Transcription
The Dutch sandwich. You are an accountant, not a billionaire's accountant. You're accountant. You were sitting at a kitchen table in March gathering receipts trying to figure out whether you qualify for a $500 deduction on your home office. You spent 45 minutes reading IRS publication 587. You will pay whatever you owe because that is how the system works.
Somewhere in a glass building in Amsterdam, a different accountant is finishing the last page of a 600-page filing. That explains how one of the largest corporations on Earth paid an effective tax rate of 2.4% on $23 billion in overseas profit. The structure is called the double Irish with a Dutch sandwich and it is not illegal. It is engineered.
Here is how it worked for Google documented in detail by Bloomberg in 2017. Google Ireland Holdings, an Irish registered company controlled from Bermuda, collected royalties on intellectual property licensed across most of the world because Irish law at the time defined a company's tax home by where it was managed and Bermuda has no corporate income tax. This company existed legally in Ireland but paid taxes in Bermuda where the rate was zero but the money couldn't flow directly from Europe to Bermuda without triggering withholding taxes.
So, it passed through a Dutch intermediary, a Netherlands holding company, because the Netherlands has no withholding tax on royalty payments to non-EU countries. The Dutch company sat in the middle of the structure doing nothing except preventing a tax event from occurring. It was a legal fiction with a street address and a filing cabinet.
In 2017 alone, Google routed $23 billion through this structure. At that year's US corporate rate of 21%, the tax owed on that income would have been approximately 4.83 billion dollars. Google paid roughly $552 million a gap of $4.27 billion in a single year from a single structure at a single company.
The Google accountants who built this did not hack a system. They read the system exactly as written and found the gaps between jurisdictions, the seams where one country's law ends and another's begins and where in that seam billions can disappear without technically breaking a rule. The EU eventually Ireland to close the double Irish loophole effective 2020. Google ended the structure and immediately deployed alternative arrangements involving Singapore and Puerto Rico that tax lawyers are still mapping. The structure changed. The principle did not.
The receipt on your kitchen table is for $47.80 in printer ink. You are going to spend another 20 minutes confirming it qualifies. The difference between your tax rate and Google's in 2017 was not a matter of resources or sophistication. It was a matter of which accounting system you were allowed to access.
Buy, borrow, die strategy. You earn a salary. At the end of the year, you have paid income tax on it, federal, state, possibly local. The government knew the moment the money appeared in your account. Your employer reported it. The deduction was automatic. You did not choose when to be taxed. The system chose for you.
Now, imagine you are Larry Ellison. Your net worth as of 2023 sits above $140 billion, nearly all of it in Oracle stock. You have not drawn a significant salary in years. You do not need to. Instead, you borrow. Major banks, Morgan Stanley, Bank of America, extend you personal credit lines at interest rates of 1 to 3%. Using your stock portfolio as collateral. You receive the cash. You spend the cash. You pay no income tax on the cash because a loan is not income. It is not income because you will, in theory, pay it back.
The loan does not need to be repaid your lifetime. When you die, your heirs inherit the stock at its current market value, a mechanism called the stepped-up basis. The original purchase price of the stock and the decades of gain between that price and the death day value is simply erased for tax purposes. Your heirs sell the stock to pay off the loan. They pay capital gains taxes only on appreciation since your death, not on the entire lifetime gain. The loan is retired. The tax on 50 years of wealth accumulation is close to zero.
The strategy has a name in estate planning circles. Buy, borrow, die. It is not a secret. It is taught in wealth management programs. It is not illegal. It is structural.
In 2021, ProPublica obtained a trove of confidential IRS data covering the 25 wealthiest Americans over a multi-year period. Their analysis found that these individuals, with combined wealth of roughly $1.1 trillion, paid a true tax rate of 3.4% on wealth growth. Jeff Bezos paid a true rate of 0.98% in certain years. Elon Musk paid 3.27%. Warren Buffett, who has publicly stated that his tax rate is lower than his secretary's, paid 0.10%.
These numbers are not the result of fraud. They are the result of a tax code that was written by people who understood the difference between income and wealth and that shows repeatedly over decades of legislation to tax income and largely not to tax wealth. The people who wrote those laws received campaign contributions from people with wealth. The people with wealth hired lobbyists who explained in precise technical language exactly what language the lawmakers should use.
Your salary appeared in a box on a form. Their net worth expanded by $200 billion. Only one of those events triggered a tax obligation.
