Can we agree that…(more on fiscal policy, and hurdle rates)

Ryan Avent puts forward some propositions on fiscal policy that he hopes we can agree on (I agree with most of them, dissenting on the multiplier discussion by wanting explicit time horizon considerations, the overall importance given by multiple mentions of ZLB, and being confused but not disagreeing with the wording of the first part of #13).  And in general I am supportive of such attempts to find common ground.

I wish, however, to add to the list.  I wish to nominate a few more items, noting that this does not exhaust my wish list.  Here goes:

1. Debt-financed government spending must eventually be paid off and the estimated deadweight loss of taxation is at least twenty percent.  That immediately puts a hurdle rate of twenty percent or more on projects, even when real borrowing rates are very low.

2. Private companies, when making investment decisions, often use hurdle rates as high as twenty or thirty percent, even when their cost of capital is much lower.  Often it is believed this is to constrain overeager empire builders, or “cowboys.”  For sure, the agency problems in the public sector differ considerably from those in the private sector, but arguably there should be a “cowboy premium” for the public sector as well, even if that premium should be lower for the public sector.

3. The real risk of public sector investment is not measured by the borrowing rate, but rather by the covariance of the value of public sector outputs with a very broad notion of the market portfolio.  I call this the Jensen premium, since Michael Jensen first outlined this argument clearly.

This is all standard stuff, none of it is like reading the “he said, she said” debates over the proper size of the fiscal policy multiplier.

To do some adding up, we have twenty percent plus the cowboy premium plus the Jensen premium.  Plus of course the real time preference rate, if that should be positive too.  It is hard to get a good sense of the size of the cowboy and Jensen premia but still we are running clearly over twenty percent, perhaps a good deal over twenty percent.

I also believe (more controversially, this point is not consensus) that the cowboy premium is considerably higher for debt-financed expenditures, relative to balanced budget expenditures.

Now let’s come to my complaint (which is not directed against Ryan).  I have read dozens — or is it now hundreds? — of blog posts arguing that low borrowing rates make for a very strong case for fiscal policy.  I do not often if ever see these posts admit that the hurdle rate for government investment still can be quite high and still is likely quite high.  I do not see these posts discussing the DWL of taxation, the cowboy effect, or the Jensen effect.  I see only the mention of a very low borrowing rate.

Of course you should adjust the social costs of the project downward to the extent it is mobilizing specific unemployed physical resources (which is quite distinct from the existence of a low borrowing rate per se; for one thing, it is easy to have projects which reshuffle resources and a low borrowing rate, plus the crowding out may come on the labor side).   I see this point about unemployed resources mentioned many, many times, though not with the proper caveats, and I see the points about hurdle rates hardly mentioned at all.  And note by the way that the DWL of taxation premium should be applied to all upfront pecuniary costs, whether or not they are true social costs.

Addendum: Brad DeLong comments.  His #1, if I understand it properly, confuses “spending at all” with “taxing vs. borrowing” calculations.  #2 ignores agency problems and #3 assumes a rather definite view on the understanding of risk.  You will note that my post is quite agnostic on the size of #2 and #3 in any case.


What is the counter factual premium you deduct from your different hurdle rates, ie, the cost of doing nothing. Moreover, if you make a capital investment in period 1 that would have been made in period 2 when there is recovery, do you give a credit for investment that is not made in period 2.

Also, how do you do your accounting for tax cuts instead of investments? What is the hurdle rate for tax cuts.

> What is the hurdle rate for tax cuts.

Perhaps I'm misunderstanding, but wouldn't it be negative? Cutting taxes eliminates existing deadweight loss.

Bill really means the hurdle rate for the spending not financed by the taxes not collected.

Yes, thank you Yancey, thats a good description. I would also add that empirical research showed that the 2007 and 08 tax cuts had little effect, and that persons just saved their money, thinking of it as windfall--so the multiplier was negative with tax cuts. And, of course, if you apply Tyler's other discounts, even more negative.

