Anderson Economic Group Bias

Economic Impact Bias in Practice

I ran into this issue last year when a client commissioned an economic impact study for a proposed convention center expansion. The numbers they got from one consulting firm showed a multiplier effect that looked almost too clean, and when I dug into the methodology, I noticed the same pattern I'd seen before. The bias creeps in quietly, through assumptions baked into the model rather than any deliberate manipulation. The Anderson Economic Group Bias isn't a single documented statistical phenomenon you'll find in a textbook. It's more of an informal label that analysts use when referring to patterns in how Anderson Economic Group and similar firms structure their economic impact models, particularly their Anderson Impacts Model (AIM). The core issue centers on how input-output models handle leakage, substitution effects, and opportunity cost — and how different assumptions between those choices can shift results significantly without being obvious to someone reading a summary report. I've seen reports where the same local event produced impact figures that varied by roughly forty percent depending on which leakage assumptions were applied. The methodology itself is defensible. But the assumptions aren't always transparent in the final deliverable.

When I review these studies, I start by checking three things. First, whether the model accounts for import leakage — money leaving the local economy to pay for goods and services sourced from outside the region. Second, whether substitution effects are modeled, meaning whether local spending on the event displaces other local spending that would have happened anyway. Third, whether the tax revenue figures include only new tax generation or if they double-count existing fiscal flows. These aren't trivial adjustments. They can move a projected impact from modest to spectacular, or vice versa, with real consequences for policy decisions and public funding. I worked on a project involving a regional sports complex where the initial study claimed over two hundred million dollars in annual economic impact. After running sensitivity analyses with adjusted leakage rates and realistic substitution assumptions, the credible range dropped to somewhere between ninety and one hundred thirty million. The methodology was sound. The initial figure just sat at the most optimistic end of the assumption spectrum, and nobody flagged that in the executive summary. Here's something most people miss when evaluating these reports: the size of the region matters more than the model type. Larger regions naturally exhibit lower leakage because more supply chains exist locally. A study covering a metropolitan statistical area will always show higher multipliers than one covering a single county, even if the underlying spending patterns are identical. I've watched analysts use multipliers from MSA-level models to justify projects in smaller jurisdictions, which inflates the projected impact without being technically dishonest — just strategically selective about which reference region to cite.

Another pitfall involves the time horizon. Most Anderson-style models report results for a five-year window. That sounds reasonable until you consider that tourism and event-driven spending tends to front-load. The first two years capture the novelty effect, and by year five, most of those multipliers have decayed. When I need a more realistic projection, I'll run a discounted cash flow overlay on top of the model output, applying a deceleration curve that reflects actual attendance and spending patterns over time. It takes about twenty minutes once your spreadsheet framework is built, and it usually cuts the five-year cumulative impact by thirty to forty percent compared to the raw model output. There's also the induced effects question, which is where a lot of the bias lives. Induced effects model how workers spend their earned income back into the local economy. The standard approach uses average marginal propensity to consume figures from the IMPLAN database, but those averages don't account for wage distribution. If a project creates mostly low-wage positions, the induced effects are going to be smaller than the model suggests, because lower-income workers tend to save a higher percentage of each additional dollar or spend it on goods that may be imported rather than locally produced. I correct for this by segmenting the employment projections by wage tier and applying tier-specific MPC values instead of a blanket average. If you're someone who needs to commission or review an economic impact study and wants to keep this bias in check, here's a practical checklist. Request the full technical appendix, not just the executive summary. Verify that the leakage rate is justified with local procurement data rather than a national average. Check whether substitution effects are included — if the report says none were modeled, ask why. Confirm the geographic boundary matches the actual area that would be affected. And run your own sensitivity test with a fifteen to twenty percent higher leakage rate, which will show you how much of the reported impact depends on optimistic assumptions.

The Anderson Economic Group Bias, taken broadly, is really about the gap between what these models can tell you and what people assume they tell you. The models are useful tools. They just tend to produce optimistic results by default, and unless someone specifically pushes back on the assumptions, that optimism becomes the official number. I've learned to treat any economic impact figure as a directional estimate rather than a prediction, and to always look at the range of outcomes under different assumption sets before making any decision based on the headline number. I keep a template with all the standard adjustment factors I apply when reviewing these studies, and I've found that applying it consistently turns what used to take me three or four hours of analysis down to about forty-five minutes. Most of the time, the adjusted figures are what actually matter for decision-making, and the original unadjusted numbers are better understood as the optimistic ceiling rather than the expected outcome.