The Other Stressors: Real Estate, Employment, Inflation and the Consumer, Ranked by What Could Actually Bring the Economy Down

• The question, the method, and the aggregate that misleads
• The two real estates: a market that has stopped moving, and one already in default
• The banks: a paper loss that is a constraint, not a crisis
• Employment: the widest channel in the economy, at the lowest proximity
• Inflation: cooling underneath, with the war on top
• The consumer: the price at the ninety-first percentile, the buffer at the fifth
• Extended without a trigger: valuations, the buildout’s financing, and the fiscal position
• The ranking, and the case against it
• The checkable markers, and what would falsify the ranking
• The close

The most alarming readings in this economy are not the ones with the widest transmission channels, and the widest channels are not the ones closest to breaking — so what could bring it down has to be ranked by scoring transmission, severity and proximity separately, and saying in advance what would prove the ranking wrong.

DeepDives · 2026 · An AI–human collaboration that weighs every source, mainstream or alternative, under strict rules of truth and logic.

§The question, the method, and the aggregate that misleads

Two episodes ago this examination took three markets apart — the price of oil, the price of government debt and the price of private credit — and asked what would have to change for each of them to get better. This one widens the frame, because those three are not the whole picture and they are not necessarily the worst of it. The question, which came from a listener, is the more useful one: what else is under strain, where does each of those things sit relative to its own history, and which of them could actually bring the economy down.

A list of alarming numbers is not an analysis, so the method is set out first. Every stressor is asked three questions: where it sits relative to its own history, expressed as a percentile; how wide its transmission is, meaning how directly it reaches household income, employment and the supply of credit if it breaks; and how close it sits to the level at which it has historically broken, which is a maturity wall, an auction that fails, an arrears rate crossing its own median. Each answer is scored from one to five, the three are multiplied, and the product is a weight out of a hundred and twenty-five. The arithmetic is not the point. The discipline of stating the assumptions is, and every four that follows says what should be seen if the score is right.

The first measurement is the one that most misleads people, and it points in two directions at once. The spread on American high-yield corporate bonds — the extra interest paid to lend to a risky company rather than to the government — is two hundred and seventy-six basis points, the twentieth percentile of the past quarter century, near the tightest it has ever been. On that number alone the corporate credit market is priced for serenity. Now look at the bottom of the same market. The spread on triple-C rated bonds, the borrowers closest to trouble, is one thousand and eighty-five basis points, and that is the ninety-ninth percentile: as wide as it was in the worst weeks of the financial crisis and the pandemic. The average is calm and the tail is screaming, and any account that quotes one of those numbers without the other is not describing this economy.

§The two real estates: a market that has stopped moving, and one already in default

Real estate follows, and there are two real estates in completely different conditions. Houses first, because that is the one people feel. Prices are at record levels: the national Case-Shiller index rose one point seven per cent over the past year and sits at the top of its history. The thirty-year mortgage rate is six point seven six per cent, the forty-second percentile of all the data going back to the early 1970s and the eighty-sixth percentile of the past quarter century — the more meaningful comparison, because nobody under sixty has lived through a high-rate mortgage market. Housing starts are running at eight hundred and eight thousand single-family units a year, the twenty-first percentile. Existing home sales are frozen at levels associated with recessions. Mortgage delinquencies are one point eight six per cent, the twentieth percentile, historically low.

So the stress in housing is not default but immobility: the people who own homes at three per cent mortgages will not sell, and the people who would buy cannot afford the payment at six point seven six. The damage is not a wave of foreclosures; it is a slow subtraction from construction, from mobility, and from every business that lives on transactions.

Commercial property is the opposite case: the defaults are already here and the prices have not fully admitted it. Office vacancy in the major central business districts is near or above twenty per cent. Delinquency on office loans inside commercial mortgage-backed securities is above eleven per cent, and in the early part of this year it went above twelve. Those are numbers that describe a sector in default. The banks report their commercial real estate delinquency at one point five three per cent. Hold those two numbers together, because the gap between them is the whole story: the securitised market has marked its loans to reality and the bank loan book has not. That is not a claim about fraud; it is a claim about accounting and incentives, because a bank loan is carried at its book value unless the borrower stops paying, so a building whose value has halved but whose loan is still being serviced shows up as a perfectly good asset.