But shouldn't all spending have to meet whatever hurdle rate is applied, not just additional spending? All spending is paid for by taxes, either now or in the future, so we'd eventually still have the same issues of deadweight loss when we collect taxes later rather than today. That's why there's no real benefit from a temporary tax cut (without even a temporary spending cut to go with it). As long as spending is high, we eventually have to pay for it, so the present value of the deadweight loss may not be changed.

Excellent questions here.

Also, I take issue with #1.Why assume that debt needs to be paid down from taxes? At least analytically, it can also be paid by printing money, or can be kept on the books forever and funded out of new debt. I am not saying these are always great options (though if you agree with Advent that monetary policy is the first resort, why not print your way out?). But if you think that you need to treat the deadweight loss from taxes as such a big deal, aren't you under some obligation to at least argue that it's higher than the deadweight loss from additional debt issuance or from adding to the money supply?

Agree, #1 is simply false.

If you're going to argue that government debt should be paid off then you might as well argue that all private debt should be paid off. See how silly that is?

Good point, James. First, if the government purchases capital goods with an expected life of more than a year, you should probably pay that as debt over the life of the asset.

Your other point about private sector debt is exactly on point. I particularly like how persons use the term "unfunded liability'. An IBM bond is an "unfunded liability". You don't hear anyone talking about a bond as an unfunded liability, but it is.

Bill and/or James,

I believe you are exactly the kinds of people from whom I would like to borrow $1 million. Call me.

Also Bill, I'm glad to see you grokking sturdy accounting concepts like the 'matching principle'. Feel free to apply your new-found knowledge to programs like pensions and retiree health care.

Brian, You have nothing to contribute with ad hominems. I am sorry, because sometimes your comments make sense, or, if they don't, at least you can respond to them. See if you can do a rational argument without the ad hominems.

Sure, but inflating away debt is not different from taxing savings. Also, carrying debt is not free.

Why assume that debt needs to be paid down from taxes?
Because your assumption that you do not need to pay it back would lead to the conclusion that you should move taxes down now. You could go all the way to zero taxes. Until taxes are zero we are paying debt down.


I'm trying to understand #1. Isn't it true that the US never paid off the debt incurred from fighting WWII? I was under the impression that it was made irrelevant because the economy grew so much the debt/gdp dropped to irrelevant levels. If this is true, doesn't this well-known factoid blow a huge hole in your entire premise? Or is this some sort of popular misconception?

Please help educate me. Thanks!

'That immediately puts a hurdle rate of twenty percent or more on projects, even when real borrowing rates are very low.'

In my recollection (back in the 80s), Virginia did not allow road bonds, meaning that current (budgeted) revenue had to be used to buld roads - which also led to the fact that not so many roads were built, and the ones that were took longer, since current (budgeted) revenue was very much less than the total cost of a major road project. But if the deadweight loss of taxation is 20%, then Virginia's model would seem to be preferable to the one used by other states - if the point was to save a bit of money in terms of additional taxation caused by debt service.

However, if the actual point was to build roads which people could use instead of contributing to the traffic hell that Northern Virginia was in the 1980s, I would submit that such concerns just might be a tad misplaced.

A bit of Commonwealth history -

'The official state agency, the Department of Highways, was established in 1927. By this time the citizens of the Commonwealth had rejected a bond referendum for road construction and instead on the advice of Governor E. Lee Trinkle instituted a "pay as you go" policy that relied on a 3 cent per gallon tax on gasoline for road construction.

The 1932 Byrd Road Act established a unified State Secondary Road System. This permitted each county to give the responsibility for its secondary roads to the Highway Commission. At the time four counties chose to keep this responsibility including, Warwick, Nottoway, Arlington and Henrico, currently only two remain, Arlington and Henrico. One economist estimated that the Byrd Road Act would reduce rural taxes by $2,895,102 annually, which was certainly Byrd's aim.
As a result of the 1986 Commission on Transportation in the 21st Century, the General Assembly created the Transportation Trust Fund along with a series of innovative financing tools, they also changed the name of the Department to the Virginia Department of Transportation and added three at-large members to the newly named Commonwealth Transportation Board. At least two of the at-large members were to be from urban areas and at least two from the rural areas.'

Twenty percent? Is that based on anything?

I think it's based on a combination of factors ranging from comparisons with the private sector to Tyler's oft-stated distrust of public sector spending. Which is to say, "voodoo."

