How to Cheat with Cost-Benefit Analysis: Double Count the Benefits
Because my economics courses focus on public policy, I often deal with benefit-cost analyses (BCA) in them. While little discussed, the central idea is simply to identify and include all the relevant benefits and costs of a decision, do our best to estimate their values, then choose the option that provides the greatest net benefits. Hardly a radical idea. It can be useful in disciplining our thinking to be more consistent. Benjamin Franklin employed a version of it.
The problem is that in disciplining our thinking and identifying the logical principles to be applied to make better decisions, BCA also teaches those determined to mislead others about public policies how to do that better, because knowing how to do it “right” also provides a template for how to be wrong in someone’s desired direction.
In fact, BCA today may be more helpful to such motivated mistakes than to appropriate application. The term itself suggests that it is difficult, technically complicated, and uninteresting, so the prospect of work and boredom deters careful thinking by those not specialists in the field. And that is reinforced by the imagery that those doing such analyses are doing so as dispassionate scientists, so that their conclusions can be trusted. Very few people, as a result, pay enough attention to act as an effective constraint on abuses of the technique.
That is why I have extended my classroom BCA discussions to include various ways to cheat on the correct principles. Not so that they can cheat better (in fact, I threaten them with signing a “superhero oath” on their final exams, where they promise they will only use what they know for good), but so they can better detect others’ cheats.
Our discussion starts with the correct principles on what should and should not be included (and why) and how magnitudes might be estimated in various cases. But then we detour to the incentives of those doing analyses for public consumption. Those who are trying to “sell” a policy proposal will be tempted to overstate benefits and understate costs. Those they employ who fear that they “won’t work in this town again” if they don’t get the answers politicians want to justify their proposals face the same incentives. On the other hand, opponents will be tempted to overstate costs and understate benefits.
Knowing which direction someone has an incentive to cheat, then, informs us of which red flags to look for in evaluating their claims. It tells us which mistakes we need to be most alert to. Those warnings are particularly useful because those misusing BCA in their preferred direction either don’t know enough for you to trust their analysis, or they are intentionally misleading you, so that you can’t trust their analysis.
Double Counting Benefits
There are many ways to overstate a policy’s benefits. With the constant drumbeat of proposals to expand government, that is the dominant direction of being wrong in the politically preferred direction. But here I wish to emphasize just one type of such cheating—double counting.
Alleged benefits can be inflated by double counting. That violates a very basic principle—when doing accounting, you would like to do the counting correctly. And when this is violated, it increases the caution we must exercise in looking for other cheats as well.
Suppose there were an irrigation project that would increase the productivity of affected agricultural land. It sounds plausible that you should count the value of the increased crop production, and since it will increase those agricultural land values, we should count those as well. But that would be double counting. The land prices rise to reflect the higher productivity, but that is not a benefit in addition to the higher productivity.
If the project were not irrigation, but a better highway system that lowered the cost of shipping a region’s goods to customers, agriculture would become more productive net of the transportation costs, increasing output and agricultural land prices. But again, land prices are capitalizing the benefits; they are not additional benefits to be included.
Similarly, projects that create better schools, better parks, better flood control, better fire protection, better views, etc., also produce benefits that increase land values. But just because there are two ways you might calculate the benefits—increased benefits produced or the increased property values that capitalize them—does not mean they are separate benefits.
Another very common double counting sin is treating both jobs created and income earned as different benefits of a project. Both income and jobs can be generated (although when you include the costs imposed by raising the funds for the project, rather than ignoring them—another common cheat—it is really more accurate to say that income and jobs are moved from one area to another rather than created). But to workers, the added income is the benefit. More work is not an added benefit. It represents the costs one must bear to receive the income. To the degree that is true, added work is a cost rather than a benefit. So this represents not just counting benefits twice rather than once, but it represents counting a cost as a benefit. Just ask yourself which you would prefer: a given lifetime income that requires you to do your job to earn it or the same lifetime income without having to do that job?
This jobs and income double counting is further overstated by combining it with perhaps the most common form of BCA cheating—multiplier effects. Since each dollar of government spending becomes someone’s income, that income creates more purchases, which creates more income, etc. Program advocates push to include all those added effects on the benefit side—and they often claim both multiplied income effects and multiplied job effects. But since government has no resources of its own, every dollar extracted from Americans to finance a program reduces someone’s disposable income, which reduces their purchases, which further reduces income, etc. So if you wish to count multiplied effects on the benefit side, you must, to be consistent, also count multiplied effects on the cost side. But you could search long and hard without finding a BCA example that does this correctly. Advocates seem to always prefer comparing multiplied benefits to costs that are not similarly multiplied.
The Social Security Trust Fund, as it was growing (to a peak of roughly $3 trillion) after the 1983 Social Security reforms, was also utilized to allow double counting. In years when current Social Security and Medicare “contributions” exceeded outlay—i.e., when there was a cash flow surplus, the excess was invested in special Treasury bonds, whose revenues the government then spent. That reduced how much the federal government had to borrow from elsewhere. And that was true even when Social Security was increasing its actuarial deficit because promised benefits were growing even faster than forecast revenues.
The double counting problem arose, because Social Security was treating those Treasury bonds as assets, but the federal government had already borrowed and spent the funds. The same dollars cannot both be already spent and still available for Social Security benefits. There were no assets beyond those IOUs beyond a government promise for one of its parts to make good on its promises to another of its parts. And the only way that could happen is with increased taxation or borrowing (deferred taxation).
As an analogy, imagine you are saving up to buy something by putting $100 in a cookie jar each week. But you quickly take the money back out to spend, replacing it with a $100 IOU from you to yourself. Eventually having enough IOUs in the cookie jar will not let you make your intended purchase. Those dollars you have already spent are not also available for your purchase. The only way your fund can finance your intended purchase is if you make the IOUs good from other sources of income (including any interest you credit yourself on your “investment”).
This problem was severe enough that in 1998, incoming House speaker Bob Livingston announced that the first House bill to be introduced in the 106th Congress would call for the federal government to stop including the Social Security Trust Fund in calculating budget deficits or surpluses.
As James C. Capretta recently wrote in RealClearPolicy, such double-counting mistakes have not withered away with time, but rather seem to involve even more zeros after the dollar sign: “A tax and Medicare maneuver championed by the Biden administration…would double-count trillions of dollars by using the same source of revenue both to pay for the social welfare spending bill now being assembled in Congress and to forestall the coming insolvency of the Medicare Hospital Insurance (HI) trust fund.” He spells out the complications involved and reaches a shocking bottom line: “In total, the administration wants to double count $3.4 trillion.”
Double counting is just one of many cheats involved in benefit-cost abuses. It is not a sophisticated cheat (the sophistication comes in when abusers try to make cheats appear legitimate), yet there are many examples involving astounding amounts of money. And those examples have an object lesson to teach: if those in government who promise us an ever-growing cornucopia of benefits from ever-expanding power over our lives can’t even count accurately, the likelihood that their involvement will produce the good they promise at the cost they portray is zero. Given the gargantuan scale to which such deception is ballooning, why would anyone think such fantasies can justify so severely compromising America’s future?