Ep. 418 Steve Keen Sides With MAGA over MMT on Trade Deficits
Steve Keen is an iconoclast economist who is sympathetic to MMT claims that right-wingers worry too much about government budget deficits. However, in a recent post Keen says that Trump Admin economists (such as Stephen Miran) make more sense on trade deficits than Warren Mosler does.
Mentioned in the Episode and Other Links of Interest:
- The YouTube version of this conversation.
- This episode’s sponsor, PersistSEO.com.
- Steve Keen’s free book and subscription-based online course.
- Steve’s post on trade deficits, Trump, and MMT.
- Murphy and Hendrickson give qualified support to Miran’s paper on tariffs.
- Help support the Bob Murphy Show.
About the author, Robert
Christian and economist, Chief Economist at infineo, and Senior Fellow with the Mises Institute.
It’s very bizarre that Steve so effectively dismantles the simplistic reasoning about trade deficits but then concludes that the government should just force us to all trade differently rather than that the government should just stop interfering with trade entirely.
He makes a big deal about recent Chinese stats, but has Steve never looked at early American stats?
I think I get your point Dave, but can you spell it out? (About “early American stats.”)
Although detailed stats weren’t compiled at the time, estimates show that early America, which relied mostly on free markets, blew modern China out of the water.
You can also see it with Argentina under Milei.
If Steve wants to say an economic policy is good or bad based on the stats it produces, he doesn’t get to just promote China and ignore places that do even better than China.
That’s fine, though early America “relied on” tariffs for revenue.
A coin toss is necessarily always ergodic because there is no mechanism by which one coin toss can communicate or coordinate with a different coin toss. Each toss is an entirely separate operation, regardless of space or time.
Borrowing Steve Keen’s example … suppose we take it that all bets are exactly $100 as the original problem proposed, and you have a guy (call him “Investor X”) who starts out with $1000 and then splits this down into 10 simultaneous bets, which are all uncorrelated, then the most likely outcome is $1050 but only slightly less likely outcome would be either $960 or $1140 and then a bunch of low probability outcomes.
Then you have another guy (call him “Investor Y”) working longitudinally, who starts with $1000 and then makes a single bet of $100 and takes whatever return he gets and puts that back into the pool, makes a second bet of $100 and puts those returns back into the pool … so on and so on … for a total of 10 bets. Well, the second guy has exactly the same probability of outcomes as the first guy. They both made 10 bets each of $100 and they both got the same outcome. You can do any mix you like to a total of 10 bets … you can have 2 bets of $100 running simultaneously and then do it 5 times longitudinally … you can have 5 bets running simultaneously and do it 2 times.
Then Keen comes up with this concept out of nowhere that “Investor Y” somehow needs to always make a single large bet, which for no particular reason takes away the original rule that all bets are $100 … but hey if you carefully compare apples with orangutans you can convince yourself of the spurious result that a coin toss is non-ergodic.
Let’s not get started on the MMT idea that government can spend money without taxation … which is just too bonkers for further analysis.
“… and in my model, I state all market products are a sphere, and go on to prove that…”
“… and in my model, I state that hummingbird song frequencies are stock market prices…”
“… and in my model, bees are fish…”
Of course they can spend money without taxation. The people will just be made to suffer in some other way.
Tel, I get your point about apples and orangutans, but check out this post where I tried to state the argument more clearly.
It’s a good article … but where I’m getting stuck is, what (if anything) the Kelly Criterion is optimizing for.
Suppose the only objective is to make money … then diversification is an advantage, and your own results with the “100 Agents” strategy demonstrates that if the option to diversify is available, then always diversify as much as possible. You could go to the investment world and say, “Hey I’ve discovered that diversification is a great idea!” and they would tell you they already know that … and anyway don’t worry about Kelly’s formula to understand diversification.
So presumably, the Kelly Criterion is optimizing for a situation where some other factor is imposing a limit … and the kind of hidden assumption of this would be that we are optimizing to make the best return with a limited number of bets … except that no one seems to bother saying this is part of the original problem description, and anyway why would there be a limit on the number of bets?
I looked up on Wikipedia and found a link to an interesting article … which perhaps you want to add as a link … here – https://arxiv.org/pdf/1701.01427
They did a real world experiment, and I quote the problem description as follows:
OK, the experiment designers claim that the Kelly Criterion is optimal strategy, and unfortunately they didn’t publish their website’s internal code to reproduce the experiment … but I put it forward that there are much more reliable ways to max out the game than by even thinking about the Kelly Criterion. Perhaps having some vague idea of geometric series might help … but really you are optimizing against time and you win by playing the maximum number of bets within the 30 minute limit … that’s the real thing you are optimizing against … maximum diversification in the time provided.