Carried interest. You are a surgeon. You work 60 hours a week. You have $1.2 million in student debt from medical school. After expenses, malpractice insurance, and the administrative overhead required to work within the hospital system, you take home roughly $280,000 a year. The federal government taxes that as ordinary income. At your bracket, you pay 37% on the top portion. You have done this for 20 years.
Stephen Schwarzman, the co-founder of Blackstone, the world's largest alternative asset management firm, earned $1.1 billion in 2021. He paid an effective rate significantly lower than yours, not because of legal gray areas, not despite the law, because of a specific provision of the tax code called carried interest. And because that provision has survived 14 separate congressional attempts to eliminate it over the past 20 years.
Here's the mechanism. When a private equity fund makes money for its investors, the fund managers take a carry, typically 20% of profits. This is their compensation for managing the fund. It is functionally wages, a fee for a service rendered. But the tax code classifies it as a capital gain, not ordinary income, because the fund itself holds capital assets. The who got this classification means the managers pay the long-term capital gains rate, 20% rather than the ordinary income rate of up to 37%. The gap on a $1 billion pay package, approximately $170 million. Not a rounding error. Not an incidental benefit. A structural advantage for a specific profession written into the permanent tax code.
Senator Joe Biden called the carried interest loophole the biggest scam in the tax code during the 2008 campaign. The Obama administration proposed eliminating it in 2009, 2010, 2012, and 2013. The Trump administration's 2017 Tax Cuts and Jobs Act contained a modest modification extending the required holding period from 1 year to 3 that tax lawyers circumvented within months using a restructuring technique called a cliff vest. The Biden administration proposed elimination again in 2021. The Inflation Reduction Act of 2022 included a carried interest provision that was stripped out in final negotiations at the insistence of Senator Kyrsten Sinema, who received $500,000 in campaign contributions from private equity interests in the year before the vote. The provision survived.
It has survived every serious attempt since 1993 when it was first identified as a potential reform target. In 30 years, it has transferred an estimated $180 billion in tax liability away from private equity managers and onto the general tax base, meaning predominantly onto wage earners. You finished your surgery at 11:00 p.m. on a Tuesday. You saved a life. The system classified that as ordinary income. The people who classify things decided what your work was worth.
The Zuckerberg GRAT. Your grandmother wants to leave you money. She has $200,000 in savings. The federal estate and gift tax exemption covers a substantial amount, but above certain thresholds, the government takes 40 cents on every dollar she transfers to you. She can give you $18,000 per year tax-free under the annual gift exclusion. At that rate, transferring $200,000 takes 11 years. And the interest on the money she holds barely keeps pace with inflation. She does not have a team of attorneys. Mark Zuckerberg does.
In 2008, when Facebook was not yet a public company and its stock was privately valued, Zuckerberg used a structure called a grantor retained annuity trust, a GRAT, to transfer a substantial number of shares to an irrevocable trust. Here's why this maneuver is specifically designed around the timing of a company's growth.
A GRAT works as follows. You transfer assets, stock in this case, into the trust. You receive annuity payments back from the trust for a fixed term. Those payments are calculated using an IRS-set interest rate called the 7520 rate. If the assets in the trust appreciate faster than that rate, the excess appreciation passes to your heirs tax-free. The IRS rate in certain years has been as low as 0.4%. If you put in stock valued today at one unjed and it becomes $100 in five years, nearly 90 lined perishably and ideal is asset of that gain moves to the next generation at essentially zero gift tax.
The strategy is called zeroing out the GRAT. You set the annuity payments to return almost exactly the initial value plus the IRS interest rate, making the taxable gift effectively zero at the time of transfer. The gamble is on appreciation. If the stock goes up, the gain escapes taxation. If it doesn't, you just get your annuity back.
Sheldon Adelson, the casino magnate, transferred an estimated $7.9 billion to his heirs using GRAT structures. Bloomberg's 2013 investigation found that at least nine members of the Forbes 400 had used GRATs to move combined fortunes exceeding $100 billion at near zero gift tax cost. The IRS has been aware of the strategy since it was first employed in the 1990s. Congress has repeatedly proposed a 10-year minimum GRAT term, which would reduce the ability to zero out by timing transfers around anticipated stock spikes, but the provision has not passed.
The GRAT is legal. It was created by a provision of the 1990 Tax Act, ostensibly to address a different problem. Wealthy families and their lawyers found the loophole within months and have used it consistently ever since. The Treasury estimates it cost the federal government $3.4 billion per year in lost revenue. Your grandmother's $200,000 will be taxed on transfer. The $7.9 billion was not. The difference was not effort or intelligence. The difference was access to the page of the tax code where the engineers left the door open.
