Archive for April 2023


Recent studies of the identification problem

My new book focuses on the question of how best to think about the stance of monetary policy. This is closely related to the “identification problem”, which is the problem of identifying monetary policy shocks.

Some of the earliest attempts to estimate the impact of monetary shocks resulted in what’s called a “price puzzle”. It seemed like easy money led to lower inflation, and vice versa. That makes no sense. (A similar problem occurs with fiscal policy, where deficit spending is often correlated with slower economic growth.) The trick is to find the exogenous part of monetary policy, not the response to economic conditions. Are interest rates rising because of tight money, or because of higher NGDP growth expectations?

Some of the best work in this area has been done by Eric Swanson. In a new paper, Michael Bauer and Eric Swanson estimate structural VAR models in two ways. In the set of graphs below, the results on the right show a naive model, that doesn’t fully account for feedback effects. The response functions show the impact of a 25 basis point rise in the 2-year Treasury yield.

You can see the price puzzle in the left column (second row), where the CPI seems to rise in response to an increase in interest rates (assumed to be a tighter monetary policy.)

In the right column, the monetary policy shock (mps) is orthogonalized. This involves regressing the shock on previous macro data, and then using the residual (the unexplained portion) as the exogenous monetary policy shock. Now industrial production, the CPI, credit spreads (EBP) and interest rates all move as expected.

Another innovation in the Bauer-Swanson paper was to look beyond FOMC announcements, and also include market responses to speeches by the Fed chair. The combined effect of all of these innovations was to produce much more reliable estimates of the impact of monetary policy. Policy was shown to have a four times larger impact than observed in previous studies.

The following graph also shows the response of commodity prices and unemployment. Notice that the (sticky price) CPI barely moves at first, while (flexible) commodity prices immediately plunge by about 100 basis points. But after 50 months, both are down roughly 20 basis points, which is consistent with long run money neutrality:

Bauer and Swanson do a nice job explaining why the private sector has a difficult time predicting (and identifying) changes in the stance of monetary policy:

There are a number of plausible reasons to think that private-sector learning about the Fed’s monetary policy rule would be quite slow in practice, with the result that changes in αt would cause a persistently large discrepancy αt − at . First, learning about a persistent component (αt) from a noisy time series (it) is difficult and happens only gradually, with longlasting biases in beliefs; see Farmer, Nakamura and Steinsson (2021) for a recent discussion. Second, the private sector in reality faces a multidimensional learning problem: Realistic policy rules are of course multivariate, requiring the public to learn about several parameters at once, which greatly slows down the learning process (Johannes et al., 2016). Third, the private sector must form beliefs about which macroeconomic and financial variables enter the Fed’s monetary policy rule, i.e., about its functional form. Fourth, the Fed’s monetary policy rule could contain nonlinearities—which we have also abstracted from here—so that, in practice, the Fed responds most aggressively to the economy when the economic data is most extreme. These extreme events occur only very rarely, so it is extraordinarily difficult for the private sector to learn the Fed’s true responsiveness to the economy during these rare episodes.

Interestingly, Bauer and Swanson found that Fed speeches had a surprisingly small impact on the stock market:

The last two columns of Table 3 report the estimated effects of Fed Chair speeches on financial markets. Two-year and five-year Treasury yields respond almost identically to Fed Chair speeches as they do to FOMC announcements, while ten- and 30-year Treasury yields respond even more strongly. The R2 for Fed Chair speech effects are also even higher than those for FOMC announcements. Together, these observations confirm the general point in Swanson and Jayawickrema (2021) that speeches by the Fed Chair are even more important for the Treasury market than FOMC announcements themselves. By contrast, the response of the stock market is substantially weaker, with an R2 around 3 percent. The modest stock market response to Chair speeches is somewhat puzzling in light of the fact that monetary policy typically has pronounced effects on the stock market (Bernanke and Kuttner, 2005; G¨urkaynak et al., 2005). One possible explanation is based on information effects: Speeches by the Fed Chair could potentially have larger information effects than FOMC announcements, given the extensive conversations the Chair is having with the public or Congress about the Fed’s outlook for monetary policy and the U.S. economy. . . . Another explanation is that other news besides the Chair’s speech could have moved interest rates and stock prices during the event window.

