Cost-Benefit Analysis (CBA) is the dominant mechanism for weighing the utility of proposed regulation, but it is not without its critics. One major alternative is so-called "feasibility analysis". Eric Posner presented today his current paper (co-authored with fellow Chicago professor Jonathan Masur) which provides the first sustained, comprehensive attack on feasibility analysis. They conclude that feasibility analysis is unacceptably vague and has no substantial normative foundation justifying its use as an alternative to CBA.
Feasibility analysis proceeds in three steps. First, the regulatory agency identifies a risk. Second, it identifies the relevant industry that it will seek to reduce the risk in. And third, it tries to reduce the risk to the greatest degree possible, consistent with two restraints: technological, and economic. Because one of the main justifications for using feasibility analysis is that it can better account for concentrated harms, the sort of economic result that would render a proposed regulation "unfeasible" is often one that causes entire plants to close (as opposed to industry-wide layoffs scattered across every factor). Alternatively, OSHA has adopted a standard which says a regulation cannot cause more than either a 1% drop in revenue or 10% drop in profits.
The problem is there is considerably ambiguity throughout the process. The decision on what counts as an industry, for example, is fraught with considerable peril. A factory that sprays paint onto automobiles and one that sprays paint onto airplanes may be seen as sufficiently similar industries for a proposed regulation requiring workers to wear respirators. But of course, these two factories may have entirely different business models, meaning that the cost of the regulation might be easily absorbed (or passed onto the consumer) by one, but lethal to another. A cost-benefit analysis, because it leans more towards an "all-things-considered" model, can better weigh out the distinctions between the airplane and automobile factories. By contrast, a feasibility analysis will lead to drastically different results depending on whether the factories are combined together or disaggregated.
The issue gets even worse on the subject of welfare loss. Feasibility analysis, as noted, tends to focus heavily on concentrated harms like plant closings. It pays little to no attention to the harms regulation may impose on consumers. From a welfarist standpoint, this blindspot in feasibility analysis is difficult to justify. That's an example of where feasibility analysis leads to overregulation compared to the socially optimal result, but Posner and Masur stress that feasibility analysis can lead to underregulation as well, particularly of smaller and/or less profitable companies or industries, which would face proportionally higher costs at complying with socially optimal regulation. Similarly, the emphasis on concentrated harm, rather than harm generally, has distorting effects of its own. Dispersed harms shouldn't be ignored just because they are diffuse (overregulation), and a regulation that saves many lives shouldn't be rejected even if it causes several plants to fail (underregulation).
Perhaps because of this, Posner and Masur observe, agencies undertaking feasibility analysis seem to smuggle in some cost-benefit analysis on the side -- a task made easier by the ambiguous nature of the project. Even the relatively solid 1/10 rule was "waived" so many times by OSHA that the authors said the agency was less making an "exception" than it was simply repeatedly ignoring its own findings. But this merely underscores feasbility analysis' lack of a normative foundation -- it is constantly forced to refer back to CBA to fill in its gaps.