On August 19, 2026, Target told its shareholders that $994 million had come back from the Treasury. Then it told them the quarter was better measured without the money. The next day Walmart, with about $2.9 billion back, said its refund was going into lower prices.
It was the same kind of money both times. A tariff is a tax paid at the border. The Supreme Court struck down one set of tariffs in February 2026, and the government is refunding them to the companies that paid.
This idea originated when I reviewed the publicly filed complaint of Cavallaro v. Target Corporation, filed July 21, 2026 in the United States District Court for the Southern District of New York (7:26-cv-06165). A complaint is the document that opens a lawsuit, and I read new ones for fun. It comes with my day job as CEO of Round Table Group, which helps lawyers find expert witnesses. The plaintiff alleges that Target had already recovered the tariff from shoppers through its prices, so that the refund is a windfall. That is a legal question for a court, and I take no view of it. It left me with a business question. When a retailer has already raised its prices to cover a tariff, and the government then hands the tariff back, who ends up with the money?
I looked at four large retailers that received refunds: Walmart, Target, Costco and Best Buy. Walmart and Costco announced that refund money would go into lower prices. Target and Best Buy announced nothing of the kind. Now take each refund out of each company's gross margin, the share of each sales dollar left after paying for the goods. The two retailers that announced refund-linked price cuts are the only two of the four whose margins did not rise.
The refunds are large and arrive once. Target's added $1.65 to a quarter's earnings of $4.11 a share, about 40 percent of the total. At Walmart and Costco, price cuts paid for with refund money are holding down today's margins. If the cuts end when the refunds do, the margins come back. If they stay, a one-time gift has become a standing cost. I own shares of Walmart, so weigh what I say about its margins with that in mind.
My guess is that the cuts end with the money. We wrote that guess down as a test before the quarters that answer it were reported, and it is settled by August 31, 2027. We think there is about a 60 percent chance that it comes back confirmed, if it reaches a verdict. If the cuts do end, the test cannot say why. Two explanations, one from psychology and one from plain finance, predict the same ending.
When a container arrives at a port, somebody signs the customs declaration and pays the duty, which is another word for a tariff. That party is the importer of record, and a refund goes back to it. A buyer further down the line has no direct claim against the government, as trade lawyers have been telling their clients since March. Each of our four brings some of what it sells into the country itself.
Customs, the federal agency that collects tariffs, opened a system to pay the refunds on April 20, 2026. By the fall the four had theirs. Walmart received about $2.9 billion. Target booked $994 million in the quarter that ended August 1. Costco reports $184 million received by the end of its fiscal year on August 30, $174 million of it refunds and $10 million interest on them, and the margin arithmetic here uses the $174 million. Another $155 million arrived afterward. Best Buy received $34 million in the quarter that ended August 1 and $41 million on September 2.
Set each refund against the sales of the quarter it landed in and the four come into proportion. Target's was by far the largest relative to its business, about 3.7 percent of a quarter's sales. Walmart's was about 1.6 percent. Costco's $174 million came to about 0.19 percent of its sixteen-week fourth quarter. Best Buy's $34 million was about 0.35 percent. Fractions of a percent sound small, but margins are read in them. A basis point is a hundredth of a percentage point, so Costco's 0.19 percent is 19 basis points, and our test turns on moves of 20 and 30.
Say you sell toasters. Last year you paid a duty on a shipment. You sold the toasters. Now a check for the duty arrives from the government. If you want another toaster to sell, the supplier will charge what the supplier charges today, and Customs will collect whatever duty is in force. The check changes neither cost. It also gives a shopper no fresh reason to buy a toaster. In the textbook, you choose the same price you would have chosen the day before the check arrived, and the check goes to profit.
Economists call such a check a lump sum. Dire Straits had a better name for it in 1985, "Money for Nothing," and as far as the price of a toaster goes the textbook agrees with the band.
Change the duty on the next shipment and the calculation changes. That happened too. In 2025 the United States imposed tariffs under a law called the International Emergency Economic Powers Act, IEEPA for short. On February 20, 2026 the Supreme Court held that the law did not authorize them. Customs stopped collecting them on February 24. The same morning a temporary 10 percent duty under another law, Section 122 of the Trade Act of 1974, took their place on most imports. From July 24, duties under Section 301 of the same act replaced that one, at 10 or 12.5 percent depending on the country.
So the tariff did not end. It changed its name and, for many goods, its rate. A retailer could have had reason to change a shelf price before it saw a dollar of its refund. Whether any of these tariffs is wise, and whether the Court was right, are questions I leave alone. A reader of either persuasion can use what follows.
The textbook, then, makes two predictions for 2026. Prices on imported goods move toward the new duty. And the checks change nothing on the shelf.
