Intro
Dan Rasmussen, founder of Verdad Advisors, explains why the private equity unwind runs through software deals rather than car washes, why he would rather own the credit than the equity on those companies, and what he thinks today's AI capital spending shares with earlier blowups. Kai Wu brings his own research on intangible value, software moats and job postings to it.
Guest: Dan Rasmussen, founder and managing partner of Verdad Advisors and author of The Humble Investor
Host: Kai Wu
Published: 27 August 2026 on Excess Returns
Episode page | 57 min
Key Takeaways
Private equity was the ultimate consensus trade and consensus trades end badly
"purchase prices in private equity reached almost probably 20 times EBITDA when public markets were trading at probably 15 and small caps at 12"
Rasmussen calls the result a massive train wreck, and says it was eminently predictable
On the same assets, Rasmussen would rather own the private credit than the private equity
Companies bought at 20 times with half of it debt, paying 10 to 14 percent interest
The listed LP fund interests destroy private equity's low-volatility story
"their volatility is like 24, 25% annualized, which is like just a little bit more than small cap"
They now trade at what he puts at a 30 to 40 percent discount to NAV
Private equity became a software business without anyone deciding to
"By my estimate probably 40% plus of private equity by 2021 or so was software or health care technology which was basically healthcare software"
Wu's test for which software survives AI is what the moat was made of
"if your if your moat was your code, then I'd be worried about you"
No AI job losses have shown up yet, and Rasmussen thinks the capex needs them to
Waymo is his one clear displacement; call centers and cleaning are not
The AI build-out is being run by people whose worldview already produced blowups
"everyone always overinvests. The returns are always lower than people anticipated. The returns always come later than people thought"
AI is brilliant at what intellectuals want and unreliable at what companies need
Value broke in the US because the US is the most intangible economy in the world
Japan's book value is tangible book value, and it still predicts returns there
Japan's governance reform is a mechanical push back to book value, and it is happening
"you can build portfolios that trade at like 0.5 or 0.6 times book that are actually profitable, cash flow generative, and growing"
The TOPIX gives you almost none of it, because everything below book is small cap
Biotech can be screened on R&D spend, specialist ownership and short interest
Both of Rasmussen's big non-consensus calls have become consensus, and he says he needs new ones
Where the Private Equity Unwind Actually Stands
Rasmussen treats private markets as a textbook fad rather than a special case: "investment fads rarely end well", and when everyone agrees on something you can almost bet the agreement is followed by subpar returns
"private equity became the ultimate consensus trade" — every endowment, foundation, RIA and family office parroting the same talking points about the best-performing asset class, and the money flowed in
The prices are the whole story: "purchase prices in private equity reached almost probably 20 times EBITDA when public markets were trading at probably 15 and small caps at 12"
The mechanism worked only while money kept arriving — "when money stopped flowing in, you had to sell it to somebody real. And turns out nobody else wanted to pay those crazy prices"
There is now a backlog of assets the funds cannot sell, held together by continuation vehicles and what he calls other weird financing tricks, so investors from the peak years cannot exit either
"I think the whole thing has become a massive train wreck. And I think it was eminently predictable" — though he adds it was also highly non-consensus at the time
Wu asks how an observer would work out the true mark on these assets, and how long it takes before the vehicles have to end and the money goes back
Why He Would Rather Own the Debt Than the Equity
The debt stack is the first problem: "they didn't just pay 20 times — they paid 20 times and half that was debt"
That equity sits subordinated to private debt paying 10, 12, 14 percent depending on how risky the asset was, and Rasmussen says there are not many micro-cap companies earning a 10 to 14 percent return on assets to cover it
His actual bet: "if I had to bet which is going to do better over the next five years, private credit or private equity, I'd be putting my money in private credit" — at least the credit is getting cash flow back
He is watching debt servicing and refinancings for what he calls creditor-on-equity violence, and thinks the debt side holds the leverage right now
The London-Listed Funds That Broke the Volatility Story
Part of the appeal was reporting: quarterly marks, no sudden drawdown, never down more than the market, and delivered a month after the LP's own reporting to their boss was due
On reported numbers private equity looked about as volatile as investment grade bonds, which Rasmussen says is why everyone thought it was safe
