Archive for October 2020


Moore’s Law and hyperinflation

In pre-modern times there was almost no hyperinflation, at least for extended periods of time. That’s because kingdoms mostly relied on some sort of commodity money. Thus you did not observe the sort of inflation we see under fiat money regimes.

The following table shows long run money growth rates and inflation during various periods, but mostly 1950-90:

In Brazil and Argentina, prices were doubling approximately every 15 months, for decades on end.

In pre-modern times, the value of various goods tended to bounce around, rising one year and then falling the next. That’s actually still true for most goods, if you define “value” in relative terms. Thus even in countries with high inflation persisting for decades, the relative price of commodities like copper or oranges doesn’t show any dramatic (long run) upward or downward trend.

But whereas this pattern was true for essentially all goods in pre-modern times, in modern times we see two examples of goods that experienced a very rapid decline in value, which persisted over many decades. These two goods are computer chips and fiat money.

For many decades, the price of computer chips has been falling in half every 18 months, on average. This tendency is called “Moore’s Law.” In some developing countries during the 20th century, the price of their currency has been falling in half every 18 months, on average, for many decades. By 1990, one unit of Brazilian or Argentine currency could be purchased at a far lower price than in 1950, almost regardless of what you used to buy the currency. The price of their currency fell in terms of apples, oranges, coal, iron, US dollars and Japanese yen.

Over time, computer chip makers found it possible to produce a billion chips at the same cost that they had previously produced a million chips. Fiat money countries found that they could produce a trillion or even a quadrillion pesos at the same cost as they formerly produced a million or billion pesos. That cost reduction was a necessary but not sufficient cause of hyperinflation. You also needed to actually produce the money. Unlike computer firms, not all central banks are profit maximizers. (Thank God!)

It’s possible to measure NGDP using any numeraire. Thus you sometime see people claim that China has the world’s second largest GDP. In yuan terms, China’s GDP is much larger than the US GDP in dollar terms. So when people say China has the second largest GDP, they are implicitly describing what China’s GDP would look like if priced in terms of US dollars (without a PPP adjustment). Similarly, you could describe China’s GDP if priced in terms of ounces of gold or silver or one pound bags of rice. Or computer chips.

If you described the US or China’s GDP in terms of computer chips, it would be rising astronomically, it would look like a country experiencing hyperinflation. But the US and China don’t have hyperinflation. Venezuela has hyperinflation. And that’s because hyperinflation occurs when a country’s price level is rising very rapidly in terms of the thing in which prices are actually denominated.

Imagine I drew 100 graphs; each showing the US GDP measured using a different numeraire. One graph showed GDP in dollar terms. Another in terms of gold. Another in terms of apples. Another in terms of toasters. And one graph showed US GDP in terms of computer chips.

Now suppose I asked you to explain why America’s GDP in terms of computer chips rocketed much higher, year after year, for decades. Would you make reference to the Phillips Curve? To an “overheating economy”? Obviously not. Indeed I’m attacking a straw man here.

When we look at actual inflation and NGDP growth in money terms, almost nobody uses the Phillips Curve to explain inflation in a country experiencing hyperinflation. During the famous German hyperinflation of 1920-23, the world’s most famous non-monetarist economists, people like Keynes and Wicksell, suddenly switched to describing inflation in entirely monetarist terms. And then when the hyperinflation ended, Keynes quickly went back to non-monetarist explanations of inflation.

It’s always been known that hyperinflation is a special case, which can only be explain by looking at what’s going on with a country’s currency. Similarly, hyperinflation of prices when measured using computer chips as a numeraire can only be explained by referring to things like Moore’s law, not Phillips curves. Hyperinflation is nothing more than a currency losing value at a very high rate. The debate is over how to explain “normal” inflation, normal NGDP growth rates.

The following equation is always and everywhere precisely true, to the very last decimal point:

M*V = C + I + G + (X-M)

Understanding what this equation means and more importantly what it doesn’t mean is the key to understanding money/macro.

The equation reminds me of two large hoofed male animals with curved horns, bashing their heads into each other, over and over again. NGDP is determined by M*V!! No it’s determined by C + I + G!!

LOL. It’s an identity!

We keep arguing about the wrong things

Matt Yglesias has a very perceptive tweet:

There are various ways of reading this tweet, but here’s my take:

There are pros and cons of the herd immunity strategy.

There are pros and cons of challenge trials for a vaccine.

Perhaps we should do both, perhaps neither.

But the case for challenge studies is clearly far stronger than the case for herd immunity.