Something like this has happened before, and the people it happened to kept good records. In 1933 the New Deal paid for its farm program with a processing tax, so many cents on every hundred pounds of hogs slaughtered or bushel of wheat milled. The Supreme Court struck the scheme down in United States v. Butler on January 6, 1936. In the weeks that followed, more than $300 million of taxes, held by the courts or never paid, stayed with the processors. That is the Department of Agriculture's figure, from the side that wanted the money back. In 1936 the tax truly ended, where in 2026 another duty took its place. Joseph Thorndike set the two years side by side in April, in "Who Gets The Tariff Refund? A Lesson From 1936," which Tax Notes and Forbes both ran. We have read the part of it that is open to us and credit the overlap.
The Department's economist, L. H. Bean, measured the processors' gross margins, the gap between what a processor paid for the commodity and what he sold the product for. The wheat margin had been 38 cents in 1932, the last full year before the tax. With a tax of 30 cents it was 68 cents in 1934 and 67 in 1935. In January 1936 it was back to 36. The hog margin was "almost immediately reduced to its former amount." Bean wrote that the processors had "collected the equivalent of the impounded taxes from consumers or producers." So the margins fell back within weeks, and that gain went to buyers, farmers or both, while the impounded and unpaid taxes stayed with the processors. The lump sum of returned taxes moved nothing that Bean measured. That is the textbook's two predictions, ninety years early.
Congress answered in June 1936 with a windfall tax of 80 percent on processing taxes a processor had passed on in his prices and then never paid or got back. The budget counted on $82 million in the first year. The Treasury collected $6,073,351.02, about 7 percent of that. The records do not say why it raised so little.
Walmart and Costco announced that they would break the second prediction on purpose. Another economist came to mind, one who has spent a good deal of time asking why a dollar can feel different depending on where it came from.
Richard Thaler, the Nobel Prize-winning economist who was on the faculty at the University of Chicago Booth School of Business during my MBA studies there, called it mental accounting, and Vista's Library, our collection of models from economics and the sciences, has an entry for it. The textbook assumes a dollar is a dollar whatever its source. Thaler found, over decades of experiments, that people keep money in labeled accounts instead. In his 1999 paper the labels fall on the sources of funds as well as their uses, "regular income versus windfalls" among them. Money in one account is "not a perfect substitute" for money in another, and a windfall is often spent more freely than a paycheck.
Thaler ran experiments. I ran one myself, without a control group. When my partners and I sold the company I run today, in 2010, the money arrived with a name on it. (We bought it back in 2019.) Bob and Chris saved and invested their share, as far as I could tell. I recorded my own approach in my memoir, The Scholars' Treasure: "How fun it was to spend the proceeds of the acquisition so cavalierly" (ch. 13, p. 161). Looking back, I wish I had deferred more of that gratification.
Applied to a company, the model predicts a timing. The money comes in, the cuts follow, and the spending is finished when the pot marked for the customer is empty.
Walmart's account of its quarter has that shape. The company has more than 10,900 stores and had $713 billion of revenue in its last fiscal year, and it keeps about a quarter of each sales dollar after paying for the goods. John Furner is its chief executive and John David Rainey its chief financial officer. Walmart calls money spent on lower prices a price investment, and it calls a temporary price cut a rollback. It counted more than 11,000 rollbacks in the quarter. Furner said, "A rollback has a start date, it has an end date." Rainey said a large portion of the refunds had been invested at the end of the quarter. He said the financial effect of the refunds and their reinvestment would be "largely contained within the current fiscal year," which ends January 31, 2027.
Walmart also left itself room. Furner said of the rollbacks, "It's probably a bit too early to call how many of these will be permanent."
Costco said the same at a smaller scale. It is a membership warehouse club with a margin near 11 percent, less than half of Walmart's. Ron Vachris is its chief executive and Gary Millerchip its chief financial officer. Its annual report says "a portion of these refunds was invested in reducing prices for members" and that it intends to "continue to invest the majority of the refunds" still arriving in value for the member. The refund was worth about 19 basis points of the quarter's sales. Millerchip put its net benefit to the margin, after the price cuts it paid for, at 9. The difference suggests about 10 basis points went into prices. The cuts fell in produce, meat and beverages, in home furnishings and hardware, and, as Millerchip put it, in "nonfood areas where tariffs would have had an impact."
Target and Best Buy disclosed no such program. Target sells style and design at discount prices in more than 2,000 American stores and keeps near 30 cents of each sales dollar. Michael Fiddelke is its chief executive. Target's guidance, the company's own forecast of its results, excludes any future refunds. Target did cut prices this year, but the timing does not fit the refund. It announced lower prices on about 3,000 items in November 2025, before the Supreme Court had ruled. It announced about 3,000 more on March 11, 2026, before a dollar of refund had arrived.
Best Buy began as a stereo shop, and the low-price, no-frills "Tornado Sale" it held after a tornado hit its store in Roseville, Minnesota, became the business. Jason Bonfig became its chief executive on November 1, 2026, succeeding Corie Barry, who led it through the refund quarter. Best Buy's refund sits in its gross profit and in its guidance, which assumes "a similar level of IEEPA tariff refunds in the third quarter" alongside continued investments in major appliances. Barry did not call those investments refund-funded.
