HARTFORD, Conn. (AP) — For anyone betting on the Northeast's casino bonanza, the odds are long on projects hitting financial expectations.

In the last several states to open casinos — Ohio, Maryland and Pennsylvania — overall revenue is coming in below baseline forecasts, according to a review of state tax data. Officials blame miscalculations of spending habits and competition, but some also question how much the projected numbers reflected wishful thinking.

The casino industry has grown exponentially over the last decade as revenue-hungry states have moved to claim business that once went across state lines to Atlantic City, New Jersey, or the tribal-owned megaresorts in Connecticut. After Nevada, Pennsylvania has emerged as the country's No. 2 gambling marketing, overtaking Atlantic City, where four of 12 casinos closed last year.

As Massachusetts and New York prepare for a new round of casino building, they have added new levels of financial scrutiny, enlisting consulting firms to vet revenue projections. But the industry's growth in the Northeast's tight geography has made modeling more complex, and experts warn there are no guarantees.

"This isn't a science," analyst Alan Woinski said.

Projections are developed through so-called gravity models, premised on the concept that bigger casinos draw more people from farther away. They are used by developers and regulators to estimate how a property will perform based on factors including the affluence of surrounding towns.

The track record shows big margins for error. With access to the same data, developers regularly come back with higher projections than regulators who run the numbers themselves, especially when companies are competing for bids. A recent study by Cummings Associates, a Massachusetts-based consulting firm, found that projections done for the same project were, on average, 20 percent apart and, in cases where the casinos were actually built, almost always were proved too high.

Casinos generally remain big moneymakers, and some projects have far exceeded predictions, but state averages have been below forecasts that set expectations for tax revenue.

One of the biggest misses came in Ohio, where the state Department of Taxation weighed in on a proposed constitutional amendment in October 2009 with an estimate that casinos would generate at least $470 million in annual tax revenue. In 2014, tax revenue from the casinos totaled $267.5 million.

"In retrospect, we were guessing," said Mike Sobul, who was the tax department's director of research.

Sobul said the gravity model run by Ohio officials used industry assumptions that were overly optimistic, and, more significantly, officials underestimated how the recession would affect consumer spending. Sobul, who tracks the numbers now as a financial officer for an Ohio school district, said schools that once were expected to receive more than $80 per pupil statewide in casino money are now expecting about $51 per pupil.

In Maryland, casino revenue has been hundreds of millions of dollars short of a December 2008 projection by the Department of Legislative Services.

Even in Pennsylvania, the casino boom's success story, it took the addition of table games in 2010 to bring revenue beyond levels projected by a state task force that assumed they would offer only slot machines. Doug Harbach, a spokesman for the Pennsylvania Gaming Control Board, said the projections were not badly inflated.

"You have to take into account that the market in surrounding states now has much more competition," he said.

For all the attention to revenue projections, many industry insiders say the real test is whether banks, which conduct their own research, will loan money for a casino.

"If a bank is going to put up a hundred million dollars, that's a tacit show of feasibility," said Michael Ross, president of the Innovation Group in Winter Park, Florida, one of many consulting firms in the forecasting business.

Where some earlier states were often guided by in-house analyses, Massachusetts and New York have brought in more outside experts to conduct their own market research, assess the numbers from the developers and guide state officials who acknowledge their limited experience with the gambling industry.

Even so, estimates made during a bidding process can vary so widely that even rival developers say they are sometimes mystified. The chief executive at Connecticut's Mohegan Sun, which lost out to Wynn Resorts in a competition for a Boston-area casino license, said the Canadian consulting firm working with the Massachusetts Gaming Commission should have challenged their rival's projections for spending by international high rollers.

"I don't think enough questions were asked by the consultants," CEO Mitchell Etess said. "If you're in meetings with the gaming commission, and you get the sense they like Wynn, you're not going to ask a lot of questions."

Massachusetts Gaming Commissioner Enrique Zuniga said Wynn promised lower numbers overall but from a broader area. He said officials are well aware that projections are only estimates, and they look at many other factors, including details of the proposed property, the developers' established clientele and market research.

"We're not taking any numbers to the bank," Zuniga said.

___

Michael Melia can be reached at https://twitter.com/MikeMeliaAP