Measuring ROI of just about anything is subjective because it often depends on context. For example there are many test automation ROI calculators and formulas running around the Internet, but all assume that we all automate our tests the same way, and they fail to take into account intangible value from automated tests (e.g. confidence, stability, etc.)
I and others have done case studies to evaluate benefits and limitations of various testing approaches in controlled settings. Of course, controlled settings don't necessarily mimick 'real-world' environments which is why some people reject these types of studies. But, sometimes even similar studies reveal a pattern that is rejected in favor of more commonly accepted folklore and emotional attachment.
Bottom line, I don't think there is a one size fits all way to measure ROI of any testing approach; however, I do think that we can assess various testing approaches in different situations and derive patterns of benefits and limitations of those test approaches within appropriate contexts.