Be careful what that silence proves. A company that announces no refund-linked program has told us what it announced. It has not told us what its prices would have been without the refund, and nothing in its filings can. So we sort the four by what they said.
All four recorded their refunds as a reduction of the cost of goods sold, so each refund sits inside the gross margin and flatters it. Take the refund out of gross profit. Divide by the quarter's sales. Set the result against the same quarter a year earlier.
Start with Target, which did the exercise itself. Its gross margin rate for the quarter was 33.7 percent, against 29.0 percent a year earlier, and the refund supplied 3.7 points of it. From the dollars Target reports, the rate without the refund was 29.93 percent against 28.99. That is a rise of about 94 basis points, which Target rounds to approximately 100. Target says where it came from. Last year's margin had been held down by heavy markdowns and order cancellation costs, and advertising and other non-merchandise sales kept growing. None of that has to do with the refund in either direction.
Walmart's rate was 25.41 percent against 24.45, up 96 basis points by the company's count. Walmart publishes no gross margin rate without the refund, so the next figure is ours. The $2.9 billion was about 1.56 percent of the quarter's sales. Walmart's quarterly report says the refund was recorded in cost of sales. Assume all of it sat in that quarter, and take it out of gross profit. The rate that remains is 23.86 percent, with rounding. Against the year before's 24.45, that is about 60 basis points lower.
Up 96 as reported. Down about 60 without the check.
The 60 basis points are for the whole company. Walmart's own explanation is for Walmart U.S., the segment for its Walmart stores, online sales and advertising in the United States. There the refunds and a better mix of business pushed the margin up, and price investments and higher fuel costs pulled it down. Advertising, part of which Walmart books as a reduction of cost of sales, grew 38 percent. So the 60 basis points hold price cuts, fuel, mix and advertising together, and Walmart gives no number for any one of them.
Costco's rate was 11.02 percent against 11.13. Take out the $174 million and it sits about 30 basis points below the year before. The gap is gasoline. Higher gasoline prices swell a warehouse club's sales at a thin margin and pull the average down. By Millerchip's figures they took about 31 basis points off the rate. Set those 31 aside and the shortfall of 30 becomes a gain of 1. Costco's rate without the refund was about level, with a larger inventory accounting charge still in it.
Best Buy reports figures for its Domestic segment, where the refund sits. Without the $34 million the rate was 23.66 percent against 23.37, about 29 basis points up. The company credits its online marketplace and its advertising business, partly offset by lower product margins.
Of the two that said they would spend it on price, Walmart's margin is below last year's and Costco's is level once its gasoline is allowed for. The two that disclosed no program are above. The shape fits refund-funded price cuts. But every company has handed us other reasons for its number, and this is the quarter the money arrived in. One quarter settles nothing. The next clue is what Walmart's and Costco's margins do after the refunds stop.
A screen that ranks retailers on reported margins this year is ranking refunds.
The labeled pot is one way to pay for a price cut. The second needs no psychology.
Walmart had promised investors a range of earnings growth before it knew what the quarter would bring. On the same call where Furner explained that rollbacks have end dates, he said this: "But we always need to balance our price investments relative to what we're seeing in the market, and with the commitments we made on our forecasts, including operating income over time." In this account, a merchant has cuts he would like to make, and the finance team has a forecast to meet. Then the check arrives.
Walmart's adjusted operating income, the profit from running the business with one-time items removed, grew about 17 percent in constant currency. About seven and a half points of that was the net benefit of the refunds after the price investment they paid for. Without it, the company said, growth sat at the top end of its guidance of 7 to 10 percent. So the cuts cost Walmart nothing against its forecast.
In Thaler's account the refund is a pot marked for the customer. In the finance team's it is temporary room in the earnings budget, with every dollar in one account. In each, the money is spent and the cuts end.
The third way is ordinary competition. The cuts were coming anyway, and the refund gave them a name. Those cuts stay after the money is gone, because the money was never their cause.
Our test reads two things: the margin with the refund taken out, and the company's own word that its refund-linked price program is over. It cannot tell the pot from the room. Nor can it see a shelf price, so it cannot say any price went back up.
The place I would look is the shelf, especially at goods made here. A change in the import duty moves the cost of imported goods, at every retailer alike. A program funded by a refund can cut the price of an American-made item just as easily. Economists call the comparison a difference-in-differences, the design that Card and Krueger's minimum wage comparison of New Jersey and Pennsylvania made famous. Take the change in one group and subtract the change in a group that did not get the treatment, over the same weeks.
effect of the duty change = (imported price after − imported price before) − (American-made price after − American-made price before)
The subtraction rests on one assumption, that the two groups of goods would have moved together had the duty stayed put. If it holds, what is left is the effect of the duty change, because a cut that falls on both groups alike drops out. The second bracket is where a refund could show: American-made prices that fall when the checks arrive and come back when the money is spent. Watching the same items week by week, through the duty changes and after the last refund, would give us a better chance of seeing which pattern we have. The same items at a retailer with no refund-linked program, over the same weeks, would give the refund its own comparison. The item-by-item record is a piece of evidence we still need.
