Kamis, 08 Juni 2023

Cost Benefit Comments

The Biden Administration is proposing major changes to cost-benefit analysis used in all regulations. The preamble here, and the full text here. It is open for public comments until June 20

Economists don't often comment on proposed regulations. We should do so more often. Agencies take such comments seriously. And they can have an afterlife. I have seen comments cited in litigation and by judicial decisions. Even if you doubt the Biden Administration's desire to hear you on cost-benefit analysis, a comment is a marker that the inevitable eventual Supreme Court case might well consider. Comments tend only to come from interested parties and lawyers. Regular economists really should comment more often. I don't do it enough either. 

You can see existing comments:  Search for Circular A-4 updates to get to, then select “browse all comments.” (Thanks to a good friend who sent this tip.) 

Take a look at comments from an MIT team led by Deborah Lucas here and by Josh Rauh. These are great models of comments. You don't have to review everything.   Make one good point. 

Cost benefit analysis is useful even if imprecise. Lots of bright ideas in Washington (and Sacramento!) would struggle to document any net benefits at all. Yes, these exercises can lie, cheat, and steal, but having to come up with a quantitative lie can lay bare just how hare-brained many regulations are. 

Both Josh and the MIT response focus on the draft proposal's use of ultra-low discount rates, ranging from historic TIPS yields to arguments for zero or negative "social" discount rates. Josh emphasizes a beautiful compromise: always show the annual stream of costs and benefits. Then it's easy enough to apply different discount rates. No Black Boxes. 

Discount rates seem like a technical issue. But they matter a lot for climate policies, or for policies with substantial cost but putatively permanent benefits, because of the long horizons. For example, climate change is alleged to create costs of 5% of GDP in 100 years. So, let's assume a 0% discount rate -- treat the future just like the present. How much is it worth spending this year to eliminate additional climate change in 2100? Spend means real spending, real reductions in everyone's standard of living, not just funny money billions on twitter. 

If you answered "5% of GDP" (roughly $3,500 per person) that's wrong, for two crucial reasons. First, the economy grows over time. At a modest 2% real growth, US GDP will be 7.4 times as large in 100 years as it is today, or 640% greater. (e^2=7.4). Thus, 5% of GDP in 100 years, discounted at 0%,  is 7.4 x 5% or 37% of today's GDP, or $17,500 per person today. Second, the gain is forever -- 5% of 2123 GDP, but 5% of 2124 GDP, and so on. Discounted at a zero rate, 5% of 2123 forever after that is worth... an infinite amount today. But GDP keeps growing after 2123. If you discount at anything less than the growth rate of GDP -- 2% in my example -- 5% of (growing) GDP forever is worth an infinite amount!  So what if $250 billion subsidizing huge battery long range electric cars made by union labor in the USA from hypothetical US made lithium mines might, all in, save a thimbleful of carbon per car (is it even positive?), if the benefits are infinite, go for it. 

If you discount by a low, but somewhat more reasonable number like 7%, then a dollar in 100 years is worth 0.09 cents today (100 x e^-7). Now you know where to put your thumb on the climate scales! 

You might be wondering, if our great grandchildren are going to be so fantastically better off than we are, let them deal with it. Or you may be wondering that maybe there are other things we could do with money today that might speed up this magical growth process and do 5% better. For an infinite amount of money, is there nothing we can do to raise the growth rate from 2% to 2.05%? 

The latter opportunity cost question is, I think, a good way to think of discount rates. The average real return on stocks is something like 5%, at least. The average pre-tax marginal product of capital is higher; pick you number but it's in the range of 10% not 1%. The right "discount rate" is the rate of return on alternative uses of money. Josh and the MIT team are exactly right to point out that using the rate of return on risk free government bonds is a completely mistaken way to discount the very risky costs of climate damage -- that 5% is a very poorly known number -- and the even riskier benefits of the government's shifting climate policy passions. But I think phrasing the experiment in terms of opportunity costs rather than proper discounting of risky streams makes it more salient, despite the decades I have spent (and an entire book!) on the latter approach. Businesses can take $1 today and turn it in to, on average, $1.07 next year. Why take away that money for a project that yields $1.00 or $1.01 next year? 

The former question has a deeper consequence. Why should we suffer to help people, even our grandchildren, who will be on average 7.4 times better off than we are? How much would you ask your great grandparents to sacrifice to make you 5% better off than you are today? 

Here the low discount rate clashes interestingly with another part of the proposal: equity and transfers. 