Every criticism of challenge studies applies 10 or 100-fold for herd immunity.

People like Alex Tabarrok keep banging the drums for challenge studies, but is anyone paying attention? (Maybe its too late.)

Cash and T-bills are not close substitutes

Tyler Cowen Alex Tabarrok recently linked to an interview of Eugene Fama. His discussion of the EMH is brilliant, as you’d expect. He’s right about “bubbles”. But Alex chooses to quote from his discussion of monetary policy, which is almost completely inaccurate:

It’s not just the Fed, around the globe central banks are flooding the system with liquidity like never before. Is this a reason for concern?

Frankly, I think this is just posturing. Actually, the central banks don’t do anything real. They are issuing one form of debt to buy another form of debt. If you are an old Modigliani–Miller person the way I am, you think that’s a neutral activity: You’re issuing short-term debt to buy long-term debt or vice-versa. That’s not something that should have any real effects.

Then again, the financial markets sure seem to love it. At least it looks like that the S&P 500 is moving upwards in tandem with the expansion of the Fed’s balance sheet.

Every day we hear a story about the movement of stock prices. But the story is different each day. So basically, these stories are made up after the fact. But when we look at it systematically, we don’t see a big effect of Fed actions on real activity or on stock prices or on anything else. That’s why I use to say that the business of central banks is like pornography: In essence, it’s just entertainment and it doesn’t have any real effects.

Let’s consider the applicability of Modigliani–Miller to monetary policy. Is it true that the Fed was just swapping one liability for another similar government liability, which should not have much impact on relative prices?

Conventional economists might try to disprove Fama by pointing to the fact that the Fed can use OMOs to control interest rates. But here I’d agree with Fama; their control of rates is greatly exaggerated. Unfortunately, he’s still wrong about OMOs not having much effect.

Fama’s been making this argument for a long time, well before the zero bound situation arose. So let’s go back to 2007, when the monetary base was 98% composed of currency. At the time, one-month T-bills yielded 5% and currency yielded 0%. How can yields have been so different on two assets that are supposedly close substitutes?

[Update: In the comment section, John Cochrane says Fama is speaking to the post-2008 period, so the preceding paragraph may be in error. The sentence beginning “Actually” kind of sounded like he was making a general proposition about central banks in various times and places. The central banks of India? Turkey?]

And how many times in your life have you stood in a checkout line at a store behind someone trying to pay for purchases with T-bills? Zero? Same here.

So both common sense experience and market prices confirm that currency and T-bills are not at all close substitutes. Don’t like 2007? In 1981, T-bills yielded 15% while cash yielded 0%. In Switzerland, cash yields more that government bonds. How does M-M explain that fact?

Currency is more like “paper gold”, an asset with idiosyncratic uses that are almost unrelated to our financial system. Most currency is used for small transactions and tax evasion, areas in which T-bills are useless.

So Fama’s basic approach is wrong. Now it’s true that modern interest bearing bank reserves are much closer substitutes for T-bills than is currency. And it’s useful to think about how to model this. But you need a general model that applies to both bank reserves and currency. The M-M theorem is not a useful starting point, even in the world of 2020.

Fama’s also wrong about market responses to Fed announcements. Fed announcements often have a powerful effect on financial markets within seconds of the announcement. The odds of this being random are far less than 1 divided by the number of atoms in the universe. Perhaps Fama hasn’t studied this area.

To be sure, there are plenty of problems with empirical work on the impact of monetary policy, especially its longer run impact on the macroeconomy. Economists have done a poor job of addressing the identification problem, and this may partly explain Fama’s skepticism. But there’s no doubt that the Fed has a huge impact on asset prices; anyone that follows the markets closely knows this.

In one respect it is like pornography; I know it when I see it. Here’s the impact on the Dow of the unexpectedly contractionary Fed announcement of December 11, 2007, which occurred at 2:15pm:

That’s more disgusting than any picture of a naked lady.

There are dozens of similar examples. Modigliani–Miller has no implications for monetary economics.

Here’s why my dispute with Fama is so ironic. Who convinced me to look at monetary policy this way? Fama did! Once I understood the EMH, I realized that other macroeconomists were doing things all wrong. The way to do macro is to look at the immediate impact of policy surprises on asset prices. I quickly saw that economists were grossly underestimating the impact of policy shocks like December 11, 2007, which triggered the Great Recession.

So I’ve used Fama’s EMH model for monetary economics, and ended up as far away from Fama as it is possible to be.

Why is the press so easy on Trump?