In all fairness, lots of macro-economics is voodoo, when you get right down to it. But it would be surprising if Tyler's voodoo did not produce high barriers to public-sector investment.

Also, shouldn't the hurdle rate be annualized? If the taxation occurs many years in the future and results in 20% loss of principal, the hurdle rate would be quite low.

Suppose you borrow for 10 years at 2%. Assuming a 20% haircut at the end, that's still only a 4% hurdle to achieve a positive ROI over 10 years. (0.98^10*0.8)^0.1-1. Seems like inflation and ordinary economic growth ought to cover it.

I'm pretty sure that the 20% hurdle is typically in terms of IRR.

low borrowing rates make for a very strong case for fiscal policy.

As Sachs pointed out, as things stand now we are looking at an interest cost exceeding defense + discretionary in the medium term. That should scare the hell out of any sane person. (Had we implemented the Krugulus, we would be there today.) And what happens if growth picks up and interest rates rise? The Fed is focusing on buying long-term debt, the bill to gov't for rolling over the short-term will be ruinous.

Sumner has it right: if you're relying on inefficient fiscal stimulus to boost NGDP, then your monetary authority is not doing it's job properly.

Let's get a definition straight. In reality 'public investment' means paying a non productive person a pension or paying to have their hip replaced.

Presumably, even those non-productive people give their money to productive people, who replace hips and make the things which pensioners consume.

That doesn't make it either (a) investment or (b) the most productive use of the money.

How exactly does one compute return-on-investment for something like building a new free public park?

Tyler's additional points are valid, but not on the issue of how unempoyed resources (of which low borrowing rates are but a symptom of unemployed capital) shifts the timing and amounts of public sector investments (any activity with near term costs and longer term benefits). And of course one looks at opportunity costs of specific resources in a proper cost benefit analysis but these are exactly what are strongly affected by AD conditions.

My point is, a hurdle rate of 20% or whatever is irrelevant in an organization where profit isn't the goal, let aside major goal.

NB: This point about the hurdle rate relates to social welfare, not profit.

I agree with what you're saying on the whole, but have to nitpick that you probably could compute the value of a park (or estimate it) using increases in home prices, usage statistics, or whatever.

But, I agree in spirit. How *do* we quantify giving someone a flu shot or repairing the broken arm of an unemployed dude?

PS I post under a variety of names here, and I feel like there is a Mulp who used to post (but it may have been me, I don't remember). If you are that Mulp, do let me know.

There are two Mulps? In this universe? What are the odds?

I knew you were one of the good guys.

One other problem you have with infrastructure especially is that spending is a lot less efficient than it used to be, mostly due to regulation. Megan looked at this a while back; for what it cost (in real dollars) to build the Tappan Zee in the 1950s, you could not even pay for the impact studies today. Even if we all agreed every single regulation was totally worthwhile, they still reduce the return.

By the way, I disagree with Avent's list, though I do enjoy seeing him hold his nose and choke out the final item through clenched teeth:

"(17) It is very difficult to say with any certainty whether a particular level of debt-to-GDP is bad or should be lower. Whether or not debt levels above certain thresholds constrain growth, there is a risk that high levels of debt may limit the responsiveness of both fiscal and monetary policy to future crises. In general, widespread hand-wringing about the level of debt is a decent indicator that debt is approaching levels at which the ability to respond to future crises faces political constraints. Subject to all the factors discussed above, then, stabilising and reducing the debt-to-GDP level is a legitimate policy goal at such times."

The wording here ('widespread hand-wringing', 'at such times') is hilarious. The guy is still unable to understand that this could be anything more than a 'political' problem. Whatevs, way to eat your vegetables Ryan!

So to add to my comment above...

If you think that monetary stimulus is better than fiscal, and you are worried about the debt, isn't the perfectly ideal government reaction to print money and use it to buy and cancel government debt?

I get that the US has institutional constraints on this, but set those aside. Why not tell the Fed to print money, buy a bunch of T-Bills and then set them on fire? Doesn't that get you to 5% NGDP growth AND lower government debt?