From the preamble p. 12: 

A standard assumption in economics, informed by empirical evidence (as discussed below), is that an additional $100 given to a low-income individual increases the welfare of that individual more than an additional $100 given to a wealthy individual. Traditional benefit-cost analysis, which applies unitary weights to measures of willingness to pay, does not usually take into account how distributional effects may affect aggregate welfare because of differences in individuals’ marginal utility of income. Related to the topic of distributional analysis is the question of whether agencies should be permitted or encouraged to develop estimates of net benefits using weights that take account of these differences.26 The proposed revisions to Circular A-4 suggest that agencies may wish to consider weights for each income group affected by a regulation that equal the median income of the group divided by median U.S. income, raised to the power of the elasticity of marginal utility times negative one.

Now wait a darn-tootin' minute. The "standard" doctrine in economics is that you cannot make intra-personal utility comparisons. Utility is ordinal, not cardinal. Here cardinal-utility utilitarianism with equal Pareto-weights is about to be carved into federal stone. (To decide social benefit of taking from A and giving to B, you construct a social welfare function \(u(c_A) + \lambda u(c_B)\). This needs you to use the same \(u()\) for A and B, and agree on a Pareto-weight \(\lambda\) implicitly one here.) 

Imagine a simple regulation: take a dollar from Joe ($100,000 income) and give it to Kathy ($50,000 income). By this standard such a straightforward transfer passes a cost-benefit test.  

But this does not get applied over time. Taking a dollar from you and me, and at a discount rate of 0% giving it to our great grandchildren who will be 7.4 times better off should set off massive inequity alarm bells. Nope. 

Indeed, you can deduce a discount rate from the inequality goal. Pure undiscounted intergenerational equity requires a discount rate proportional to the economic growth rate. 

(With power utility, an intervention that costs A $1 to give B $\(e^{rt}\) just passes a cost-benefit test if \[c_A^{-\gamma} = e^{rt} (c_B)^{-\gamma}.\]  If B is \(e^{gt}\) times as well off as A, \(c_B=e^{gt} \times c_A\) then we need  \(r=\gamma g\). \( \gamma\) is usually a number a bit bigger than one.  The preamble's discussion of \(\gamma\) values is pretty good, settling on a number between one and two. However, they haven't really heard of the finance literature: 

Evidence on risk aversion can be used to estimate the elasticity of marginal utility. In a constant-elasticity utility specification, the coefficient of relative risk aversion is the elasticity of marginal utility. There are numerous different estimates of the coefficient of relative risk aversion (CRRA), using data from a variety of different markets, including labor supply markets,29 the stock market,30 and insurance markets.31 Relevant estimates vary widely, though assumed values of the CRRA between 1 and 2 are common.32

30 Robert S. Pindyck, “Risk Aversion and Determinants of Stock Market Behavior,” The Review of Economics and Statistics 70, no. 2 (1988): 183-90 uses stock market data and estimates the CRRA to be “in the range of 3 to 4"

Since then, of course, the whole equity premium literature sprang up with coefficients 10 to 50. Shh. That would justify insane levels of equity. 

The draft also encourages all sorts of unquantifiable non-economic "benefits," but I'll leave that for another day. 

Read and comment. 

BTW, despite my negative tone and picking on these elements, much of the draft is quite good. Here is a particularly nice piece, from p. 26 of the full text 

j. A Note Regarding Certain Types of Economic Regulation

In light of both economic theory and actual experience, it is particularly difficult to demonstrate positive net benefits for any of the following types of regulations:

 price controls in well-functioning competitive markets;

 production or sales quotas in well-functioning competitive markets;

 mandatory uniform quality standards for goods or services, if the potential problem can be adequately dealt with through voluntary standards or by disclosing information of the hazard to buyers or users; or

 controls on entry into employment or production, except (a) where needed to protect health and safety (e.g., Federal Aviation Administration tests for commercial pilots) or (b) to manage the use of common property resources (e.g., fisheries, airwaves, Federal lands, and offshore areas).

Well, FAA tests and rules for commercial pilots is not actually quite so obvious and really needs a cost benefit test. "Commercial pilot" does not mean "airline pilot," it means can you do anything in an airplane and get money for it. But leave that for another day, these principles if applied could clean out a lot of mischief. Well, I guess many on the progressive left or nascent national-conservative right would deny there is such a thing as a "well-functioning competitive market." 


I should have added: It's insane to make a once and for all cost benefit analysis, especially for projects with 100 year horizons. All regs should be re evaluated every 5 to 10 years, and use experience to update costs and benefits. 



Tidak ada komentar:

Posting Komentar

Fiscal inflation and interest rates

Economics is about solving lots of little puzzles. At a July 4th party, a super smart friend -- not a macroeconomist -- posed a puzzle I sho...