I wonder if this difference might also reflect the fact that FOMC announcements primarily move rates via the “liquidity effect”, that is, where higher interest rates represent tighter money. Perhaps Fed speeches also impact rates via the Fisher effect. Thus the speech might convey information about the chair’s longer run views on the appropriate monetary regime. In that case, higher rates can be associated with easier money.

Here it might be useful to review a couple graphs from my new book, which show two possible expected exchange rate paths after a monetary shock:

These two money announcements have an identical impact on interest rates. Due to the interest parity condition, both reduce the nominal interest rate for a period of time. But in the first case the shock produces long run currency depreciation, and hence is expansionary. This is Dornbusch overshooting. In the second case, there is long run currency appreciation, and hence the shock is contractionary.

If you think of monetary policy as moving the actual interest rate relative to the natural rate, in the first case the market rate has been reduced, without any necessary change in the natural rate. In the Swiss case from 2015, the sharp currency appreciation caused the natural interest rate to fall even more sharply than the actual interest rate, hence policy became tighter.

The ECB recently published a very interesting paper by Marek Jarociński, which looks at these issues using a slightly different approach. I’m pretty weak at econometrics so I’m not qualified to provide an overall evaluation of the paper (or Bauer-Swanson), but I like the way Jarociński thinks about these issues. Here’s the abstract:

Fed monetary policy announcements convey a mix of news about different conventional and unconventional policies, and about the economy. Financial market responses to these announcements are usually very small, but sometimes very large. I estimate the underlying structural shocks exploiting this feature of the data, both assuming that the structural shocks are independent and relaxing this assumption. Either approach yields the same tightly estimated shocks that can be naturally labeled as standard monetary policy, Odyssean forward guidance, large scale asset purchases and Delphic forward guidance.

I like this. It’s not just a question of how much the interest rate changes, there are qualitative differences between various types of monetary shocks. Again, if we use the market rate/natural rate language, monetary shocks can move both the market interest rate and the natural interest rate.

In my research on the Great Depression, I concluded that extremely large shocks provided an unusually important source of information about the structure of the economy. Here’s Jarociński:

To identify the structural shocks I exploit a striking, yet hitherto neglected feature of the data. Namely, financial market reactions to FOMC announcements are usually very small, but sometimes very large, i.e. they have very fat tails, or excess kurtosis. This feature implies that the data may contain information about the nature of the underlying structural shocks. Given the importance of the Fed policies, it is vital to exploit this available information as well as possible. Previous literature has ignored it, treating the shocks explicitly or implicitly as Gaussian. This paper is, to my knowledge, the first attempt to tap this valuable source of information.

And here’s how he identifies the four types of shocks:

More in detail, the baseline model expresses the surprises (i.e., the high-frequency reactions to FOMC announcements) in the near-term fed funds futures, 2- and 10-year Treasury yield and the S&P500 stock index as linear combinations of four Student-t distributed shocks. It turns out that these four shocks are very precisely estimated and ex post have natural economic interpretations. The first shock raises the near-term fed funds futures, with a diminishing effect on longer maturities, and depresses the stock prices. It can be naturally labeled as the standard monetary policy shock. The remaining shocks do not meaningfully affect the near-term fed funds futures. The second shock increases the 2-year Treasury yield the most and depresses the stock prices. It can be naturally labeled as the (Odyssean) forward guidance shock. The third shock increases the 10-year Treasury yield the most and plays a large role in some of the most important asset purchase announcements. It can be naturally labeled as the asset purchase shock. The fourth shock has a similar impact on the yield curve as the Odyssean forward guidance shock, but triggers an increase, rather than a decrease, in the stock prices. Therefore, this shock matches the concept of Delphic forward guidance introduced by Campbell et al. (2012).

I welcome all of these studies, partly because they support my intuition that monetary policy (including forward guidance and QE) are much more powerful than many people seem to believe. We have a monetary regime that when combined with the dominant theoretical framework (roughly IS-LM) is almost ideally suited to making monetary policy appear less effective than it actually is.

On the other hand, I don’t think we’ll ever be able to resolve the key macro policy problems using this approach. Policy is too complex—it’s about both Fed errors of commission and omission, what are often extremely hard to identify.

Instead, I favor moving toward a regime where we collapse the monetary policy indicator, instrument and goal into one variable—NGDP futures prices. Ideally, NGDP futures prices would be the policy instrument (the thing we target), the policy indicator (the thing we look at to determine the current stance of policy) and the policy goal (with the goal being say 4%/year growth in NGDP futures prices, level targeting.)

