In a recent column entitled “Recession was inevitable, economists said. Here’s why they were wrong,” Gary Smith writes:
In an August 2022 CNBC interview, Steve H. Hanke, a Johns Hopkins University economics professor, predicted: ‘We’re going to have one whopper of a recession in 2023.’ In April 2023, he repeated the warning: ‘We know the recession is baked in the cake,’ he said. Many other economists also anticipated a recession in 2023. They were wrong.”
I am not an expert on monetary policy or economics. Rather, this story interests me as a political scientist, in that policy recommendations sometimes rely on academic arguments, and also as a student of statistical workflow I am interested in how people revise their models when they learn that they have made a mistake.
Along those lines, I sent an email to Hanke asking if he had written anything addressing his error regarding the recession prediction, and how he had revised his understanding of macroeconomics after the predicted outcome did not come to pass.
Hanke replied:
Allow me to first respond to your query of January 23rd. No, I have not written up why my longtime colleague John Greenwood and I changed our forecast concerning the timing of a likely recession. But, given your question, I now plan to do that. More on that below.
In brief, Greenwood and I employ the quantity theory of money to diagnose and predict the course of the economy (both inflation and real GDP growth). That’s the model, if you will, and we did not change our model prior to changing our forecast. So, why was our timing on the likely onset of a recession off? After the onset of the COVID pandemic, the money supply, broadly measured by M2, exploded at an unprecedented rate, resulting in a large quantity of excess money balances (see Table 2, p. 49 of the attached Greenwood-Hanke paper in the Journal of Applied Corporate Finance). We assumed, given historical patterns, etc., that this excess money would be exhausted and that a recession would commence in late 2023. Note that economic activity is typically affected with a lag of between 6 and 18 months after a significant change in the money supply. The lags are long and variable, sometimes even shorter than 6 months and longer than 18 months.
We monitored the data and realized that the excess money exhaustion was taking longer than we had originally assumed. So, we changed our forecast, but not our model. The attached Hanke-Greenwood article contains our new forecast and the reason why we believe a recession is “baked in the cake” in late 2024.
All this is very much in line with John Maynard Keynes’ quip, which has become somewhat of an adage: “When the facts change, I change my mind. What do you do, sir?”
Now, for a little context. After thinking about your question, I will include a more elaborate answer in a chapter in a book on money and banking that I am under contract to deliver by July. That chapter will include an extensive discussion of why the quantity theory of money allowed for an accurate diagnosis of the course of the economy and inflation during the Great Financial Crisis of 2008. In addition, I will include a discussion of how Greenwood and I ended up being almost the only ones that were able to anticipate the course of inflation in the post-pandemic period. Indeed, in 2021, we predicted that U.S. headline CPI would peak at 9% per year. This turned out to be very close to the 9.1% per year CPI peak in June 2022. Then, the Fed flipped the switch on its monetary printing presses. Since March 2022, the U.S. money supply has been falling like a stone. With that, Greenwood and I forecasted that CPI would end 2023 between 2% and 5% per year. With December’s CPI reading coming in at 3.4% per year, we hit the bullseye again. And, in this chapter, I will also elaborate on the details of why our initial prediction of the onset of a recession was too early, and why the data have essentially dictated that we move the onset forward by roughly a full year. In short, we have moved from the typical short end of the lag for the onset of a recession to the long end.
Again, macroeconomics is not my area of expertise. My last economics class was in 11th grade, and I remember our teacher telling us about challenges such as whether checking accounts count as “money.” I’m sure that everything is a zillion times more complicated now. So I’ll just leave the discussion above as is. Make of it what you will.
P.S. Since writing the above I came across a relevant news article by Jeanna Smialek and Ben Casselman entitled, “Economists Predicted a Recession. So Far They’ve Been Wrong: A widely predicted recession never showed up. Now, economists are assessing what the unexpected resilience tells us about the future.”
You people robbed us just like you robbed the Greeks. That theft is powering your economy for a little while for now.
Regarding Hanke’s reference to Keynes:
“All this is very much in line with John Maynard Keynes’ quip, which has become somewhat of an adage: ‘When the facts change, I change my mind. What do you do, sir?’”
An adage/quip it may be, but I have seen several references which claim the adage/quip is apocryphal:
https://quoteinvestigator.com/2011/07/22/keynes-change-mind/
On a very personal note, as I get older, changing facts hardly change my mind at all.
I think the sentiment in the referenced quip is telling, though. Hanke’s models were not based entirely on “facts”; they were also based on assumptions. I think many an economist/author would be well served by acknowledging those assumptions when making predictions and not be so self-assured in their pronouncements.
I don’t find Hanke’s response very reassuring. A model that predicts a recession within “6 to 18 months” is pretty broad; so broad as to be close to useless. And it turns out that “longer than 18 months” is not so rare as to cause us to gasp. Color me unimpressed with his willingness to consider what might have gone wrong with his predictions.
Financial markets participant here: Forecasting inflation in a range between 2% and 5% for 2023 is like holding a dart while standing three feet away from the side of a barn and saying you’re going to hit it. It’s not a serious or useful forecast.
His answer kind of reminds me of the people predicting the end of the world on a certain date — when it turns out to be wrong, they usually just discover a miscalculation and carry on confidently regardless.
Note also that he didn’t say a recession was “very likely” in 2023, he said it was “baked in the cake.” Now, he says that it was always less than certain? What’s the joke? Economists have successfully predicted nine of the last two recessions.
To pile on a bit: Hanke’s response talks about how successful his model was at predicting inflation. But the prediction at issue is a recession (ie GDP growth). This is a lack of candor on his part.
The only quantity that can possibly matter is a dimensionless one. Neither GDP nor CPI or any similar index matters by itself.
Such a quantity might be defined for example by GDP/population /(cost of a basket of family goods * avg personal basket consumption rate )
This is dimensionless and therefore has a hope of being meaningful, and also reflects on the population of interest (everyday families).
Then the question is did this increase or decrease? I have a post in moderation which shows it’s very likely this quantity hit a peak in 2021 and declined after then rose and may be about back where it was in 2021
So the real question is why are economists so oblivious that they don’t even know that the recession DID occur?
Could you say a little about why you think only dimensionless magnitudes stand a chance of being meaningful?
Yes. This is a fundamental symmetry law of the universe.
If two people observe something and each one measures the thing with a different set of units they should still agree that the same thing occurred. In order for that to be the case, the thing that occurred when measured quantitatively must be invariant to what units they use for measurement.
The formal statement of this is known as Buckingham’s Pi theorem and was first proved mathematically in 1878 according to wikipedia: https://en.wikipedia.org/wiki/Buckingham_%CF%80_theorem
To understand why this is and why it still matters for economics consider the following:
GDP is measured in dollars per year. If this number matters, we can make it bigger by simply measuring the amount of dollars exchanged in let’s say 2 years, and we would double the size of the number. If the size of the number makes a difference in people’s lives, then any method of doubling the size would by assumption improve the economy. Since it’s possible to double this number by just redefining the time interval of measurement, its numerical level can’t possibly matter.
Consider the alternative dimensionless measure:
((Money value of final goods / one year of measurement time) / population ) / (price of a basket of rent, food, clothing, transportation, childcare, healthcare and utilities utilized by the average person / timescale over which they utilize that size basket)
This is dimensionless, we’ve got GDP/capita in dollars / person / year, and we’ve got (dollars/ basket * baskets/person / year) in dollars / person / year
if we change the timescale of measurement to say 6 months, nothing changes about this number, if we change the unit from person to “household of 2.5 people” nothing changes about the number, if we change the currency from dollars to pennies… nothing changes. The number is invariant to any change of units, and therefore is capable of representing something real in the universe.
Daniel –
You lost me there. Dollars per year is a rate. If GDP is 2 x 10^12 dollars per year in 2022 and 2 x 10^12 dollars per year in 2023, and you measure 4 x 10^12 dollars changing hands in the 2-year period 2022-2023, you still get 2 x 10^12 dollars per year.