Imagine you used to pay $3 for an item. One day it goes up to $3.40. You flinch, and within a season $3.40 is simply the price. When the tariff goes, nothing on the shelf says which part was the tariff, and nobody expects a check. We borrow the idea from Harry Helson, a psychologist who showed that people judge each new experience against a running average of recent ones. Philip Brickman and Donald Campbell drew the consequence: a change is felt once and then absorbed. Reading a shelf price and a tariff that way is our application, not theirs.
Now suppose the bill says $3 for the item and forty cents for the tariff. The forty cents keeps its name. A customer looking at that bill after the tariff is struck down knows to the cent what to ask for.
Companies that billed a separate line have behaved accordingly. Varex Imaging, a maker of X-ray components, billed its customers what its filings call "IEEPA tariff surcharges previously billed." By July 3, 2026 it had received about $17.0 million of refunds. It set aside $6.6 million, about 39 percent, to return to the customers who had paid the surcharge. Dame Products, which had added a tariff surcharge at checkout, is refunding every customer who paid it. FedEx said in February that any refund it received would go back to the shippers and customers who had borne the charges. Amazon described doing both. Where it can trace import charges it passed on to particular customers, it will refund them. Otherwise, its finance chief said, it will use the refunds "to continue to invest in low prices for customers."
FIGS, a maker of medical scrubs, recovered $20.5 million and told analysts it was "not earmarking it for anything specific." Whether its earlier price increase was the tariff folded into list price, its statements do not say, and we do not say it for them.
The law drew this line ninety years ago. In 1934, in United States v. Jefferson Electric, the Supreme Court considered a manufacturer, the American Chain Company, that had stamped its invoices to show how much of each bill was an excise tax. An excise is a tax on the sale of particular goods. The Court held that by its invoices the company had collected the tax from its buyers. That the company had cut its price by the amount of the tax was, the Court said, "of no importance."
Trade lawyers have argued over itemized surcharges against folded prices since the spring, and the invoice line gives us a second thing to test. Among publicly traded importers, those that itemized a tariff line should report a customer refund, credit or rollback far more often than those that folded the tariff into their prices.
The method travels to any company that receives money with a name on it. Strip the windfall from the margins, then watch whether the spending tied to its label ends when the money does.
Williams-Sonoma recorded $167.8 million of IEEPA refunds against its cost of goods sold. It set aside $47.5 million to reimburse vendors that had given it tariff concessions, and made a one-time $10.0 million 401(k) contribution it called tariff-related employee recognition. Kohl's received about $150 million and said part was shared with its vendors and "invested to deliver greater value to our customers." Mettler-Toledo, which makes laboratory and industrial instruments, booked $52.4 million of refunds from the government and, against them, $27.8 million of tariff-related refunds to its own customers. Microsoft, which sells Xbox consoles and Surface devices and names tariffs among its risks, has disclosed no IEEPA refund at all.
Older windfalls carried labels too. Delta Air Lines received $5.6 billion of federal payroll support in 2020 under a program whose money was "to be used for employee wages, salaries and benefits." The question has prior owners. James Hines and Richard Thaler described in 1995 the flypaper effect, the finding that money granted to a government sticks where it lands, and saw it in firms' windfalls too. Olivier Blanchard, Florencio Lopez-de-Silanes and Andrei Shleifer asked in 1994 what firms do with cash windfalls, in eleven companies that had won or settled lawsuits.
Andrew Leahey, writing in Bloomberg Tax on September 22, 2026, made the core prediction before us: that price relief paid for with refunds will disappear once the money is spent. What we add is a test with a date on it.
We froze the rules at 7:30 in the evening, Pacific time, on October 24, 2026, before any of the quarters they read had been reported. We amended them on November 6 and November 11, each time with a dated note, and each time before any of the four retailers had reported a new quarter. The thresholds have never moved. The original and both amendments are printed in "For the record," with the fingerprint of each block of the scoring program, a string that changes if any character of the block does. The order matters more than the lines, and the Library has an entry that says why. An explanation found after the outcome is weak evidence, because one is nearly always there to be found. Only a prediction written before the outcome, with a result that would prove it wrong, counts as a test.
The claim is the one that can be observed: refund-linked price investment ends when the refunds do. The measure is the gross margin rate with any disclosed refund taken out, against the same fiscal quarter a year earlier measured the same way. Fuel or gasoline is set aside only where the company itself states its effect in basis points, as Costco does and Walmart so far has not.
