Midrange Jumpers for the Middle Market
Power laws feast on the center until what's left is either scale or soul
This is Part IV of Midrange Jumpers for the Middle Class. A series examining different-yet-connected economic dislocations, social imbalances, and structural forces that erode the center.
A lot of artwork and two different songs are included. The early sections of this essay were written listening to this. Enjoy.
Prelude
A unifying phenomenon appears across sports, retail, beer, banks, app stores, Hollywood, music, academia, the middle class, the list goes on. Unrelated domains all undergoing the same process and outcome.
The difference between narrative and structure: one attributes retail’s collapse to Amazon, the other explains why competitive systems produce Amazon in the first place. Narratives are perishable; frameworks compound. Top-down stories obscure this; bottom-up bedrock reveals it.
If this only happened to newspapers, we could blame Craigslist. If it only happened to music, we could blame Spotify. But when it happens everywhere, we ought to look deeper.
Give a man a fish, he eats for a day; teach him to fish, he eats for a lifetime. Give a man a story, he sees one thing; teach him a framework, he sees for a lifetime. Stories are fish. Frameworks are fishing rods. Feed someone a narrative about Amazon destroying bookstores, and they understand one industry’s death. Teach them the game theory of optimization, and they see the skeleton beneath retail, banking, entertainment, and their own career.
Frameworks endure because they’re upstream of opinionated narratives that treat situations as unique, novel snowflakes: no two patterns alike! To systematize is to realize each snowflake is basically the same, and nominal differences distract from their structural parallels. Once you see mechanisms, anecdotes become unnecessary.
This isn’t an opinion piece bemoaning NBA shot selection or how streaming ate media; those are symptoms and descriptions. What you’re about to read consists of diagnostics and fractal competitive patterns: something you can use repeatedly in different contexts and scenarios. It decomposes underlying mechanisms that convert spectrums into dichotomies, bell curves into barbells, and stable middles into hollowed cores.
Do you have opinions on the working class, income inequality, globalism, or why your hometown is a shell of its former self? You’re in luck, this essay is for you! I believe Midrange Jumpers for the Middle Market encompasses the most useful, informationally-dense-yet-digestible ~14k words you may ever encounter. A great deal of work went into this, with each section earning its place and building into the next. I know attention is valuable and focus is fleeting; if you commit to this essay, it will disproportionately give back to you.
If you’d like to absorb the reasoning and writing you’re accustomed to in my other work, you can skip Section V’s industry breakdowns and the Research Appendix at the very end and reduce read time by about 60%. They’re case studies that empirically reinforce the thesis and logic, but the framework stands without them.
However, if you read the case studies (recommended!) you’ll holistically internalize — with timelines, data, and some pizzazz — how entire swathes of the US economy (and basketball courts) have been eviscerated and will continue to be unless dramatic action is taken. We conclude with atypical answers about what that action should entail: taking no half-measures and providing solutions that look at the root of the issue, not its leaves.
By the end, you’ll understand why optimization doesn’t smooth things out, but sharpens them into points. You’ll recognize the forces humming around you that quietly polarize, consolidate, and impose a choice between extremes. If you derive value from this, please give it a like and share it with someone else.
Outline
I. Optimized Poles, Exposed Core
When optimization runs long enough, it eats its own bell curve
II. Patterns of Extremes
The forces that breed fat tails
III. The Smile Curve
How supply chains grin and eat the center
IV. Is My Center Disappearing?
Diagnostics that suggest your core is on its way out
V. Midrange Jumpers for the Middle Market
Each case study demonstrates the same transformative pattern across different domains and timescales. Six examples are provided mid-essay to illustrate mechanism, process, and outcome, with more provided in the Research Appendix.
NBA
Local News
Hollywood & Movies
App Stores
Department Stores & Retail
Banking
VI. The Natural State of Extremes
Extremes are the steady state, parity a peculiarity
VII. Individual Application
Technology feasts on centers until what’s left is either scale or soul. Which way, Western worker?
VIII. Organizational Application
Businesses either weaponize scope or worship distinction
IX. Civilizational Application
How to sustain a functioning society when human economic value varies by five orders of magnitude
X. Societal Implications: Tradeoffs All the Way Down
Purchasing social stability with economic productivity: unique solutions to preserve the center
XI. Universal Basic Income
Societal volatility and the rage of the structurally obsolete
XII. Concluding
The middle is a garden requiring enduring effort to sustain
Research Appendix
Additional Business Case Studies
Music, CDs, & Streaming
Book Publishing
Casual Dining
Beer & Breweries
Condensed Case Studies: Cities, Politics, Fitness, & Academia
Counterexamples: Extremists in Middle Disguise
I. Optimized Poles, Exposed Core
Unassociated competitive domains converge on dominant strategies with a consistent theme: the middle dies.
When systems optimize, intermediate positions become unsustainable. This pattern devours industries and blue-collar life alike, transforming colorful variability into black-and-white division. What was once a spectrum splits into binary, transforming imperfect art into precise algorithm.
We assess these situations individually, relying on bespoke top-down analysis as if they’re peculiar aberrations: “People don’t like malls anymore”, “Hollywood makes superhero movies because culture is stagnant”. This is superficial and misleading; these are not isolated events. Culture doesn’t create circumstances; circumstances create culture.
These stories are symptoms of the same mechanical architecture. Bottom-up environments produce similar outcomes under similar conditions.
Almost without exception, our first impulse is to assume we acted upon the world; we never consider that the world is acting upon us. Laws of nature impose a physics of coordination and information flows. We are subject to this, not separate from it.
Nature breeds extremes across all walks of life. This isn’t capitalism’s pathology or democracy’s failure, but the constitution of resource dissemination and information propagation: applicable to all creative and adversarial domains. Power laws emerge anywhere performance compounds and selection intensifies. The longer domains adopt performance filtering under pressure, the more their outcomes resemble one another: skewed, polarized, efficient, brutal.
This happens regardless of political system, personal beliefs, or whether we think it’s fair.
Soviet institutions had exaggerated prestige hierarchies via the nomenklatura, with elite privileges rivaling Western contemporaries. Chinese state-owned enterprises exhibit Pareto-esque productivity ratios despite central planning. Academic citations have fat tails in every country studied. When selection operates on multiplicative (exponential) rather than additive (linear) returns, skewed curves are the organic end state. ‘Capitalism’ is simply a politicized label for people acting in their interests through markets; it neither causes nor fixes these outcomes.
When a system tunes for specificity and scale, the intermediate becomes inefficient and exposed. Optimization eats its own bell curve if it runs long enough.
Baseball and venture capital operate in different worlds, yet both discovered identical strategies. Baseball neglects .270 singles hitters for power/patience homeruns. VC funds eschew steady 2x returns for portfolio allocation embracing twenty failures to find one Airbnb. Convergent evolution operates on competitive systems as surely as it does on species.
The middle, by its essence, is unoptimized: a compromise between price points, use cases, and audiences. As adversarial pressures intensify, dominant strategies align on targeted tactics that manufacture extremes and feast on centers.
To survive and thrive, you must choose a side: expand or specialize. Engineer for efficiency or experience. Become the cheapest to produce or the hardest to replicate. What’s left is either scale or soul.
As the middle class goes, so goes the market: bifurcating into luxurious top and discount bottom. The middle market becomes a contradiction in terms without a working class to sell to: no market is there anymore.
Steve Jobs called technology a “bicycle for the mind”: you get somewhere faster pedaling than on foot, just as you get somewhere quicker with technology than without. Charming, but incomplete. Tech is also a centrifuge: separating what was once mixed and spinning outcomes to the peripheries.
Geographic barriers guarding regional business vanish with digital distribution and globalized supply chains. Information asymmetries fragmenting markets evaporate with price transparency. Technology is both a bike for the mind and rocket ship for power laws.
The middle pays full freight and gets no favors. Scale players with generalized objectives overpower it; niche players with specific audiences outflank it.
Poles amplify when you optimize for pure efficiency or pure differentiation; the middle optimizes for nothing.
The middle is neither fast nor strong, cheap nor premium, simple nor sophisticated. In basketball, these players get cut. In markets, these companies get crushed.
There’s a physics to human coordination and resource circulation. Whether app stores, breweries, movies, capital, or countries, selection effects in open arenas yield Pareto’s 80/20 curve. Nature says inequality is the rule, parity the exception.
The tragedy of the center: if you try to be everything to everyone, you’re nothing to no one.
Terms
Power law: A mathematical relationship where small inputs produce outsized effects (y = x^k). When one variable changes, another responds asymmetrically.
Heavy (fat) tail: Right-skewed distributions where extreme outcomes are far more probable than normal distributions would predict. Power laws commonly produce heavy tails.
I use ‘power law’ throughout as shorthand for “extreme concentration at endpoints”.
Pareto distribution: The 80/20 rule. A heavy-tailed probability distribution where a small share of inputs (<20%) account for a large share of outputs (>80%). Outcomes cluster at one end with a long tail at the other.
Equilibrium (stable state): The point where a system naturally settles once its internal elements stop changing.
You can identify non-stable states by their need for consistent intervention against equilibrium-seeking dynamics.
A rock halfway up a hill rolls down unless you hold it there. Remove your hand, and it returns to the base (equilibrium).
Linear, Exponential, and Multiplicative:
Linear growth: Adds a fixed amount each time: e.g. 10 → 20 → 30 → 40 (constant +10, declining percentage gain of 100%, 50%, 33%).
Exponential growth: Multiplies by a constant rate each time: 10 → 20 → 40 → 80 (constant 2x and 100% increase).
Multiplicative growth: Each step multiplies the previous one, but the rate isn’t necessarily constant.
It might multiply by ×3, then ×1.4, then ×2.1: still compounding, but not strictly exponential.
General note: Power laws, Pareto, and fat tails overlap but aren’t identical. I use them interchangeably because the practical implication is the same for this essay’s purposes. I do the same for ‘multiplicative’ and ‘exponential’. The technical distinctions matter to mathematicians; the shared conclusion matters to everyone else.
When I say ‘optimization’: I mean systems converging on a measurable objective under competition: points per possession, cost per unit, engagement per user, return on capital. Add feedback loops and enough time, and anything that doesn’t help that metric either mutates or dies. The middle is mostly made of things that are good for people but bad for that objective.
II. Patterns of Extremes
Power laws are nature’s ancient directives, appearing wherever systems compete for attention, resources, or influence… which is to say, everywhere.
Markets are calculators, not caretakers. They don’t weigh dignity, cohesion, or ideology, only throughput and efficiency. Competition’s gravitational pull adheres to predictable patterns pointing the same direction: away from the center.
