The basic premise of benefit-cost analysis is that society should make the most of its scarce resources by maximizing the benefits it obtains from them.  In theory, B-C analysis is pretty straightforward.  The analyst calculates the benefit and cost schedules for each candidate policy over the appropriate time horizon, discounts these, and sums them to obtain comparable present values of benefit streams versus the cost streams for each option.  In choosing between a set of discrete policy options, the usual decision process involves calculating a B-C ratio for each option and immediately ruling out options with B-C < 1. At this point, the optimal choice may be the option with this highest B-C ratio (if that option is scalable), or the option with the largest B-C difference (if the options are not scalable).

In practice, unfortunately, B-C analysis is rarely straightforward, particularly in environmental policy decision-making.  In fact, the practical difficulties of doing B-C analysis have created serious political mistrust of the process.  Many Federal environmental statues include wording that actually discourages the use of B-C analysis.   The following discussion lays out some important limitations of B-C analysis, and summarizes the challenges facing economists who undertake to do such analyses.

One hindrance to public acceptance of B-C analysis is political unwillingness to acknowledge that there are limits to what a policy can achieve.  The public doesn't like to be told that environmental risks cannot be eliminated entirely, and politicians are sensitive to irrational risk perceptions.  Indeed, most people believe they are entitled to a risk-free, pollution-free environment, and are unwilling to countenance any political compromising of these rights.  These unrealistic expectations are rarely challenged..  Consequently, we have had various economically inefficient policies such as the Delaney Amendment (1958, only recently repealed, which banned food additives that could be shown to cause tumors in lab animals at any dosage) that rule out any comparision of benefits versus costs.  Although the costs of the Delaney Amendment (preventable obesity, nutritional deficiencies, spoilage, illnesses and deaths from food-borne pathogens) almost certainly exceed its benefits, it took 40 years for Congress to muster the political courage to repeal it.

A related problem involves uncertainty.  B-C analyses can only be as good as the scientific data they rely on, and the chain of causality between the policy implementation and the desired outcome can be long and tenuous.  For example, the link between a policy (reduce emissions from a company's smokestack by fifty percent, say) and its intended results (reduce the incidence of lung cancers by 2 in 10,000) depends on understanding (1) the climatologic and engineering processes explaining how the smokestack controls will reduce ambient concentrations of the pollutant, (2) the geographic and socioeconomic links explaining how reducing ambient concentrations will reduce human exposures, and (3) the epidemiologic links between reduced exposures and reduced incidences of cancers.  Experts may be required to elucidate the causalities at each step, and the uncertainties get compounded at each step.

In the absence of data, everyone is an expert.  True experts will generally provide appropriate caveats with their findings, disclose areas of uncertainty, and concede the technical limitations of their analyses forthrightly.  Unfortunately, this style may not carry the day in a higly politicized debate.  Public trust in experts appears to have declined over the last two decades.  US courts have been slow to reject the junk science, bad statistics and other nonsense peddled by so-called "expert witnesses" who are paid to support one side or the other in civil cases.

Public risk perceptions are often inconsistent with experts' risk assessments.  People naturally trivialize familiar risks and risks they undertake voluntarily, while exaggerating the significance of statistically trivial risks that are unfamiliar, are incurred involuntarily, or provoke dread.  They ignore experts who say they should focus less on carjackings and terrorist attacks, and more on quitting smoking and wearing seat belts.

The B-C framework is supposed to compare all benefits and costs associated with policy choices in monetary units, but some elements such as the values of saving habitats for endangered species, preventing cancers or saving human lives are difficult to translate into money terms.  In fact, attempts to do so often elicit expressions of moral outrage, and some opponents of B-C analysis misrepresent it as morally corrupt.  The economic valuation of lives does not involve any moral judgment by the analyst.  Rather, it is based on observations of statistical risks people incur voluntarily.  Voluntary acceptance of risk implies a self-valuation of one's health or life in statistical terms.

Another  problem with B-C analyses is that they often go beyond purely positive economics, and enter the realm of normative economics.  In discussing any policy, we have to be mindful of who benefits and who pays, and these are politically sensitive questions that politicians instinctively prefer to avoid.  B-C analyses typically compare costs and benefits without explicit regard to these distributional issues, but wrangling over such issues often frustrates political implementation of policies that would clearly be good for society.  Public mistrust of the redistributive effects of government policies creates substantial political inertia that impedes adoption of obviously beneficial policies.  B-C analysts may identify winners and losers, but typically  remain silent on whether the losers should actually be compensated.  Compensation may be appropriate in order to counter or neutralize organized political opposition from the losers of a policy.

A final controversial element in B-C analyses is the choice of discount rate.  In some cases the choice of discount rate largely determines the outcome of the analysis.  Here's an illustration:.  Suppose we are trying to determine the optimal use for a vacant parcel of public land. The local power company wants to use it for a nuclear waste dump, a local developer wants to build condos on it, and a local environmental group wants to preserve it for recreation and wildlife habitat. These uses would generate the following net social benefit streams:
 

The low discount rate (r = 0.04) favors the preserve option, since this yields the largest net benefits in the distant future.  The high discount rate (r = 0.08) favors the condo option, since its net benefits are realized soonest.  And the intermediate discount rate (r = 0.06) favors the nuke dump.  The graph below shows NPV for each option as a function of the discount rate applied.  In this example the preferred options change at discount rate thresholds of about 0.047 and 0.069.