Over the past 4 years, I’ve consistently argued that the press has gone easy on Trump. The criticism for his botched handling of the Puerto Rico hurricane was mild compared to the criticism Bush received for his handling of Katrina. (Perhaps neither president should be blamed, but there is a clear double standard.)

Trump outrages that would cause a normal president to be hammered by the press for months are barely even mentioned by the US news media. How much attention was paid to Trump’s praise of war criminals? Or his encouraging Xi to put Muslims into concentration camps? Or his claim that we should have stolen Iraq’s oil? Almost none. Why does he consistently get off almost scot-free?

Now the media seems to be acknowledging its shameful double standard:

He is held, by necessity, to a more forgiving standard than any modern president. But however low the bar is set, Trump continues to trip over it.

Here is the question by Guthrie that gave Trump the most difficulty:

“Let me ask you about QAnon. It is this theory that Democrats are a satanic pedophile ring and that you are the savior, of that. Now can you just, once and for all, state that that is completely not true?”

Answering this query should have been extremely simple. Trump couldn’t do it.

Imagine if Biden had been asked the same question. He would have had little difficulty saying, ‘No, the Democratic party is not the cover for a satanic sex ring.’ The reason reporters don’t pose this question to Biden is not because it would be too difficult for him to answer, but because it would be too easy.

Another example:

Guthrie asked Trump why he tweeted “a conspiracy theory that Joe Biden orchestrated to have SEAL Team Six, the Navy SEAL Team Six, killed to cover up the fake death of bin Laden.” If Biden had tweeted out a claim that Trump had killed somebody, and that person was in fact alive, he would probably be asked about it — a lot. Indeed, if Biden had tweeted a ludicrous murder accusation, it would represent a crisis for his campaign so dire that the press would likely talk about little else. His allies would be pressed to denounce him, and Democrats would be discussing ways to force him off the ticket. For Trump, it was just another item on the list of questions.

It’s not the first time Trump has made a baseless murder accusation.

PS. I like Matt Yglesias’ take on the Hunter Biden scandal:

One corrupt son is better than two.

PPS. This headline caught my eye:

Ardern Storms to Historic Election Victory After Crushing Covid

Wait . . . you mean voters like it when you prevent thousands of deaths?

Maybe that’s why Merkel is so popular.

The September retail sales shocker

Remember how people used to say, “This isn’t your grandpa’s recession”? Well this isn’t even your slightly older brother’s recession.

The government just announced that retail sales surged 1.9% in September, way above expectations. In a normal year, that would be 6-months worth of growth.

Some will argue that when rebounding from a deep slump it’s levels that matter, not growth rates. And to some extent that’s true. But retail sales had already fully recovered by August, so this figure pushes us well above trend.

It’s almost impossible to overstate the weirdness of this recession. In a normal recession, retail sales plunge sharply and take years to fully recover. Look at the Great Recession, for instance. This time around sales fully recovered in just 4 months, and just 5 months from the April trough sales are already above trend:

And yet despite the surge in retail sales, the overall economy remains deeply depressed. RGDP is down sharply, and total employment is down by roughly 10 million. What gives?

Obviously, this is not a normal demand-side recession. That’s why the fiscal cliff at the end of July did not affect retail sales. The economy is being held back by Covid-19, not a lack of disposable income. Covid is obviously a problem in the service sector, but more surprisingly is also a problem in manufacturing. But how could manufacturing remain deeply depressed while retail sales booms? Who builds the stuff being sold in stores?

American manufacturing is increasingly focused on investment, not consumer goods sold in stores (many of which are imported.). The slump in travel affects everything from fracking to aircraft manufacturing to hotel/restaurant construction. With less investment, there is less manufacturing of inputs used in investment, like oil pipelines for frackers. We’ve seen other manufacturing mini-slumps when oil prices tanked, even when the broader economy was OK.

We don’t need fiscal stimulus. We do need a fiscal relief package to help the many people who are being hurt by Covid-19. The difference between fiscal “stimulus” and a fiscal relief package is that the latter would not include $1200 checks to middle class Americans with jobs.

PS. This post is not a forecast. Hospitalizations are now entering a third wave, and this may slow the economy. I cannot predict Covid and thus I cannot predict the economy. I’m also not denying that demand still has some effect, even in a supply-side recession. Nor am I denying that it would be better to have a more expansionary monetary policy, expansionary enough to raise inflation expectations up to 2%. That’s all true. Nonetheless, this recession could end quickly if we get a widely available vaccine or a cure. It’s mostly supply-side, which means it’s nothing like 2009.