## 37000 Feet Above the Utah-Nevada Border on EDI-SFO, Fueled by Two Moscow Mules...

@Noahpinion: On Twitter: Brad DeLong will find your algebra errors: http://www.bradford-delong.com/2017/10/note-to-self-greg-mankiw-providing-backup-for-kevin-hassett-department-larry-summers-one-last-time-on-who-benefits-fro.html

**delong**: Every day, a new fresh hell...

Today's new fresh hell is chasing links on the way back from Edinburgh to find Mulligan, Cochrane, Mankiw all marveling over how "gorgeous" and "striking" it is that the ratio of the wage gain to the "static" revenue loss is always 1/(1-t)—how it is a remarkable and important insight that the production function parameters drop it. With not a single one saying:

Wait a minute—the excess wage gain over the tax revenue reduction comes from extra investment and higher productivity from making it cheaper to invest via the tax cut. How much extra investment and how much higher productivity there will be

mustdepend on the production function parameters. Theycannotdrop out. The fact that the algebra says they drop out tells use that there is a mistake in the algebra somewhere.

You test your models with edge cases: does it still make sense for wage gain > revenue loss if production is Leontief, and hence k does not rise? No. Does my algebra say wage gain > revenue loss? Yes. Is that a big problem with my algebra? Yes.

No control over one's algebraic model, and little desire to gain enough control over one's model to understand where and why its results are coming from.

Plus, of course, **IT'S THE WRONG MODEL FOR THE QUESTION**.

Plus, of course, **IT'S NOT A COMPLETE MODEL. ANYTHING THAT SAYS "POLICY X IS GOOD BECAUSE IN SPECIFYING POLICY X I HAVE ADDED LUMP SUM TAXES TO MY PANOPLY AND ARMAMENTARIUM OF POLICY TOOLS!" IS NOT A SERIOUS ANALYSIS BUT A CON GAME!!!!**

I must have been very bad indeed in a previous life for Lachesis to have woven me here and now, 37000 feet above the Nevada-Utah border, ranting and raving about the state of public sphere economic discourse!!!!

@raywthompson: Replying to@delong @Noahpinion: WOW. Just. Wow.

@JustinWolfers: It’d be more effective without the ad hominem, but it appears @delong found an error in Mankiw’s corporate tax math: http://www.bradford-delong.com/2017/10/note-to-self-greg-mankiw-providing-backup-for-kevin-hassett-department-larry-summers-one-last-time-on-who-benefits-fro.html

@jonathanchait: "without the ad hominem?" That's probably the nicest criticism @delong has ever written about anybody.

@Meanwich: Replying to@delong @jonathanchait @JustinWolfers: Justin is a centrist dork, even more than you, but he should know it hurts the analysis to pretend Mankiw makes these errors in good faith...

**@delong**: The math error is a little thing—& happens to everybody who blogs. Modeling the United States as a small open economy is a bigger thing. Assuming that the tax rate cut is financed by lump sum taxes is the biggest thing.

We know better than to model US as SOE. We know better than to use an incomplete model that implicitly gives your policy access to lump sum taxes. Both of these are bad economics. To say "Kevin would be half-right if the US were a SOE" or "Kevin would be half-right if we replaced CIT with lump-sum taxes" are not additions to but subtractions from the public discourse.

We have a very bad downward spiraling intellectual Gresham's Law going on here...

@NoahpinionReplying to@AlanMCole: Everyone makes algebra errors...

@jasonfurmanReplying to@Noahpinion:@AlanMCole: It [Mankiw's calculation] is the right algebraic answer to the well-defined question he poses—which is static. The interpretation is where he goes awry.

**@delong**: No, [Mankiw's calculation] is not “static”. Static does not freeze asset valuations. Static freezes the path of production. Freeze the path of production, rig the model so 0% of tax incidence falls on owners of capital, and do the calculation right and you find that d(TR)/dw = -1 always. What comes out of tax revenues is diverted to wages: it can go no place else for there is no place else to go in the model.

Dividing a fixed pie (that is the "static" part) into 3 pieces, one of which you fix in size, is not rocket science. dTR/dw cannot be other than one.