1. The Matthew Effect (Cumulative Advantage)
The strong disproportionately gain through multiplicative returns. Those with more get more; those with less lose what little they have. This principle embodies cumulative advantage: success breeds success, wealth attracts wealth, fame begets fame.
This appears biblically in the Parable of the Talents, where servants entrusted with resources were rewarded or punished based on how they multiplied what they were given. Compound interest as a metaphor for divine justice.
“For whoever has will be given more, and they will have an abundance. Whoever does not have, even what they have will be taken from them.” — Matthew 25:29
In network science, this is known as Preferential Attachment:
The probability that a node gains new connections is proportional to its existing connections.
This is a subset of cumulative advantage, and the principle generalizes beyond networks.
Wherever coveted resources are sought, those who already possess them (citations, downloads, wealth, attention) experience Matthew’s preferential attachment.
Rewards don’t add but multiply. The paper with 10 citations gets cited ~100x more than the single-citation one. An app with 100k downloads gets exponentially more visible through algorithmic amplification than its 50k counterpart.
Benefits don’t stack linearly like bricks; they compound like interest. The rich get richer at an accelerating rate. The signature shape of Matthew distributions isn’t a bell curve centered on average, but a long tail where the top 1% commandeers more than the bottom 50% combined.
Concentration at the poles is the standard; equality is the anomaly.
2. The Barbell Distribution
Barbells illustrate outcomes clustering at poles while the middle empties. Normal distributions concentrate around the mean; barbell distributions evacuate it.
When systems prioritize a single metric (minimize cost, maximize engagement, increase points per possession), intermediate positions functionally become dominated strategies. They aren’t specific enough to monopolize niches nor general enough to prevail at scale.
In game theory, a ‘dominated strategy’ is inferior because it produces competitively worse outcomes to alternatives, regardless of what opponents do.
Single-objective optimization yields barbell distributions:
The premium end optimizes for quality, experience, or status
The discount end optimizes for peak efficiency and price
The middle optimizes for nothing. You either specialize, scale, or starve.
The middle must prioritize a single approach to persist. Competing on price weakens premium positioning; competing on quality weakens pricing. You can’t defend both flanks simultaneously! No half measures. Either go massive or microscopic.
3. Inverted-U Cost Curves
An Inverted-U cost curve: when per-unit costs peak at intermediate scale.
Giants achieve economies of scale in marketing, operations, and procurement (AWS data centers, Walmart supply chains). The tiny either avoid fixed costs or compete on dimensions other than price. The middle bears high fixed costs without scale to offset them: the worst of both worlds.
Digitally native businesses often have inverted-U cost curves:
Small: Minimal fixed costs (indie developers with laptops and cloud services)
Medium: Fixed costs (office, staff) without major volume
Large: High fixed costs, but massive scale drives per-unit costs to near-zero (Visa, Cloudflare serve billionth customer for pennies)
Traditional U-shaped curves (applicable to atoms-based businesses) show per-unit costs initially declining via economies of scale, then eventually rising with bloat and excess complexity (diseconomies of scale). The middle does well here. Technology is inverting the U, raising the ceiling where diseconomies occur and promoting scale.
Amazon, Google, and Apple attain gargantuan capacity without diminishing returns because bits coordinate better than atoms. Automation reduces marginal costs and digital distribution evades geographic boundaries. Technology streamlines scale and privileges size. The middle is squeezed between differentiated small players and tech-enabled giants.
4. Preference Bifurcation
Widening income gaps produce widening markets. When the middle class shrinks and wealth concentrates, markets bifurcate in kind. Income inequality (where people either have a lot of money or not much) drives consumer demand to the same extremes! The middle market goes the way of median incomes.
The premium segment chases the affluent minority
The value segment chases the cost-conscious majority
The middle market serves who, exactly? A shrinking demographic with diminishing purchasing power.
Middle market contraction exacerbates trends: mid-size failures eliminate middle-class jobs, which further depletes demand for middle-market products. A doom loop for the blue-collar man.
5. Network Effects and Algorithmic Amplification
Call these Turbo Matthew: the biblical dynamic, boosted by technology. What took decades in the physical world takes months in the digital realm. Things move faster with the computer’s bits vs reality’s atoms.
Network Effects: Each additional user makes the network more valuable. Digital platforms breed radical concentration through Matthew loops where early success informs future success.
Algorithmic Amplification: The algo cares not for who deserves visibility, only who already has it. Blessing engagement with more engagement.
Network effects shower connected nodes with more connections. Visibility elevates visibility. Positions become self-reinforcing.
Technology manufactures heavy-tailed ecosystems with industrial efficiency.
6. Threshold Cliffs
These are step-function thresholds, where minor growth triggers major punishment. When compliance, ranking, or eligibility requirements jump discretely, the middle is strangled.
Giants clear the cliff and then some, capturing disproportionate value.
The small and focused evade the cliff, mitigating expenses and punitive red lines.
Examples:
Regulations where slightly crossing a barrier carries burdens:
Dodd-Frank’s (banking) $10B asset threshold:
Middle institutions right below the threshold face competitive liability (less nimble to specialize, still lack scale); those just above are saddled with crushing compliance expenses.
You’re ideally way above $10B or comfortably beneath it to keep costs proportional to profits.
Obamacare’s 50 full-time employee threshold for mandatory health coverage:
Companies with 45-55 employees face a binary: stay under 50 or expand enough to make the regulatory burden worthwhile. Modestly above 50 is max compliance with minimum revenue relief.
Environmental regulations with size-based triggers push firms to stay deliberately little or get big.
Regulations sterilize the middle not through intention but incentive. Policymakers design for average while the average systematically fades.
III. The Smile Curve
Supply chains have vastly different margins at each stage; smile curves map how value is captured.
In multi-entity supply chains, value clusters at endpoints: IP/design/brand on one side; distribution/platform/customer access on the other.
The middle, comprised of manufacturing/assembly/coordination, is left commoditized, undifferentiated, fungible. Vitiating margins and reaping the least value in the chain.
This pattern emerges because endpoints control irreplicable scarcities:
Upstream: IP, R&D, brand, and design aren’t fungible
Downstream: Customer access, platform control, delivery networks resist commoditization
Middle: Manufacturing, assembly, and coordination all standardize, facing relentless price pressure and margin compression.
Smartphones are an archetypal case:
Apple feasts on 45%+ margins through design and brand
Qualcomm gets 35%+ margins on chip IP; 70%+ on licensing
Foxconn, handling the difficult work of actually making the phone, operates on emaciated ~6% gross margins.
The firm that does the most labor earns the least, with less than one-tenth the profits of the endpoints. Assembly is hard but replaceable. Fungibility means whoever does it cheapest wins.
The pattern reproduces across mature-growth industries with multipart assembly:
Sneakers: Nike takes ~45% gross margins with brand and design. Asian manufacturers making the shoes get 8-15%.
Pharmaceuticals: Pfizer and Merck enjoy 70-90% margins on patented medications or blockbuster drugs through R&D and IP. Manufacturers managing complex chemical processes creating the drugs: 15-25%.
Coffee: Starbucks sips on ~27% margins via brand and experience. Processors and bulk distributors/logistics: single digits.
The smile bares its teeth as industries mature. Manufacturing standardizes, competition piles into the middle, margins compress. Meanwhile, endpoints accumulate benefits that defy commoditization.
The middle is saddled with fixed costs and fungibility while the ends harvest rents. That gap widens until the curve grins back at you :)
IV. Is My Center Disappearing?
Power laws leave fingerprints. Here are harbinger diagnostics that your industry’s center is on borrowed time. If you see three or more, you probably have 5-10 years.
Are returns superlinear with size?
Do businesses have naturally high operating leverage?
Is the middle segment of a supply chain getting highly competitive or considered fungible?
Are search/discovery/information frictions abating (ranked feeds, price transparency, more algorithms)?
Were threshold cliffs (regulatory, ranking, eligibility) recently introduced?
Are there network effects?
Senior leadership has begun using the phrase “platform economics”
Does adding one more user make the product better for everyone?
Is the objective function converging on a single-metric (points per possession, COGS, etc.)?
Is your trade association discussing industry consolidation? A lot of M&A recently?
Does geography matter to customers? Is it starting to matter less?
These heuristics are corroborated by the subsequent case studies in Section V and the Research Appendix.
V. Midrange Jumpers for the Middle Market
The following reviews how competition’s telos and technology’s ethos consume the core across business case studies. If you’d like to take my word for it, you can cut read time and skip ahead to Part VI The Natural State of Extremes.
However if you’ve gotten this far, then this topic clearly interests you. You’ve absorbed the mechanics of middle death and can now witness how it manifests in real life. If you have time, stick around. You’ll find historical nuance and deconstructions of industries that have touched your existence in one way or another.
The middle class itself has already been addressed holistically in Part I Midrange Jumpers for the Middle Class and expressively in Part II Touching Soul, Touching Senses: The Art You Deserve. Now we’ll observe the same dynamics playing out across media, retail, sports, finance, entertainment, restaurants, and manufacturing: disparate domains with the same patterned skeleton supporting them.
As tech proliferates (algorithms, borderless logistics, financial engineering) and geographic barriers are removed, the center’s disintegration accelerates. Most barbell transformations (from balanced beginnings to skewed endings) span 15-35 years; digitally native ones can manifest in ~8-15 years.
The more a domain lives in the atoms-based, physical world and opposes technological enhancement, the longer it takes to reach full barbell. But resistance buys time, not immunity. Rejecting technology is like playing clean in a sports league that allows steroids. Choose to stay doughy and slow on romantic principle, and you get blown out. Markets don’t factor in noble motivations.
Geographic constraints were a Chesterton’s Fence of sorts.
“Don’t ever take a fence down until you know the reason it was put up.” — G.K. Chesterton
When borders disappear, discovery automates and metrics dominate.
We thought land boundaries merely limited consumer choice. They also limited winner-take-all concentration. Computers removed the geofence, and we found out it was a load-bearing column for the American core.
Consider what Amazon did to retail: your local bookshop’s only moat was its down-the-street convenience. Physical proximity made it viable. Remove that land-based advantage, and thousands of shops and hundreds of quirky Main Streets collapse into one website. Now a single omnipresent Everything Store in Seattle serves 300 million Americans while your downtown is hollowed out.
Technology clusters strategies and concentrates power in whoever leverages its best. As it propagates, you curiously find yourself with no midrange jumpers, and no middle market.