And the calendar is not sympathetic. Roughly a trillion dollars of American commercial real estate debt matured across 2024 to 2026, and of the hundred billion dollars or so of securitised loans maturing this year, more than half are expected to default. The most exposed vintages are the five-year loans written in 2021, at low cap rates and high leverage, so the appraisal that justified them no longer exists and the refinancing has to be done at today’s rates against today’s rents. The phrase the industry uses is that extend and pretend is ending: the extensions granted in 2024 have been rolling forward, and they now land in this year and the next one. The cohort where this clears is the regional banks with commercial real estate exposure above three hundred per cent of their risk-based capital — a specific, identifiable list of institutions, and the place where a slow-burning property problem becomes a banking problem.

§The banks: a paper loss that is a constraint, not a crisis

The real estate stressors meet the banking system at one number. American banks were carrying three hundred and twenty-six point seven billion dollars of unrealised losses on their investment securities at the end of the second quarter, according to the Federal Deposit Insurance Corporation, and that figure has risen for two consecutive quarters: three hundred and twenty-one billion at the end of last year, three hundred and twenty-five billion in the first quarter, three hundred and twenty-six point seven billion in the second. The industry is simultaneously profitable, around eighty billion dollars in the quarter, with strong capital ratios.

So this is not a solvency crisis. It is a constraint. A bank with large paper losses has less room to sell securities to raise cash, less room to grow its loan book, and every incentive to hold loans and hope rather than to mark them and take the loss. That is why the commercial property problem is best understood not as a price event but as a slow withdrawal of credit from exactly the borrowers who need to refinance.

On the method set out above, those two stressors score eighty and sixty out of a hundred and twenty-five. Commercial property is firing already, which is why its proximity score is high, and its severity is high because banks are the channel through which property losses become economy-wide losses. The banking stressor is loaded rather than firing: the losses are real but paper, the profits are real, and nothing has yet forced the marks. What would force them is deposit flight or a credit event, which is precisely what happened in 2023, when three mid-sized banks failed in a fortnight.

§Employment: the widest channel in the economy, at the lowest proximity

Employment is the stressor everything else has to travel through, because wages are what pay the debts described above. On the headline numbers the American labour market is tight by any historical standard. Unemployment is four point one per cent, the twentieth percentile of the post-war record. Initial claims for unemployment insurance are running at two hundred and six thousand a week, the fifth percentile — lower than in almost any week of the past fifty years. Continuing claims, the people still drawing benefits, are one point seven seven million, the thirteenth percentile. On those three numbers alone there is no labour market problem at all, and that reading is right about the level.

Then look at the shape, and a different picture appears. The median spell of unemployment is eleven point four weeks, the eighty-sixth percentile of the modern record — the longest typical job search since the aftermath of the financial crisis. Quits are running at one point nine per cent of employment, below their median, which says workers are not confident enough to leave for something better. Job openings are still elevated but a fifth below their peak. Temporary help employment, the classic leading indicator because firms hire temps before they commit to permanent staff, is only back to the fortieth percentile of the past quarter century. Youth unemployment, for people between twenty and twenty-four, is nine point one per cent and rose six tenths of a point in a single month. The accurate description is low hiring and low firing: firms are not letting people go in numbers and they are not hiring them either. That is comfortable while it lasts and cruel to anyone who loses a job, because the queue they join does not move. It explains a statistic that otherwise looks like a contradiction — a labour market at the fifth percentile for claims and at the eighty-sixth for how long it takes to find work.

Weigh it, then. If this breaks, transmission is as wide as it gets, because employment is income and income is consumption, and American consumption is about sixty-eight per cent of the economy; severity is maximal for the same reason. Proximity is the lowest score given to anything, because unemployment at four point one per cent is not near the level at which labour markets historically crack, which is closer to five per cent and rising. So the weight is twenty-five out of a hundred and twenty-five — a low score for the most dangerous channel in the economy. The honest way to hold it is that employment today is not a stressor but the hinge that determines what every other stressor becomes. It also gives the cleanest marker in the examination: while claims stay below two hundred and fifty thousand and unemployment stays below four and a half per cent, this stays a one, and if either breaks the entire ranking has to be redrawn.