Therefore, it finally depends on how fast you can click the mouse to trigger the next round of betting … the size of bet only matters slightly, in as much as trying not to blow the whole initial stake … and bet at least enough to reach the earnings cap within the time limit. Other than that, small bets will beat big bets, not because of Ergodicity, but because of diversification and the Central Limit Theorem. If you slow yourself down calculating an “optimal” bet, you are already losing thanks to the time you waste making the calculation.
Further explanation – https://plato.stanford.edu/entries/bounded-rationality/
Summary: Bounded rationality is the idea that people make decisions with limited time, information, and mental resources. Instead of always finding the best possible solution, we often settle for one that’s “good enough” because figuring out the perfect answer might take too long or be too complicated. This concept was introduced by Herbert Simon, who argued that real-world decision-making is shaped by practical constraints, not ideal logic. The article explores how this affects economics, psychology, and artificial intelligence, showing that humans and machines alike often rely on shortcuts, rules of thumb, or simplified models to make choices quickly. It also discusses how researchers try to model these limitations to better understand behavior and improve decision-making systems.
Now if, and only if, the game involves some fixed cost per round of betting, then perhaps the Kelly Criterion is worth consideration. However, if there is a fixed cost that operates a bit like brokerage does in a stock market, then you must also consider this fixed cost for the “100 Agent” strategy because he pays 100x that cost. At any rate, such a cost needs to be declared upfront in the game definition … else we are not answering the original problem posed.
OK I’ll let this sit for a bit and maybe get back to you, Tel. My quick reaction is something like, “Yes, of course people recognize that you should diversify your portfolio, but even there, if you model it as ‘the S&P500 has a 50% chance of giving you a +50% return and a 50% chance of giving you a -40% return” then you want to use Kelly, or at least, Kelly makes more sense than “putting your entire net worth into your diversified portfolio.”
What I was thinking about after I put up my post, was that maybe Spitznagel’s way of looking at it helps clarify: He doeesn’t view it so much as “reducing risk” as a distinct goal from “ending up with as much money as possible.” In his book, Safe Haven, he starts out by rejecting the conventional paradigm that says investors like return but dislike risk (volatility). That conventional paradigm is the way people in econ/finance have generated standard results showing that you want to hold a diversified portfolio.
The ludicrousness of the proposition that posits an equivalence for averaging outcomes of market processes over space versus time – explodes any explanation.
It is such bat-sh** insanity that there is no possibility for discussion with an entity that unironically expresses it. This kind of non-sense is specifically satanic. It’s numerology in economics.
Are you caught-up on your gene-ja.. ahem, ‘vaccines’, Mr. Keen?
You know Steve’s point was that the mainstream subscribes to this equivalence, and he was saying it was a fallacy?
The export deficit isn’t a disease in itself, but in the case of the USA it’s a symptom of multiple illnesses.
These illnesses are destroying the human-capital stock, the manufacturing capital stock, and far more importantly and under-appreciated: The Hayekian embedded knowledge that is built-up over *generations* within a network of production. If you’ve got a dollar-empire, these losses don’t show-up in dollar value today, or tomorrow, but rather over generations. THAT is the harm.
That is not addressed at all by Steve’s facile observation that one nation’s export is another nation’s import. We already knew that. You’re not teaching anything, Steve.
Without the ability to even *maintain* the embedded knowledge of how to produce the stuff you need, within your sovereign political unit, you’re vulnerable to forces and conflicts which economists, (including Dr. Murphy) spend their careers ignoring. The ‘unseen’ in this case is … the entire existing world of geopolitics which stands knee-deep in the blood of history.
In this domain, the methodology of the historical school stands without alternatives. And that is the kind of stuff makes the kids who liked math and avoided fights really uneasy. In this doman, some groups of people enslave, poison, torture and kill other groups people — without requiring economic gain to do it.
In the current instance of targeted, intergenerational, maleficent, systemic national genocide, the name of the game is to eradicate our nations potential for transforming lower value inputs to higher value outputs. That has been achieved by the monopolization of media, the cartelization of banking, the destruction of independent schooling, the imposition of woke corporate culture, the draining of market talent to ‘defense’ projects, the draining of talent to ‘financial markets’ (casino).