The test he ran with Richard Ennis was deliberately simple: London lists real LP interests in private equity funds — HarbourVest and Partners Group among them — rather than management companies like Blackstone or KKR, so they are mark-to-market private equity
"their volatility is like 24, 25% annualized, which is like just a little bit more than small cap" — which he says makes sense, since levered micro cap should be more volatile
Wu adds that the same vehicles show what buyers think of the marks, and Rasmussen agrees: "now they trade like closed end funds with a you know 30, 40% discount to NAV which is probably where private equity should trade in aggregate"
Those vehicles are doing horribly this year, alongside private credit
How Private Equity Talked Itself Into Software
Rasmussen relays a line he credits to someone else: "investment committees are momentum investors of the three-year lag" — allocators add to whatever has the best trailing three-year returns and cut the worst, which is why he expects private equity allocations to keep falling
Fund partners run the same momentum on a longer clock: making partner means doing a deal and exiting it inside four or five years, so power on the investment committee belongs to whoever did the deals that now look like the fund's biggest winners
The people who did the bad deals get fired, which lets the firm fundraise on pro-forma numbers that exclude them
Trace that cycle back and 2014-15 was energy private equity — SCF and Lime Rock, money into shale — which blew up in 2015 and 2016 and was recast afterward as a carveout that was never core strategy
Allocators then looked at Thoma Bravo and Vista's 2014-16 returns and decided software was the thing to copy; those deals looked good in 2018 and 2019, and COVID pushed the multiples higher still
"By my estimate probably 40% plus of private equity by 2021 or so was software or health care technology which was basically healthcare software" — bought at what he calls crazy prices under a software-is-eating-the-world narrative
The catch is what they were actually buying: "private equity wasn't taking private Salesforce or Adobe" — the targets were subscale, like a company that had cornered car dealership software in the tri-state area
He doubts those companies had an IP moat or a talent moat at all, as against being generic software businesses with high share of a niche
The Moat Question: What Happens When the Moat Was Code
Wu says he is not a private markets expert, but the roll-up story sells easily to an LP: if you roll up car washes, why not roll up the software the car wash uses
The public comparables have already repriced: "these stocks are down 50 to 80%" — and Wu's examples, Salesforce and Adobe, are the best-positioned names in the group
Wu's one-line answer, from research he did in May on the software selloff: "if your if your moat was your code, then I'd be worried about you"
The survivors are the ones whose moat was the customer relationship — the old IBM position, with tendrils in enterprise America
Some niche vertical players may hold a regulatory moat built through lobbying and regulatory capture, which he says is hard to displace
He expects size to protect: software serving the biggest enterprises, where workflows are deeply embedded, is safer than a consumer app anyone can vibe code in a day
He compares it to a recession like 2008, when small caps get hit first
Rasmussen applies the same arithmetic to the levered version: "if the public software companies are down 50 to 80%. And these private equity backed things are 50% levered"
"Like that's bankruptcy basically, right?" — which is why he thinks private credit is better positioned than private equity on the same assets
He thinks the cash flows may be fine at some of these businesses for a while; it is the equity value and the valuation multiples that got smoked
Private Credit, ARR Loans and a Collateral Mismatch
Wu's framing: intangibles are hard to use as collateral — nobody lends against your patents, your human capital or your brand — so traditional banks would not do much debt financing for software
He asks whether private credit is what made the migration into software possible at all, because someone had to be willing to lend against recurring revenue
Rasmussen agrees the tip-of-the-spear product was exactly that: "ARR loans annual recurring revenue loans", pitched at software businesses banks would not touch
His objection is four words long: "newspapers were recurring revenue" — it did not stop them blowing up, and people can cancel a contract or stop paying if they no longer want the product
Wu calls the structural problem a mismatch: debt capital is meant to seek stable, predictable cash flows, while technology is by definition the frontier of innovation and exposed to obsolescence
He thinks the revenues probably will be stable and bleed out slowly rather than vanish overnight, which is what makes the risk easy to mask
He separates the structure from the outcome: everything may get bailed out and look like a false panic in hindsight
Measuring Obsolescence Risk Before It Reaches the Numbers
Rasmussen turns the question back on Wu: is there a quantitative way to tell whether a company's intangible value is about to become obsolete
Wu says the obvious cases are easy — a ChatGPT wrapper could go away in a day, and Wispr Flow, which he uses heavily, could become a feature on iOS