§Inflation: cooling underneath, with the war on top

Inflation has a split personality worth explaining carefully. Headline consumer inflation is three point four per cent over the past year and core, which excludes food and energy, is two point four. Both are above the central bank’s two per cent target and neither is alarming on its own. Under the surface the trend is softer than at any point in this cycle: the median consumer price change across the whole basket is two point one per cent, the nineteenth percentile, and the sticky-price measure that strips out volatile items is two point seven per cent. Set the energy shock aside and the underlying inflation process is close to normal and decelerating.

Then put the energy shock back, because it is the largest single input in the picture. The gasoline component of the consumer price index is up twenty-three per cent over the past year, on the war and the blockade examined in the last two episodes, and diesel went above two hundred dollars a barrel — not only a motorist’s problem but a cost that enters the price of everything moved by road, sea or rail. The reason headline inflation is three point four while the median basket is two point one is the war. That frames the honest risk in both directions. If the conflict eases, energy drops out of the arithmetic and measured inflation falls toward the underlying trend, which helps every other stressor here. If it does not, a supply shock keeps working its way into wages and expectations.

Expectations are where the tension shows. Consumers expect four point two per cent inflation over the coming year, the eighty-first percentile and far above the underlying trend, while the market’s own five-year forward measure is two point three one per cent, the fifty-fourth percentile. Ordinary people expect a great deal more inflation than the bond market does, and one of them is going to be wrong. Weight: forty-five.

§The consumer: the price at the ninety-first percentile, the buffer at the fifth

In the consumer, three separate things are true at once and they are easy to confuse. The first is the price of debt. The average annual percentage rate on a credit card is twenty point nine four per cent, the ninety-first percentile of the modern record, against a long-run median of thirteen point eight. That is the highest price households have ever paid to borrow on plastic, and it applies to balances that are themselves at records: total consumer credit is five point one nine trillion dollars, of which one point three six trillion is revolving or card-like debt, growing at three point seven per cent a year.

The second is the buffer. The personal saving rate is three point zero per cent, the fifth percentile. Consumer sentiment, as measured by the University of Michigan, is fifty-five point two, the second percentile of a series that began in the 1950s — lower than in the depths of the financial crisis. Household debt service is eleven point one six per cent of disposable income, the twenty-sixth percentile and genuinely moderate, but it is moderate in the aggregate because so much mortgage debt was refinanced at three per cent, a tailwind that expires with each passing year rather than renewing.

The third is the part that surprises people: the arrears have not moved. Delinquency on all loans at commercial banks is one point four two per cent, the eighth percentile, narrower than at any time in the past thirty years bar the last few. Credit card delinquency is two point eight five per cent, the forty-second percentile since 2000 and below its own median. Consumer loan delinquency is two point six two per cent, also below its median. Mortgage delinquency, as already noted, is at the twentieth percentile. So the price of household debt is at the ninety-first percentile, the buffer is at the fifth, and the arrears are below average.

There is one exception and it is a genuine record. Student loans: twenty-five per cent of federal borrowers are now delinquent and about ten point six per cent of the balances are ninety days or more past due, with roughly nine million borrowers in default, after the pandemic-era protections ended. That is the only consumer line item where the damage has already arrived in the data, and its transmission is narrow, because there is no house to repossess and the government bears the loss. It is why the most extreme reading in the table scores only twenty. Hold the picture together and the interpretation is this: the arrears are low because employment is holding, and employment is holding because the level of activity is still decent. The buffer is gone, so nothing absorbs a shock, and the price of the debt is at a record, so any household that has to borrow pays more than at any point in the modern record. That is not a crisis by itself. It is fragility: a cushion consumed before the stress arrives, in a labour market where losing your job now costs you eleven weeks rather than six. The exhausted consumer scores forty-eight; credit cards score twenty-four.

