

  Lerner index, L

  0.02

  Elasticity, e

  65.6

  Producer surplus

  $0.4B

  Consumer surplus, a band over two demand shapes

  $0.2B to $0.4B

  $0.2B if demand is a straight line, $0.4B if its elasticity is constant. The gap is how much the answer rests on a shape no filing shows.

• UPS

  Partial fit

  Packages are counted, but only by the day: the 10-K prints average daily packages and operating days, so the year's packages are the two multiplied. Its segments print no cost of revenues, so the cost is every segment operating cost. The line starts in 2014 because UPS twice changed how segment profit treats pensions; 2014 to 2017 money is as restated in the fiscal 2018 10-K, so the whole line is on one definition.

  Fiscal 2025. U.S. Domestic Package only: segment revenue against segment revenue minus segment operating profit. Scored revenue $59.5B, 67% of the company's.

  Units, Q

  4,412.5 million packages delivered

  Price per unit, P

  $13

  MC per unit

  $13

  Lerner index, L

  0.07

  Elasticity, e

  15.2

  Producer surplus

  $3.9B

  Consumer surplus, a band over two demand shapes

  $2.0B to $4.2B

  $2.0B if demand is a straight line, $4.2B if its elasticity is constant. The gap is how much the answer rests on a shape no filing shows.

• AT&T

  Partial fit

  Only the wireless business has a unit, subscribers: the direct analog of Netflix's paid memberships. AT&T prints no cost of revenues for it every year, so the cost is every wireless operating cost, selling and depreciation included. The count includes connected devices such as tablets and car systems, whose fast growth pulls revenue per subscriber down over the years.

  Fiscal 2023. Domestic wireless only: its operating revenues against those revenues minus its operating income, as printed under three names over the years - the Wireless segment (2007-2014), the AT&T Mobility supplemental table (2015-2017) and the Mobility business unit of the Communications segment (2018-2023). Scored revenue $84.0B, 69% of the company's.

  Units, Q

  241.5 million wireless subscribers

  Price per unit, P

  $348

  MC per unit

  $241

  Lerner index, L

  0.31

  Elasticity, e

  3.2

  Producer surplus

  $25.9B

  Consumer surplus, a band over two demand shapes

  $12.9B to $37.4B

  $12.9B if demand is a straight line, $37.4B if its elasticity is constant. The gap is how much the answer rests on a shape no filing shows.

• Verizon

  Partial fit

  Only the wireless business has a unit, retail connections: the direct analog of Netflix's paid memberships. The cost is every wireless operating cost, selling and depreciation included, since no cost of revenues is printed for it. The count leaves out wholesale lines that the revenue includes, and the segment is all of Verizon Wireless, 45% owned by Vodafone until 2014.

  Fiscal 2018. The Wireless segment only: its operating revenues against those revenues minus its operating income. Scored revenue $91.7B, 70% of the company's.

  Units, Q

  118.0 million retail wireless connections

  Price per unit, P

  $777

  MC per unit

  $499

  Lerner index, L

  0.36

  Elasticity, e

  2.8

  Producer surplus

  $32.8B

  Consumer surplus, a band over two demand shapes

  $16.4B to $51.1B

  $16.4B if demand is a straight line, $51.1B if its elasticity is constant. The gap is how much the answer rests on a shape no filing shows.

• T-Mobile

  Partial fit

  T-Mobile is wireless service end to end, so the whole company is scored and the unit is customers, the direct analog of Netflix's paid memberships. The cost is every operating cost, selling, depreciation and the cost of the phones it sells included, since no cost of revenues is printed for service alone. The revenue includes equipment sales, which pulls revenue per customer up. Until 2019 the count includes wholesale customers (machine-to-machine and resold lines); from 2020 it does not.

  Fiscal 2023. The whole company: total revenues against total revenues minus operating income, as printed in each fiscal year's own Form 10-K. From April 2020 the figures include Sprint. Scored revenue $78.6B, 100% of the company's.

  Units, Q

  119.7 million customers

  Price per unit, P

  $656

  MC per unit

  $537

  Lerner index, L

  0.18

  Elasticity, e

  5.5

  Producer surplus

  $14.3B

  Consumer surplus, a band over two demand shapes

  $7.1B to $17.4B

  $7.1B if demand is a straight line, $17.4B if its elasticity is constant. The gap is how much the answer rests on a shape no filing shows.