Here are examples that illustrate:
Note: There are many data points in here and the Research Appendix, and despite my diligence, errors happen. If you catch any, let me know and I’ll review and change as needed. I’ll probably continuously update this essay as an ongoing point of reference for this concept. If you can’t read it all now, bookmark it, share with a friend, smash the like button, and return to it later.
1. ENTERTAINMENT & COMMUNICATION
Fastest: bimodal in 8-20 years
Additional case studies in Research Appendix:
Music, CDs, & Streaming
NBA Moneyball
THEN (2000):
What the middle looked like: Balanced shot chart across court geography. Post-ups, midrange jumpers, three-pointers all viable strategies. The midrange game (15-22 feet) was ~35% of shots; three-point attempts ~17%.
Shot selection reflected coaching philosophy and player strengths rather than expected-value data analysis. Defenses guarded all areas equally, creating strategic diversity.
Why it seemed stable: The middle represented sound decision making and tactical variety. Eye test suggested 15-foot jumpers were good shots. Traditional basketball wisdom: attack basket first, take open midrange if defense collapses, shoot threes selectively.
NOW (2025):
What happened to the middle: Three-pointers jumped from ~17% (2000) to >42% (2025): more than doubling. Midrange fell from 35% to ~12%.
Modern superstars (Giannis, Luka, Curry) avoid midrange unless defense forces it. The “midrange specialist” is an extinct species for role players.
Defensive schemes adapted: Defenses actively funnel opponents into midrange, treating 15-foot jumpers as a success. The middle went from guarded territory to intentionally conceded space.
Current structure: Extreme barbell. The court is now bifurcated: either take restricted area (at-rim) or beyond the arc. The middle a suboptimal desert.
In 2000, the Lakers won a championship with ~16% of shots from three. In 2024, the Celtics won with ~45% three’s. No team has won a title with a three-point rate under 35% since 2014.
The transformation: Points per possession became the only metric that mattered. Strategic diversity died when everyone discovered the same math.
Analytics revealed midrange as a dominated strategy in points per shot (PPS) terms.
A good 40% three-point shooter generates 1.20 PPS; equivalent to a superb 60% two-point shooter.
Even excellent midrange shooters (50% from 15-22 feet) produce only 1.00 PPS.
Three-pointers are worth 50% more (3 vs 2 points), and when shooting percentages are comparable, this point premium makes them decisively superior:
Three-pointers by good shooters: ~1.05-1.20 PPS
Restricted area attempts (dunks, layups) have highest efficiency per possession at ~1.20-1.30 PPS
Midrange jumpers by good shooter: ~0.84 PPS (dominated strategy)
Time to collapse: 15-20 years (2000-2020) from balanced shot charts to bimodal poles.
MECHANISM: Barbell Distribution + Single-Objective Optimization
When everyone calibrates for the same metric, strategies converge. The midrange isn’t slightly worse, but utterly inferior:
A team shooting 45% from midrange (elite) yields 0.90 PPS
That same team shooting 30% from three (below average) delivers 0.90 PPS
But improving three-point shooting from 30% to 35% (achievable) yields 1.05 PPS: 17% more effective than elite midrange
Daryl Morey’s Houston Rockets pioneered NBA Moneyball. From 2013-2019 they took around 40 three’s per game while averaging ~3-6 midrange shots. In 2018 they went 65-17 attempting ~45 threes and ~3 midrange per game; a 15:1 ratio deemed insane by any conventional standard. The whole league noticed.
Once Moneyball-style data quantified shots with precision (widespread by 2013), every team adopted the same expected-value approach: maximize restricted area and three pointers, minimize midrange. Within eight years, the league average jumped from 20 to 34 three-pointers per game. Coaches who resisted (like Jeff Van Gundy) got replaced by analytics respecters.
The feedback loop was merciless. Analytics-driven teams won games. Winning teams got copied. The shift spread rapidly through coaching, front office, and player development. All franchises calibrating for the same mathematical goal and strategy.
Every NBA team is now attempting more 3’s per game than Steph Curry’s 2016 Warriors. By 2024, zero teams attempted fewer than 30 per game: a figure only three teams exceeded in 2013. In 2025, the team average is 37 per game.
The midrange is an abandoned wasteland on NBA heat maps:
Local News
THEN (Pre-2005):
What the middle looked like: Thousands of metro and small-city dailies formed a robust middle tier between national brands and hyperlocal weeklies. Metro coverage employed newsrooms of 50-200 journalists covering regional politics, business, crime, education, sports.
2000 was the peak of classified ad revenue at $20B
Market structure: Fairly balanced three-tier system with national papers (NYT, WSJ, USA Today), metro dailies serving cities/regions, and local weeklies covering neighborhoods. Geographic fragmentation prevented winner-take-all dynamics.
Who thrived in the middle: Metropolitan newspapers. The Cleveland Plain Dealer peaked at 400 journalists in the 90s. San Jose Mercury News had 400. The Boston Globe: 550. Seattle Post ~200. More than jobs, they were careers with pensions.
Local retailers bought display advertising to reach regional audiences. Classified advertising (employment, real estate, automotive) drove 20-40% of revenue and was the profit engine. Print advertising funded majority of revenue (up to 80%) with high margins.
A quarter-page “Help Wanted” ad cost $2k and took 15 minutes to place. Real estate listings ran $400 each. Pure margin, minimal production cost. These boring rectangles of text funded investigative journalism.
Why it seemed stable: Geographic monopolies formed news citadels with limited metro competition. Classifieds were highly profitable with minimal production costs. Dual-revenue model of subscriptions and advertising provided solid base.
NOW (2025):
What happened to the middle: 3,300 newspapers closed since 2005, creating news deserts across America. Metro papers hemorrhaged revenue as advertising migrated to digital platforms.
Google, Meta, and Amazon estimated with 56% of the ad market; some estimates high as 75%.
Out of 3,143 US counties: 208 have no local news source (6.6%) and 1,563 (50%) have only one. That’s about 55M Americans with either very limited or no local news. Source.
Classifieds went to specialized tech players:
Craigslist (run by ~50 people) disintegrated $20B in newspaper classified revenue. Indeed took employment listings. Zillow occupied real estate. AutoTrader, automotive. Each is faster and either free or cheaper.
The $2k help-wanted ad became a $100 Indeed posting. The $400 real estate listing, free on Zillow. Craigslist posts cost nothing.
Current structure: Extreme barbell. Platform leviathans (Google, Meta, Amazon) seized digital advertising through scale and targeting. Few national big-brand papers (NYT 11M+ subscribers, WSJ 4M+, WaPo ~2.5M) achieved digital subscription scale.
Local operations survive on volunteerism, philanthropy, or subsidy. The metropolitan middle (50-200 person newsrooms) was eviscerated.
Middle casualties: Since 2008, newsroom employment plummeted 57% (from ~71k to ~31k) Pulitzer-winning papers leaked readers and resources to digitally adversaries.
Denver Post down from 250 to 60 journalists. Baltimore Sun and Philadelphia Inquirer comparably decimated. In 2004, the Plain Dealer sold 350k papers daily, by 2020: 80k.
Print circulation declined 60-80% from peaks.
Time to collapse: ~20 years (2005-2025)
MECHANISM: Smile Curve Disaggregation + Matthew Effect + Preference Bifurcation
Local news was razed by smile curve dynamics within the attention economy.
Pre-internet, metro newspapers controlled both content production AND distribution, capturing full-stack attention chain.
Digital disaggregation severed this: Google/Facebook annexed distribution (downstream endpoint), national brands like NYT captured premium content positioning (upstream endpoint).
Metro dailies got stuck in the commoditized middle: serving content without controlling discovery and no brand premium.
Preferential attachment in algorithmic discovery amplified this handicap. A NYT article on urban crime appears in the same Facebook feed as a Denver Post article on Denver crime, but the algo’s engagement optimization favors the brand with more prior engagement, more social proof, more accumulated authority.
Advertisers who previously bought Denver Post ads now access hyper-targeted Facebook audiences. The affluent (who pay for subscriptions) consolidated on a few national outlets. The price-conscious (who supported ad-revenue models) migrated to free social platforms. The Denver and Cleveland readers advertisers targeted? Both buyer and seller evaporated.
Geographic monopolies dissipated with digital access.
Hollywood & Movies
THEN (1990s-2000s):
What the middle looked like: Studios regularly greenlit mid-budget, star-driven films, anchoring the middle between indies and blockbusters.
$20-70M films were large share of releases, targeting theatrical audiences seeking stars without spectacle (i.e. not superhero slop).
Market structure: Balanced distribution with indie films (<$10M), mid-budget ($20-70M), and blockbusters ($100M+) each serving distinct market.
Critical: DVD backend sales provided profit a safety net for theatrical risk.
Who thrived in the middle: Star power plus DVD backend equaled profit even if theatrical disappointed.
Michael Clayton cost $21.5M, made $92M box office, then added ~$20M in DVD sales. Taken cost $25M, grossed $227M, then another $85M in DVD sales; that’s an additional 37% of an already excellent box office from discs. Wedding Crashers invested $40M, returned $283M in theaters, then $147M in DVDs sales! 52% of theatrical revenue.
I was shocked learning how much money DVDs made. I thought they’d be an afterthought revenue source. Not the case.
Why it seemed stable: DVD backend materially de-risked mid-budget films. Star power alone provided enough marketing to recoup investment without requiring franchise IP.
NOW (2025):
What happened to the middle: Box office and marketing economics shifted toward $150-300M IP with franchise potential, or platform-distributed indies. Either you go full Batman or full Theater Kid, and Vince Vaughn gets left behind.
Theatrical middle disfavored as DVD backend evaporated. Total DVD sales fell ~94% from ~$16B peak (not counting rentals) to <$1B in 2024.
Current structure: Creeping barbell distribution. Studios blockbusters ($100-200M+ production plus $100M+ marketing) with global franchise potential are economically en vogue.
Streaming platforms distribute micro-budget indies and occasional mid-budget acquisitions. Theatrical middle viewed like a midrange jumpshot.
A mid-budget drama that underperformed theatrically used to recoup on DVD; now it just loses money. Marketing costs increased. Theatrical windows shrank. The Owen Wilson rom-com became much riskier.
Fat-tail supremacy: Marvel films routinely gross $1B globally, Avatar sequels break records, franchise IP (Star Wars, DC, Fast & Furious) delivers predictable returns.
Micro-budget artsy indies succeed through streaming and festival prestige: films under $5M find audiences on platforms without theatrical costs.
Both prioritize single objectives: maximize global franchise potential OR minimize production risk.