John, what I mean is this, measure dollars per foo where foo is a unit of time equal to 2 years. Now suddenly the numerical value in your variable is twice as large, because per one foo… You have twice as much money value changing hands.
In an equation describing the world using numbers say
F(a,b,c) = 0
An alien who uses weird units of time, length, mass, or currency can verify this equation describes some experiment if a,b,c are dimensionless but if a,b,c are not dimensionless then you must also transmit additional and irrelevant information to them about the way *your culture* measures time, length, mass, and currency in order to have them verify it.
Unless we are talking specifically about the NIST and measurement technology, the information about the way you measure the dimensions is extraneous to the activities you are measuring.
For example, if you tell me that the length of a belt divided by the circumference of your waist is 1.2 then I know the belt fits you. If you tell me that the belt is 39 inches long, and your waist is 32.5 inches long, I can verify your assertions only if I know how big an inch is. For example maybe you used a cloth tape that shrank in the wash, and so although you assert these measurements are true, in reality your belt is 37.1 inches and your waist 30.95 inches.
However even if your measuring tape is incorrectly marked, if your measurements were made with the same tape I can verify the ratio and it’s **only the ratio** that matters for whether the belt fits.
@Daniel Lakeland:
‘The only quantity that can possibly matter is a dimensionless one.’
Could you explain what you mean by the word, ‘matter’? Whether something ‘matters’ is usually a matter of perspective (e.g., the non-dimensionless number of euros in my bank account matters to me, even though it may not matter to you). Therefore, I think you are talking about a kind of ‘objective meaningfulness’, something that transcends opinion. Could you briefly explain how you derive your ‘objective meaningfulness’?
Does it really matter to you, or do you only think it matters because secretly you know the price of stuff?
What matters to you is the number of dollars in your bank account divided by the number of dollars it costs to buy the stuff you care about. If the govt said “On July 1 we are putting an extra 999 dollars into every bank account for every dollar currently in the account”, you might THINK you had just won the lottery and were rich, but when you went out the next day you would find that you couldn’t buy any more stuff than the day before despite having 1000x as many dollars.
> The only quantity that can possibly matter is a dimensionless one. Neither GDP nor CPI or any similar index matters by itself.
For what it’s worth in the context of predicting recessions GDP growth means CPI-adjusted GDP growth.
It seems that the dimensions of CPI-adjusted GDP are dollars/dollars even though there is a constant factor linked to the arbitrary choice of origin. (What is the physical dimension of the dollar anyway? (answer: 6.14 x 2.61 inches)).
In any case, the dimension of a the growth rate of anything is 1/T.
> Consider the alternative dimensionless measure: [….]
This is dimensionless, we’ve got GDP/capita in dollars / person / year, and we’ve got (dollars/ basket * baskets/person / year) in dollars / person / year [….] if we change the timescale of measurement to say 6 months, nothing changes about this number,
The number changes. Unless by some unlikely coincidence the result happens to be the same. (It wouldn’t change if everything was constant but then we would not be discussing how to measure the changes.)
Aside: you use “person” of a dimension. Blindly following your “the only quantity that can possibly matter is a dimensionless one” dictum it seems that a demographic quantity like “population” cannot possibly matter. And yet aliens can count people – a long as they know what does *your culture* count as people, I guess. “42 people” is dimensionless with an implicit “42 persons/1 person”.
You’re right that person is a question as to whether it’s a dimension or a count of discrete things. For questions involving small counts it’s probably best not to consider it as a dimension, for questions involving populations of 350M people it’s best to consider it as a separate dimension. For example your unit of population of the US could be persons, or dozens of people, or “population multiples of story county Iowa” and no confusion would arise, measures would be well described by smooth curves etc, unlike saying that the population of my house is currently .41667 dozen but once my friend walks through the door it jumps instantaneously to 0.5 dozen. There it’s best to describe as a count. Similarly we treat kilograms or pound masses as valid units of mass, even though mass is actually discrete molecules or even electrons.
Real GDP per capita is a poor substitute for a proper dimensionless ratio, as is my uncertain index of 3-4 categories, it’s just what I have access to plot. I’m just one guy not an entire govt bureau. The BEA has no such excuse.
Real GDP/capita suffers from several problems
1) the least important, is that it’s got a dollar dimensioned constant. This is equivalent to the “2019 dollars” issue. I’d prefer they give CPI as basically a dollar/person/yr quantity for the goods consumed by a “median person” but they don’t.
2) It is based on CPI all items, this is an important problem. It suffers from the issue I mention elsewhere of shifting consumption. Some shift in consumption represents people’s preferences changing through time. Few buggy whips are bought because people really don’t care about them today. Other shifts represent people being poorer off rather than merely preferring different things. Rent gets too expensive and people move under a bridge and buy Fentanyl and you consider this the same as the buggy whips issue… It’s not! Similarly people moving into crowded households. A proper index would focus on those goods which are necessary to sustain life and to grow children into functioning safe adults and based on a fixed standard of living that is neither extravagant nor impoverished. Hence, housing, food, utilities (avoiding heatstroke or freezing, cooking food etc), transportation (to and from work and school), healthcare, childcare, clothing, home maintenance and insurance for example. These are appropriate costs to include in an index which is supposed to be about widespread economic well being. To the extent that people have income that extends above those costs that’s a good thing and it’d be expected they could shift those around a lot… Cable TV vs streaming services, iPhone vs Android, eating out at Thai food vs Chinese, concert tickets vs movie tickets vs plays vs sporting events… Whatever. Including “disposable income” is problematic. In some sense a recession in power to keep your family housed is different than a recession in power to attend jazz festivals. There should be separate recession concepts for disposable vs “required” income. I shouldn’t have to say this out loud, it should be in the first chapter of every Econ textbook… But it’s not.
3) The middle importance issue is that it’s got time involved. The proper way to handle this is to index it to an internal variable of consumption rate. For example if climate change necessitates that people do a lot more house maintenance then the consumption rate of lumber and roofing and such to maintain the SAME standard of living increases. This implies a recession even if nominal GDP remains constant. Similarly if for example violence increases from protests and people need to consume more hospital services than they used to. If GDP stays constant then the newly necessitated increase in consumption for the same standard of living implies a recession. If educational quality declines… Recession… If durable goods last less long… Recession. If Private equity firms buy up all the bowling alleys (Bowlero) and stop maintaining the lanes so they’re warped and sticky and your league has to go to a luxury alley to keep the quality the same… Recession. If Private equity buys up hospitals and strips then for parts and leaves sewage seeping out of the walls and 5000 bars living on the 5th floor… Recession.
What I’m saying is, Economists have major major problems with measurement and some of them are contrary to basic ideas that have mathematical proofs from the 1870s and others are derived from a complete failure to align measurements with the underlying concepts of interest, in part because many economists fail to recognize fundamental aspects of physics, such as there is no way to substitute for caloric intake, or substitute for protection from heatstroke or freezing to death, or diabetes meds etc.
Stupid autocorrect… That’s should read 5000 bats
The bats reference https://doctorow.medium.com/when-private-equity-destroys-your-hospital-d3e6c290b1eb
The Bowlero reference
https://jacobin.com/2024/05/private-equity-bowlero-ruining-bowling
@Daniel Lakeland (10:50 AM):
Your last comment in this thread reads a bit like you are blaming the economics profession for all the government mismanagement in the world. Economists are responsible for the advice they give, not the policy decisions. Politicians are responsible for that.
Many of the issues you discuss are well known to economists in their respective fields, e.g. the failure of privatisation to make hospitals more efficient in health economics. (Of course, not every economist will know the details of each sub-field, e.g. my knowledge of ideal monetary policy is only very basic).
A big part of the problem is that (1) the real experts in a field may not even be consulted, or they are dismissed with ‘we don’t have the money for that’, (2) people fall for the easy stuff. GDP is easy. Anyone can understand GDP. Is it highly problematic? No doubt about it! But it is difficult to find replacements for these oversimplified measurements that everyone can agree on*. The EU Commission, led by a panel of experts, has come up with a number of new indicators and hopes to promote them. For an overview, see page 11 of the report “Sustainable development in the European Union – Monitoring report on progress towards the SDGs in an EU context – 2023 edition”:
https://ec.europa.eu/eurostat/web/products-flagship-publications/w/ks-04-23-184
Details on individual topics (e.g. poverty, page 40) can also be found in that report.