The shortest way to read the lines is to imagine last year's rate was 24.5 percent. A rate of 24.3 percent or better in either of two quarters confirms, provided the program has ended. A program counts as ended only on an affirmative indication: the company says so, or a dated end to the program passes. Silence on a later call leaves the program's status unknown, and an unknown status cannot confirm. A rate of 24.2 percent or worse in both quarters falsifies. That counts against the claim that the cuts end with the money, and it cannot say why the margins stayed down. Between those lines, the test cannot tell. It also cannot tell if tariff rates on these retailers' imports change inside the window, since that would move the cost of the next toaster.
Walmart's own quarterly report expects its price investment to continue through the fiscal year ending January 31, 2027. So Walmart is read in the two quarters after that, those ending April 30 and July 31, 2027. Costco is read in the two quarters after the last quarter in which it reports receiving a refund. Both are settled by August 31, 2027. If a late refund pushes Costco's second quarter past that date, Costco is "untested" and Walmart is scored alone. The second test, the surcharge split, reads importers' filings through June 30, 2027. Its lines are in the record.
Where both Walmart and Costco are scored, the claim is confirmed only if both confirm and falsified only if both falsify; a split cannot tell. Where Walmart is scored alone, its result is the verdict, and we will say that it rests on one company. Where an unaffirmed end would decide the verdict, the result is "untested." Target and Best Buy are read beside the test as a comparison, in the quarters frozen in October.
The rule asks less of the evidence than I would ask of the person who set the prices. For Walmart's quarter ending July 31, 2027, the base is the quarter ending July 31, 2026, which already carried refund-funded price investment. So cuts that simply stayed in place could read near level against it. Costco's refund-funded cuts are about 10 basis points, inside the 20 basis point band. And a company can affirm an end simply by saying the refund money is spent. A falsification is the stronger result.
The other limits are smaller. The duty in force differs between the quarters compared, which can move the comparison either way. Walmart's margin also moves with mix and advertising, which the test never sets aside. Varex sits among the itemizers today, marked as having returned money; FIGS, with no statement about how it priced the tariff, is not yet sorted.
We think there is about a 60 percent chance that the test, if it reaches a verdict, returns "confirmed." That is our judgment, and it is our estimate of the test, not a forecast of any company's earnings or shares. Walmart's rollbacks carry end dates by design. Walmart also says some may become permanent, and competition may hold the cuts in place whatever paid for them at the start. We think it quite possible that the test reaches no verdict at all, because no company affirms an ending.
Margins can take us only so far. The data that we're missing is how the refund was budgeted, as a separate pot for price or as slack in the year's earnings arithmetic. Likely, we can get some of it from primary research, which is what Vista is all about.
I would ask a former retail pricing or merchandise finance leader, as a matter of general past practice, who decided what a one-time recovery was for. I would ask a customs broker or trade compliance lead what clients generally did once a claim was paid. I would ask a pricing or channel leader who ran a tariff surcharge through distributors what generally became of an itemized surcharge, and of a tariff folded into list price, after the Court ruled. And I would ask a historian of the processing taxes for a line we did not find: how much was actually refunded to processors who proved they had borne the tax. Every conversation would draw on past, general experience only: no current employees of the companies named, nothing confidential, and everyone screened for conflicts before a word is said. For a client engagement, Vista would combine this inquiry with those interviews.
The textbook says the name on a check is a decoration and the check is cash. Target told its shareholders to measure the quarter without its $994 million. Walmart put a name on $2.9 billion.
By the end of August 2027 I expect to have Walmart's margins in front of me. If they sit 30 basis points or more below the year before in both quarters, that will count against my guess. If they are back and Walmart has said its program is over, I will know that much and no more. I would still like to ask the person who sets the prices what happened to the number on the toaster's tag when the money ran out.
This section holds what the body summarizes: the test as first frozen, its two amendments, their fingerprints, and the 1936 record in full.
The test as frozen on October 24, 2026. The paragraphs and the table that follow set out the frozen rules in the wording we settled on October 25. That wording was edited after the freeze, and one sentence about Varex and FIGS now sits with the test's limits in the text above. No rule or number differs from the code fingerprinted below, except two the October wording left out: the code also excludes Amazon and Microsoft from the second test's population, and it rounds each change to the nearest whole basis point. Where the text says "spenders" it means Walmart and Costco, and "the pot reading" is the claim as we first worded it.
The first test is the pot reading, and its measure is the one we have used throughout: the gross margin rate with the disclosed refund benefit taken out of gross profit, against the same fiscal quarter a year earlier, in basis points. It is read for each company in the two fiscal quarters after the company's last disclosed refund quarter. The companies keep different calendars (Walmart's quarters end on the last day of the month, Target's and Best Buy's on a Saturday near it, and Costco's year ends on the Sunday nearest the end of August), so the quarters are, for Walmart, those ending October 31, 2026 and January 31, 2027; for Costco, whose last refund so far falls in the quarter ending in November 2026, the twelve-week quarters ending about February 14 and May 9, 2027, with a later refund moving them later; for Target the quarters ending about October 31, 2026 and January 30, 2027; for Best Buy those ending about January 30 and May 1, 2027. Dates the filings do not state are our calendar arithmetic.