• Boeing

  Partial fit

  Airplanes are counted: the 10-K prints commercial deliveries every year. It prints no cost of sales for the segment, so the cost is every segment cost, research and development included, measured by program accounting, which spreads a jet's expected cost over its whole production run. The line starts in 2003, the first year the 10-K measures segment earnings that way; figures are as each year's own 10-K filed them.

  Fiscal 2025. Commercial Airplanes only: segment revenues against segment revenues minus segment earnings from operations. Scored revenue $41.5B, 46% of the company's.

  Units, Q

  600 commercial airplanes delivered

  Price per unit, P

  $69,156,667

  MC per unit

  $80,955,000

  No markup read - model does not apply.

• Marriott

  Partial fit

  A looser fit. Most Marriott hotels are franchised or managed and earn it only fees, so only its few owned and leased hotels are scored; the unit is a room (capacity at year-end, not a night sold), and the revenue line also carries non-hotel items such as termination fees and design fees, so price per room is only roughly a price.

  Fiscal 2025. Owned and leased hotels only: 'Owned, leased, and other revenue' against the owned, leased, and other direct expense line (hotel operating costs, excluding depreciation); total revenue includes cost reimbursements. Starts 2018, the first year on the current revenue standard after the 2016 Starwood merger. Scored revenue $1.7B, 6% of the company's.

  Units, Q

  14,406 owned and leased hotel rooms

  Price per unit, P

  $116,549

  MC per unit

  $101,416

  Lerner index, L

  0.13

  Elasticity, e

  7.7

  Producer surplus

  $0.2B

  Consumer surplus, a band over two demand shapes

  $0.1B to $0.3B

  $0.1B if demand is a straight line, $0.3B if its elasticity is constant. The gap is how much the answer rests on a shape no filing shows.

• Hilton

  Partial fit

  A looser fit. Most Hilton hotels are managed or franchised and pay Hilton only fees, so only the hotels Hilton owns or leases are scored; the unit is a room (capacity at year-end, not a night sold), and revenue also includes food, drink and other hotel sales, so price per room is a yearly revenue per room rather than a price.

  Fiscal 2025. Ownership segment's consolidated owned and leased hotels only: 'Owned and leased hotels' revenue (called 'Ownership' from 2025) against the matching expense line (hotel operating costs including rent, excluding depreciation). Scored revenue $1.2B, 10% of the company's.

  Units, Q

  13,657 owned and leased hotel rooms

  Price per unit, P

  $90,283

  MC per unit

  $80,105

  Lerner index, L

  0.11

  Elasticity, e

  8.9

  Producer surplus

  $0.1B

  Consumer surplus, a band over two demand shapes

  $0.1B to $0.2B

  $0.1B if demand is a straight line, $0.2B if its elasticity is constant. The gap is how much the answer rests on a shape no filing shows.

• Apple

  Does not fit

  No unit disclosed. Apple stopped reporting how many iPhones, iPads and Macs it sold after fiscal 2018, and it never split its cost of sales by product.

  No current reading: the latest filing gives the model nothing to read. The years it could read are charted above.

  Form 10-K, fiscal 2026

  Item 7, Products and Services Performance: net sales by category, with no unit counts

• Walmart

  Does not fit

  No unit. A store selling a huge mix of goods has no single thing with a price; the 10-K gives net sales by segment and says what drove comparable sales, never a count of units sold.

  No numbers: the model has nothing to read.

  Form 10-K, fiscal 2026

  Item 7: net sales for Walmart U.S., Walmart International and Sam's Club U.S.; comparable sales described by transactions and average ticket

• Amazon

  Does not fit

  Multi-segment with no unit: North America, International and AWS, none with a unit count. AWS is software-like, with marginal cost near zero, so a gross-margin Lerner index would mistake fixed cost for pricing power.

  No numbers: the model has nothing to read.

  Form 10-K, fiscal 2025

  Item 7, Results of Operations: three segments, North America, International and AWS

• Microsoft

  Does not fit

  Multi-segment with no unit: three segments, none with a unit count. Software costs almost nothing per extra copy, so a gross-margin Lerner index would read near 1 and mean nothing.

  No numbers: the model has nothing to read.

  Form 10-K, fiscal 2026

  Item 7, Operating segments: Productivity and Business Processes, Intelligent Cloud, More Personal Computing

• Alphabet

  Does not fit

  Multi-segment with no unit: Google Services, Google Cloud and Other Bets, none with a unit count. Ads and software cost almost nothing per extra sale, so a gross-margin Lerner index would mean nothing.