It is easy as pie to come up with new, better indicators than we have now. But getting everyone to accept them is the hard part.
*There is a ‘cottage industry’ (Walter Radermacher, former head of eurostat) producing their own indicators. Some of them do a good job, but they have not been widely accepted.
I’m not a monetary economist, but as I understand it, the core problem is this: Hanke is an old-school monetarist who, as he says, thinks changes in “the money supply”, as measured by M2 (which includes those checking accounts) drive economic growth and inflation. There aren’t many of these beasts left, for a variety of reasons.
A big reason is that it has become clear to most of the rest of us that monetary aggregates are endogenous; they are not outside the system, driving it, but a part of it. Most money in a modern economy is a reflection of credit creation, and credit markets are surely not outside the system. Of course, being endogenous, money supply measures are related to other measurables, but not in a simple cause-and-effect fashion. One way to link this to what Hanke said is to point out that the factors that influence the length and variability of money-“real” relationships, which he takes to be mysterious, are the main objects of macro/monetary analysis for most other economists.
I ended up doing my PhD in a different area of econ, but I did take all of the macro courses my program had to offer so I’m well-versed in it. I think you’re basically correct; the spiritual successors to this type of thinking would instead look at something like nominal GDP growth rates (which has done very well as an analytical tool, incidentally).
Another big weak point of the old-school money supply guys is the “long and variable lag” narrative. In practice the saying is little more than a rationalization for making unfocused or incorrect predictions — it’s a way of hedging a prediction to have any timing you’d like. And as an economic matter, it de-emphasizes how monetary policy changes things immediately due to changes in expectations arising from the change in monetary policy changes. (Just look at how stock markets react or exchange rates change. There’s even evidence that policy announcements can have better immediate effect than the implementation thereof).
To be fair, a lot of people, including New Keynesians with their obsessive focus on interest rates, suffer from the same blind spot even though their own models put the importance of expectations front and center. “Long and variable lags” is one of those things that have become common wisdom to certain cliques. Good news is fewer people are accepting it.
The deeper issue here is that while these types of guys are very much at the fringe of the profession, their loud predictions get an outsized response in the media. Most macroeconomists work within the paradigm that recessions are fundamentally unpredictable, but that’s a hard headline to sell. The fact that they are making predictions at all should basically disqualify their opinion.
Peter, from a monetary perspective, money comes from deficit spending. Steve Keen does the accounting on this in a simplified form (note, I’m writing not necessarily directly for you, but rather for the blog readership overall which includes a lot of NON economists please don’t feel I’m talking down to you)
https://profstevekeen.substack.com/p/its-a-mixed-credit-fiat-world-e3f
https://profstevekeen.substack.com/p/why-are-economists-trying-to-hide
Deficit spending and money creation works like this (here Money = M2 = more or less the sum of all checking account deposits and some cash bills)
0) Total money balance is D0 to start
1) Govt needs to pay a bureaucrat’s salary
2) Govt sells $1000 bond to a bank
3) Bank gives some reserves to the govt
4) Govt deposits reserve money into bureaucrats’ checking account (money is created, total money is D0 + 1000)
5) Bob goes to bank to get a loan
6) Bank has $1000 bond to use as assets against which it can lend, so lends $800 to Bob placing the money in his checking account (money was created money is D0 + 1800)
7) Bureaucrat has to pay taxes, pays $200 to the govt (money was destroyed total money was D0 + 1800 – 200)
8) The Fed agrees to buy bonds at higher prices thereby adjusting interests rates, Bank decides it needs more reserves, sells the bond to The Fed for $1002
9) Bank has $1002-800=202 to lend against, and lends out $200 (money created, total money is D0 + 1800 – 200 + 200)
Note that when the govt did its deficit spending of $1000 to pay the salary it created $1000 of money immediately. Furthermore, as things went along the bank converted $800 of the $1000 bond into present money through loans. Furthermore eventually the fed converted a $1000 par value bond into $1002 via adjustment to the market price it pays for bonds.
Deficit spending creates money. There’s a time lag and certain amount of decoupling because banks need not lend against assets every possible amount they could, but you can never have M2/TotalPublicDebt be very far from 1, at the moment it’s about 0.6 https://fred.stlouisfed.org/graph/?g=1olHK in 1990 it was 1.0
most likely some of that imbalance is due to exporting money to China and such.
I think the role of money creation is quite complex, and I believe it requires a dynamic and realistic model. In particular, deficit spending drives economic growth in sectors that provide services to the govt, such as for example Halliburton or defense contractors or sometimes stuff like the Solarwinds software that so many govt agencies used, and the Russians completely hacked for a decade and spied on the govt with.
I consider much of that not “goods and services” but rather “bads and hindrances”. For example if we drop bombs on people and destroy buildings we destroy economic value, but GDP as measured by us goes up (because depreciation is not actual it’s an estimate and the estimate is unaffected by reality of dropping bombs or whatever).
So whether deficit spending and therefore money creation drives economic growth or inflation or economic decline or whatever we need a real model of what is going on and such a model must be 1) dynamic and 2) make value judgements about what is a good vs what is a bad.
This brings to mind J. Bronowski’s quip about the fatal reasonableness of Adam Smith keeping economics from becoming an empirical science.
Very cute to use the phrasing “So, why was our timing on the likely onset of a recession off?”
Just wait till the next recession (6 months? a year? 2 years?) then declare victory stating that your models were right, but the timing was off just a bit!
Yes, exactly.
I don’t have an economics forecast, but I do have an economist forecast: whenever we next have a recession — sometime in 2024, 2025, 2026, whenever — Hanke will say “see, the recession we predicted has arrived. We admit that we got the timing wrong.”
By 2022 when he was quoted the recession had already been going for a year…
Economists drive me crazy!
https://statmodeling.stat.columbia.edu/2024/05/31/whassup-with-those-economists-who-predicted-a-recession-that-then-didnt-happen/#comment-2373142
Daniel, suddenly you’re Humpty Dumpty! A word means whatever you want it to mean.
You can argue that we need a different definition of “recession” in order to capture what you want it to capture, and you may well be right. I could make that argument about a bunch of words. But all you’re saying here is “if we change the definition of ‘recession’ in a particular way then we’ve been in one for a while”… well, sure. If we change the definition of “pig’ in a particular way then there are flying pigs.
Phil, I’m saying the only meaning any normal person takes for the statement “we are in a recession” is that “economic conditions for most people are worsening through time” and if some official measure fails to detect when that occurs because it heavily weights groups with abnormal amounts of wealth who are taking advantage of the worsening conditions to extract rent and it assumes shifting preferences are always just a good thing then it’s a bad measure and economists never should have been using it.
If you admit uncertainty across the population, create a metric that is targeted at the core of the economy and well being, and almost every spaghetti plot curve from the uncertain distribution has a downturn, then you should have no problem saying “we are in a recession”
What I’m saying is a recession is not defined by some official measure that is a bad measure, it’s a more fundamental concept and the official “real GDP per capita” doesn’t have metric validity for that concept.
One of the major concepts that’s wrong with all-items CPI is that as important stuff becomes too expensive, people shift their consumption to other stuff. So if instead of bothering with rent people live under a bridge and buy fentanyl, economists would say that their preferences were shifting and now they’re actually richer because fentanyl is way cheaper than rent and they have more disposable income, and I would say economists who say that are idiots.
Phil, definitions of recession:
investopedia: A recession is a significant, widespread, and prolonged downturn in economic activity
Whitehouse.gov The National Bureau of Economic Research (NBER) Business Cycle Dating Committee—the official recession scorekeeper—defines a recession as “a significant decline in economic activity that is spread across the economy and that lasts more than a few months.”
So, recessions are not *defined* by “official real gdp graph declines for longer than a few months” it’s just that “real gdp graph” is mostly used as the measured proxy for what they’re trying to determine.
I’m saying as a proxy it has major problems.