Existing Treasury debt is skewed towards short maturities. If bond buyers believe the sovereign intends to inflate its way out of debt, interest rates rise, and Treasury has to refinance its short-term debt at the higher rates.

(If the sovereign takes the next step to printing money to finance the debt service on the refinanced-at-higher-interest debt, welcome to hyperinflation.)

"the estimated deadweight loss of taxation is at least twenty percent. That immediately puts a hurdle rate of twenty percent on projects..."

You seem to be mixing up rates of return with NPV. For instance, if the bonds are fully paid off after 30 years via taxes (instead of just being refinanced) then the effect on the "hurdle rate" (aka annual required rate of return) would be less than 1%, even if you require an additional 20% at the end.

Yeah, I thought that too. He's using "hurdle rate" incorrectly, or at least colloquially.

There is no way to lower the overall net DWL by paying off the debt later rather than sooner, provided the proper discount rate is showing up in all other parts of the analysis. No mistake on my end.

You may be right about the DWL, but you're still misusing the word.

Few projects return 20%/year.

Well, Tyler is right about it not mattering when you pay off the debt for his analysis, people are right that he's misusing "hurdle rate," and in private firms I doubt there are many projects undertaken that _aren't_ anticipated to return >20% per year. You need projects that offer some premium over the expected rate of return of your investors, and you need that premium to be significant because you're probably optimistic and may be systematically wrong across projects in a bad year. You don't want a confluence of problems to bring down the firm, hence you use a hurdle rate that gives you a margin of error.

It's possible I'm misunderstanding you, Tyler. It sounds like you are comparing a 20% hurdle rate ("The project must return in excess of 20% per year to be worth undertaking") with DWL ("Paying for this project via taxes involves a one-time hit of 20% because of the problems the taxes cause in distortion, compliance costs, etc."). One is annualized, one is one-time.

@Fitch -- are we talking about one time taxes of taxes supporting on-going government activities? I don't think that's been clarified in the discussion.

You are confused. First, if the deadweight loss due to taxation is 20%, you only get 80 cents on the dollar from taxation which means you have to tax at 125% to get 100%. Second, if you pay the bonds off at the end of 30 years (no coupons in the interim) you will have to pay that 25% with interest for 30 years.

For a government project, I'm not sure it makes sense to amortize the 20% deadweight loss over the life of the bonds used to pay for the project. I suppose you could amortize it over the period that the project is producing economic benefits, but (at least for a federal government project) it's unlikely that would have anything to do with the tenor of the bonds.

On 1) Let's say we need to invest money to repair a broken bridge.

The question now is either a) repair it now when interest rates are low, or b) repair it later when interest rates are high. Assume the deadweight loss of taxation is the same in either case. Which is more cost effective, a or b?

Now relax the assumption that deadweight loss of taxation is the same in either case. Assume tax rates will rise gradually in the future to meet debt-service needs. In this case, when is the DWL of taxation higher, now or later? What does this imply about your answer to the first question?

This is a fine way to frame the question, but why is it necessary to borrow money to do expected routine maintenance? The deferred maintenance was used to buy some political favor or support, and now money needs to be borrowed to make sound a pension fund. Illinois is issuing bonds for the sole purpose of making their pension fund contributions.

Considering the length of US Treasury debt, you may get low rates now but you will also pay the high rates later when you have to roll over the debt.

To add to your point b), the choice can also be repairing it now while borrowing is cheap or repairing it later when interest rates are high AND the bridge has further deteriorated or even collapsed, making the necessary repairs more expensive.

see also:

You can argue about the values of the deadweight loss due to taxation and the appropriate hurdle rate, but Tyler's analysis is spot on. The critics seem determined to resist any attempt to submit government "investments" to any cost/benefits analysis.

The Grand Poobah of debt drug pushers weighs in:

Excerpt: "The Congressional Budget Office expects the deficit for fiscal 2013 (which began in October and is almost half over) to be $845 billion. That may still sound like a big number, but given the state of the economy it really isn’t."

He's right. In the current year, putting to one side the $16 trillion in debt we started with the US government will only spend $2,694 per person more than it takes in in revenue. For my family of four, this means a mere $10,777 this year in additional "accrued but unpaid" taxes. This, of course, assumes taxes are paid "per capita". Get a clue, dude.