Dynamic pricing on the poor. You need car insurance. You fill out the online form. You enter your zip code, your vehicle, your driving history. You have had no accidents in seven years. You are a careful driver. The algorithm returns a number. You pay it because you need the insurance and because you have no meaningful way to evaluate whether the number is fair.
20 miles away in a zip code with slightly higher median income, someone with an identical driving record and identical vehicle gets a quote that is $400 per year cheaper. The difference has nothing to do with their risk. It has to do with their zip code and more specifically with a data model that has determined how likely you are to comparison shop. This practice was documented extensively by ProPublica in its 2017 investigation into auto insurance pricing across six major companies in California, Illinois, Texas, and Missouri. The investigation found that in some zip codes, drivers in minority communities were charged as much as 30% more than white neighborhoods with statistically equivalent or higher accident rates. The variable was not risk. The variable was market power. The algorithm's estimate of how likely the customer was to leave.
The academic framing comes from Kaiser Fung, the statistician who wrote Numbers Rule Your World. Fung documented the broader principle. When companies have access to behavioral data, they use it to price discriminate not only by risk, but by price elasticity, your sensitivity to price increases. People who comparison shop get the competitive price. People who don't because they don't have time or don't know the market or don't have internet access or live in an area with fewer agents pay the extracted price.
In 2020, a study by the Consumer Federation of America found that a good driver in a low-income urban zip code paid on average $681 more per year than an identical driver in a higher income zip code from the same insurer. Extrapolated across tens of millions of low-income drivers, the excess premium represents a multi-billion dollar annual transfer from the people least able to pay to the shareholders of the companies with the best data infrastructure.
The system is not a conspiracy. There is no meeting room where executives decide to charge poor people more. There's a pricing algorithm that has been trained on behavioral data and the behavioral data reflects that low-income customers, for structural reasons, are less likely to switch. The algorithm found the signal. The algorithm is acting on the signal. The signal is you. The insurance policy you just bought is priced to the limit of what the model believes you do.
The Walmart poverty subsidy. You are a taxpayer. You file your return each April. You pay for roads, public schools, the military, Medicaid. You do not choose where those dollars go. The government allocates them based on need and policy. This is the social contract. The community pools resources to support things individuals cannot provide for themselves. You are also, whether you knew it or not, subsidizing the labor costs of one of the world's most profitable corporations.
A 2014 report commissioned by Americans for Tax Fairness, using data from the House Committee on Education and the Workforce, calculated that Walmart employees in Wisconsin alone received approximately $900,000 per year in Medicaid benefits, $678,000 in food stamps, and $105,000 in subsidized housing assistance per store. Nationwide, the organization estimated that Walmart's workforce received $6.2 billion per year in government assistance programs.
The logic is this. Walmart's average wage at the time was approximately $8.81 per hour for full-time associates, below the federal poverty line for a family of four. Employees earning below poverty wages qualify for Medicaid, food assistance, and housing subsidies. Those programs are funded by tax revenue. Walmart employs 1.5 million workers in the United States. The gap between what Walmart pays and what its workers need to survive is filled by the public treasury.
Meanwhile, the Walton family, the heirs to Sam Walton, held a combined fortune of approximately $250 billion in 2014. The six Walton heirs collectively held more wealth than the bottom 40% of American households combined, as documented by the Economic Policy Institute. Walmart's dividend payments to shareholders in 2013 totaled $6.1 billion. The government subsidy to Walmart's workforce was $6.2 billion. The numbers were within rounding error identical.
Walmart disputes the characterization. Company representatives have argued that they provide jobs to workers who would otherwise have none, that their wages are competitive with the retail sector, and that workers are free to seek employment elsewhere. These arguments are accurate and miss the point entirely. The point is not that Walmart broke a law. The point is that the design of the wage floor, combined with the availability of public assistance, creates a system in which below subsistence wages are viable for a corporation because the gap between wages and subsistence is paid by people who don't work at Walmart. The subsidy is structural. It is invisible. It is paid by you. Every time you bought something at Walmart because it was cheaper than the competitor. Part of that price difference was funded by your tax return.
The Berkshire Hathaway tax strategy. Warren Buffett is the most publicly self-critical billionaire on the subject of taxation. In a 2011 New York Times op-ed, he wrote that his effective federal tax rate was 17.4% lower than any of the 20 people who worked in his office, whose rates ranged from 33 to 41%. He called for higher taxes on the wealthy. The op-ed was widely praised. The strategy that produces his 17.4% rate remains intact.