> By 2022 when he was quoted the recession had already been going for a year…
As you write later recessions are not *defined* by “official real gdp graph declines for longer than a few months” – and they are even less *defined* by you.
But people had indeed noticed the two-quarters decline in real GDP in 2022:
https://www.npr.org/2022/07/28/1113649843/gdp-2q-economy-2022-recession-two-quarters
> The White House has taken pains to remind people that just two quarters of negative growth doesn’t automatically mean the economy is in a recession.
Hi Daniel,
I want to reply to your comment below. Specifically:
” So if instead of bothering with rent people live under a bridge and buy fentanyl, economists would say that their preferences were shifting”
I am not an economist and I can not comment on the specifics of the CPI. However, as I understand it, economists do not simply say that preferences shifted if a observed behavior is different. If I remember correctly, one of the assumptions of rational choice is the opposite: Preferences are constant.
This assumption is important to prevent arbitrary and tautological explanations by which any change in behavior can be explained by a shift in preferences; as you demonstrate with your fentanyl example. If you do assume constant preferences, however, this gives you a powerful tool to explain how changes in conditions lead to changes in behavior. For example if the price for a Chinese EV increases suddenly due to tariffs , consumers by one from the US instead. What explains this change in behavior is not that they suddenly prefer the EV from the US but that the Chinese EV is suddenly more expensive.
Btw. I am an silent reader since many years and I absolutely love your comments. You are on fire lately!
Best, huan
Huan,
You are very kind, and I really appreciate the feedback because I sometimes feel like “old man shakes fist at cloud” and to know people are reading my stuff and thinking about it is valuable.
The truth must be somewhere in between obviously, at some point in the past buggy whips and say boots had similar utility to people, and now, boots still have utility and lots of people buy them but buggy whips are extremely niche and almost no one buys them compared to boots. Preferences MUST have shifted. Also these days we are buying cell phone plans and broadband internet, not land-lines plus analog phone modems. Kids want tablets and phones and laptops, not desktop computers, streaming music services not CDs. Fish have been decimated from the oceans and so people just don’t buy it much anymore, etc.
And on the other hand, Economists must know that it’s meaningless to say every shift in purchasing is a shift in preferences and tautologically therefore the world is the best of all worlds…
But how do they get the CPI all-items? Well, clearly they can track the prices of goods by sending people to stores and recording prices. You can find some of that data in the followup comment (I’m tired of having my comments held because of links so I’ll just attach another comment with all the links)
But how do they get the weights to give to each category? I’m not a specialist here but they describe some of it at the CPI site, (see followup links) and most likely some of it is based on Consumer Expenditure Survey (also see followup)
Under “hedonic quality adjustment” at the CPI link for example.
My biggest point is this… for the purpose of a price index the all-items methodology may have some value, but in general, price indexes are a mistaken methodology overall, because they are irrelevant if you are properly building dimensionless ratios (where dollars/dollars cancel out). Plus for the purpose of human quality of life, which is the more important concept IMHO when it comes to “recession” including stuff like statistical adjustments for the size of the screen of a big screen tv, or for streaming services vs CDs or tilapia vs salmon or whatever is pretty much missing the point.
Choose 50 random households, different cultures, different interests in food type, higher ed preferences, entertainment, vehicle types, vacation expenditures etc… but all of them will have certain things they absolutely need: shelter, calories at some baseline quality, clothing at some replacement rate, water, baseline healthcare (diabetes meds, allergy treatments, blood pressure meds, etc), baseline transportation (number of miles traveled to work and school), etc.
They may choose more of some of those things than is in some sense “necessary”, like they might prefer higher quality of calories, or slightly more square footage at slightly higher cleanliness than what would be considered “acceptable to child protective services” or whatever, but there’s a baseline expenditure which they need… The Census and BLS made some effort to create a supplemental poverty measure, which is very related (see followup links). But “poverty” is not entirely the right baseline either. Obviously we should strive to imagine people thriving to some extent.
When it comes to a question of whether households have a widespread downturn in their economic consumption ability, it would make sense to calculate a typical value of the ratio of their income (in currency/time) to the price of consuming the “baseline plus some fixed amount of above baseline expenditure” (in dollars/time). Furthermore, when people buy durable goods, it would be good to calculate the expenditure they made now, divided by the expected lifetime over which they will consume the durable good.
And, rather than taking the average income and dividing by the average expenditure level, we should calculate individual incomes and individual expenditure values, and **take the average of the ratio**.
So that’s mean(income_[i]/baseline_expenditures[i]) not mean(income) / mean(expenditures)
And furthermore, we should report not just the mean but some samples from the entire distribution, say the deciles of this quantity. And furthermore, we should at least do some kind of matching on households, so we can also calculate the trend-line in time for families and report average trends rather than the trend of averages…
So, from the perspective of “how do you measure stuff” I say I simply don’t trust much of anything published by the BLS or Census or whoever because it seems to be extremely problematic not conforming to very basic notions of proper measurement.
It’s telling that when the Census compares traditional poverty measures to supplemental poverty measures there are potentially some very important and significant discrepancies. For example figure 4 in the comparison pdf in the followup comment shows basically every single category of people shows 3-10% points increase in poverty when moving to supplemental measure. For example “female reference person households” have 10.9% points increase in poverty.
Figure 5 shows child poverty rates for black children and hispanic children **doubled** between 2021 and 2022 from about 10% of children to about 19%, non-hispanic white children went from about 2.5 to 7.2%
Anyone who says we “aren’t in a recession” while childhood poverty is doubling or tripling in one year is misinformed by failure to measure properly, or an idiot, or a shill for big capital.
There are not just hints but outright screaming problems with Economic measurement at a fundamental level.
Well, my comment got held anyway even with no links… Sigh. Here are the links to stuff I referenced in previous comment which hopefully shows up soon
Price timeseries:
https://download.bls.gov/pub/time.series/ap/
CPI site:
https://www.bls.gov/cpi/factsheets/common-misconceptions-about-cpi.htm
Consumer Expenditure Survey:
https://www.bls.gov/cex/pumd.htm
Supplemental poverty measure:
https://www.census.gov/topics/income-poverty/supplemental-poverty-measure.html
comparison of poverty official vs supplemental for 2022:
https://www.census.gov/library/publications/2023/demo/p60-280.html
Daniel, everyone agrees there was a recession in 2020-2022. https://en.wikipedia.org/wiki/COVID-19_recession
Phil, apparently maybe everyone but Fred and the official BLS or whoever, because Fred shows a recession in 2020 that lasted maybe 6 months or so and no official recession since. (Official Recessions are shown as grey bands in the graph)
https://fred.stlouisfed.org/series/A939RX0Q048SBEA
> everyone agrees there was a recession in 2020-2022.
“Majority of Americans wrongly believe US is in recession – and most blame Biden”
https://www.theguardian.com/us-news/article/2024/may/22/poll-economy-recession-biden
Carlos, Phil,
I think everyone agrees right at that major dip in the “real GDP per capita” caused by the stay at home orders etc there was a recession. This is maybe March 2020-august 2020 in the US, or something like that. It’s reflected in the thin grey band on the Fred graphs.
The question is from late 2020 to mid 2024 were there additional recessions?
Based on graphs I posted elsewhere I’d argue from late 2021 through at least late 2022 if not mid 2023 we were also in a recession
https://fred.stlouisfed.org/graph/?g=1olmB
And given delays in data etc and revisions that sometimes come out it’s possible we are starting another one or not.
Normal people’s purchasing power of income declined and they spent down cash reserves and hiked up credit card debt like crazy to compensate.
https://fred.stlouisfed.org/series/RCCCBBALTOT
“Borrowing heavily on high interest rate credit cards in order to afford food and transportation and rent” is not a sign of a healthy economy in a real growth period.
Of course possibly aging baby boomers with their own homes or even investment real estate are cleaning up and taking cruises and things, subsidized by earlier major govt transfers to otherwise bankrupt businesses (cruises, airlines, vacation hotels).
