Who Pays, Who Collects

The view is that a city provides services to people who do not pay for them. A commuter uses the roads, the police and the transit of the city where she works, and pays her income tax to a state and her property tax to the town where she sleeps. A fan who flies in for a game uses the same streets as the residents. An incentive brings more of both, and the government that wrote it pays to serve them. Whether the investment pays off depends on whether that government can collect from the people it serves.

Use matters. A driver passing through dozens of jurisdictions on the way from Virginia to New Jersey pays for none of them directly, but each uses only a little of the road. A commuter uses one city every day. Congress bars the District of Columbia from taxing the income of the people who work there and live elsewhere. [1]

This piece started with a map. FilmHedge (an Osparna portfolio company; the full disclosure is below) published a guide to film and television tax credits in 38 states, and the map sorts the states by how the money reaches the producer, not by how large the credit is. In 12 states the credit is refundable, which means the state pays cash. In 13 it is transferable, which means the producer has to sell it to a buyer. Nine pay a cash rebate. Four do something in between. [2]

The percentages alone mislead. Take a $10 million film with $8 million of qualifying spending. In Georgia the base credit, 20 percent, has a face value of $1.6 million. In Massachusetts it is $2 million. Face value is not cash. FilmHedge's own tool says the credit and the cash "remain different measures," because a transferable credit must find a buyer and the sale price comes off the top. [3] In Louisiana, a producer who hands the credit back to the state gets 88 percent of its face value after the fee. [2]

Then there is the cap. Indiana's credit can reach 30 percent, but the state limits the whole program to $2 million. Georgia has no cap at all. [3]

What the state gets back is a different question. Georgia's auditors estimated that the credit returns 19 cents to the state for each dollar it costs, and that it supports about 34,000 jobs, where an industry-funded study had claimed 59,700 jobs for fiscal 2022. The credit costs the state about $1.3 billion a year at face value, counted when the credits are issued.

What comes back is smaller. In fiscal 2024 the auditors estimated that the film activity brought the state $225 million in tax revenue. Take out the $80 million the state would have collected had it spent that money elsewhere, and $145 million is left, against $763 million in credits claimed that year. [4]

The auditors left out wages paid to people who do not live in Georgia, because that money does not stay in the state's economy. That is the commuter's problem seen from the other side: the state paid for the production, and part of what it produced crossed the border. [4]

In my reading, film is the plainest case of what this piece is about. The same word, incentive, covers deals that cost a state very different amounts and pay a producer very different amounts. The difference is in the contract, not the percentage. It is also in whether the state can collect from the people the production brings.

A stadium shows the same bet over time. A government gives land, pays part of the construction or gives up taxes, and expects jobs and spending to return the money before the bonds come due. Suppose the public share is $500 million, financed at 4 percent over 20 years. The payment is about $36.8 million a year. Suppose also that each new job returns 5 percent of its wage in taxes to the government that paid the bill, after the cost of serving the worker. A $60,000 job then returns $3,000 a year, and paying the bill takes about 12,000 new jobs. Where there is no local income tax, taxes on earnings go to the state. A city or county collects property and sales taxes, which depend on where the worker lives and spends, so 5 percent is generous for a city. It gets muddier at a border. The author lived in New Jersey and worked in New York, filed in both states, and had to apply what one state refunded against what the other state asked for.

A stadium employs a few thousand people, and many of those jobs are part-time or moved from other work nearby. The bill goes to one government, and the taxes the jobs produce may go to another.

