The Number Factory

“There are industry advocates saying the [ethanol] industry accounts for maybe a quarter of a million total U.S. jobs,” says Swenson, who is a critic of current economic impact studies of the ethanol industry. “By my methods and accounting, the best I can come up with is about 35,000 jobs, and that is being somewhat generous.” If 35,000 total jobs are created nationwide by ethanol, it’s unlikely that nearly 50 percent of them are located just in Minnesota, he notes.

In his research paper, “Input-Outrageous, the Economic Impacts of Modern Biofuels Production,” Swenson analyzed economic impact assessments from a number of states, and found employment multipliers of well over 20 in many of them, and even some over 50—suspiciously high, even given Lindall’s concession that employment and income multipliers can trend higher than output multipliers do.

Swenson asserts that economic impact projections from ethanol-producing states are grossly overstated because of an absence of accurate industry data (that would be the interindustry trade relationships) in both Implan’s and the Bureau of Economic Analysis’s RIMS II multipliers for this new and rapidly changing industry. Results are also skewed by volatile costs and revenues in the industry, along with changing federal, state, and local subsidies. In addition, he says, people running the models are double-counting inputs.

“They are counting all of the labor and production inputs needed to produce corn as part of their economic impacts,” Swenson writes. But “the corn was already there—that’s why the [ethanol] plants are locating there. So the net job gains to the economy should exclude all the corn production–related numbers. Skilled model operators know this and discount existing production when compiling economic impacts. Unskilled modelers or people trying to promote an industry are prone to ignore this critical accounting step,” he concludes.

Is This Even the Information We Need?

I-O models do have inherent limitations that are stated up front, say Lindall and other users.
 
Bill Lazarus, a professor and extension economist in the University of Minnesota’s Department of Applied Economics in St. Paul, has used Implan to study certain aspects of Minnesota’s agricultural markets. Like Swenson, he points out that fixed proportions in the models don’t capture economies of scale. When you change demand in a model, it automatically increases your supply of labor by the corresponding amount, even though that doesn’t necessarily happen in the real world. For example, an ethanol plant that’s undergone an expansion may be buying twice as much corn and other inputs as it used to, but that doesn’t mean it needs twice as much labor.

I-O models can’t tell you whether or not a local industry that appears to have a big impact (and thus, a big multiplier effect) is also profitable, and therefore viable, Lazarus notes. They can’t readily adjust for price spikes, or a sudden shortage of available labor or supplies. And they absolutely can’t tell you whether the effects they project will actually occur in your economy.

“You just have to look at [the model] and see if it looks reasonable,” Lazarus says. But because the economy is constantly shifting, “you never really know.”

Given these uncertainties, perhaps Implan’s biggest limitation is the expertise of its users—how knowledgeable and skilled they are in framing a study to suit the region they’ve chosen, whether they know the right questions to ask, and can adjust for variations in data.