The lines. The pot reading is confirmed if each of the two spenders, Walmart and Costco, comes within 20 basis points of the year-earlier rate, or above it, in at least one of its two quarters, and its refund-tied price program is over, judged from the company's own words: a program is over when the company's release, filing or call for the second of the two quarters no longer describes it as continuing. It is falsified if both spenders sit 30 basis points or more below the year-earlier rate in both of their quarters: then the cuts outlived the money, and they were ordinary competitive repricing that the refund merely labeled. Between those lines the test cannot tell. It also cannot tell if tariff rates on these retailers' imports change inside the window. The Section 301 duties of July 24, 2026, at 10 percent for some countries and 12.5 percent for the rest, are taken as given; a change to that schedule, or a new tariff of 10 percent or more on consumer goods, would move the cost of the next unit, and we would say so rather than score it. Results are read through releases published by June 30, 2027.
The second test is the surcharge split. The population is publicly traded American importers whose quarterly or annual reports filed from March 1, 2026 to June 30, 2027 disclose an IEEPA refund received or receivable, found by a full-text search of filings with the Securities and Exchange Commission. Each is sorted by its own documents into itemizers, whose filings, releases or published price notices describe a separate tariff surcharge or tariff line, and folders, who describe tariff recovery through list price or pricing actions only; and each is marked returned if the filing records a customer refund liability, credit, rebate or surcharge rollback tied to the refund. The test needs at least eight companies in each group, or it cannot tell. It is confirmed if the share of itemizers that returned exceeds the share of folders that returned by 30 percentage points or more, falsified if the gap is under 10 points, and between those it cannot tell.
| Test | Measure | Confirms if | Falsifies if | Read by |
|---|---|---|---|---|
| The pot | Spenders' gross margin rate without the refund, against a year earlier, two quarters after the last refund | within 20 basis points of the prior year, or above, in either quarter, and the program over | 30 or more below in both quarters, both spenders | June 30, 2027 |
| The surcharge split | Share of itemizers reporting a customer return, less the share of folders | gap of 30 points or more, with 8 or more in each group | gap under 10 points | June 30, 2027 |
The terms for Target and Best Buy, as the October text set them and as neither amendment changes them: each is read in the two quarters named above, and our expectation is contradicted if Target announces a price program it ties to its refund, if Best Buy ties its appliance investment to its refund, or if either company's margin without refunds (for Best Buy, its Domestic gross profit rate) falls below the year-earlier rate in both quarters.
Amendment 1, dated November 6, 2026. Made after the fact-check found limits in the frozen test and before any of the four had reported a new quarter; we checked EDGAR that evening, and none had. Walmart is scored on the two quarters after its price investment ends, which its own quarterly report expects to continue through fiscal 2027: the quarters ending April 30 and July 31, 2027, reported, by our estimate, about May 20 and August 19, 2027. A disclosed refund is taken out of both quarters compared, so Walmart's second quarter is set against last year's refund quarter without its refund. Fuel and gasoline adjust a quarter only as each company itself states them in basis points: Costco's change excluding gasoline price inflation is used; Walmart has given fuel only as a direction, leaving its figures unadjusted. Costco is scored from the last refund it reports, and if that pushes its second quarter past the final date the test's verdict is "untested," a test not taken, whatever Walmart shows; that is likely, since Costco has about $329 million of some $500 million back and says more will come. Walmart's second quarter alone moves the first test's final date, to August 31, 2027. The second test, Target's and Best Buy's terms, and every line and threshold stay as frozen. The amendment is fingerprinted and committed like the freeze, and the scoring program stops if a character of it moves.
Amendment 2, dated November 11, 2026. Committed before any of the four retailers had reported a new quarter. It changes three things and no threshold. First, a price program counts as ended only on an affirmative indication: the company says so, or a dated end to the program passes. Silence on a later call leaves the program's status unknown, never "ended." The frozen text had read a program as over when the company's documents for the second quarter no longer described it as continuing, which would have let silence confirm. Second, if Costco's last refund comes too late for its window, Walmart is scored alone and Costco is "untested." Amendment 1 had made the whole first test "untested" in that case, so that one late refund would have voided Walmart's result as well. Third, the claim is restated as what the test can observe: whether refund-linked price investment ends when the refunds do. Mental accounting is one possible explanation of that answer, beside a refund relaxing an earnings constraint for a time, and the test does not tell them apart. To match, the verdicts are now called "confirmed" and "falsified" on the same lines as before; a falsification counts against both explanations, since each predicts the cuts fade. A date through which a company expects a program to continue is not an end date, so Walmart's expectation that price investment continues through fiscal 2027 does not by itself end its program, and as of today neither company has affirmed an end. Where an unaffirmed end would decide the verdict, the verdict is "untested"; silence can hold back a confirmation, never a falsification. The 20 and 30 basis point lines, the quarters, the final dates and the second test stay as they were.