Economists who have studied this have not found that the bet usually pays. A review of the evidence by John Charles Bradbury, Dennis Coates and Brad Humphreys concludes that stadiums are poor public investments. Public spending on them keeps rising. The St. Louis Fed's explanation for why they don't pay is substitution: the money spent at a game is mostly money not spent at a movie or a restaurant nearby. [5]

The Dallas Cowboys come closest to a counterexample. Arlington, Texas, put up $325 million toward AT&T Stadium, which cost more than $1.2 billion, and paid for it with three taxes voters approved in 2004 and set aside for that debt: a half-cent sales tax, a 2 percent hotel tax and a 5 percent tax on rental cars. The Cowboys covered the rest. The city paid off the bonds in August 2025, about a decade early. [6] The team's value has gone from the $140 million Jerry Jones paid in 1989 to $17 billion this year, according to Forbes. [7]

The conditions were as good as they get. The Cowboys are loved, and fans fly in for games from all over the country. The author knows season-ticket holders who flew in from New Jersey for decades. A bigger stadium was almost certain to bring in more money, because the demand was already there and the seats were the limit. People who did not live there paid almost all of the hotel and rental-car taxes that repaid the bonds.

Even so, most of the new revenue went to the team. A study of Arlington's stadium financing found that the city exported much of the tax burden to its neighbors and lost taxable spending inside its own borders. [8]

Others have not done as well. Hamilton County, Ohio, paid about $455 million to build the Bengals' stadium, and local reporting put its total spending on construction, interest and operations since 2000 at more than $920 million. [9] Miami-Dade County borrowed about $409 million for the Marlins' ballpark, on public land the team did not have to buy, and the team pays about $2.3 million a year, rising 2 percent annually, to repay a county loan. [10]

Glendale, Arizona, financed the Coyotes' arena with $180 million of city debt, and in 2021 its city manager said an economist the city hired had found the city was better off financially without the team as a tenant. [11] In 2019, the Raiders' last season in Oakland, the city and Alameda County, which own the Coliseum together, still owed about $65 million on its renovation bonds, and neither the Raiders nor the Athletics owed any of it. [12]

In each, the public held the debt, and the team could leave.

Most of what a stadium produces goes to the people nearby. The author stayed at a budget hotel across the street from AT&T Stadium that was charging premium prices because of where it sat. The rooms, the cars, the restaurants and the airport capture the money, and a government sees only the part it taxes inside its own borders.

A private college does the same on a smaller scale. On parents' weekend, families rent hotel rooms and eat in the college town, and some of them fly in from across the country. No subsidy was needed to bring them, and the visitors come because of the college. In my reading, the subsidy at the stadium bought seats, not fans.

A factory is the hardest of the three. In 2017 Foxconn said it was considering seven states for a plant, and Wisconsin won with an incentive package of up to $3 billion for a promised 13,000 jobs. The governor of Wisconsin said at least one other state was prepared to give Foxconn more. [13] The flat-screen factory was never built. In 2021 the state revised the contract to up to $80 million for 1,454 jobs, and by late 2023 Foxconn had earned about $43.7 million in credits, with 1,029 employees in 2022. [14]

Studies of the whole market find the same pattern. Timothy Bartik reviewed 30 studies and concluded that incentives change a firm's decision in roughly 2 to 25 percent of cases, so for at least three in four firms the money bought nothing. [15] Cailin Slattery and Owen Zidar found that the average deal gives $178 million for 1,500 promised jobs, that the jobs do arrive, and that there is little evidence of broader growth beyond them. [16]

A government taxes everything that grows around an incentive, so it counts the money. It cannot say how much of it the incentive caused. A company that buys advertising on television, radio, online and billboards has the same problem: sales rise, and no one can say which channel did it.

A pastry shop opens near a studio, and no one can prove it would not have opened anyway. The audit counts that shop's jobs. Only about 4,900 of the roughly 37,000 jobs it attributes to the film credit were employed directly by production companies; the rest are an estimate of what the production set off. [4]

The 19 cents is an attribution number. To get it, Georgia's auditors had to choose how much of the film activity to credit to the incentive, and they took 92.1 percent from a film industry study, which says nearly all of it would vanish without the credit. [4]

Bartik asked the same question across 30 studies, and his answer for most firms was that the incentive changed nothing. [15] Neither number is a count. Each is a judgment about cause.