I'm surprised that you're agnostic on size of 2. 2 is by far the meat and juice of the entire thing. 1 is simply untrue, because whether there's a need to tax or not is precisely what the bond market signals.

What you could say is that the market signals demand for a tax cut, not a spending increase.


If the DWL is 20%, that means each project has to yield a cumulative 20% over the governments funding rate. Not 20%/year. That's all Delong is saying in his #1. All of this confusion is about you calling it a hurdle *rate*. It's not a rate. It's a 20% hurdle on the NPV of the project.

Also, your assumption that government debt has to be paid off is not at all uncontroversial. You are assuming the dynamically efficient debt ratio happens to be exactly zero. Tobin '78 gave us lots of reasons to believe that it could be greater than zero, not least of which is that lots of people don't have kids or care less about their descendants than they do about themselves. This causes systemic risk aversion which, in the absence of sufficient government debt, depresses the natural rate below the rate of growth.

K. You are absolutely correct, at least IMHO. I am surprised that no one here has questioned Tyler's claim about the deadweight cost of taxation. In modern economies, the deadweight costs of taxation, while subject to even greater uncertainty than enforcement costs, are usually assumed to be larger – 20-40 times collection costs (GAO, 2004) – given the high tax rates characteristic of modern economies and the progressivity of their tax systems. As a rough rule of thumb, the deadweight burden of taxation is approximately equal to .5 times the average effective rate squared times the tax base, with an appropriate adjustment for the progressivity of the tax system in question. This simple rule of thumb reflects the presumption that the absolute price elasticity of demand for articles that are subject to taxation is approximately unitary (see Diewert, Lawrence, and Thompson, 1997). Hence, France, with taxes equal to about 55 percent of GDP and a proportional tax system, would have a deadweight tax burden approximately equal to 15 percent of GDP. The U.S., with much lower taxes, but a more progressive tax system (Guner, Kaygusuz, and Ventura, 2012a,b,c), would have a deadweight burden approximately equal to 2.4 percent of GDP, about $336 billion in 2007 (which is within the range of most estimates, see GAO, 2004). That implies a hurdle of between 10 and 17 percent.

Can someone unpack this phrase into a couple sentences? "the covariance of the value of public sector outputs with a very broad notion of the market portfolio" Do you mean: "public sector investment will tend to do well (or poorly) when private sector investment does well (or poorly); therefore when the former does well it will tend to have crowded out the latter, and of course when it does poorly ... it does poorly." What's "the market portfolio?"

Thanks for posts like this; they are very helpful in meliorating the usual Dunning-Kruger trainwreck that is me attempting to understand your judgment. I wonder if your emphatic conclusion that the long-term-unemployed are simply unproductive is a linchpin for your portfolio of macro-political-economic views as a whole. If it were not true, would you have to change your views about everything else? Relatedly, to what extent is concern over deadweight loss displacing concern over the human disaster that is extended unemployment? Is thinking about suffering a psychological threat to believing in the mathematical foundations of economics?

I haven't read every paper estimating a fiscal multiplier, but the ones I have do not separate out these named effects. They just observe changes in GDP and changes in policy. So these effects are inside the system that is being observed. In other words, when Christina Romer tries to measure the multiplier, she does not estimate it before the subtraction of the whozits effect. She just measures it after all the effects have cumulated. Maybe we should ask her if she thinks changes in the whozits affect would change her estimate of the multiplier. If she says yes, then this line of argument is absurd. If she says no then Tyler is onto something.

Tyler, I'm a long-time reader, but I usually avoid commenting. However, this post is so deeply mistaken, as in it would earn a high-school econ student a poor grade, that it has obliterated your credibility with me. At least to me, though I doubt it much matters, your blog is now nothing more than a link-aggregator.

I do thank you for your time and effort in finding interesting econ links.

TC admitted his mistake:

What's up, I check your blogs daily. Your story-telling style is witty, keep up the good work!

I always used to study post in news papers but now as I am a user of web so from now I am using net for content, thanks to web.

Comments for this post are closed