Economists attempt to make the fudge factor as grounded in reality as possible, but instrumental variable analysis inevitably requires the analyst to give short shrift to the hard data and make assumptions about the differences between two populations, and how external factors (like insurance status) affect them.