Fingerprints. Each block of the scoring program is identified by its SHA-256 fingerprint, a string that changes if any character of the block changes. The freeze of October 24, 2026: d6f3307765ca6c36bb8593547b4b6dfd13c4db8648f998a21fb0bbbdc12fb00f. Amendment 1: cb4fc4b50a70c31f8481f0cf176207be309ae6edc64d5eabcbb99fd305fa3dca. Amendment 2: 4076971e27c1ce9c4e86994e8bacfcf19ca656a8b01a1ffd7b8e0e0fa694064a.
The 1936 record. In 1933 Congress passed the Agricultural Adjustment Act, the New Deal's plan to raise farm prices by paying farmers to grow less, and it paid for the plan with a processing tax: a levy on the first processing of a commodity, so many cents on every hundred pounds of hogs slaughtered, every bushel of wheat milled, every pound of cotton spun. The Supreme Court struck the whole scheme down in United States v. Butler, decided January 6, 1936. The Court's words for the tax and the spending it funded were short: "They are but means to an unconstitutional end." By then the processors had been fighting the tax in the lower courts for months, and large sums sat impounded, held by the courts pending the decision. In the weeks after Butler the courts returned them to the processors, and unpaid taxes went uncollected. A report prepared inside the Department of Agriculture six weeks later put the amount at over $300,000,000, about $180,000,000 impounded and released and more than $120,000,000 never paid, ninety percent of it in the hands of three groups of processors; the Secretary of Agriculture, Henry Wallace, called the return "an outright gift of public money." That report is the administration's own advocacy, written by the side that wanted the money, and its figures are the Department's own; I use it for what it measured and leave its verdict to it.
What it measured was the margins. L. H. Bean, the economic adviser of the Agricultural Adjustment Administration, set the processors' gross margins, the gap between what they paid for the commodity and what they sold the product for, beside the tax, month by month. In hogs the margin had run about 61 cents a hundred pounds in the first halves of 1931 to 1933; with a processing tax of $1.93 in the first half of 1934 it rose to $2.52, and with a tax of $2.25 in 1935 to $2.91. Then the tax ended, and in January 1936 the margin was "almost immediately reduced to its former amount." Wheat and cotton goods told the same story. Bean's point was that the processors had collected the tax from their customers and were now being handed it a second time. His tables measure the processors' margins and not shelf prices, and by his own account the tax had come out of consumers or producers, so its end reached the shopper, the farmer or both.
Congress answered on June 22, 1936, in the Revenue Act of that year. Title VII governed the refund of processing taxes. Its section 902 said that no refund would be made "unless the claimant establishes to the satisfaction of the Commissioner" that he "bore the burden of such amount and has not been relieved thereof nor reimbursed therefor nor shifted such burden, directly or indirectly," including "through inclusion of such amount ... in the price of any article," or else that he "has repaid unconditionally such amount to his vendee (1) who bore the burden thereof," the vendee being the buyer. Title III, styled a tax on unjust enrichment and known at once as the windfall tax, levied in its section 501 "a tax equal to 80 per centum of the net income from refunds or credits to such person from the United States of Federal excise taxes erroneously or illegally collected with respect to any articles, to the extent that such net income does not exceed the amount of the burden of such Federal excise taxes with respect to such articles which such person shifted to others." A subcommittee had wanted 90 percent; a hundred, it was thought, would be confiscation.
Keith Bondurant, writing in the Michigan Law Review in January 1939, called the two titles complementary, and the purpose, in a phrase he quotes from the Harvard Law Review, was to "tap the windfall in its every retreat." In his words the titles were "stripped of any pretense of protecting the purchaser." A processor could escape the windfall tax by having repaid his customers in full, but only by June 1, 1936, three weeks before the act became law. The buyers who went to court for a share were turned away, on the ground that the processor, not the buyer, had paid the tax; a practical defense of that result, which Bondurant quotes from an earlier note, was that if the wholesaler could recover, so could the retailer who bought from him, "and finally the individual consumer," and the allocation would be too intricate for a court.
The President's revised budget for the fiscal year 1937, summed up in a statement of September 2, 1936, counted on $82,000,000 of windfall tax, which it called non-recurring. The Treasury's annual report for that year shows the unjust enrichment tax collected at $6,073,351.02. Some 17,000 returns were filed; about 55 percent of them showed no tax owed; and roughly 170 suits were brought asking the courts to stop the tax as unconstitutional. The next two fiscal years brought in about the same, $6.2 million and $6.7 million, so that by June 1939 the tax had collected about $19 million, against the $82 million counted on for its first year alone. The records I read do not say how much of the gap the accountants won and how much the lawsuits and extensions delayed. It has been done another way since: in 2006, after five federal appeals courts held that the excise tax on long-distance telephone service no longer applied to service as it was billed, the Internal Revenue Service let the people who had paid it, on bills from March 2003 through July 2006, claim it back with interest on their 2006 income tax returns, so that the money went around the companies whose bills had carried the tax.