So the economy is complicated and any single line is problematic, we need to understand the distribution, but I would argue given the situation of 2022-2023 if you ask college graduates 10 years from now they will say their family struggled in the years right after COVID.
I can’t tell you how many people have posted to Reddit in the last year that they’ve been scammed by their own parents who opened up credit cards in their name, maxed them out, and defaulted. It’s like more than one a month, and then in the comments sections, it’s “yeah this happened to me too” from hundreds of people. So that spike in credit card debt is at least partly driven by people having to report their parents for felony identity theft or suck up a decade of lost access to credit.
If that doesn’t point towards a recession I don’t know what does.
I guess Economists might say “well yes but they did spend that $50k in fraudulent credit card loans” and I think if they say that we should laugh at them. Fraud does not a healthy economy make.
Recessions are declared by NBER according to the opinion of a committee:
https://www.nber.org/research/business-cycle-dating
Not only that, it requires comparing a peak AND a trough. So it is necessarily post-hoc. I don’t see why people think this concept is useful.
Hi Daniel, thank you for your reply!
Regarding our discussion of constant preferences: You are obviously
right that preferences do change. The point of assuming that
preferences in rational choice models are constant however is the
following: If you want to argue in rational choice models that a shift
in behaviour is caused by change in preferences as opposed to a change
conditions, you can do this . Then, however, you have to provide
independent evidence. This is to guard against useless explanations
whereby any change in behavior is explained by a change in
(unobserved) preferences. The evidence for this in preferences has to
come from an independent source, a survey for example. (I think you
agree with this, I just wanted to point it out). If an economist argues
that people live on the streets and buy fentanyl because they want to,
he has to provide evidence that they really prefer this over renting a
flat and drinking coffee. I doubt that such evidence exists and I also
doubt that economists argue for this.
Regarding your point of measurement issues in economics, I try to
summarize your points: You say economics has a measurement problem
because important measurements do not align with reality. More
precisely, they do not reflect that which is to be measured. Your
examples are measures for inflation and recession. I focus on your
example of inflation. You say that inflation as measured using the CPI
gives a distorted view because it is an average price increase where
many products are included that do not matter for many, especially
poor, people. Are you saying then, for certain purposes another
measure for inflation should be used? For example, say there are
policies designed specifically to help the poor curb with
inflation. The agencies implementing the policy should use a different
measure of inflation then agencies helping the middle class. Did I
understand you correctly?
Furthermore, I followed your link
https://www.bls.gov/cpi/factsheets/common-misconceptions-about-cpi.htm
Under section “When the cost of food rises, does the CPI assume that
consumers switch to less desired foods, such as substituting hamburger
for steak?” they write: “the BLS is not assuming that consumers
substitute hamburgers for steak. Substitution is only assumed to occur
within basic CPI index categories, such as among types of ground beef
in Chicago. Hamburger and steak are in different CPI item categories,
so no substitution between them is built into the CPI-U or CPI-W.”
Seems pretty reasonable to me! Also they say they align their methods
with agencies in OECD countries. Also seems reasonable to me.
I agree with you that for poor people the “true” (for them relevant)
inflation is likely higher than the officially reported rate. Also
your point of calculating individual income/expenditure levels and
then averaging seems a good idea. Then you also can average only over
some relevant part of the income distribution. This sounds nice! I
think economists might even agree! I think it would be great for you
to publish your thoughts in some econ journal. What do you think?
Daniel,
I write subject to correction but I think the “recession” is often or usually defined as the period over which the economy is in rapid decline. No matter how far it drops, once the recovery starts it is said to be “no longer in recession”, even though things can suck for years after the recovery starts.
As with inflation, economists are talking about the first derivative of the quantity that directly affects most people.
While I’m here I’ll mention that I although the stay-at-home orders obviously had a huge effect on the economy, not having those orders would also have had a huge effect on the economy. Countries like Peru that tried to continue business as usual ended up with mass graves and still had a recession. I’d blame the pandemic, not the response, at least for the first year. (After that there’s no question that the response was pretty dysfunctional).
Phil, huan, Anon… thanks for your messages!
I agree with Anon that the actual official post-hoc recession declaration is largely political garbage. But it does represent the interaction between Economists and politics to some extent.
huan,
>Are you saying then, for certain purposes another measure for inflation should be used?
CPI as published by fred at least, is actually a dimensionless ratio, its the ratio of a basket price at one time to a basket of prices at a previous set time (times 100). Unfortunately, it’s usually used improperly. For example, let’s look at “real GDP”. Take GDP in nominal dollars and divide it by CPI. This is still a dimensional ratio in dollars/time because it now has a constant factor (the basket price at a fixed time). But the TIME dimension has been ignored! The problem isn’t solved. But Economists pretend it is.
Let’s see what a real measurement would look like, (GDP/population) / (price of basket of goods used by 1 person / duration of time for one person to consume the basket)
Now we have (dollars/time/person) in the numerator and (dollars / time / person) in the denominator. Let’s call this “dimensionless income”
Now it’s irrelevant how much inflation we’ve had, this ratio measures the same thing through time for all time! Furthermore, when technology innovations increase the time it takes to consume the basket of goods… dimensionless income goes up! and if tech changes make our durable goods break faster… the duration goes down… and dimensionless income goes down.
Another way to say it is if we fixed the time over which we measure people’s consumption as say a year or so, that the quantity of stuff we need to consume could change. For example innovations in HVAC and insulation reduce the cost of heating and cooling. Or, for example, increases in global temperature increase the number of heating/cooling degree-days we must consume more!
>The agencies implementing the policy should use a different measure of inflation then agencies helping the middle class. Did I understand you correctly?
Yes, this kind of thing also is true, inflation as experienced by different groups is different, and Economists know this, but it doesn’t enter into official economic usage by the govt etc.
>I think it would be great for you to publish your thoughts in some econ journal. What do you think?
I have no faith in the journal system and especially in Econ being interested in hearing it. I may be motivated to write a book or something though. I’m working on a math modeling book, and it has some econ examples, and it talks about dimensionless ratios. So perhaps that book will start a conversation among some economists.
This brings me to Phil: if you look in some of my posts, I argue that an alternative measure of inflation, focused on core expenses of households like rent, food, utilities, transport whatever… will almost regardless of what you use for weighting, show that we recovered from the downturn of 2020, things climbed back up… and then in late 2021 *they turned around again* and went down rapidly for a year. So there are two separate recessions. However the “official” GDP/CPI doesn’t show that it has a merely small downward blip. The reason is multi-fold I guess, but in part due to CPI all items not focusing on the core expenses of families, and also because there are certain populations which managed to increase consumption by taking massive credit card loans and such. Plus perhaps some groups taking capital gains and using them to buy stuff (which was mentioned in a link from ineteconomics which you can ctrl-f search for on this page)
But during that time period as mentioned in one of my posts poverty increase MASSIVELY, tripling among white children, doubling (to ~ 20%) among black children. Sure, at some point you have to decide who do you care about and what kind of average are you going to rely on. But I don’t think it’s useful to talk about “not in a recession because the richest people are getting richer while poverty doubles and everyday households are declining in purchasing power”
So it’s a little like averaging between a large group that is declining moderately, and a small group that’s exploding in wealth and rolling in capital gains, and coming up with an average of fairly flat and saying “everything is fine”… not really.
Man, you’re really pushing all my buttons the last few weeks Andrew… I love it!
The real answer is that the recession happened and Economists methodologies are too stupid to let them know it.
Let’s take a look at GDP/capita divided by a price index relevant to consumers. I would also prefer fully dimensionless ratio by including a required rate of consumption in the ratio but let’s assume rate of consumption is fairly constant in the last few years, that’s actually not true because what we consumed changed in 2020 during pandemic etc but I don’t have access to the full dataset and time and funds which would allow me to do the right analysis, the way that BEA does.
we’ve been through this a bunch in the blog the last few weeks. I’ve got a Julia script that generates such an index using uncertain weights to give a spaghetti plot, but let’s use as an example (0.2 * all items + 0.4* rent + 0.3*food + 0.1 * childcare ). I’ll set 2020-01-01 as the reference value for all the indices
https://fred.stlouisfed.org/graph/?g=1olmf
Using this plot purchasing power of GDP/capita peaked in Q4 2021 and declined until Q2 2023, then rose slowly to be still below the peak by Q1 2024.