The best evidence on the other side is a study by Michael Greenstone and Enrico Moretti. They compared counties that won a large plant with counties that came second. Earnings in the plant's industry rose 1.5 percent and property values rose 1.1 percent, with no sign that local government finances were hurt. They concluded that subsidies for large plants do not reduce residents' welfare. [17]

In my reading, the jobs usually arrive. What is in doubt is what they cost, and whether anything grew beyond them. That depends less on the size of the incentive than on how the deal was written.

The contract matters more than the size of the deal. Two states can spend the same money and get opposite results, because the contract decides who pays, when the money moves, and what happens if the promise is missed. Deals are made in good faith, and contracts are still sometimes poorly drafted, by people who were rushed or did not know better. A good contract is drafted to protect both sides.

Arlington put the debt on taxes set aside for it. Wisconsin paid Foxconn only as jobs and investment appeared, and when Foxconn hired 178 people against the 260 it needed in 2018, it earned no credit for that year. [14] Economists who count these deals say the link is weakening: as subsidies have become more common, they are no longer as closely tied to what a firm invests. [16]

A good deal structure can be written down. It has nine terms. Most of them answer one question: how does the government that pays get paid back?

Pay as the jobs appear. A grant paid up front cannot be taken back.
Dedicate the money that pays the debt. Arlington's taxes could be used for the stadium bonds and for nothing else, so the general fund was never at risk.
Count the land at what it would sell for. If a buyer would pay fair market value, a gift of land is a cost, and it belongs on the bill. If no one would buy it, the gift costs little. That one question separates a bad gift from a good one.
Write the clawback before anyone signs. A promise with no penalty is a hope.
Write the rules for changing the deal. A 2019 University of Texas at Austin study by Nathan Jensen and Calvin Thrall of 165 Texas Enterprise Fund awards found that 46, about a quarter, had been changed after signing. The changes generally lowered job requirements or loosened deadlines, and many came shortly before a penalty would have applied. [18] A deal gets rewritten when the party that can leave asks. No change without public notice and a vote, and every amendment published.
Cap it. Georgia's film credit has no ceiling, and the state's cost grows with every production. Most states set one. [2]
Ask whether the incentive changed the decision. For at least three in four firms, Bartik found, it did not. [15]
Tax the people who benefit, and count only the benefit that lands inside your borders. A fan who flies into one city and spends in another pays the taxes of the other. New York City taxed what out-of-state commuters earned in the city, at 0.45 percent for most of them, until it was eliminated for them in 1999 after a court ruling. Proposals to bring it back have followed, including one in 2012. [19]
End it when the debt ends. Arlington's bonds were paid off in 2025, but the taxes stayed in place. [6] In 2026 the city approved using $273 million of those same venue taxes to help pay for improvements to the stadium. [20] A tax that outlives its purpose finds a new one.

In my reading, a state that writes all nine terms into the contract has made a bet it can check. A state that writes none has made a gift.

The same goes for the reader. The terms on paper are at face value, like a credit's, and what the public pays and gets back is set later. Do not take them at face value. Ask who can change them, and when.

At AT&T Stadium, fans could buy and sign bricks, and the bricks are set into the sidewalk outside. [21] The author gave a friend a brick with his name on it, and there are two, one at the stadium and one on his desk. Finding the first meant walking the stadium past thousands of names. Every one of those people had paid to be part of the building. In my reading, a state can write a contract for the building, but it cannot write one for the names.

Sources

Disclosure. FilmHedge, whose map started this piece, is a portfolio company of Osparna, the author's company. FilmHedge did not see this piece in advance. What is drawn from FilmHedge comes from its public guide, and the interpretation is the author's. The payback figures are an illustration with stated assumptions (4 percent financing over 20 years, 5 percent of a job's wage returned in taxes to the government that paid), not a forecast. The counts of states by credit structure are the author's tally from FilmHedge's guide. The comparisons and the terms of the deal are the author's interpretation. This is not investment advice.