His analogy is the hundred auto companies that launched to exploit the invention of the car, 99 of which failed
The hard case is the newspaper case, where an incumbent is disrupted rather than out-competed
Reading what CEOs said during the internet is unreliable, because many simply refused to acknowledge it, which is part of why they were disrupted
He goes to third-party and sell-side research instead, to see what people outside the company say about its exposure
Wu's two-step test: first work out which sectors are exposed — freelance web development and outsourced IT in India are in the crosshairs of AI, laying cement much less so — then sort the companies inside them into survivors and casualties
His survivor example is the New York Times: one of the top three or four papers in the world, with a business model it changed for the internet, the games, and management that navigated the period
Walmart plays the same role in e-commerce, a company that started behind and caught up
Job postings are his live indicator: a company's mix of substitutable versus complementary roles, tracked through time, shows whether it is shifting hiring toward adaptation
Rasmussen adds the capital constraint that makes this a private equity problem: pivoting requires investment, and "if you've got massive amount of debt, where's the money going to come from to like build new AI tools"
The AI Job Losses That Have Not Happened
Rasmussen starts from a model he credits to Wu: the benefit of an innovation has to reach the people consuming it, after paying the producer's margin, and for AI the promised benefit is replacing people and taking out cost
"we just haven't seen any AI-related job losses yet" — and he wonders when, and whether, that changes
The one clear displacement he can name is driving: "take Waymo which went from like 0% of car rides to like 20 or 30% of car rides in San Francisco" for Uber-type rides, which he says has to have put some drivers out of business
Against that, cleaners have not been displaced by robot vacuums and call center employment has not fallen, and he argues the capex being spent economically requires that impact to arrive
Wu says the data shows a shift rather than a fall: hiring is moving away from roles AI can do and toward roles it cannot, with the cuts concentrated in junior positions while senior people are handed more leverage
"AI is pretty jagged right it's very uneven where it's really really good at coding" — Wu says Claude Code and Codex give him the equivalent of ten junior developers, so he builds things that would otherwise have sat on the shelf
Work expands to absorb the efficiency gain: call center wait times might fall from an hour to a minute rather than the jobs disappearing
The historical anchor is stability: labor force participation and unemployment have been remarkably stable through the agrarian revolution, the industrial revolution, the internet, globalization and the entry of women into the workforce
"the MIT economist David Autor right who found that I think it was like 60% of jobs didn't even exist in 1940" — Wu relaying Autor's work, with airline pilot as the example
Wu still expects a large disruption in the composition of the workforce even if the totals hold
The Other Side of the K: Crowded Hardware, Sold-Off Software
Wu sets up the hardware leg: trillions of dollars flowing into AI data centers, reaching memory and chip makers, utilities and power, and everyone else in the infrastructure complex
"the software index is down 50% or whatever and you know relative to the market the semiconductor index is up by 200%" — Wu's description of technology as a K-shaped economy
The crowding is his real point: he names Situational Awareness, the San Francisco hedge fund that was long chips and short software, which he says was doubling down on the same trade rather than hedging it
"the Goldman Sachs VIP index. I think like it's the portfolio is like 90% correlated with the AI trade. Basically, everyone is in on the same bet" — Wu on how crowded the positioning has become
He asks Rasmussen, as someone who leans against consensus, what he makes of the bullish side
The Mindset Behind the Capex
Rasmussen answers by describing the worldview rather than the numbers, and says it has two elements: futurism, and rationalist effective-altruist thinking
The futurism treats Popular Science outcomes as normal and imminent — Mars colonies, robots smarter than people — and concludes that they should be built
The rationalist half is where the risk comes from: evaluate everything statistically, maximize expected value, and accept that nine bets go to zero if the tenth is worth a billion dollars
He points to two blowups from the same milieu: Sam Bankman-Fried, and Leopold Aschenbrenner, who worked for him — "they're all part of this current Silicon Valley San Francisco thinking and they're making sort of some of the same mistakes"
He puts the leadership of Anthropic and the other big tech companies in the same category: believers in artificial general intelligence, ten years left of people doing things, universal basic income, data centers in space, and half the money given away afterward
"Same mindset, same viewpoint. You think they're immune from the probability of blow up? No."