Who’s left: Studios dominate megahit tier (Disney/Marvel with $200M+ per film). Streaming platforms acquire micro-budget passion projects and select mid-budget content. Theaters mostly occupied by spectacle-slop franchises and micro-budget breakouts.
Bifurcation economics: Marketing costs spend better on global franchises; $100M only justifiable for $1B turbo-hit potential.
Streaming gave indies cheap path to audiences:
Mid-budget films have theatrical risk sans scale economics or micro-budget nimbleness.
The middle bears theatrical and marketing costs without franchise potential, economically suboptimal.
Time to collapse: ~15-20 years (2005-2025), with DVD death and streaming catalyzing transformation.
MECHANISM: Scale Economies + Preference Bifurcation
Here’s Mr. Matt Damon with the details:
App Stores
THEN (July 2008):
What the middle looked like: iOS App Store launched with 500 apps and revenue distribution where middle-tier developers could build sustainable businesses. Top-10 developers claimed $9M of $21M total revenue (43%), leaving viable middle.
Freemium model not yet prevalent, just regular purchases.
Market structure: Early platform with manageable competition, multiple paths to visibility, flat discovery curve. Independent developers and small studios could thrive. Category diversity allowed niche apps to find audiences.
Why it seemed stable: Low customer acquisition costs made organic growth feasible. Algorithm neutral, no sophisticated engagement optimization yet. Early audience willing to explore and pay for quality apps.
In 2008, you could launch an app, get featured by Apple’s editorial team, and hit top 100 through word-of-mouth. Customer acquisition cost: essentially $0.
Angry Birds didn’t even advertise and became a global franchise. Instagram had 13 employees when Facebook bought it for $1B. The platform was a discovery engine, not a slot machine. Users browsed categories, tried apps, paid $2.99 without thinking.
NOW (2024):
What happened to the middle: Eight years later, the bottom fell out. By Q1 2016 top 1% of publishers got ~94% of App Store revenue ($1.34B of $1.43B). Out of millions of apps, ~3-4k earned above US median income.
Candy Crush alone earned ~$1.3B in 2016. The bottom 99% of developers split ~$100M.
Gaming follows an equally radical Pareto distribution.
Current structure: Radical fat tail with mega-hits at top and long tail of struggling apps. Big apps with budgets dominate paid acquisition. Long tail survives on organic/niche discovery with near-zero costs.
Middle casualties: Apple algorithm changes weighted engagement metrics, ensuring early success predicted future success (Matthew). Customer acquisition costs skyrocketed.
The middle couldn’t reach escape velocity: too small to fund scaled user acquisition, too large to survive on organic discovery alone. Professional development demands investment but has insufficient scale to justify it.
Time to collapse: 8 years (2008-2016) from balanced to peak Matthew, the fastest transformation across all domains examined.
MECHANISM: Network Effects + Algorithmic Amplification (Turbo Matthew)
App network effects created moats (users joined where friends already were). Each download improved chart positioning, generating more discovery. Freemium monetization (introduced 2009) with in-app purchases favored those with retention, amplifying early strength into a stranglehold.
The rank-to-revenue curve is exponential, not linear. E.g. jump from rank 10 to rank 5: revenue triples. Jump from rank 5 to rank 1: revenue up tenfold. Drop from rank 5 to rank 11: revenue craters 70%. Position isn’t everything, it’s the only thing.
The distribution has a heavy tail, and the top annexes disproportionate value. Turbo Matthew.
Mobile operating systems:
Similar dynamics:
2010: multiplayer field with Symbian ~30%, iOS ~24%, BlackBerry ~19%, Android rising, plus Windows
By 2016: iOS + Android owned ~99.6% of smartphone share. Mid-share operating systems couldn’t attract enough developers once app ecosystems tipped. Two-sided network effects (users and developers) locked in two winners while mid-tier platforms (BlackBerry, Windows Phone) died despite technical merit.
2. PHYSICAL RETAIL
Medium Speed: bimodal in 15-35 years
Additional case studies in Research Appendix:
Book Publishing
Casual Dining & Restaurants
Department Stores
THEN (1990):
What the middle looked like: Department stores recorded 14.5% of US retail sales, serving as anchors of American shopping malls with ~20k locations nationally and sales peaking at $233B in 1998.
Regional chains had geographic edge alongside national players.
Market structure: Balanced three-tier system with budget (discount stores), middle (department stores), and luxury.
Who thrived in the middle: Sears captured 1% of the entire US economy at peak, with two-thirds of Americans shopping there quarterly. JCPenney and Dillard’s catered to middle-income families. Macy’s and Nordstrom had middle-to-upper positioning.
The Sears catalog was 1.4k pages and reached 30M households
Macy’s Thanksgiving Day Parade had cultural aura
Why it seemed stable: Geographic monopolies in malls provided captive traffic. Breadth of merchandise (clothing, home goods, electronics, furniture) under one roof was convenience edge. Middle-class economic stability was a reliable base.
NOW (2025):
What happened to the middle: In twenty years, department stores went from dominant to decorative: no longer anchors but artifacts. Department store share crumbled from 14.5% to below 2% by 2024: an ~87% decline from 1990 peak over 34 years.
Store count: 20k to ~4.6k (76% reduction). Retail employment down 40%+ over two decades.
Number of department store brands fell from around 20 to 8.
Current structure: Significant barbell. Growth accrues to value giants (Walmart, Target, dollar stores, TJ Maxx, fast fashion) and premium/luxury (specialty stores, boutiques). Department store share of general merchandise plummeted from 59% (2000) to 18% (2020).
Ulta Beauty’s market cap exceeds Macy’s, Kohl’s, Nordstrom, and Dillard’s combined. One category-specific retailer is worth more than four major department stores with 150 years of history.
Meanwhile, Shein floods the bottom with $5 polyester while Hermes prints money selling $10k handbags to ladies who don’t check prices (women be shopping).
Walmart’s market cap: $470B; Amazon: $1.8T
Middle casualties: Sears declined from 1% of US economy to under 30 stores from 4k peak. JCPenney filed 2020 bankruptcy. Gap, J.Crew, Banana Republic all floundering. The middle market needs a middle class to sell to.
Time to collapse: 34 years (1990-2024)
MECHANISM: Preference Bifurcation + Scale Economies
E-commerce eliminated regional superiority for physical retail. Generic merchandise became available everywhere, and department stores lost differentiation. Amazon’s unlimited products plus 24/7 availability were unsurmountable. Fast fashion (Zara, H&M, Shein) claimed cost-conscious consumers. Luxury brands claimed the wealthy.
Income polarization drove customer bifurcation: affluent minority traded up for luxury goods, cost-conscious majority traded down for low-price giants. The department store isn’t cheap enough to compete on price nor special enough to justify premiums.
The middle paid full fixed costs (stores + employees) without brand power or discount leadership.
3. REGULATED & HEAVY REGIONAL
Slowest: bimodal in 40-60+ years
“Heavy Regional”: businesses serving a specific geography by distributing physically hefty products with significant logistics considerations.
Additional case studies in Research Appendix:
Beer & Breweries
Banking
THEN (1970):
What the middle looked like: US banking featured ~13k independent banks in highly fragmented market. Insured depository institutions at ~20k.
Regulations prevented interstate banking: only 12 states allowed unrestricted statewide branching, and 38 (76%) did not. Thousands of regional banks thrived serving specific geographies.
Market structure: Balanced distribution across institution sizes. Top-10 banks controlled ~16% of entire banking system. Deposit bases were sticky and loyal to local institutions.
Who thrived in the middle: Regional banks with ~$10-100B assets maintained strong positions combining relationship banking with solid scale for basic technology and compliance. Large enough to offer business lending, wealth management, and treasury services, small enough to know the construction company owner who banked three generations with them.
A $50B regional could fund a $20M commercial real estate project that JPMorgan would ignore. They occupied the Goldilocks zone: big enough to help regions, small enough for relationships.
Note: All $ figures are inflation-adjusted for today
Why it seemed stable: Regulatory fragmentation created protected local monopolies. Geographic regulations prevented concentration. Technology hadn’t yet enabled economies of scale in IT, risk management, and operations.
NOW (2020):
What happened to the middle: Number of banks fell 86% from 13k to ~4.4k. Industry underwent four decades of relentless consolidation eliminating mid-tier players.
Insured depository institutions from 20k to 5k (-75%).
Most states restricted branching into the 1970s; by 1992 all but four allowed some statewide branching.
From an all-time high of 30,456 in 1921, the vast majority of commercial banks are gone. Since 1934, there’s been a 71% decline.
Current structure: Major barbell distribution. The five largest banks control 50% of commercial bank assets and the top 10 have 1.3x the assets of all others combined (versus 1/4 of the assets in 1970).
As of Dec. 2020, four mega-banks dominate: JPMorgan $3.4T, Bank of America $2.8T, Citigroup $2.26T, Wells Fargo $1.9T
Community banks survive on relationships and stay under regulatory thresholds. Dodd Frank’s $10B enhanced compliance was onerous.
Of note: since 2019 some requirements were raised to $250B.
Middle casualties: Regional middle ($10-100B assets) faces structural disadvantage: too large for relationship banking, too small for scale economies.
Unable to compete on lending rates (funding costs higher than megabanks), service (technology), or distribution (branches/ATMs).
Strong midsize players either acquired or imploded during the GFC:
Washington Mutual had $328B in assets in 2007, crumbled in 2008, acquired by JPM. Wachovia: ~$750B in assets, forced into Wells Fargo merger. National City: ~$140B in assets, sold to PNC. These weren’t small banks.
Time to collapse: 40+ years (1970-2020) from fragmented to concentrated, with acceleration post-1994 deregulation and post-2008 regulatory burden. Merger wave in 1990s-2000s eliminated thousands of mid-tier institutions.
MECHANISM: Scale Economies/Inverted U + Threshold Cliffs + Network Effects
1994 Riegle-Neal Act removed geographic restrictions, enabling interstate banking and national consolidation. Once branch networks went digital, IT, funding, and compliance favored giga scale.
Network effects in payments and ATMs fostered winner-take-all dynamics; largest banks secure disproportionate value as hubs. And ‘too big to fail’ implicit guarantees provide funding benefits to mega-banks.
Post-2008, Dodd-Frank imposed onerous compliance: stress testing, risk management, reporting mandates, brutal overhead for mid-tier institutions. These regulations were written by people concerned about Citibank, not First National Bank of Topeka.
Either be huge (spread costs over trillions) or be tiny (exempt from enhanced standards). The middle was in no man’s land.