Ok, but what about those weights I offered above? why are those the right weights??? Well they aren’t but every family has different weights, so let’s just try a couple different weights and see how samples from an uncertain weight scheme would look. How about (0.2 * all items + 0.3*rent + 0.3*food + 0.2*childcare)
https://fred.stlouisfed.org/graph/?g=1olmB
Yep, it declined, but now it’s climbed about back up to be even with the peak.
Ok, how about another mix:
https://fred.stlouisfed.org/graph/?g=1olmG
Yep, it declined, but now much more shallowly and is now just a little above the peak.
Here’s a final one:
https://fred.stlouisfed.org/graph/?g=1oloA
Yep, it declined and then rose and is a little above the peak today.
If you use those items and have any reasonable weight scheme, you will find that the peak was Q4 2021, there was a decline and it may or may not have climbed back to its peak today.
Ok, you might ask, “why not just use the official all-items measure?” why should we care about Dan’s measure? and the answer is multi-fold, but first I make the claim that families spend a LOT of their money on rent, food, and childcare (as well as healthcare, transportation, and food away from home), plus I make the assertion that families matter more for the health of the economy than non-families. This is for a number of reasons.
The first is that it’s widely believed and I agree both through personal experience and indirectly, that experiences in childhood have long lasting consequences for those individuals and the economy. And literally every adult 20 years from now will be someone who experienced the economic conditions of *families today* not of single people, not of retirees, not of childless couples, but *families today*. For a little context, how many of you know people who were children in the great depression who suffered throughout their lives from problems such as hoarding or inability to spend money on their house (and so live in a crumbling houses), or similar? I personally know multiple instances.
How about black kids who lived through the 80’s and 90’s crime and drug related problems? A huge fraction of black males wound up incarcerated. Economists might argue “those are social problems not economic” which if they did so I would laugh in their face. “The best perceived economic options for these kids is black market economy with high violence” is not an economic problem? Get real.
How about children of divorced parents where divorce ramped up massively starting in 1960 and peaked in like 1979 or so Again, Gen X kids suffered considerably from this and it influenced their future economic conditions. Think of all the “slacker” tropes and such. Nihilism is a Gen X trait. Gen X kids who are 45 now still make jokes about how they’re still waiting around at softball for their parents to remember to come pick them up. It affected the economic and social milieu for 30 years easily.
Consider an economy where family purchasing power of food, childcare, and rent declines, families wind up on the street, or moving into crowded conditions, cutting back on education, pulling out of the labor force to provide childcare etc, but purchasing power of elder care, luxury second homes by childless couples, and yachts by the mega-rich increases enough to more-than-compensate, is this a sign we’re “not in a recession”? I would say no, this is a sign we’re cannibalizing the future.
So, anyway, you can make more sophisticated indices and add Dirichlet priors on the weights and do whatever, but if you focus on the question “did the power of the economy to make everyday lives of people and their children decline?” the answer will be yes, it declined starting in 2021 and is only barely recovered maybe today.
And ECONOMISTS DON’T EVEN KNOW IT.
Like, it’s right there, in their damn faces, but they have created a religion in which certain idolatry of certain measures give “the one true answer” and they don’t accept variation they are unwilling to make value judgements, they stick to what the high priesthood says… and that makes them unscientific and we should laugh at them. It’s as much ridiculous as power-posing or beautiful people have more girls or whatever else we’ve seen here.
Dan, I’m certainly not an expert on this, and thank you for this nice discussion. Looking at your graphs I just have the potentially very naive thought that the as-dimensionless-as-possible indexes actually say that the US is much better off now (and was in 2023) than almost all the time before, and whether that’s not a reasonably good thing, having seen a drop in 2023 or not. The conventional understanding of “recession” is based on a downward trend of non-dimensionless indicators. These indicators arguably don’t measure how well off people actually are (this is your point I think), but the philosophy there may be implying that if these indicators drop, and be it at a high level, in fact the “real” economy (as far as it matters to the people) will be in serious trouble, and that may well mean more serious than what your mild 2023 drop shows. So may it be that a slight drop at high level of your indicators isn’t really as much of a problem of what the economists would actually call a recession? Could it be more appropriate to look at the absolute level of your “dimensionless” indicators, rather than at their derivative?
If you look at “real GDP per capita” you’ll see that Fred calls out a recession whenever that indicator declines for more than a brief period of time, so the decline I’ve shown which occurs for at least a year in every version of the graph, would conventionally be called a recession. The problem is really that the measure “real GDP per capita” is disconnected from the experience of the important main group of people in the country. This is actually the point of the article Peter links to below
https://statmodeling.stat.columbia.edu/2024/05/31/whassup-with-those-economists-who-predicted-a-recession-that-then-didnt-happen/#comment-2373146
What’s going on in the US is that inequality is so severe that the bulk of people can be suffering and the richest people so skew the conventional measures that they still increase.
If you look at median household income spaghetti plotted with similar indices it’s actually the case that we are well below the level of the 1990s today, so have been in a recession as far as typical people are concerned for about 30 years. And economists are not jumping up and down on the street corners and yelling this information from megaphones… Which means they are as a group complicit.
It’s annoying that linking to Fred graphs puts your message in the moderation queue… I posted median household income etc but it’s being held
Just to make it easier to look at here’s median household income divided by one reasonable sample of a normal-family-household CPI index:
https://fred.stlouisfed.org/graph/?g=1onet
It’s peak was 1999, declined continuously throughout the dot-com bust, the housing baloney, and the wars in Iraq and Afghanistan until 2012, bobbled around a bit, then in 2014 increased up to 2019, where it was still only about 95% of the 1999 level, turned around and declined until 2022 where it’s now only about 90% of its 1999 level.
Household size changed a bit but it’s in the 2nd decimal place or something, it’s been close to constant at around 2.5 for a long time.
https://www.statista.com/statistics/183648/average-size-of-households-in-the-us/
Thanks, appreciated!
I’ve got a post held in moderation because I linked to 3-4 Fred graphs. But here’s a relevant pew article for some of the issues I raised there. I basically shows how important various categories of spending is for families, a bit older data though, but still relevant
https://www.pewtrusts.org/en/research-and-analysis/issue-briefs/2016/03/household-expenditures-and-income
And this piece by Tom Ferguson and Servaas Storm on the importance of top decile spending financed by cap gains is worth a look: https://www.ineteconomics.org/perspectives/blog/the-second-coming-trump-vs-biden
Peter, love that summary and generally agree with it. 40 years of heavy deficit spending has manufactured a lot of money which concentrated into the accounts of the very wealthy and they spent it on buying up companies and reducing competition and etc to enable them to control prices and collect even more money, driving the real incomes of typical people downward. This article seems to get that and it’s rare to see in economics. The inetecononics people are one of the few that seem to be onboard.
Just this morning read this thread on Mastodon by an anonymous guy on the internet who I have enormous respect for because of his demonstrated intelligence and insightful commentary. It’s about the disconnect between what institutions tell us (like, we’re not in a recession, or the wonders of the modern world are the best of all possible economic conditions) vs what everyday people experience and many of the statistics on environment and soforth tell us in our lived experience…
https://kolektiva.social/@HeavenlyPossum/112537297155718352
Why doesn’t he mention the obvious thing that happened in 1971 though? Ie, the US suspended the last vestiges of the gold standard (essentially went bankrupt).
He has mentioned money creation as a form of “enclosure” by the state before so I’m pretty sure that’s relevant to him, just not mentioned here.
I quit economics when my macro prof informed us in a review session that consumers smooth their consumption across time periods because of “the consumption smoothing motive”. Working on my E&M problem sets in the back of the class until the end of the quarter struck me as a much better use of my time after that.