Berkshire Hathaway has not paid a dividend since 1967. This is not an accident. It is the foundational tax strategy of the entire enterprise. When a company pays a dividend, that payment is taxable income for shareholders in the year they receive it. When a company retains its earnings and reinvests them, the shareholder does not receive cash and therefore has no taxable event. The value of the shareholder's stake increases, but unrealized gains are not taxable income under US law. They are not taxed until the shares are sold.
Berkshire's Class A shares were trading at approximately $8 in 1965. They traded above $500,000 per share in 2023. A shareholder who bought in 1965 and held has paid no federal income tax on the appreciation of that $8 into $500,000 for 58 years. The gain is deferred indefinitely. If they die holding the shares, the stepped-up basis rule eliminates the entire gain from the tax base.
This is Warren Buffett's strategy. It is not a secret. It is explained in Berkshire's annual letters, in business school curricula, and in every serious analysis of Buffett's investment philosophy. The tax efficiency of the buy and hold forever strategy is documented and central. The company is structured deliberately to never distribute taxable income to shareholders.
The ProPublica analysis of Buffett's own tax situation found that between 2014 and 2018, his wealth grew by an estimated 24.3 billion. He paid $23,207,000 in federal income taxes during that period. A true tax rate of 0.10%. The 17.4% figure he cited in his op-ed was his rate on the income he did report, salary and some dividends from other investments. His rate on total wealth growth was a fraction of a fraction of a percent. Buffett is correct that tax rates on the wealthy are, as a policy matter, arguably too low. He is also correct that he personally benefits from those rates more than almost anyone alive. Both of these things are simultaneously true, and neither is a contradiction. They are the system working exactly as designed. He wrote the op-ed in 2011. His effective rate on wealth accumulation in 2023 is functionally unchanged.
Payton Trolling. You are building a mobile app. You have a team of six engineers. You have been working for two years. You have raised $3,000,000 in seed funding. You are three months from launch. A letter arrives from a law firm you have never heard of representing a company called something like Unified Communications Technologies LLC or Acacia Research Corp or Marathon Patent Group. The letter informs you that your app infringes on US Patent 7,349,921, a patent for methods of transmitting and receiving data in a networked computing environment, and that you owe licensing fees. The licensing fee demanded is $400,000. Alternatively, you can defend yourself in federal court, which will cost your startup approximately 1.5 to 3 million dollars in legal fees over two to three years.
You have not infringed on anything. The patent is written in language so broad that it could apply to almost any software that transmits data over a network, which is nearly all software, but proving this in court costs more than settling. Your investors are watching the legal risk. Your runway is 18 months. You pay.
Nathan Myhrvold, the former chief technology officer of Microsoft, founded Intellectual Ventures in 2000 with the stated mission of funding and monetizing invention. By 2012, the company had accumulated a portfolio of over 70,000 patents and had generated $3,000,000,000 in revenue, primarily through licensing fees and litigation threats, not through developing any products based on the patents it held. Myhrvold has argued that he is simply creating a market for intellectual property, enabling individual inventors to monetize their work by aggregating patents into a portfolio that can be licensed efficiently.
Critics, including former business partners, academics, and the companies that paid licensing fees, described a different function, systematic extraction of money from operating companies using the cost asymmetry of patent litigation as leverage. A 2012 study by Boston University School of Law economists James Bessen and Michael Meurer estimated that patent assertion entities, entities that hold patents to license rather than to build, cost the US economy $29,000,000,000 per year in direct costs and potentially hundreds of billions more in suppressed innovation. The study documented 5,842 defendants in patent troll lawsuits in 2011, up from 1,401 in 2005. The majority were small and medium-sized companies. Fewer than 25% were large corporations with resources to litigate.
Congress passed the America Invents Act in 2011, which modified the patent system in several ways. It did not meaningfully reduce patent assertion activity. The rate of NPE filed lawsuits continued to increase through 2015 before stabilizing. The fundamental economics, settlement cheaper than litigation, remain unchanged. Your app launched late. You spent the settlement money instead of hiring two engineers. The patent that stopped you was filed in 2003, has never produced a product, and is owned by a Delaware LLC whose principals are not disclosed in its public filings.
The art storage strategy. Somewhere in a climate-controlled warehouse near Geneva's Cointrin Airport, in a facility that spans the equivalent of 13 city blocks underground, there is a painting. It is a masterwork. It may be a Picasso, a Modigliani, a Basquiat. It was purchased at auction for perhaps $100,000,000. It has not been displayed publicly. It has not been seen by anyone except the logistics staff who processed it into storage and the security teams who rotate shifts in its vicinity. It is not for sale. It is not for display. It is a financial instrument.
