Atikon That depends on what your comparison is. After developing an error metric, edge (or margin of safety) is dependent on an appropriate normalized error term like standard deviation. You should calculate that for the IV and the comparison volatility. We calculate a confidence level for each expiration, so that would be the error term for the IV. We calculate historical goodness of fit for our forecasts, and that is the error term for your comparison.
Don't typically look at differences between IVs in percentage terms, but rather in the magnitude of the divergence from a meaningful average. Thinking in % terms wouldn't work for IV spreads/strategies that could fluctuate between positive and negative. When looking at overall IV (VIX), ATMs, or individual outrights, then thinking in % terms for rich/cheap valuation could apply. I would probably be one that looks for 20%+ undervalued/overvalued given that I prefer to trade the more extreme intraday and long-term IV levels.
Does anyone have the included CD-ROM for Haug's 'Complete Guide to Option Pricing Formulas'? While I have the book on a table my son managed to drag the CD across the floor...