His objection is that history does not run rationally: it twists and surprises, things take longer than they should, nobody uses robot vacuums though the technology is 20 years old, plenty of people will refuse driverless cars, and a populist revolt against data centers is possible
"everyone always overinvests. The returns are always lower than people anticipated. The returns always come later than people thought" — with a gap between the investment and the realization that is always too long for most people to survive
Wu's Defense: Individually Rational, Collectively Overinvested
Wu argues the spending is forced even on a purely rational operator: "If you're sitting on top of one of these big tech companies and Sam Altman's putting a trillion dollars into play, what are you going to do? Sit around, right? You're going to be fired instantly"
The belief driving it, true or not, is that AI has network effects like social networking or Uber, so the first mover to capture the market locks it
That belief produces collective overinvestment out of individually rational decisions
The risk he flags is timing, not direction: he says he personally believes in AI and expects it to get from point A to point B, and the question is how long and how the path unfolds
Rasmussen's illustration is Sam Bankman-Fried, whose bets he says were all right if he had been able to hold them; Wu finishes the thought — he would have been the world's best VC — and offers LTCM as the same case
Being overlevered and being overconcentrated are two forms of one mistake: you make a massive bet, and if you are wrong, or if it just takes longer, it does not matter that you were right
"there's always been a crash that punctuates the starting point and the ending point of a technological deployment" — Wu, citing the Carlota Perez model and, he says, every historical example he can find
The brake on adoption is society, not the technology: Wu points to Sam Altman on a podcast in the past week, where "he was saying he underestimated the inertia of society"
"have you ever worked at like a big company where like it takes, you know, 20 years to use Excel 2007 or whatever"
On the doom-loop worry that AI moves fast enough to destroy incomes, Wu lists the natural governors: compute, institutional constraints, and "there's labor unions, which you know, most people don't like them, but like they have a purpose, I guess, which is to slow down the pace of this deployment"
His base case is that the median person is not AI-pilled, will adopt slowly or leave it to their children, and that the S-curves take far longer than people inside the bubble expect
Why AI Seduces Intellectuals and Fails at the Call Center
Rasmussen's explanation for the gap is that smart people live in a highly verbal world of ideas, and AI is very good at meeting them where they are
"I really want to know about like the history of the Ottoman Empire in the 12th century, right? Like AI can tell you." — his list runs on through solving an Erdős problem, vibe coding an app, and photographing a garden to be told what to plant in it
That is the seduction: everything an intellectual wants to learn about, AI is there to talk to them about
"have it actually reliably do something like answer calls at a call center for like 24/7 for two weeks processing insurance claims and like you're going to go off the rails with disaster like very very quickly"
The gap between those two things, he says, is what creates the delay
Wu summarizes Rasmussen's recent written piece back to him: "you can still be long AI but you just don't have to be long today" — if there is a crash, there are better times to get in
Wu links it to consumer surplus: historically most, if not all, of the value created by technological advance has gone to the corporate adopters and individual users rather than to the investors funding it
He wonders aloud whether the capex is a gift from the people making the investments to whoever survives to use it
Rasmussen's version of the same idea is a joke: "You and I are going to do so much with all that excess data center compute supply when the prices crash to zero. It's going to be amazing."