§Extended without a trigger: valuations, the buildout’s financing, and the fiscal position

Three stressors remain, and they are extended without being close to a trigger, which is a distinction worth being precise about. First, equity valuations. The cyclically adjusted price-to-earnings ratio, which averages ten years of earnings to smooth out the cycle, is forty point six — the ninety-eighth point eight percentile since 1881, a level reached only three times before, in 1929, in 1999 and in 2021. The index itself is up eleven point seven per cent over the past year. But here is the discipline the method imposes: a valuation is not a stressor in the way a maturity wall is, because nothing has to happen on a date. Expensive assets do not break; they amplify. Transmission is real, through the wealth effect on spending and the credit that finances the boom, and severity is high, because a repricing would remove household wealth at a scale that feeds straight back into consumption. Proximity scores two, because there is no calendar and no threshold. Weight: forty.

The financing of the artificial intelligence buildout is the sharper version of the same risk, examined in detail in the episode on where the money comes from: data-centre construction, vendor financing, private capital assembled at scale, and debt issued against cash flows that are projected rather than realised. Proximity is higher, because debt has dates attached even when valuations do not, and if the projected revenues do not arrive the refinancing is not optional. Weight: forty-eight.

And the fiscal picture, briefly, because the last episode described it and it has not changed. Federal debt is one hundred and twenty-two point six per cent of the economy, the ninety-eighth percentile. Interest payments have reached one point two five trillion dollars a year, the highest share of the economy in the modern record. Customs duties are running at three hundred and twenty-six billion dollars a quarter, the ninety-ninth percentile, and the monthly deficit is in the region of a hundred and sixty-seven billion. Its weight is thirty-two, and the reason it is not higher is that proximity scores two: auctions are clearing with strong demand, and there is no scheduled moment at which this becomes a crisis. It is a slow tax on everything else.

§The ranking, and the case against it

Here is the ranking, all fifteen, from the heaviest weight to the lightest, out of a hundred and twenty-five. The energy shock from the war, one hundred. Long-term government yields and the term premium, eighty. Commercial real estate, eighty. The banking system’s unrealised losses and property concentration, sixty. Private credit, forty-eight. The financing of the artificial intelligence buildout, forty-eight. The exhausted consumer, forty-eight. Inflation, forty-five. Equity valuations, forty. Housing activity, thirty-six. The fiscal position, thirty-two. Mortgage arrears, twenty-five. Employment, twenty-five. Credit cards, twenty-four. Student loans, twenty.

That ordering contains a claim worth stating plainly, because it is the answer to the question this episode was asked. The stressors with the highest weights are not the ones with the most alarming levels. Student loans have the most extreme reading in the table and the lowest weight, because the losses fall on the government and there is no asset to seize. Employment has the widest transmission channel in the economy and one of the lowest weights, because its level is historically tight. Commercial real estate has a delinquency rate at the bank level of one point five per cent, which sounds fine, and a weight of eighty, because eleven per cent of the securitised market is already in default, the maturity calendar is not moving, and the banks are the channel through which a property loss becomes a credit contraction. Level is what people argue about. Proximity is what determines whether a stressor becomes an event.

Now the case against all of it, made properly rather than as a formality, because there is a serious version. Almost every record here is a record because the economy is larger than it has ever been: nominal debt, prices, household balances and federal borrowing all reach new highs in every expansion, and quoting them as alarming is a category error. The stress measures that matter most — delinquencies, defaults, charge-offs, bank capital — are below their historical medians almost everywhere, and the two that are extreme, student loans and office property, are both ring-fenced: one by the government, the other inside a banking system that is profitable, well capitalised and has had four years to prepare. The labour market is tight on every level measure. Financial conditions are loose, with the Chicago Fed’s index in the thirtieth percentile, meaning easier than normal. And the credit market’s apparent disagreement can be read the other way: as healthy discrimination, the market finally charging the weakest borrowers what they are worth, rather than as the eve of a crisis.