The figures in this inquiry are computed in code from public documents, except six that are our own arithmetic: the 40 percent of Target's earnings per share ($1.65 divided by $4.11), the 7 percent of the windfall tax estimate ($6,073,351.02 divided by $82,000,000), the 39 percent of Varex's refunds set aside for customers ($6.6 million divided by $17.0 million), the $19 million the windfall tax collected by June 1939 (the three fiscal years added), Costco's refund-funded cuts of about 10 basis points (its 19 basis point refund less the 9 its finance chief gave as the net benefit) and Costco's $329 million back ($174 million and $155 million). The refunds, the gross profit, the net sales and the cost of sales come from each company's earnings release and quarterly or annual report as furnished to the Securities and Exchange Commission; the margin rate without the refund takes the refund booked in cost of sales out of gross profit and divides by net sales as reported, for the quarter the refund arrived in, against the same quarter a year earlier. Target's own stated rates are reproduced to a tenth of a point, and its "approximately 100 basis points" comes out at 94 on its reported figures; Walmart's, Costco's and Best Buy's margins without the refund are ours, derived from their statements, as "What was left at the register" says where they are computed, with Best Buy's computed for its Domestic segment. Costco's gasoline and inventory accounting (LIFO) effects are the company's own figures from its call. The refund as a share of sales divides each refund by the quarter's revenue. The two tests' thresholds are constants in model.py, frozen at the time stated in the text and committed to the inquiry's repository then; Amendment 1 of November 6, 2026 and Amendment 2 of November 11, 2026 are further blocks, each written after the freeze check and committed the same way, naming only what it changes; the program checks a fingerprint of each block and stops if a character of any has changed, the frozen rules score everything the amendments do not name, and the verdict rules pass known-answer checks. The 1936 figures are transcribed from the Statutes at Large, the Senate document carrying the Department of Agriculture's report, the Treasury's annual reports for fiscal 1937 and 1939, the President's budget statement of September 2, 1936 and Bondurant's article. The examples in "Other money with names on it" are quoted from each company's own filing or call and are not scored. The other calculations are in model.py, charts.py and amzn_msft.py, and the facts behind each input, with links, are kept with the inquiry's working files.
The Supreme Court struck down one set of tariffs in February 2026, and the government is refunding them to the companies that paid. Walmart, with about $2.9 billion back, and Costco announced that refund money would go into lower prices. Target, which reported $994 million, told shareholders the quarter was better measured without the money. Target and Best Buy announced no price program paid for with refunds.
It takes the disclosed refund out of gross profit and compares each spender's gross margin rate with the same quarter a year earlier, in the two quarters after its last refund. Our guess, that the cuts end with the money, is confirmed if Walmart and Costco each come within 20 basis points of the year-earlier rate, or above it, in at least one of the two quarters and each one's refund-tied price program is over. If they sit 30 basis points or more below in both quarters, that counts against the guess. The test is settled by August 31, 2027. We think there is about a 60 percent chance it comes back confirmed, if it reaches a verdict. It cannot say why the cuts end, or why margins stay down.
When the Supreme Court struck down the New Deal's processing tax in January 1936, the processors' gross margins fell back within weeks. The wheat margin had been 38 cents in 1932 and 68 cents in 1934, with a tax of 30 cents; in January 1936 it was back to 36. More than $300 million of taxes held by the courts or never paid stayed with the processors. The windfall tax Congress passed in June 1936 collected $6,073,351.02, about 7 percent of the $82 million the budget counted on.
This inquiry is research, not investment advice. It states what the filings and models show under labeled assumptions; nothing here is a recommendation to buy, sell or hold any security, and the decision belongs to the reader. It takes no side in the question whether any refund is or is not owed to anyone, and no view of the Target complaint or of any related case. Russ W. Rosenzweig, Vista's founder, owns shares of Amazon, Microsoft and Walmart and holds options on Amazon and Microsoft; he owns no shares of Target, Costco, Best Buy, Varex Imaging, FIGS, FedEx, Williams-Sonoma, Kohl's, Mettler-Toledo or Delta Air Lines. Holdings through mutual funds and exchange-traded funds are not counted. How this inquiry was made: written with the help of AI tools. The facts were gathered from the four retailers' releases and filings, the statute, the Senate document and the Treasury's reports, the models were built and the figures computed in code or, for the six named in "How the figures were made," by our own arithmetic, each public source was checked against its original, and the text was fact-checked and read for plainness. I read it and listened to it before it was published. Why I write with AI. Why I publish these inquiries.