VI. The Natural State of Extremes
These aren’t market failures. This is what unimpeded optimization, measured purely by output, looks like. The barbell is destination, not distortion. The question isn’t whether a domain will bifurcate, but when and how.
Cost minimization and premium maximization cannot simultaneously preserve the core; the math doesn’t allow it. Moneyball logic and Matthew Principles colonizes domains sharing no superficial relation; any system that optimizes measurable performance under constraints will rediscover these strategies independently.
‘Capitalism’ is an ideologically manipulative term for what natural selection gestates in open markets. When resources circulate in a decentralized way, with no central authority dictating it, we call that a ‘capitalism’. Clearly a misnomer because it’s not a governance or personal belief system, but the absence of one. It’s what organically happens when human nodes consume and allocate as they see fit.
You don’t need to ‘believe’ or ‘support’ a forest for it to grow; it does so all on its own if you don’t stand in the way. Beavers accumulate and stack wood because that’s what they do; humans accumulate and stack capital (resources, land, property, money, energy, etc.) because that’s what they do.
There’s a physics to how information and resources disseminate. Capital flows like water down a hill when left unencumbered, taking the path of least resistance as it pools at the base. If we want to preserve the middle, tradeoffs must be made. Dams will have to be built that stop the water from pooling at the base in the same way.
Power laws exist independent of man, not due to him. Appearing in nature and social systems (earthquakes, wildfires, city size, rainfall, animal foraging, river networks, etc.) without human involvement, intent, or awareness.
An emergent force, they coalesce not through dysfunction or ideology, but as the steady state. The optimal solution to competition under constraint: the natural shape of extremes.
They manifest fractally across individual, business, and nation.
VII. Individual Application
Imagine explaining to a time traveler from 1950 that one Google AI engineer earns multiples of an entire high school faculty. Then tell him that engineer’s code impacts a billion users while the faculty only touches 1,000 students. The leverage chasm is obscene, and the engineer’s scarce skillset facilitates it. The compensation tracks accordingly.
How do you sustain a functioning, cohesive society when this scenario ceases to be anomalous and society-wide human economic value stratifies by some five orders of magnitude between top and bottom?
Companies optimize for revenue per employee with the same ruthlessness NBA teams do points per possession. The peddler of modestly complex knowledge work will find AI accomplishes it faster, cheaper, and without vacation days.
His profession is technologically commoditizing. He exists in the middle of a nation’s labor Smile Curve. What can he do?
The scale path:
Calibrate for leverage, influence, or network effects:
Tech careers where one engineer’s code serves millions
Finance roles where one trader’s decisions move billions
Entertainment where one creator’s content reaches tens of millions
The specialization path:
Become locally irreplaceable and trusted through deep expertise or personal relationships, resisting automation and circumventing algorithmic subjugation.
The master plumber in a small town with 20 years of reputation
The physical therapist whose patient relationships foster loyalty and family referrals
The orthopedic surgeon specializing in hip replacements that only 200 others nationwide can perform
These positions harvest value through personalized relationships, loyalty, and human trust: impervious to algorithmic replication and fungibility.
The fatal middle:
Neither hyperscales nor carries unique distinction, too tiny for capacity and too generic to protect. It can’t outwork the algorithm or out-human the artisan.
Professional services, trashbag-of-facts knowledge work, middle managers (automated by project management software and AI).
Common lawyers and accountants (LLMs send their regards)
Regional salespeople without personal relationships
All consultants (good riddance)
Positions like these face dual pressures: tech automation from above and commodification from below (outsourcing, freelancers, global competition).
Successful adaptations position at tails:
The middle manager who evolves into:
Engineering leader embracing AI and tripling productivity
Executive coach with deep expertise in family business succession (i.e. highly personalized rich-people services)
The regional accountant who transitions into:
CFO at high-growth tech startup leading venture funding
Specialized tax strategist for real estate syndications
The journalist who becomes:
Star reporter at national publication with big social following
Indie newsletter writer with 2k paid subscribers in specific industry (Substackcore)
The middle manager is to the LLM what the switchboard operator was to automated routing: blissfully, ephemerally residing in an untouched center during a technological grace period with an expiration date.
Technology feasts on centers until what’s left is either scale or soul.
I mostly listened to this song on loop while writing the remaining sections.
VIII. Organizational Application
The individual and institution are fractally subject to the same pressures. The middle is a transient mode born of either geographic barriers or technological virginity. Healthy centers are likely to be found in nascent industries that haven’t yet discovered their barbells.
To position for the long term: either weaponize scale or worship differentiation.
The scale path:
Pursue winner-take-most games with network effects, data supremacy, and compounding operational edge.
Platform economics: Each user makes the platform more valuable (network effects), emits more data (improving product), and spreads fixed costs over a larger base.
Example: each Uber rider increases value to drivers (shorter wait times), generates more data (improves routing algorithms), and spreads fixed costs (AWS infrastructure, legal, insurance) over a larger base.
A platform is readily identified when more value is created on top of it than the value of the platform itself.
The strategy involves accepting years of losses to achieve escape velocity, then extracting disproportionate rewards after establishing a stronghold.
Scale economics: If the marginal cost of serving an additional customer approaches zero while marginal value increases with each one, you’ve found high-leverage territory. Digital services like software, content, and finance share this property.
Netflix’s cost to stream to the 100 millionth subscriber is functionally nothing. Visa’s cost to process the billionth transaction is infinitesimal.
In accounting, this is known as “high operating leverage”
Not everyone can win here; power laws anoint ~2-4 major players per market. You gotta try to be one of them.
The specialization path:
Deliberately stay diminutive. Defend through specialization, brand, or premium experience that cannot be copy/pasted at volume.
Craft breweries succeed despite Big Beer controlling 80%+ of market; their revenue per barrel is 3-5x higher
Boutique hotels at $400/night thriving while mid chains recede
Premium fitness studios like SoulCycle charge +$40/class while $30/month common-man gyms decay (elaborated further in Research Appendix)
This strategy demands discipline and refusing expansion if it dilutes distinction. Developing moats through quality, community, or prestige that repel commodification.
In-N-Out Burger has ~420 locations across West Coast states and could easily open 4,000. Yet they don’t. Regional scarcity preserves the brand’s cult status and quality control.
Many can win here. Long tail of custom, esoteric niches supports thousands of smol specialists.
The fatal middle:
Either modestly scales or partially differentiates. It hedges rather than commits.
Mid-market retailers competing on selection/convenience (vs Amazon), price (vs Walmart), and experience (vs boutiques): failing at all three
Regional banks attempting consumer lending (vs megabanks’ low costs) and relationship banking (vs community banks where the president knows your business personally)
Mid-budget game studios with worse graphics than AAA blockbusters and higher costs than indies
Department stores offering meh selection at meh prices with meh service: a ho-hum business all around
Mid journalism trying to be both authoritative (but not as credentialed as NYT) and accessible (but behind a paywall). Substack indies eat their lunch from below while MSM hoovers up prestige from above.
Successful pivots entail monomaniacal repositioning: weaponize scope or worship distinction.
Walmart realized it couldn’t beat Amazon online and invested in omnichannel capabilities, leveraging its physical stores as a competitive strength.
Luxury brands (LVMH, Kering) overtly avoid commoners. They acquire high-class names, raise prices, and limit distribution to preserve status.
Publishers either become platforms (Penguin Random House consolidation) or micro-presses for distinct audiences.
IX. Civilizational Application
In tech-saturated industries with sophisticated players — be it venture capital, music, or app stores — the top 1% eats 70-95% of the riches. A nation is not immune to these forces.
A society must act to prevent the same outcomes, otherwise they will accumulate naturally. Either actively commit to preserving the middle or prepare for civilizational extremes. A half measure is no measure.
Power laws manifest fractally:
The employee’s career barbells (hyperscale or hyperspecialize)
The company barbells (platform dominance or premium distinction)
Then society barbells (oligarchs or peasants)
Which way, Western man?
With AI permeating all facets of work, this will only accelerate. The more computation enters the system, the faster Matthew goes.
Societies either capitulate to bimodal class structure or intervene permanently against power law equilibrium. The middle will not be sustained with wishful thinking and campaign slogans.
This is not new. Technology doesn’t cause this, capitalism isn’t the catalyst; they’re only bicycles for the process, getting you there quicker.
Pre-industrial societies exhibited similar effects. In 1700s France, the top 1% owned 60% of wealth (top 10% held 90%). The Romans had their patricians and plebeians with scant affluence in between. Medieval guilds (labor union parallels) guarding artisan middles gradually dissolved as open trade and mass production (factory systems separating owners from workers) proliferated. Every technological revolution (printing press, steam engine, internet) champions democratization but delivers consolidation.
There are of course many positives to this. The standard of living in absolute terms has risen dramatically since then; the poor have smartphones, cars, AC, and aren’t only not hungry but more likely to be fat. Credit goes to technology for this rotund accomplishment.
However it’s the wealth percentages, the ratios, the relative differences, that are schismatic, destabilizing, and likely to get far worse. “Economics isn’t a zero-sum game!” Oh, well do you know how percentages work? Because that is zero sum, and it ain’t a game.
Absolute living standards can rise while relative gaps expand. Economics isn’t zero-sum, but power and influence are: a 5x or 10x spread between classes materially changes how societies organize. Contentiously factionalizing, regardless of how many people own iPhones.
In nature, it’s large temperature and pressure gradients that fuel the hurricane’s destructive, dissipative formation: even if the whole atmosphere is ‘richer’ in energy. Likewise, it’s the steepness of the wealth gradient, not just the size of the economy, that destabilizes societal structure.
Extreme capital gradients are buoyed by technology’s waves: like water down a steepening hill, it gets there quicker. Whether there’s more liquid or less of it, the ratios of where it pools persist.
The post-war industrial boom temporarily flattened tails as three serendipitous preconditions aligned:
1. Geography mattered (transportation costs defended local producers)
In the 1960s, shipping containers carrying 10,000 tons cost ~$700k in today’s dollars (inflation-adjusted $70/ton). Local producers reaped default protection from afar. Distance carried an embedded tariff.
2. Technology created more jobs than it eliminated (e.g. factory work absorbed agricultural surplus)
US manufacturing peaked at 19.6M in 1979. Factories hired agricultural workers displaced by machines. For every farm job lost, 1-1.5 factory jobs appeared. Creative destruction for the low-skilled didn’t on net eliminate their jobs, only shuffled them around.