Then we can stop doing all these SVAR models.

HT: Ben Southwood.

PS. I also comment on the Bauer/Swanson paper over at Econlog.

And you wonder why we are unpopular

Bloomberg has an article on the growing unpopularity of the US:

Former Treasury Secretary Lawrence Summers warned of “troubling” signs that the US is losing global influence as other powers align together and win favor among nations not yet aligned. . . .

“Somebody from a developing country said to me, ‘what we get from China is an airport. What we get from the United States is a lecture,’” said Summers, a Harvard University professor and paid contributor to Bloomberg TV.

Obviously, lecturing other countries is not the best way to win friends and influence people. Better to lead by example. But it’s actually far worse than Summers suggests.

Over the past 4 decades, many if not most of our “lectures” have been US officials arrogantly telling less developed nations (and even developed places) that they needed to follow the “Washington Consensus”. You remember the Washington Consensus, the idea that countries should refrain from protectionism and industrial policies.

Now the US has abandoned the Washington Consensus and decided to go all in with protectionism and industrial policy. And that’s because we supposedly need to do this to keep from falling behind. But weren’t we told that these policies slow economic development?

It’s annoying when you get lectured to by more successful countries. It’s especially annoying then the lecture comes from self righteous societies that don’t follow their own advice. Is it any wonder that developing countries have lost respect for the US government.

I have too.

Before you call me a moron . . .

You might wish to consider whether you and I hold the same views. Here’s a brand new commenter named kipd:

stats are made to be manipulated and then lied with you moron. try some time out here in the real world. EVERY business I can see needs people. I’m building a house. EVERY sub is begging for people. this crisis did not exist 3 years ago. you exemplify the “elitist” leftists of the phantasy utopia of big cities and college campuses.

Of course I agree that there’s a tremendous shortage of workers. I’ve discussed that problem in dozens of posts, well before most other pundits. So why does kipd assume that I disagree with him or her? (Based on the tone, I’m guessing “him” is the preferred pronoun.)

I presume it’s due to my recent post about the myth of the “missing workers”, the idea that workers left the labor force during Covid and never returned.

There are no missing workers; employment is at an all time high. The very real labor shortage is due to the huge surge in aggregate demand triggered by a recklessly expansionary Fed monetary policy. Full stop.

We have more than enough workers to produce an equilibrium level of output. But due to the high level of demand, firms are trying to produce a level of output far beyond equilibrium. We have a worker shortage, despite not having any missing workers. So tell me kipd, who’s the moron?

Another commenter linked to this graph:

I hate it when people draw trend lines over cyclical data; they seduce us into seeing patterns that are not there. It’s like drawing a line from the 2009 trough to the 2019 peak for “trend RGDP”. This is done all the time, and probably deserves a new post. But I’m too lazy.

The problem here is that the trend growth in “not in labor force” is gradually accelerating over this entire period, due to the increasing retirement of boomers. During economic booms such as the late 2010s, the rate of growth slows briefly. But the underlying trend is rising, and is steeper than the line shown on this graph. Draw a slowly rising trend line through the very first and the very last point, and you’ll have a more accurate read of the actual trend. The late 2010s are below the actual trend line due to the cyclically strong economy. (Just as RGDP growth was above trend.)

The Economist sees the light . . . er . . . darkness

The day after the 2020 election, I pointed out that Trump would be back. In contrast, most of the media wrote him off after January 6. But once a country becomes a banana republic, it’s hard to go back. If Trump were to die tomorrow, another authoritarian demagogue would replace him.

The Economist has a new article entitled:

Why do Democrats keep helping Trump?

The charges in Manhattan have boosted his political prospects by restoring him to his favourite role

They point to the deterioration in America’s political culture:

It could be hoped, once, that the challenge to decency posed by Mr Trump would bring out the best in America. Instead, Mr Trump has made zealots and quislings out of Republicans while inflaming and dumbing down Democrats and polarising, which is to say corrupting, the news media. He promises to make America great but he keeps making it worse, and he is far from done.

Exactly. History suggests that authoritarian nationalism doesn’t end well. The worst is yet to come.

PS. It’s in Biden’s interest to face Trump in 2024. But it’s not in America’s interest. Bragg is helping both Biden and Trump (and hurting DeSantis).