The Geneva Freeport, officially the Geneva Freeports and Warehouses, is a duty-free zone under Swiss federal law, operating since 1888. Goods stored there are considered to be in international transit and are therefore exempt from import taxes, value-added tax, and capital gains taxes. The facility currently holds an estimated $100 billion in art, plus gold, wine, and other high-value portable assets.
The strategy, documented by The New York Times in 2016 and by art market analyst Claire McAndrew in multiple industry reports, works as follows: A collector purchases a work in New York. Rather than importing it to their home country and paying import duties and sales taxes, they ship it to Geneva, where it enters the free port in transit status and incurs no taxes. They hold it there, sometimes for years, sometimes decades. When they wish to sell, the transaction may occur within the free port itself as sale in bonded storage that passes title without the artwork ever crossing a taxable border. The buyer receives the work, the seller receives payment, no country receives tax revenue because the artwork never technically left transit.
But buyers Rybolovlev, the Russian billionaire, purchased Amedeo Modigliani's Nu Couché for $118 million in 2015. The work was stored in Geneva. He later sold it in a private transaction routed through Singapore. Tax paid in any jurisdiction disputed, but experts citing the transaction structure estimated it at or near zero.
The art market is estimated by Artnet at approximately $67.8 billion in annual global sales. A substantial fraction of high-end transactions are structured through free port channels. The Swiss Federal Customs Administration has acknowledged that the volume of assets stored in free ports makes comprehensive auditing functionally impossible. The European Parliament passed a resolution in two thousand and nine hundred six calling for greater transparency. The Swiss government has implemented some reporting requirements. The fundamental structure, goods in permanent transit tax perpetually deferred, title changing hands inside a bonded warehouse, remains intact. The painting in the Geneva warehouse has not been seen in eight years. It is, by some measures, one of the most important works of art created in the 20th century. It is also, by any measure, one of the most efficient tax vehicles money can buy.
Opportunity zone exploitation. The 2017 Tax Cuts and Jobs Act contained a provision that its architects described as a way to direct capital investment into economically distressed communities. The mechanism, opportunity zones, allowed investors to defer and reduce capital gains taxes by reinvesting profits into designated low-income census tracts. If held for 10 years, gains on the new investment would be completely tax-free.
The stated purpose was to address a documented problem. Investment capital concentrates in already prosperous areas and economically distressed communities struggle to attract private funding. By providing a tax incentive specifically tied to geographic investment in low-income areas, the provision was designed to redirect some portion of the $6 trillion in unrealized capital gains held by American investors toward places that needed development.
In 2019, ProPublica published an investigation into how the opportunity zone designations had actually been applied. The findings were specific. In Baltimore, one designated opportunity zone contained the headquarters of Under Armour, a Fortune 500 company with $5 billion in annual revenue. Under Armour had announced a major campus expansion before the zone was designated. The investment that would have happened anyway now qualified for billions in tax deferrals. In New Orleans, a designated zone contained a luxury hotel development in a rapidly gentrifying neighborhood whose median income had been rising for years before designation. In Story County, Nevada, a county with a population of fewer than 4,000 people, a zone was designated that encompassed a massive data center campus being built by technology companies. The investors building the campus were already committed. The tax benefit was retroactively attached to an investment that required no incentive.
The Treasury Department's Inspector General found in a 2020 report that the IRS could not track whether opportunity zone investments actually benefited low-income residents because the law contained no requirements to measure community impact. An investor who builds a luxury condominium tower in a designated zone qualifies for the same tax treatment as one who builds affordable housing or a community health clinic.
Senator Tim Scott of South Carolina, one of the provision's architects, has defended opportunity zones as generating significant investment in distressed areas. Independent analyses from the Urban Institute and the Brookings Institution have found that investment has concentrated disproportionately in zones that were already gentrifying or that contained large commercial developments and that demonstrably distressed neighborhoods, the ones the provision was ostensibly designed for, have received a smaller share of total capital.
The mechanism is not broken. It is working exactly as the incentive structure shapes behavior. It rewards capital deployment in designated geographies without specifying what kind of capital or what kind of development. People with capital responded rationally to the incentive. They deployed it where returns were highest, which was, in most cases, not in the most distressed neighborhoods. The provision was designed with good intent. It was drafted without enforcement mechanisms. The people who wrote the drafting language understood the gap. Some of them had clients waiting to use it. The community that the zone was meant to help is still waiting for the investment. The fund that used the zone to defer $40 million in capital gains already closed its books.