  No numbers: the model has nothing to read.

  Form 10-K, fiscal 2025

  Item 7: two Google segments and Other Bets

• ExxonMobil

  Does not fit

  Commodity price-taker. Oil and gas sell at a world market price, so its margin reflects the commodity price, not pricing power the model could read.

  No numbers: the model has nothing to read.

  Form 10-K, fiscal 2025

  Item 1A, Risk Factors: prices and margins depend on global supply and demand for the commodity

• Delta Air Lines

  Does not fit

  No passenger count is disclosed. Delta's 10-K measures traffic only in revenue passenger miles and seat miles, and its total operating revenue also includes its oil refinery's sales to outside buyers, so there is no unit that matches the revenue.

  No numbers: the model has nothing to read.

  Form 10-K, fiscal 2025

  Item 7, MD&A, Operating Statistics: revenue passenger miles and seat miles, no passenger count; Other Revenue: Refinery $5,077 million inside total operating revenue

• FedEx

  Does not fit

  FedEx prints packages only as an average per day. Its fiscal 2001 and 2002 10-Ks also print FedEx Express's and FedEx Ground's operating weekdays (fiscal 1997 to 2002), so FedEx Ground's packages for those years can be multiplied out as UPS's are; no 10-K since prints the number of operating days, only that a year had more or fewer than the last, and FedEx Ground's average now mixes a 5-day and a 7-day basis, so no later year has a yearly unit count.

  No current reading: the latest filing gives the model nothing to read. The years it could read are charted above.

  Form 10-K, fiscal 2024

  Item 7, FedEx Ground Segment table: Average daily package volume (ADV), footnote (1) - ground commercial on a 5-day-per-week basis, home delivery and economy on a 7-day-per-week basis; no operating-day count printed

• Coca-Cola

  Does not fit

  The unit counted is not what Coca-Cola's revenue is for. Unit case volume counts drinks sold by the whole Coca-Cola system, mostly independent bottlers, while the company's revenue is largely concentrate sold to those bottlers; the 10-K itself says a significant portion of net operating revenues is not based directly on unit case volume.

  No numbers: the model has nothing to read.

  Form 10-K, fiscal 2025

  Item 7, Beverage Volume: definition of unit case volume (sold by the Company and its bottling partners) and 'a significant portion of our Company's net operating revenues is not based directly on unit case volume'

• Starbucks

  Does not fit

  No count of units sold is disclosed. Transactions (the closest thing to a unit sold) appear only as a percentage change in comparable-store transactions, and the store count is not a unit sold, so price per unit cannot be read off the filing.

  No numbers: the model has nothing to read.

  Form 10-K, fiscal 2025

  Item 7, Overview: key metrics are store count and comparable store sales; 'Comparable transactions declined 4%' is the only transaction figure, with no count

The model

• Price, P. Revenue divided by units: what one unit sells for, on average, over the fiscal year.
• Units, Q. The unit count the filing discloses: memberships, vehicles, subscriptions.
• Marginal cost, MC. Cost of revenues divided by units, standing in for what one more unit costs. Real marginal cost is not in any filing; this is the closest line that is.
• Lerner index, L = (P - MC) / P. The share of the price that is markup. Units cancel out of it, so it is the gross margin of the line being scored.
• Implied elasticity, e = 1 / L. If the price is the monopolist's best price, where marginal revenue equals marginal cost, then L = 1/e. So e is the price sensitivity the firm acts as if it faces: near 1 is pricing power, large is a firm that must price close to cost.
• Producer surplus, PS = Revenue / e. Equal to (P - MC) x Q, the markup times units, so it is the same under either demand shape.
• Consumer surplus, a band. It depends on the shape of demand, which no filing shows. A straight-line demand curve gives Revenue / (2e); a constant-elasticity curve gives Revenue / (e - 1). The page shows both ends, and the width of the gap is how much the answer depends on an assumption nobody can check.

Assumptions

• The price observed is the monopolist's best price, where marginal revenue equals marginal cost.
• Cost of revenues per unit stands in for marginal cost. It holds some fixed cost, so it overstates MC and understates L.
• One fiscal year at a time; nothing is adjusted for inflation or currency.
• Where the markup is zero or negative there is no reading: the model does not apply that year.

Every figure, with its source

Revenue and cost are in billions of US dollars, as each filing reports them. A year missing from a table is a year the filing did not disclose; nothing here is estimated.