Three answers to that, and they hold. The first is that the bull case describes levels, and the whole purpose of the proximity column is that levels are not triggers. A bank that is well capitalised on paper and holding three hundred and twenty-six billion dollars of unrealised losses is well capitalised until it has to sell, which is what 2023 demonstrated in a fortnight. The second is the buffer: arrears are low because employment is holding, and the cushion that would normally absorb a shock — the saving rate at the fifth percentile, sentiment at the second — has already been spent. Fragility is not a forecast; it is a statement about how the system behaves when something unexpected happens. The third is that the market’s own riskiest borrowers disagree with the optimists: triple-C spreads at the ninety-ninth percentile are not a valuation opinion but the price at which people with money are willing to lend to the weakest companies in the country. When the average is calm and the tail is at crisis levels, the honest reading is that the average is describing a different population.

§The checkable markers, and what would falsify the ranking

So here is what this series will check, stated in advance and dated, so that it can be held to account. If claims stay below two hundred and fifty thousand and unemployment below four and a half per cent through the end of the year, employment belongs at the bottom of this table and the fragility case is not yet paying. If the bank credit card delinquency rate rises above its own median of three point eight three per cent within the next two quarters, the consumer stressor moves from twenty-four to somewhere near sixty and this ranking is wrong. If office delinquency inside securitised loans falls below eight per cent and bank property delinquency stays under two per cent through next year, commercial real estate at eighty was too high. If high-yield spreads remain under three hundred basis points while triple-C stays above nine hundred for two more quarters, the bifurcation is the story rather than a broad credit event. If the median basket inflation rate rises back above two point eight per cent, the claim that inflation is cooling underneath the war is wrong. If the cyclically adjusted earnings multiple falls below thirty without a credit event, the artificial intelligence financing channel was not how an equity repricing transmits. And if the central bank’s balance sheet keeps growing while the deficit stays above six per cent of the economy, the monetary backdrop for the next episode in this pair gets stronger regardless of what the stressors do.

§The close

The examination asked what else is under strain, where each of those things sits relative to its own history, and which of them could actually bring the economy down, and it found that those questions do not have the same answer. The most extreme reading in the table, student loans, is among the least dangerous, because the loss lands on the government and there is no asset to seize. The widest channel in the economy, employment, is the least proximate, because it is tight on every level measure — which is why it ranks at the bottom while remaining the hinge that decides what the others become. The calmest headline in credit, a high-yield spread near the tightest in a quarter century, sits beside the least calm number in it, triple-C at the ninety-ninth percentile. Commercial property and the banks that hold it carry the heaviest weights of anything examined here, not because the banks are insolvent — they are profitable and well capitalised — but because the marks have not been taken, the maturity calendar is not moving, and credit withdrawn from the borrowers who must refinance is how a property problem becomes an economy-wide one.

What would falsify that is set out above in dated, checkable terms, and it is deliberately not a forecast: a bank card delinquency rate above its own median, student-loan and office records that refuse to spread, a median basket that re-accelerates, a valuation that reprices without a credit event. The honest summary is the one the method was built to produce. The danger is not concentrated in the things that look worst; it is concentrated in the things closest to a date. Level is what people argue about, and proximity is what decides. Whichever stressor moves first will determine what the others become — and if a credit event does come, that will be the moment everyone finds out how much of this was cushion and how much was structure.

Sources: this article names its sources inline — the high-yield and triple-C credit spreads and the percentile readings drawn from them; the Case-Shiller house-price index, the mortgage-rate and housing-starts series, the home-sales and mortgage-delinquency data, and the commercial property vacancy, securitisation and maturity figures; the bank unrealised-loss and profitability data from the Federal Deposit Insurance Corporation, with the arithmetic of the property maturity wall and the regional-bank concentration cohort; the labour market series on unemployment, claims, continuing claims, duration, quits, openings, temporary help and youth unemployment; the consumer price measures — headline, core, median and sticky — with the energy components and the two rival expectation measures, one from households and one from the market; the consumer credit, saving-rate, debt-service, sentiment and delinquency series, and the student-loan data that followed the end of the pandemic-era protections; the cyclically adjusted valuation measure and its history since 1881; the Chicago Fed’s financial conditions index used in the case against; and our own earlier examinations of the oil and bond markets, the fiscal position and the financing of the artificial intelligence buildout. Every source is weighed with the same skepticism regardless of politics or business model. The audio version of this article, read by DeepDives, is available on Wavlake.