Why Value Broke in the US and Still Works Abroad
Wu's proposition: value's severe underperformance has been concentrated in the US while it kept working in international, small cap and emerging markets, and the reason may be that the US is the world's most intangible-intensive economy, where the best companies are asset-light and price-to-book cannot see them
Rasmussen has been running the same question by region, measuring what percentage of book value is intangible
Japan is his control case: book value there is basically tangible book value, with almost no intangible assets on the balance sheet
"if you're buying something at half book, you're literally buying like a building that cost $100 for $50", and investment securities worth $100 for $50
"book value is a really good predictor for returns in Japan because it's like an entirely tangible economy"
The same metric does not work at all in the US, which he says is no surprise given "the US is the most intangible economy"
He calls Wu's work on constructing a new way to calculate intangible book value the answer to a problem others had identified but not solved, and the path to making value investing work in the US
What a pure US value screen actually returns is the problem: shoe retailers, suppliers to commodity producers, chemical refineries — cyclical, asset-dependent businesses he thinks are priced roughly fairly
"I'm trying to bet on expectation errors, systematic expectation errors" — and value, he says, is not leading investors to those errors in the US any more
Wu agrees the danger is that an accounting metric quietly encodes a different bias, long asset-heavy and short asset-light, which he says is itself a negative factor
Japan: A Government-Mandated Walk Back to Book Value
Rasmussen says he is a skeptic of governance mattering in general — who the shareholders are, who sits on the board, all of it too airy-fairy for him — but that Japan's version is genuinely tangible
The headline reform is a value investor's dream: "every company that trades below book value needs to get to book value"
Because book value there is tangible, a company below book has an overcapitalized balance sheet — cash, investment securities and real estate — and the instruction is to return the cash, sell the securities, divest non-core property and hand the proceeds back through buybacks and dividends
He describes it as a mechanical forcing to par by asset transfer, and calls it simple and brilliant
Going down the cap scale is what makes the trade: "you can build portfolios that trade at like 0.5 or 0.6 times book that are actually profitable, cash flow generative, and growing" and are massively overcapitalized
"It's like old school like '60s style value investing" — buy cheap on book, then let the government and the Tokyo Stock Exchange pressure the companies into handing the assets over
He says it is already visible in the data: large increases in buybacks and dividends, decreases in long-term investments and in balance sheet cash, and he wants to be as long as possible while it lasts
Wu's framing is time travel: taking what we now know about investing back to the industrial economy where Warren Buffett and Ben Graham could pick stocks on price-to-book
Buying the index does not buy you the trade: "the specific exposure to the governance reform as a tailwind is basically zero in the TOPIX", because everything below book is small cap
Maybe a dozen large caps qualify, and "It's like it's Mazda and it's Nissan and it's like, you know, like fine, take those bets, but like they're they're below book for a reason"
The Yen, and the Debt Number He Says Everyone Misreads
"the other sort of big story recently has been the yen where the yen is, you know, 40-year lows" — and Rasmussen says people are now panicking about whether it falls further
The yen has been falling for as long as he has invested in Japan and he has been fine investing anyway, so he is less worried; he thinks it is probably at a bottom
The debt statistic is where he says people go wrong: they look at gross debt to GDP without netting off the assets Japan holds — "if you subtract the assets then the debt's like only 100% of GDP", which he calls totally fine and normal
He describes the Japanese government as having run a carry trade for years, borrowing at zero percent in JGBs and buying foreign equities, and says everyone else should have done the same
"it's one of the only countries where the fiscal situation is improving not deteriorating"
He expects Japan's advanced manufacturing and high-end intellectual work to benefit in an AI world, and sees Japan as a clear beneficiary as investors look for an alternative to China
Wu adds Japan's lead in robotics and precision manufacturing, a culture more open to using it, and its position as a US ally
Where Else the Below-Book Universe Sits
Wu asks whether micro caps in the US and Europe show similar discounts, and whether the discounts come with catalysts or without them
"Europe is getting pretty close to that level. And then US there's still a big discount, but the stocks are less crazy than Japan, Korea, or now Europe."
Rasmussen says his career has always pushed him toward international markets rather than the US for value, because the discount has been so much bigger there
His rough map of the universe trading below book: "it's probably 30% of them are in Japan 20% of them are in Korea another call it 30 or 40% in Europe", with the remainder in the US, Canada and Australia, and emerging markets excluded
Biotech, Where R&D Spend Is the Only Book Value There Is
Wu frames biotech as the mirror image of Japan: instead of going to where the economy still looks like the 1960s, you rebuild the metrics for an economy that does not
Rasmussen accepts the framing outright: "I'm drinking the Kai Wu Kool-Aid"
Biotech is the extreme case, he says, because there are no traditional metrics at all — no assets, no revenue, nothing but intangibles
The three measures Verdad uses, which he calls Kai Wu-light: market value against cumulative historical R&D spend, the share of stock owned by biotech specialists rather than tourists, and short interest
"the companies that are owned by specialists that are cheap relative to their historic R&D spend and that have very low short interest massively outperform the ones that are the opposite"
More sophisticated intangible-value metrics for biotech are in progress at the firm, but he says they are harder and less far along
Wu's point is that the crude version working is the reassurance: R&D divided by market cap is obvious, easy to build, public, and has been in the finance literature for years, and it works best in the most intangible-intensive industries
"when I was at GMO, we had a much fancier model that Jeremy used to run through the 70s and 80s and it was better than price to book" — but plain price-to-book would have been fine too, which he treats as a robustness check
The Closing Question: A Contrarian Out of Contrarian Views
Wu closes with the show's standard question, aimed at a self-described contrarian: what do you believe that your peers do not
Rasmussen's answer is that his two big non-consensus calls stopped being non-consensus: "I've been very bearish on private markets and I think consensus is moving towards me and I've been very bullish on Japan and consensus is moving towards me"
"I think I need to go refresh my arsenal of non-consensus opinions" — because he thinks those two are tapped out
Rasmussen's bottom line is that the machinery that pushed allocators into software deals at twenty times EBITDA is now pushing big tech into AI capital spending, and the trade he wants instead is the unglamorous one nobody has a story about: overcapitalized Japanese small caps being pushed back to book value by their own stock exchange.