Once AI saturation hits an inflection point, low-skill human capital will cease being shuffle-able; you flat out won’t need them anymore. Employing them will become a quaint, inefficient gesture done for nostalgia; like going antiquing, but for labor.
3. Policy deliberately asphyxiated winner-eat-everything environments (antitrust, labor/union laws, progressive taxation)
The 1890 Sherman Antitrust Act broke up Standard Oil. In the 1950s, top marginal tax rates hit 91%, and union membership reached 35%. Laws preventing interstate banking precluded megabank consolidation.
Make all the libertarian noises you want, but these were very ‘pro-middle’ developments.
All three are now reversing
Today, that $700k 10,000-ton shipping container runs $20k (-97% decline in shipping costs). US manufacturing employment is down to 12M. The top marginal tax rate is 37%. Union membership: 10%. The four biggest banks form a nationwide financial cartel.
The affluent American middle was a 20th-century windfall, birthed by winning two world wars that eliminated global industrial capacity and allowed the US to exist indulgently unoptimized. It was an evanescent period touted as a new normal. From 1947-1973, median household income almost doubled in real terms. High school graduates could buy homes and raise families on single incomes. No more. No wonder boomers have such a decadent worldview; they were raised in perhaps the most luxurious timeline in human history.
This was circumstance, not structure. A fortunate window we mistook for the natural order’s architecture. We did not defend it and took it for granted. Power laws reassert themselves as these fleeting factors decline.
Globalization’s emergent extremes are to be expected when technology neuters geographic boundaries, facilitates colossal scale, and nothing is done to preserve the center of the societal barbell. The middle isn’t dismantled by intention, but inaction.
Ascribe nefarious motivations to politicians or corporations if it satisfies your need for villains, but power laws require no conspiracy. When supply chains, trade, and labor compete globally with minimal friction, the bottom-up’s scaling physics and Turbo Matthew dictate outcomes.
To demonstrate, consider how monopolies form. The policy prescription is simple:
Do nothing.
That’s it. Don’t regulate mergers. Don’t enforce antitrust. Don’t tax. Just let the natural system percolate and Matthew handles the rest. Only through persistent intervention to thwart this organic phenomenon can we cultivate an artificial non-equilibrium situation with a healthy center.
Power law distributions are stable equilibria under competitive optimization. Markets engineer for gross margins and efficiency; they care not for societal cohesion, political stability, and working-class quality of life. Those are human values that oblige human interference.
Monopolies at the business level emerge when you do nothing.
Monopolies at the societal level (oligarchies) emerge when you do nothing.
We are subject to the same laws of physical reality at every substrate of existence.
If power laws at civilizational scale are inevitable without intervention, what should that intervention look like?
X. Societal Implications: Tradeoffs All the Way Down
Societal polarization and volatility rise as power laws run wild and you’re left with a ruling class, lower class, and nobody else. Dichotomy is destiny. Without a middle, all that’s left are extremes.
Permitting this is no different from not having antitrust laws for business. You just have to zoom out to see it.
Without antitrust, monopolies appear naturally. Extremes accumulate organically. Business enhanced by technology, with open borders for capital and labor, incubates monopoly outcomes for nations: giga corporations and niche micro businesses alongside peasants and patricians.
We know this doesn’t help the consumer, so why would it help the country?
When the FTC reviews antitrust action, it asks: does this substantially lessen competition and harm consumers? Swap “consumers” for “citizens” and the parallel clarifies. This is market logic at national scale. Same process, different domain. Fractals.
Apply the antitrust assessment bottom-up systemically at the aggregate level of nation, not the top-down individual review of each business.
Does an economy of multi-trillion goliaths and bodega-style indies help or harm the nation’s consuming core? Amazon offers lower prices (consumer surplus rises over the short term) while hollowing out blue-collar jobs and wages (middle class shrinks, surplus declines over the long term). You are trading one form of harm for another.
Traditional policy tools are palliative, not curative; they address symptoms rather than causes:
Progressive taxation/wealth redistribution ameliorate capital concentration but cannot stop the foundational agents that sustain it.
US top marginal rates dropped from 91% (1950s) to 37% today. The top 1% held ~23% of wealth in the 1980s and ~35% by 2020s. The tax code helps, but redistribution doesn’t confront root causes and serves mainly as a partisan distraction.
Antitrust enforcement breaks up monopolies but doesn’t eliminate network effects (or political power) that enshrine king-maker paradigms. Break up Meta tomorrow, and within five years a new social platform congeals through parallel dynamics.
These interventions manage outcomes, not mechanisms.
They trim visible branches without touching the veiled oak tree beneath. Kings remain the same kings if all they do is tithe some jewelry each year.
To concretely address distal causes, not paper over proximate ones, we must target bottom-up substrates, not shift resources around with top-down political stories.
Possible approaches:
1. Change optimization functions
The NBA modified rules before: shot clock in 1954 (to prevent stalling), three-point line in 1979 (to open spacing). It could change them again. Push 3’s back or alter point values so expected values equalize. Shot selection would rebalance within a season.
Platforms could modify algorithms to stymie Turbo Matthew results (revise engagement metrics to value diversity over virality).
Meaningful reform requires regulation. Big Tech has no incentive to cap its own growth. Asking Meta to de-escalate its network effects is like asking Exxon to cap its oil drilling.
2. Change the tax code: target dominance, not income
Tax systems could penalize market dominance at the network level, not the profit level. Don’t target net income (specious, easily manipulated), focus on Matthew signifiers.
Example: Above 25% market concentration, implement graduated levies
25-40% market share: 2% tax per percentage point on revenue attributable to that share
Above 40% market share: 10% tax per percentage point
Example: Say Amazon has 45% of US e-commerce market (irl it’s about 40%)
First 25%: No network tax
Points 26-40% (15 points): 2% × 15 = 30% tax on that revenue segment
Points 41-45% (5 points): 10% × 5 = 50% tax on revenue segment
Result: Amazon faces 30-50% tax on market share over 25%
And yes, I mean revenue, not net income. The topline is the target. You have nowhere to hide. Your accountants cannot save you.
This targets the mechanism (Matthew principles that entrench dominance) rather than the outcome (profits that may reflect genuine innovation), creating notable network-effect frictions without confiscatory taking.
This curtails platform efficiency and possibly innovation velocity. That’s the point. You’re buying political stability with economic productivity.
Yes there are flaws (tradeoffs!) to this, but the thinking is what’s important: it targets structure, not symptoms.
You can still get big, but you can’t get infinite. Expansion becomes prohibitively expensive at the roots, not branches. Dominance has a price, because the middle has value.
3. Create deliberate scarcities in the middle:
A) Intellectual property protections only for products generating under $150M annually in sales. Beyond that threshold, patents expire and become public domain.
This is a Threshold Cliff that doesn’t punish the middle, only the hyperscale. Disney’s Mickey Mouse copyright would expire. Generic drug manufacturers could produce majors immediately instead of waiting decades.
Small pharma could still patent new compounds and profit for years before hitting the threshold.
There would need to be exceptions for investment size:
Developing a new drug can cost billions and take 10-15 years; no company accepts that risk if the patent expires when the product succeeds without recouping costs + substantial returns. To address this, we could have both a minimum ROI hurdle and annual sales threshold.
B) Make geography great again. Limit procurement to producers within 500 miles. Ohio school districts buy Ohio-manufactured desks, not Chinese imports.
This involves federal legislation that every economist in Washington would despise and every voter in Pittsburgh would adore. Trump’s 2016 coalition understood this instinctively. Spreadsheet People do not bear the consequences of their decisions, Detroit does.
I’m acutely aware of the ugly tradeoffs here. I’ll acknowledge them when Ivory Tower minds acknowledge the strengths. Pick your poison, spreadsheet boy.
Geography is irrelevant, information is immediate and free; this has never existed before. Unique circumstances demand unique thinking.
C) Provide government subsidies or tax breaks for intermediate-scale operations: manufacturers with 50-500 employees, media companies with 10-100 staff, etc.
A variation of Threshold Cliff, redesigned so it’s not a discrete boundary but a gradual gradient. I shall call this the Gradient Slope.
A Gradient Slope: benefits decline smoothly as you scale, not via cliff-like step that’s punitive to intermediate players.
E.g. companies with 100 employees get 20% tax credits. At 200 employees: 15%. At 400 employees: 5%. At 500+: 0%.
Germany’s Mittelstand policies follow this model, sustaining millions of mid-sized positions. It works, at the cost of some efficiency.
Germany actively subsidizes mid-sized firms through tax policy, apprenticeship systems, and patient capital from regional banks. Protecting the middle requires intent.
The Mittelstand is not a natural exception, but a consciously crafted one.
The challenge: these solutions impede competitive dynamics. They stipulate ongoing intervention against natural market forces. Stop intervening for a decade, and the middle gets eaten. This is less a policy you implement than a commitment you maintain.
You’re buying social stability with economic efficiency. Whether that trade makes sense depends on how much you fear pitchforks relative to GDP evangelism.
There is no scientific answer here; it’s not a matter of ‘right’ or ‘wrong’. You cannot optimize for both the middle’s stability and the tail’s efficiency. You must pick your poison; you don’t get to say your path has no toxins.
4. Pros and cons of middle preservation:
Every nation faces the same power laws. You’re choosing how to respond, not whether to respond.
Con: Lower productivity. Near/mid-term efficiency weakness vs those who let power laws run unchecked.
Pro: Stronger domestic demand over time (middle supports domestic consumption), political moderation, and cultural stability.
The productivity cons at the beginning are quantitatively apparent, but the benefits over longer timelines must be qualitatively factored in. Political stability and societal cohesion don’t initially show up in Excel sheets, but eventually they impact them.
Radical inequality should be understood as a structural issue, not ideological moralizing talking point. I do not care about things being equal as any moral stance or personal belief that everyone deserves the same. We are not equal. We are not all the same. I cannot stand fictions that pretend otherwise.
My concern with inequality is one of violent volatility, not fairness; extreme wealth gradients destabilize societies the way extreme heat gradients birth hurricanes. It’s gradients that matter for system dynamics. The same cosmic forces that govern stars and snowflakes also govern societies. We must stop telling stories and start seeing structure.
Bottom up.
XI. Universal Basic Income
When power-law economies strand large populations with no economic role, you either pay/subsidize them directly or you get revolt. There’s no third option that preserves both market dynamics and civil order. Which dystopia, Western man?
Every intervention I can imagine — network taxes, regional procurement mandates, IP restrictions — will either be kinda toothless or kinda totalitarian. There really isn’t an in-between.