Full consumption smoothing is only a benchmark. I do not think that any economist would argue that any consumer perfectly smoothes their consumption. This would be impossible because of liquidity constraints on the one hand and the difficulty of reducing consumption costs on the other hand (e.g. a completely non-smoothing consumer would have to move to a new apartment or house the moment they loose their job). For this reason, a theoretical model might employ individual smoothing parameters for each consumer (heterogeneous agent models).
I think one of the reasons why macroeconomics has such a bad reputation is that people assume that macroeconomists believe it like the gospel. I know of none who do (at least personally). To them, these are theoretical models, nothing more, nothing less.
In the graduate macroeconomics courses I attended, the contrast between theory and empirics was front and centre. In my undergraduate macroeconomics classes, however, there was much less discussion of empirics. I think part of the reason for this was that we had a lot of people in these classes who had massive problems understanding the mathematics behind these models. Surely some of these poor souls thought they had successfully ditched all maths when they held their high school diplomas in their hands. Until they sat in their first economics class…
If you keep predicting a recession, you’re going to be right eventually.
Trump was just convicted yesterday, so I would like to compare this apocryphal quotation from today’s blog:
“All this is very much in line with John Maynard Keynes’ quip, which has become somewhat of an adage: ‘When the facts change, I change my mind. What do you do, sir’?”
with the famous actual statement from “Otero County Commissioner Couy Griffin, a 48-year-old former rodeo rider who helped found a group called Cowboys for Trump.” Griffin actually said, “My vote to remain a no isn’t based on any evidence, it’s not based on any facts, it’s only based on my gut feeling and my own intuition, and that’s all I need,”
Paul:
One of the irritating problems with “facts” is that they have an unpleasant habit of turning out to be wrong. Hence, I stuck with my gut (literally) regarding cholesterol & eggs, and here I am today some 20-30 years down the road from the time the horrible “truth” was revealed about eggs, basking in the sunshine of all the delicious breakfasts I have enjoyed, while who knows how many other losers gave up bacon and eggs for yogurt and grape nuts. Gross.
I hear different “facts” about coffee and booze on a regular basis. I blow them off all totally and enjoy copious volumes of coffee and now and again a tipple of booze. But I confess to have yielded to conventional “facts” about hot-dogs: they are no longer part of my regular diet; and my moderation with booze is unfortunately derived only from personal experience regarding it’s negatives, not from any mooted “facts” about it.
And meanwhile the hot hand is with us again; the great famine never came; hydrocarbons are abundant and it’s floodin’ down in texas, but that’s nothign new. I wonder: would it be surprising if four freakin’ billion years of evolution had imparted some innate abilities to nature’s most successful form of life that scientists and “fact” purveyors – even Daniel – haven’t yet grasped?
forblogs,
Can you boil that down into a simple declarative sentence? Or, for that matter, a non-simple declarative sentence? Maybe “In spite of all the evidence, I do not think Donald Trump committed a felony.” Or “I did not see the evidence, but I do not think Donald Trump committed a felony.” Something like that?
Phil! Hope you’re doing well sir!
Sorry, that was me, Chipmunk. Now you know where to spam me if you want. But it’s probably not worth your trouble. Most spam never even makes it to my spam folder, its gone before I ever see it.
No, I can’t. I have no opinion on Trump’s legal status. My point has nothing to do with Trump.
My point is that the argument about “facts” in science – often where modeling and statistical analysis are involved, but this is not a requirement – is frequently beset by the problem that many things claimed to be “facts” turn out to be wrong.
Often, as is the case with Hanke’s forecast, the claim of a “fact” is implicit. His claim that a recession was “baked in the cake” – e.g., certain – is also equivalent to claiming that all the assumptions in his model are facts, and the model itself is infallible. In this case I suspect his assumptions – his “facts” – turned out to be wrong. Historically, it’s hardly rare for scientists to claim something is a “fact” and have it turn out to be, actually, not a fact at all. As you know I point frequently to claims like the Population Bomb and Peak Oil. But surely the erroneous claims of fact – in both the mainstream press and academic publishing – must number in the thousands or tens of thousands annually.
For example, I read a while back that some academic statisticians claimed it was a “fact” that traffic forecasters were regularly overstating demand for roads. The statisticians claimed that, because subsequent traffic volumes didn’t live up to forecasts, that the forecast models were wrong and should be abandoned or revised. However, the academic statisticians’ claim about the traffic forecast models was, eventually, demonstrated to be in error – e.g., wrong. They had hardly bothered to understand the models and made a very simple mistake – they assumed that capacity imposed no constraint on volume, and that somehow the volume should increase according to demand without additional capacity. Of course, that’s impossible when capacity is very close to maxed. You’d think people with physics degrees might understand this simple relationship Phil, but it turns out belief is a very powerful blinding agent. They might have understood it if they had bothered to analyze the problem, but because of their belief, they looked at the actual problem with barely a passing glance, had an idea fit with their cognitive bias and went with it. They made an incorrect statement of fact.
So with that poignant example in mind, I’m sure you’ll agree: A sensible response to Keyen’s question would be: “Sir, when the facts change, I hold my position and wait for them to change back”.
I might even suggest, Phil, that, most of what’s claimed to be fact by scientists in both research papers and the press is bullshit. That will seem ridiculous to you, but you should understand of couse there is a natural bias to publishing things that are wrong: scientists don’t need to run around making claims in the press or papers about things that are actually established as fact. No one has to claim that gravity is real (although I admit there’s still a lot of claiming about evolution). In other words, what’s being published today is mostly *new* claims of fact that are often controversial. I’m sure you know that some ungodly proportion of research is so useless that its hardly even cited, so it’s probably safe to assume that most claims of fact made in that research are false.
So, there is a bias to publishing what’s percieved to be surprising (aka “news”), which, often, turns out to be not so surprising because it’s actually wrong.
My comment about evolution is aimed at the frequent claims of humans having “cognitive bias”. My bet is that the “cognitive bias” is every bit as powerful – if not more powerful, given the size of the egos – in academic scientists than it is in humanity in general. And I suspect that the bias in academic scientists is amplified by the many ways in which statistics can be misread but still confirm one’s biases. Humans are very well adapted to their social environment – and that’s an environment that’s rich in people who employ myriad forms of self-interested misrespresentation. Keynes be damned: it’s more than reasonable for people to dispute what is claimed to be “fact”, since it often turns out not to be so.
Chipmunk,
Your claim about traffic is interesting. Andrew and I wrote an article about traffic forecasts once! https://stat.columbia.edu/~gelman/research/published/ChanceEthics11.pdf
I guess you’re going to look at the plot and say “those forecasts were right after all; the fact that they were miles away from the actual data doesn’t indicate that there was anything wrong with the method. It’s just that the real world didn’t act the way it was supposed to.” Or something. Actually I don’t know what you can say, I’m interested to find out.
That said, you’re certainly right that people claim all the time that something is a “fact” when it is not actually a fact. Claims about what constitutes “healthy eating” are notorious.
And yeah, everybody knows that loads of claims in published papers are false. That’s one of the points Andrew harps on all the time on this blog! So it’s funny to see you announcing it as if it’s a surprising claim.
You see, it is not the forecasts that were wrong, the world is wrong. The forecasts were not for actual observable demand, they were for the unobservable quantity representing what demand would hypothetically have been at a particular point in the traffic cost curve of Chipmunk’s personal choosing.
The real interesting question is why are comments like this, from strangers of no consequence who are demonstrably incapable of parsing the answers, such attractive bait for nerds to waste their time with. Remember this is a person who is either too thick or too lazy to follow the simplest possible decision tree of their own design
https://statmodeling.stat.columbia.edu/2022/10/21/from-soup-to-bayes-make-inferences-using-strong-assumptions-not-because-you-believe-your-model-but-because-you-dont-believe-it/
“Prediction is very hard, especially about the future.” Maybe Yogi should have said, “especially about monetary policy and macro.”
Apparently what’s also very hard is admitting that your prediction was wrong. In 2010, twenty-three economists (mostly conservative) signed a letter to Ben Bernanke warning that the Fed’s quantitative easing policy – adding billions of dollars to the economy – would be disastrous. It would “debase the currency,” create high inflation, distort financial markets, and do nothing to reduce unemployment.