A theory of second best

What is the best poem? The best play? The best sculpture? The best work of architecture? I’m not qualified to say, but I think I know which work has the reputation of being the greatest in its field. (The Divine Comedy, Hamlet, David, Taj Mahal.)

Now consider second best. What is the second greatest work in each field? Isn’t that question much harder to answer?

[As an aside, I don’t believe that every field has an obvious number one. What’s the best novel? War and Peace? Ulysses? In Search of Lost Time? Don Quixote? Madame Bovary?
And what’s the greatest rock album? Heck, there are three of four possibilities for the greatest album by the Beatles, or Dylan, or the Stones. No obvious “focal point”.]

Today I’d like to search for the second best painting. (I could look for the best, but that question seems too easy to be interesting.) Let’s start with Velazquez, often regarded as the greatest painter of all time. What’s his second best painting? I have no idea. His best history painting is probably The Surrender of Breda:

His best portrait is probably Juan de Pareja:

I’m old enough to recall then the Met bought this (in 1971) for the then record price of $5.5 million, which seems absurdly low, even accounting for the effects of inflation.

His best (only?) nude is certainly the Rokeby Venus:

When you go to an art museum, most paintings from the Baroque era have a sort of “old masterish” quality, which makes them seem dated. These two paintings by Velasquez have an uncanny timeless quality. Indeed this was even recognized at the time. When Juan de Pareja was first displayed in Italy, it astounded viewers with its life-like quality. I’m not aware of any reaction to the Rokeby Venus, which presumably was displayed only in private due to its “pornographic” quality. But it certainly looks far more modern than a nude by Giorgioni, Titian or Rubens.

Alternatively, we might look for the Velazquez painting that most resembles the number one painting, say The Spinners (which was recently restored):

Like the number one painting, this doesn’t reproduce well and must be seen in person.

Notice that Velazquez tips his hat to Titian by placing a tapestry of The Rape of Europa in the background:

I recall an art expect claiming that Titian’s masterpiece was the greatest painting in the Western hemisphere. But given the disparity between the two hemispheres, that’s almost as weak a compliment as being called the greatest painting in the Southern hemisphere.

Instead of looking for the second best painting by the greatest painter, perhaps we should be searching for the best painting by the second best painter?

That task is not easy, as Vermeer produced many fewer paintings than Velasquez, and the quality tends to be more uniform. They all tend to be rather small, and near perfect. Should we look for an example of the sort of painting that he was most associated with, say a woman standing near a window? If so, which one?

Or an outdoor scene, like View of Delft or The Little Street? Or a portrait, such as Girl with a Pearl Earring?

In the end, I lean toward The Allegory of Painting:

Last year, I traveled all the way to Vienna to see this painting, and it was not on display. Interestingly, of all of Vermeer’s paintings, this is the one with a composition that most resembles Las Meninas:

You might wonder if I’m just an old reactionary, ignoring all the great art produced in recent centuries. I don’t think so. These paintings by Velazquez and Vermeer are actually more recent that the greatest works of poetry, drama, sculpture and architecture cited at the top of the post.

These two painters took realism as far as it could go, and achieved the most profound effects. After that, painting stagnated for 150 years, becoming somewhat decorative. In the nineteenth century, painters began a long series of visual experiments. When those were exhausted around 1960, conceptual art took over. But no single modern experiment, no matter how impressive, can compete with the perfection of a great Velasquez or Vermeer.

They painted at roughly the same time, for the same reason that Newton and Leibnitz worked on calculus at roughly the same time. It was the appropriate time to produce a new form of math, and the appropriate time to produce the most perfect paintings. Sorry, you can’t invent a new form of math or produce great paintings, because you were born too late.

I have a bias toward technique, but there are other ways to evaluate art, such as the emotional force of an image. Using that criterion, we can begin considering more “primitive” painters, such as Giotto. Since it’s Easter, how about concluding with the famous Isenheim Altarpiece:

PS. The Illustrated London News did survey of art experts back in 1985. This photo is from a Boston Phoenix article discussing the poll:

The Phoenix did its own poll, and American art experts had broadly similar views to those in the UK:

Note the striking changes in these two 1985 surveys from the earlier (1950) “Masterpieces” survey. The author of the Phoenix article (David Bonetti) had this to say about the differences:

I think we can detect in both 1985 lists, especially the British, a shift in taste from the blockbuster to the subtle, the complex, and the personal, a shift from which both Vermeer and Velazquez were bound to benefit.