Products, Companies & Tools Mentioned
Verdad Advisors (Rasmussen's firm — the Japan below-book portfolios and the biotech intangible-value screens described here are its work)
HarbourVest and Partners Group (London-listed LP interests in private equity funds, used as mark-to-market proxies; the source of the volatility and NAV-discount figures)
Blackstone and KKR (Named as the listed management companies, to distinguish them from the listed fund interests he actually studied)
Thoma Bravo and Vista (Their 2014-16 returns are what convinced the rest of private equity to move into software)
SCF and Lime Rock (The energy private equity managers allocators chased in 2014-15, before shale blew up in 2015 and 2016)
Salesforce and Adobe (The public software comparables Wu says are down 50 to 80%, and the size of company private equity was never able to take private)
IBM (Wu's analogy for the software company whose moat is enterprise relationships rather than code)
New York Times (His example of a newspaper that survived disruption on brand, a changed business model and the games)
Walmart (The company that started behind in e-commerce and caught up — his case that the loser list changes over time)
Waymo and Uber (Rasmussen's single clear case of AI displacing labor, at 20 to 30% of San Francisco car rides)
Claude Code and Codex (Wu's own coding tools, which he says give him the equivalent of ten junior developers)
Wispr Flow (A tool he uses heavily and expects could be made redundant by becoming an iOS feature)
Anthropic (Rasmussen puts its leadership in the same rationalist-futurist category as the other big AI spenders and says it is not immune from blowing up)
Situational Awareness (The San Francisco hedge fund Wu says was long chips and short software — the AI trade twice over rather than a hedge)
Goldman Sachs VIP index (Wu's evidence for crowding: he puts the portfolio at around 90% correlated with the AI trade)
TOPIX (Gives you Japanese equities but almost none of the governance-reform tailwind, since everything below book is small cap)
Mazda and Nissan (Among the dozen or so Japanese large caps below book — and, Rasmussen says, below book for a reason)
Tokyo Stock Exchange (With the government, the enforcer of the below-book reform he is betting on)
GMO (Where Wu worked with a model more sophisticated than price-to-book, which he says price-to-book would have matched anyway)
LTCM (Wu's second example of a bet that was right and still did not survive the interim)
Books & Resources Mentioned
The Humble Investor – Dan Rasmussen (His book, named in Wu's introduction)
Verdad's research letters (Wu credits them for combining quantitative investing, financial history and a willingness to challenge conventional wisdom across public and private markets)
Wu's May research on the software selloff (Where the question of what a software company's moat is actually made of came from)
Wu's piece on disruption (Uses the New York Times to ask which incumbents survive a technology shift)
Rasmussen's recent piece on AI (Wu's summary of it: you can be long AI without being long it today)
The Carlota Perez model (Wu's reference for the crash that punctuates a technological deployment)
David Autor's work on jobs (Cited by Wu for the finding on how many of today's jobs did not exist in 1940)
Get the latest market chatter and takes as they happen:
X | Threads | Instagram | YouTube | TikTok | Facebook
Really interesting conversation. It also makes me think about relative valuation and what happens when sentiment starts wandering too far from fundamentals. Software really did change the world, but at some point “software is wonderful” became “therefore almost any price for software is reasonable.” Those are two very different statements.
AI may create the same problem in different places. The technology can be transformative while particular investments are terrible—and, strangely enough, the companies capturing the most economics don’t always receive the highest valuations. That’s where relative valuation gets interesting to me. Sometimes comparing what the market is willing to pay for different beneficiaries tells you almost as much about sentiment as it does about the businesses themselves.