Tax network effects above 25% market share? Platforms will jurisdictionally arbitrage or restructure to stay below cutoffs. Mandate regional procurement? You’ve just imposed Soviet-style limits, sacrificing comparative advantage for parochial protectionism. Deny IP rights for titan corporations? Congratulations, you’ve decimated incentives for the kind of R&D that relies on massive investment.
Implementing Matthew Regulations means relying on government to outmaneuver physics-style forces and trillion-dollar optimization engines, and, well, we all know how brilliant and capable bureaucrats can be.
Yet another unpalatable option that may be unavoidable: allow power law dichotomization and provide society-wide safety nets, otherwise known as universal basic income (UBI). Either subsidize existence or drown in instability. There is no situation where ~30% of the population has no role, no recourse, and we have a safe place to live. The market offers no solution to this.
I viscerally hate UBI on an ideological level. However the analysis here is concerned with stability and sustainability, not ‘does this suit my political priors?’. If we let Turbo Matthew run rampant with no economic floor, what exactly is the plan when ~1/3 of society occupies unemployable poverty? Learn to code! Oh wait, the LLM does that now.
A tragic symmetry: the same networks that generate exponential upside also generate exponential human redundancy. UBI is the price of letting The Machine run hot.
I dislike UBI for the same reason it seems inexorable: it medicates the symptom (mass unemployment) without curing the cumulative-advantage disease. But any root-cause solution involves intervention so aggressive it threatens the dynamism producing wealth in the first place.
UBI represents pragmatic intellectual surrender. It concedes we can’t sustainably solve these multilayered middle-eating problems, so we’ll pay people to exist. Because the alternative — obstructing tech platforms, banning algorithmic amplification, restricting AI — requires state power that would make the New Deal look libertarian.
There is no panacea. These solutions have flaws. Yours will too. Political decisions are collective-action problems expressing moral judgements, and they all involve tradeoffs.
Prioritizing some values necessarily means deprioritizing others. All moral frameworks have shortcomings, because no set of values is comprehensive. Every political belief we espouse does this implicitly. No exceptions.
If you only understand the world in net-income terms, fine. Economic efficiency is a value, not the value. Markets excel at optimization but cannot adjudicate between competing social priorities. Whether you genuflect to GDP over political stability, innovation over community, consumer surplus over producer dignity: these are moral decisions masquerading as technical or political ones.
Every efficiency claim embeds a tacit value statement. Your personal priorities are revealed by what you promote.
If your political stance distills to “no barriers, no borders, free trade”, you’re confessing that gross margins exhaust your moral faculties. You see one metric, cost, as the remedy for human flourishing. Don’t get me wrong, it helps, but does so in a unidimensional way.
The economist who perceives productivity as the sole metric has mistaken his personal values for objective truth. Claiming ‘efficiency is neutral’ while ignoring distributional consequences is a moral stance dressed in technocratic drag. His entire personality is one big optimization function.
These are people so uncomfortable with subjective judgement they’ve outsourced their ethics to Excel. A myopic quarterly earnings mentality for a game measured in generations. A one-track mind that only knows what the calculator tells him is ill-suited to lecture on collective-action decisions for civilization. Your opinions are cool for Nike, less so for a nation.
The McKinsey answer isn’t irrelevant, just woefully incomplete. The consultant with a 47-slide deck and the New Jersey construction worker who senses something’s broken are both communicating parts of reality. One has data and theories, the other has skin in the game and ramifications. Neither has the full picture, but only one gets paid $600/hour to be wrong.
The economist’s spreadsheet measures productivity gains but not political polarization, GDP growth but not the latent rage of the structurally obsolete. Maximizing one variable while ignoring all others is how you optimize your way into revolution. Neat expected-value thinking man, shame a country isn’t a shopping mall.
Efficiency and productivity are some values, not the values. There is no ‘right’ answer, only tradeoffs that expose priorities.
XII. Concluding
What happens to the individual impacts the business; what happens to the business shapes the country. All of these elements are entwined, inseparable, and subject to the same laws and fractal contagions. The barbell cascades.
A powerhouse middle was a confluence of serendipitous historical circumstance: industrial dominance, geographic fortresses, information asymmetries, and regulatory protections. Technology has melted the first three, and only the last one remains.
You can either accept polarization in favor of optimization and growth, or implement systemic, perpetual interventions to counteract Matthew’s will. You cannot solve this once and forget it. Skewed distributions are where systems settle; maintaining healthy middles entails enduring effort. Sie müssen sich der Mittelstand verschreiben.
Don’t forget:
How do you get monopolies? Do nothing.
How do you get oligarchy? Do nothing.
Does the cost of intervention (diminished efficiency, slower growth, competitive disadvantage) exceed the social cost of severe inequality (political disruption, civil unrest, loss of middle-class opportunity)? Your answer elucidates your priorities. Not what’s right, just what you value.
Middles are a garden in the desert: stop watering them and the sand wins. The social, political, and economic center are not self-sustaining against the sand’s equilibrium.
The middle serves as more than economic category; it’s a social shock absorber, political counterweight, cultural moderator, and civilizational bedrock. It gives people stakes worth defending without forcing them to choose between predator and prey. When you treat a country not as a home but a shopping mall, running it for margins and returns, you make its social fabric brittle, fragile, and vulnerable to revolution’s deterioration. Brittle systems don’t bend under pressure, they break.
If the center dies, don’t be surprised when the population divides into zealots and nihilists. Eliminate conditions that make moderation adaptive, and moderation becomes a losing strategy. Extreme environments beget extreme beliefs, and extreme behaviors follow suit.
By definition and essence, the middle is what keeps society from splitting in two.
Subscribes and shares are much appreciated. If you enjoyed this essay, please give it a like and share with someone else.
I’m building something interesting, visit Salutary.io.
You can show your appreciation by becoming a paid subscriber, or donating here: 0x9C828E8EeCe7a339bBe90A44bB096b20a4F1BE2B
Research Appendix
Content:
Additional Business Case Studies
Music, CDs, & Streaming (15 years)
Book Publishing (12 years)
Casual Dining (14 years)
Beer (70+ years)
Condensed Case Studies: Politics, Fitness, & Academia
Counterexamples: Extremists in Middle Disguise
1. Additional Business Case Studies
Music, CDs, & Streaming
THEN (CD Era, 1990s-2007):
What the middle looked like: Healthy middle class of working artists between bar bands and superstars. Music generated ~$23B at the Carson Daly 1999 peak (inflation-adjusted), with ~$21B of it from CD sales.
Disc sales had natural price points ($12-17 per album) where artists with 50k-100k annual album sales cleared $150k-300k income: a viable middle-class career.
Market structure: Balanced distribution with megastars, mid-tier artists, and up-and-comers forming recognizable tiers. Physical retail (Tower Records, Virgin, Borders) and radio offered multiple revenue streams alongside regional touring.
Who thrived in the middle: Mid-tier artists sustained careers through album sales, radio rotation, and touring. Physical album economics ($12-17 × 50K-100K sales = $600K-1.7M revenue) supported professional musicians. Music industry facilitated artist development through multi-album deals, allowing careers to build gradually.
Why it seemed stable: Album price floors supported mid-tier viability. Multiple revenue streams (sales, touring, radio) diversified income. Retail and radio provided discovery paths. Geographic diversity in touring created stable income base.
If you read the movie industry case study, you may be sensing some parallels between CDs and DVDs…
NOW (2025):
What happened to the middle: Revenue and attention dramatically concentrated at the top as streaming replaced physical albums. By 2022, streaming was 84% of industry revenue ($13.3B of $15.9B).
Top 1% of artists garner 77% of all revenue; top 1% drove ~90% of streams. Of 2M artists on Spotify, only ~22k (~1%) earn over $50k.
Current structure: Matthew heavy tails. Megastars command overwhelming majority through algorithmic amplification. Long tail of hobbyist artists get negligible income. Blue-collar working musician crushed.
Middle casualties: Artists who sustained careers with 50k-100k album sales now need millions of monthly streams.
Musicians previously earning $150K-300K annually with albums now log $3k-5k monthly if they hit ~800K-1M monthly streams.
Artists earn ~$0.003-0.005 gross per stream, requiring ~400k monthly plays just to earn about minimum wage (.004 x 400k = $1,600 x 12 = $19.2k).
Time to collapse: Healthy normal curve to concentrated skew in 15 years (2008-2023).
MECHANISM: Preferential Attachment + Algorithmic Amplification + Smile Curve
Spotify pools all subscription revenue into a single pot, then allocates based on each artist’s share of total platform streams, not what individual subscribers actually listen to.
Example: if you exclusively stream jazz artists but 90% of Spotify streams are pop, then 90% of your subscription dollars flow to pop artists. The system socializes revenue toward the already popular, creating a subsidy from niche listeners to mainstream artists.
Physical retail allowed serendipitous shelf-based discovery and geographic diversity; whereas algorithmic discovery is mediated not by happenstance or store shelves, but black-box calculations calibrated for engagement. Winner-take-all dynamics surfaced when playlist placement led to massive amplification for top artists.
Mid-tier artists lack breakthrough hits for algo momentum and face comparable production costs as megastars. The smile curve smirks: platforms extract chunky distribution margin, stars snatch attention premium through existing fame, and the middle gets crumbs producing quality content few see. The endpoints get fat while the center starves.
Book Publishing
THEN (Pre-2007):
What the middle looked like: Publishing sustained a midlist, with authors who weren’t stars earning steady advances and royalties off bookstore shelf space (Borders/Barnes & Noble + local shops), library/coursepack licenses, and backlist sales.
Midlist authors could sustain careers through moderate sales of 5k-50k and reliable advances ($10k-75k).
Market structure: Balanced distribution with bestsellers, solid midlist, and debut authors. Physical bookstores provided shelf space for discovery and regional markets.
Who thrived in the middle: Midlist authors earning $30k-80k through advances, royalties, and licensing. Writers of literary fiction, genre novels, narrative nonfiction sustained careers without bestseller status. Publishers invested in author development through multi-book deals.
A midlist author in 2000 might earn a $25k advance, sell 15k copies, generating $36k in royalties. Add $5k from coursepacks and subsidiary rights, total: $66k
Why it seemed stable: Retail shelves created discovery that protected the midlist from algorithmic curse. Bookstore geo-diversity prevented winner-take-all concentration. Libraries purchased broadly across titles. Multiple revenue streams (retail, libraries, licensing) diversified income base.
NOW (2019-2025):
What happened to the middle: Multiple simultaneous optimization shocks (2007-2019) murdered the midlist.
