Mackinac Center analysts have long been critical of the way consultants — and frequently their clients — misuse economic modeling software.
We were reminded why with the recent publication of two papers purporting to show large job creation, if only state voters would approve a renewable energy mandate on the statewide November ballot.
The initiative, commonly known as the Michigan Renewable Energy Amendment, would mandate that 25 percent of Michigan's energy usage be provided by renewable sources by the year 2025. There is at least one major problem with each study, however: costs.
The study authors' did not include the estimated costs of the mandate in their respective models. The utility industry has estimated that the mandate will cost some $12 billion. Until these costs or some other reasonable approximation thereof are included, neither report should be taken seriously.
The two studies were produced by Michigan State University for the Michigan Environmental Council and the consultancy Hill Group Management Consultants of Pennsylvania for the Energy Innovation Business Council, respectively. Each uses input-output models to measure the jobs impact of the mandate. Both use a model known as IMPLAN, and the former also used a model known as the Jobs and Economic Development Impact model.
The first study, “Projected Job and Investment Impacts of Policy Requiring 25 Percent Renewable Energy by 2025 in Michigan,” was produced by Michigan State University scholars, and was commissioned by the Michigan Environmental Council. It should be noted that the MEC, an environment-focused nonprofit, supports passage of the 25 by 2025 mandate. The MEC press release accompanying this study reported that passage of the mandate would create a minimum of 74,495 jobs.
First, and as Michigan Capitol Confidential has already noted, the press release misled readers because the report doesn’t actually say that. The report breaks down its job count to 74,495 job years. That is, a person newly hired by a wind company and who works 25 years would count as 25 job years, though it is still just one job.
Second, the report doesn't include costs associated with this mandate, only the benefits, so the model treated all new spending as manna from the heavens. Such an assumption does not comport with reality. As economist (and former Mackinac Center intern) Daniel J. Smith remarked, "The increasingly sophisticated Keynesian models … still fail to address the most basic question, 'Where is the money coming from?'"
Costs do occur and — if actually fed into the model — would show a different and perhaps negative jobs impact.
To the authors' credit, they responsibly spell out the fact that their modeling efforts "only reflect gross and not net impacts." Unfortunately, their disclaimer is on page 23 — the last page of the written text and far from the money quotes journalists typically seek upfront. This gross versus net fact was also left out of the MEC's mandate-aggrandizing press release.
The other study, "Economic Impact of New Energy Manufacturing in Michigan," similarly employed an input-output computer software model; the authors claim, however, that only "20,791 jobs would be supported" by the mandate. This study also ignores costs. Although the report isn't as explicit on this issue as the MSU paper, the authors confirmed by phone that their calculations involved a gross and not a net impact.
It must be underscored here that the two studies come to vastly different job creation conclusions. There are probably a number of good explanations for this, most notably that one counts job years while the other just counts jobs. Regardless of the reasons, these reports show how easy it is for consultants to obtain wildly different estimates of the same study subject (renewable mandate and jobs), over the same time period and using the same impact model (IMPLAN).
Economic models are not exempt from economic incentives, which may help temper the results produced in these particular models.
This author is far from the only individual to criticize the use of these models. In his 2006 paper, "Economic Impact Studies: Instruments for Political Shenanigans," author John Crompton argues that "most economic impact studies are commissioned to legitimize a political position, rather than to search for economic truth."
In his 1993 paper, "The Misuse of Regional Economic Models," Edwin Mills demonstrates how such impact models might be abused for political gain and offers some thoughts on how this occurs:
... [T]o justify increased spending, government officials must identify some publicly desired goal to be accomplished by government spending. Creation of new jobs is among the best such goals that can be found. ... [T]hey must make it plausible that government can accomplish the goal in a way that the private sector cannot. This is where REMI is so valuable. It is a complex computer model that lay people cannot understand or evaluate, and it has important scientific merits. Thus, the frequent government claims that the best scientific model available shows that x thousand jobs will be created by the project helps to carry the day.
Michigan economist and economic consultant Patrick Anderson made a similar argument in his 2005 book "Business Economics and Finance," writing, "Because the claimed economic impact of a proposed development can affect political support for a proposed project … an incentive often exists to exaggerate the benefits."
While economic impact software can be a useful tool for measuring economic phenomenon based on policy changes, they (and the output they produce) must be used responsibly.
Until the sponsors of the 25 by 2025 studies rework their estimates to include all the costs associated with this mandate, voters should not use these studies as an assessment of the proposed mandate's economic impact.