Four years later, when it was clear that all these preditions were wrong, Bloomberg asked them if they’d changed their mind about inflation. You can read the article here. The headline was “Fed Critics Say ’10 Letter Warning Inflation Still Right.”
What? “all these predictions were wrong”… uh no. QE has been a disaster for the economy and it did cause ENORMOUS asset price inflation and drove massive inequality.
Just graph the S&P 500 / CPI all items since 1990, you’ll see it exploded in 2009. You might think that’s a good thing, but I don’t, because I don’t think it represents real economic improvements I believe it represents inflating a bubble with manufactured money, and drove tremendous consolidation of industries through exercising the power created by asset price inflation.
There’s a whole article related linked just a few messages above that discusses how many many people have had falling real income and that asset price inflation has disrupted the measurements of those things https://www.ineteconomics.org/perspectives/blog/the-second-coming-trump-vs-biden
If anything the naysayers on extended QE were 100% correct.
I’m particularly thinking of this article https://www.politico.com/news/magazine/2021/12/28/inflation-interest-rates-thomas-hoenig-federal-reserve-526177
Back in the 1980s an economist colleague of mine was asked why her prediction of a very high rate of inflation in New Zealand never came true.
She paused for a moment and then said “ …… but it could have.”
His model is wrong, but you are going to have to pry it out of his cold dead hands.
David:
I’d say it’s more likely that his assumptions are wrong, and that, contrary to dieing to defend them, he has already updated them.
Perhaps, rather than making a suprising statement about “facts” that “change,” Keynes should have suggested that when his assumptions – OH! Lets go Baysian! – when his *priors* – pretty baysian huh! :) turn out to be wrong….Like this:
“Sir! When my priors turn out to be wrong, I update them! What do you do?”
His model of how the economy works is not how it really works (in fundamental ways). But, rather than use a model that is closer to reality, he’ll tweak his model’s parameters to make it appear to be consistent with history. He wants the world to be like his model because his model is so pretty.
I am an economist, albeit not of the macro variety. I’ve never known anyone to take economic forecasting even the slightest bit seriously. We don’t even know what’s going on now, hence Varian’s “now casting”, much less a year in the future. That doesn’t stop some economists from claiming that they have it all figured out. But they certainly don’t represent the economics profession as a whole, or even a substantial plurality. More like glory-chasers throwing shit up against a wall and hoping something sticks. All I can say to people like Hanke is that if you’re so smart, why aren’t you a trillionaire yet?
If an aircraft cockpit has a good predictive model for collision with a mountain, it will always fail if the pilot believes in the model.
Similarly, a macroeconomic model can succeed only in isolation when the pilots (central bank, treasury, consumers, investors) don’t believe the outcome is likely.
I hate this argument. It’s used to justify the use of unfalsifiable and useless models in economics.
How do you know whether the aircraft crash model is good? It was tested. By crashing models into model mountains, and looking at what happened in real aircraft crashes.
How do you know whether that macroeconomic model is good? You don’t, because arguments like this are used to justify not testing them.
Another commenter said that no real economist takes economic forecasts seriously. I hope they are right.
Dear correspondents,
Many thanks for your edifying comments and remarks about “Whassup with those economists who predicted a recession that then didn’t happen?” John Greenwood and I have had an opportunity to read all your comments and respond specifically to Daniel Lakeland’s comments.
Just as a reminder, US real GDP did decline in 2022 Q1 and Q2. Although many market operators use two successive quarters of decline in GDP to designate a recession, this period was not designated as a recession by the NBER because, among other things, the labor market remained very buoyant.
US broad money growth (M2) peaked at 26.7% on a year-on-year basis in February 2021, and by January 2022, its growth rate had more than halved to 11.6% year-on-year. Such slowdowns typically lead to significant downturns in real GDP with a 6 to 18-month lag. So, the two-quarter downturn in real GDP in 2022 Q1 and 2022 Q2 was in line with the normal lags associated with significant changes in monetary growth rate.
As we have argued, the 2022 downturn was mitigated by the large overhang of excess money balances from the prior monetary expansion in 2020-21, which kept employment and other indicators more buoyant than normal. However, from December 2022, M2 turned negative on a year-on-year basis and remained negative until March 2024. In consequence of the very slow average M2 growth since December 2022 is that nominal GDP has continued to slow from 10.7% in 2022 Q1 to 5.4% in 2024 Q1.
We have looked carefully at Lakeland’s data. He has deflated the nominal GDP per capita by selected weightings of the CPI. For example, his first measure is as follows:
100*Gross domestic product per capita/(0.2*Consumer Price Index for All Urban Consumers: All Items in U.S. City Average + 0.4*Consumer Price Index for All Urban Consumers: Rent of Primary Residence in U.S. City Average + 0.3*Consumer Price Index for All Urban Consumers: Food at Home in U.S. City Average + 0.1*Consumer Price Index for All Urban Consumers: Tuition, Other School Fees, and Childcare in U.S. City Average.)
Lakeland’s calculation produced a peak of real GDP per capita in 2021 Q4, followed by six quarters of decline in real per capita GDP to 2023 Q2. Notice the heavy weighting of services in his formula.
While it might be possible to justify Lakeland’s measure, we do not want to go down this rabbit hole (re-weighting deflators) because such re-weighting is inevitably going to be subjective and subject to unproductive debate. Lakeland’s formula emphasizes rent (40%), food (30%), and tuition & childcare (10%) — components of the CPI that were strongly rising and lagging.
Using Datastream data, we have compared three measures: real GDP per capita (deflated by CPI, not GDP deflator), real GDP per capita (using the official data), and total real GDP (i.e., not per capita). Each produces the same 2-quarter decline in 2022 Q1 and Q2, but then they switch to increases.
Overall, our conclusion is that the use of per capita data does not change things very much.
– Steve H. Hanke & John Greenwood
Steve, and John.
Thanks for your comments. I actually agree with you that simply reweighting CPI is not the appropriate way to go about determining the health of the economy and families and the overall growth of economic activity etc. It’s merely a thing that I can calculate with only a few clicks on Fred and was supposed to be a suggestion of what a more comprehensive analysis might show. Also I gave multiple weights to show that the result was somewhat robust to reweighting (I have a Julia script that spaghetti plots a bunch of these weighted curves). So I agree we shouldn’t argue specifically over what the right weights are.
However, what do you think about the complaints specifically of failure to use fully dimensionless measures. This is by far the bigger issue, by FAR.
In particular the measure I actually advocate is something like the one described in this comment: https://statmodeling.stat.columbia.edu/2024/05/31/whassup-with-those-economists-who-predicted-a-recession-that-then-didnt-happen/#comment-2373307
Specifically if we like GDP (as opposed to median household income) then:
(GDP / population) / (basket of goods used by family of N people / N / time to consume the basket)
In calculating the denominator I would actually advocate calculating an average across different household sizes N, and each average would be:
1/k * sum over k goods(price of good * quantity consumed by N people / N / time for N people to consume that good)
For things like food you might use the quantity purchased weekly, for things like washing machines you might use 1 washing machine per household divided by avg lifespan of washing machines sold today. Similarly for say utilities you might sum across an entire year the expected cooling or heating costs given changing climate conditions and divide by 1 year. For things like computers you might look at turnover/upgrade lifespans of a year or three. For clothing perhaps 2 or 3 year turnover, for childcare services the annual quantity (assuming some variation in school year vs summer, etc etc.)
After calculating this quantity, you have a dimensionless measure, inflation is irrelevant, durability of durable goods is included, variation in seasons is included, changes in efficiency of machines are included, excess costs from global warming are included, etc etc.
What would this dimensionless income look like? We can only wonder, because although it is more or less an obvious way to go about measuring income to a person familiar with Buckingham Pi theorem, it is, I’m afraid, apparently novel to economics.
I would argue that if we were to calculate that quantity using the full power of govt surveys and soforth, we would have found a recession in the 2021-2023, and it’s based on a gut feeling that the “easy” CPI weighting scheme provides preliminary but not definitive evidence for.
I hope you will have the time to consider this question and tell me what you think.