By 2019, Amazon had ~77% of online book sales (print + ebook). One store = one algorithm = winner-take-most discovery.
Median author income fell 42% from 2009 to $6k by 2017. ~25% of published authors earned $0.
The Kindle launched at $399 in 2007; by 2011 it was $79. Ebook prices dropped from $12-15 to $3-10. Readers paid less, publishers earned less, authors earned least.
Current structure: Fat barbell. Publishers optimize their launch budgets for star authors and celebrity IP (Stephen King, celebrity memoirs, airport bookstore slop) that can pierce algorithmic gravity.
High-velocity indie authors exploit subscription model through rapid-release (6-12 books/year) in popular genres building direct reader communities.
Midlist authors trapped between traditional publishers who won’t market them and indie economics requiring velocity incompatible with quality literary work.
Middle casualties: In 2007, traditional publishers released 276k books; in 2018, self-publishing flooded the market with 1.68M books. Scarcity shifted from distribution to attention.
Borders filed bankruptcy in 2011, eliminating ~1,200 stores. Barnes & Noble shrunk from ~730 locations to 600. Shelf space, the physical element thwarting king-makers, vanished.
Fair-use rulings gutted the licensing revenue midlisters relied on.
Google Books and university coursepack cases meant professors could legally share excerpts they previously paid for. These were small payments individually, but collectively kept writers afloat.
Tech companies celebrated this as “democratizing knowledge”; a platitude that means “writers now get less and the delta accrues to me”.
Time to collapse: ~12 years (2007-2019) from Kindle launch through multiple optimization shocks to Authors Guild survey documenting 42% income decline.
MECHANISM: Smile Curve Disaggregation + Algorithmic Amplification
Casual Dining & Restaurants
THEN (1990s-2013):
What the middle looked like: Casual dining chains had ~$100B in sales at 2013 peak
Applebee’s, Chili’s, TGI Friday’s, Red Lobster, Olive Garden thrived serving affordable experience between fast food ($5-8/meal) and fine dining ($50+/meal).
Market structure: Balanced three-tier system with fast food (efficiency), casual dining (moderate experience), and fine dining (premium) serving distinct segments. Sector maintained growth trajectory with casual dining positioned as treat for middle-class families.
Who thrived in the middle: National chains expanded through early 2000s with ~$20-30 entrees, consistent quality, and experience unavailable in fast food.
Applebee’s peaked at 2k locations in 2007. TGI Fridays hit 600 in 2008. 700 for Red Lobster. These were community stalwarts: Friday night at Chili’s, birthday dinners at Olive Garden, Little League celebrations at Applebee’s.
A family of four spent ~$70-80 including tip. Affordable enough for regular visits, nice enough to feel like an occasion.
Why it seemed stable: The middle class visited reliably. Alcohol subsidized food margins. The model worked as long as people had disposable income and nowhere better to go.
NOW (2025):
What happened to the middle: The middle became too expensive for everyday dining and not nice enough for special occasions. Sector bled traffic for 13-14 consecutive years as of 2021. Sales crashed to $56.6B by 2023: a 43% decline from 2013 peak in just a decade. Nearly 350 chain restaurants closed in 2024 alone.
Current structure: Barbell distribution. Growth accrues to fast-casual (Chipotle, Panera) offering comparable quality at 30-40% lower prices with speed.
Fine dining and Instagram-worthy independent restaurants attract experience-seeking segment.
Value-oriented survivors reposition down market: e.g. Chili’s had three straight quarters of double-digit growth after $400M investment in discount positioning with $10.99 “3 for Me” deal.
Middle casualties: Red Lobster filed Chapter 11 in 2024, depriving the world of the Cheddar Bay Biscuit. TGI Fridays went from 600 restaurants to under 300 in six months before bankruptcy. Hooters closed ~70 locations. Applebee’s went from ~2k to ~1.6k. Each closure eliminates 40-80 blue-collar jobs.
Time to collapse: About 14 years of consecutive traffic losses (2008-2021)
MECHANISM: Preference Bifurcation + Inverted-U Cost Curve + Optimization Pressures
Increasing income inequality drove consumer behavior that followed suit: budget-conscious diners optimized for value (fast-casual), while affluent consumers traded up to experiential fine dining.
Fast-casual uses supply chain efficiency and simplified menus to deliver Denny’s quality at fast-food prices. The middle bore full operating costs (staff, facilities, table service) without pricing power of fine dining or operational efficiency of fast-casual.
Moderate prices, moderate selection, and moderate experience only sells to a moderate consumer base.
Beer & Breweries
THEN (Pre-1980):
What the middle looked like: In 1947, America had 421 macrobreweries. Most were regional: Rainier in Seattle, Lone Star in Texas, Narragansett in Rhode Island. Beer was historically local; transport costs and distribution constraints favored regional brewers.
Regional operations (15k-6M barrels/year) served specific geographic markets.
Market structure: Reasonably balanced with national brands, regionals, and hyperlocal operations.
Why it seemed stable: Transportation costs were a moat, mid-sized breweries were operationally efficient. Beer as local product and regional identity seemed inherent to category.
Beer is heavy, perishable, and expensive to ship. A barrel from Milwaukee costs more delivered to San Francisco than a local pub charged at the tap. Regional breweries owned their geography by default.
NOW (2025):
What happened to the middle: Consolidation crushed regional breweries as market bifurcated. By 2006, only 20 US macrobreweries remained, a 95% decline from 421 peak.
Craft beer revival beginning ~1980-1985 drove chasm between big and small. By 2024, 9.6k crafts operated producing 23M barrels.
Top 2 macrobrewers control ~55% of the market, top 4 control 74%, top 10 control 88%, materially exceeding typical manufacturing concentration where top 4 average 36% share (2012 data).
In Europe, the top 4, depending on industry/product, average 41-61%. Global research found average of ~45% share for the top 4.
Current structure: At one tail, multinational consolidation birthed brewing behemoths (AB InBev, SABMiller, Heineken) who approach beer as commodity logistics. At the other, craft breweries lean on nimble uniqueness: direct sales, premium pricing, and local authenticity.
Regional breweries that once dominated realized they were the wrong size: too large to credibly claim craft authenticity, too small to match industrial efficiency.
Middle casualties: Regionals died slowly then all at once. Facing horrible economics and unable to compete with macro scale or craft authenticity and $12 six-packs.
Olympia Beer (Washington’s third-largest brewery in 1980) was absorbed by Pabst. Schaefer (New York’s largest brewery in 1950) went to Pabst. Texas’ Lone Star… you guessed it, Pabst.
Time to collapse: ~70 years. 1947-1980 consolidation to 80 breweries, then 40+ years toward barbell market (1980-2025).
MECHANISM: Scale Economies + Preference Bifurcation + Distribution Network Effects
Technology in brewing and distribution (automation, refrigeration, packaging) made scaling possible, then unbeatable. High fixed costs in production, logistics, and marketing drove consolidation.
Regional brewers were too big to be charming, too small to be cheap.
2. Condensed Case Studies: Cities, Politics, Fitness, & Academia
Mid-sized cities lose population to megacities (optimizing for career, cultural amenities, network effects) and small towns (optimizing for cost of living, community, quality of life).
The middle offers neither the opportunities of big cities nor the affordability and charm of small towns. Mostly traffic, sprawl, and strip malls.
Moderate political parties/candidates get crushed in politically polarizing settings.
Coalition governments in Europe show center parties losing share to extremes. Primary systems and partisan media fan these flames.
The “sensible center” assumed to represent the largest constituency has no passionate advocates, losing to intense minorities on both sides.
Fitness Industry: Mid-market full-service gyms are being squeezed between budget chains and boutique studios. The middle bears full facility costs but lacks both price leadership and experience/status differentiation.
Budget chains design for price:
Planet Fitness scaled to 2,762 locations at $10-20/month through brute simplification: no pools, no courts, minimal staff.
Boutiques monetize community and social status:
SoulCycle’s $42/class snags the premium demographic. The Hermes of fitness.
Mid-market clubs ($40-70/month) offer neither discount nor prestige:
Gold’s Gym and Town Sports International filed bankruptcy in 2020.
Academia: Elite R1 institutions and small teaching colleges survive while mid-tier research universities struggle. The middle bears research expectations without elite resources or teaching-focused mission clarity.
Elite R1s dominate hiring and funding:
Top 5 universities (Berkeley, Harvard, Michigan, Wisconsin-Madison, Stanford) produce one in eight tenure-track faculty nationwide (13.8%).
Top 100 institutions receive nearly 80% of federal funding for science and engineering research.
Johns Hopkins alone gets 2.4% of the entire National Institutes of Health (NIH) grant pool.
Faculty trained at mid-tier institutions leave academia at nearly 2x the rate of elite-trained PhDs.
Small teaching colleges avoid the squeeze:
Liberal arts colleges and regional state schools optimize for undergraduate education, avoiding research cost structures while maintaining stable enrollment.
The same dynamics that killed department stores are now eating universities (Ivy League or community college, but not state schools), healthcare (concierge medicine or Medicaid, but not employer PPOs), and media (Netflix or TikTok, but not cable).
Each industry thinks it’s special. None are.
3. Counterexamples: Extremists in Middle Disguise
What about Costco thriving between Walmart and Whole Foods? Or Southwest Airlines surviving between Spirit and Delta?
These aren’t counterexamples so much as confirmation: each carved out a distinct specialty focus and pursued it obsessively.
Costco optimized for a single objective: maximize inventory turnover with low-margin bulk purchases, backed by membership-funded fixed costs.
Costco charges $60-130/year for membership to shop there. It has 76.2M paying members, earning $4.8B/year from just the right to walk into the store.
This subsidizes operating expenses, allowing razor-thin product margins (low prices!)
Members prepay for access, inverting the retail model.
Southwest looks middle-market only if you squint at prices. Structurally, it’s an extremist: one aircraft type, relentless turnaround times (very high asset utilization), point-to-point unique flight routes. Fanatical optimizers.
Successful ‘middle’ players are extremists who found unoccupied territory. They appear moderate only through superficial price comparison. Structurally, they’re often as focused as any tail player, calibrated for variables others ignored.
There are many more examples than what I’ve provided; even things like the stock market and liquidity are subject to it. An omnipresent gravitational force, impacting all of existence.
Thank you for reading.
What an intellectual achievement this is. Incredible work sir. I will be thinking this through for the next little while. I'm not yet sure if I agree with everything here, but this piece has already impacted my thought in a profound way.
Anybody got any bright ideas for getting the analysis + action items read by those who wield enough influence to act on them?