The curret ‘trade deficit’ is just a symptom of the political success of our enemy making US unable to compete. Absent that, the only thing feeding the American gut is the global predations of the US Empire, which the enemy also controls, as recently made blatantly obvious, again.
So sorry Bob. So sorry Steve. Nukes, Nintendo and Netflix will not save the nation.
“Our enemy” is the people calling themselves the government and telling us that we have to obey their arbitrary rules, not the people across the globe who ignore those rules and profit from doing so.
I probably shouldn’t waste so much time on these things … but Steve’s history of Toyota kind of bugged me until I went and looked it up. Turns out that Toyota started out in the Meiji Period (late 19th Century) working in the textile industry … focusing on power looms and steam engines. They branched out into cars in the early 20th Century and never make pedal bicycles … although they did briefly have a subsidiary called “Toyo Motors” which started making motorbikes in 1949, without using any parts borrowed from military surplus fighter planes. This forray into motorcycles was OK but unexceptional, never did all that well and got wound up in 1960.
I found evidence of one bike in the entire world built using part of a Rolls Royce Merlin engine taken from a Mustang fighter … the original engine (I believe) was a hulking V12 and they cut off two cylinders to make a beefy V-twin approx 300 cubic inches. Hand made custom chopper and never a commercial product … not even Japanese!
Very likely, what Steve remembers is the history of Honda, not Toyota … the parts were taken from disused Imperial Army radios … and no they didn’t fly.
https://www.honda.co.jp/powerproducts-brand/en/history/
Honda went on to specialize in small engines … while Toyota did much better in the 4 cylinder vehicle market then V8’s and recently V6’s.
I find these historical school anecdotes fascinating … but it massively helps to get the details right 😎. I know that’s difficult off the cuff … but the Bob Murphy Show is forever ya know.
Steve has put himself in a position where he is so very heterodox that he refuses to even learn the names for what other people already studied in the Economics discipline. Sure I would be the first to agree that Engineers are smarter than Economists … but you have to give the other party a sporting chance … and I have a tiny little suspicion that Steve himself might be one of them Arts grad types.
As an example, the stuff about a factory getting more efficient as it becomes better utilized is OK up to a point … at some stage the factory is possibly going to be pushed to run night shifts, pay penalty rates, etc. This phenomenon is typically called, “Economies of Scale” and it isn’t like this concept comes from way out of left field … there’s discussion on this all over the place.
Soros wrote about his “theory of reflexivity” in The Alchemy of Finance (2003) and even before that Paul Ormerod
wrote about “positive feedback” in The Death of Economics (1994). Similarly, if you use words like “non-ergodicity” then people go “Huh?” but if you say “path dependency” they go, “Oh yeah that thing” … and W. Brian Arthur was writing papers about this back in the 1980’s as well as what he called “Increasing Returns”.
OK, this is my last effort on the Kelly thing … and I was stuck at home with a Winter cold … so I achieved none of my adulting tasks … but I did end up poking around with the numbers a bit on the problem described here. Since there exists a trivial solution for the situation where unlimited bets are available, I’m only considering optimization within the scope of a fixed number of bets as follows:
* Players start with $100.
* Players have exactly 10 bets to utilize, whether that be in parallel or in serial.
* Bets can be any fraction of the current balance: from 0% to 100%.
* Heads / Tails are equally probable … Heads pays back 150% of the bet, Tails pays back only 60% of the bet.
* We try to optimize for the final balance, after 10 bets are completed.
With the number of bets limited, there’s no longer any advantage in parallel betting … any parallel strategy can also be mapped to an equivalent serial strategy, so therefore only serial betting needs to be considered.
As an initial simplifying assumption … consider only a fixed ratio strategy (always the same percentage of current balance) … which includes Kelly’s strategy. It’s easy to demonstrate that the ordering of Heads/Tails for a given player has no effect on the final outcome, and the only thing that will make a difference is the total number of Heads within those 10 bets. Thus, we only need to consider 11 different outcomes: from 10 Tails (worst), down to 10 Heads (best), with the median (and most likely) outcome being 5 Heads and 5 Tails. This distribution is symmetric, so it is easy to find the median.
Once we understand that there are only 11 different outcomes, the well known Binomial Theorem gives the probability of those outcomes. You can get the table quickly by asking your nearest AI calculator. Here it is:
Count Heads / Percent chance.
0 / 0.097656%
1 / 0.976562%
2 / 4.394531%
3 / 11.718750%
4 / 20.507812%
5 / 24.609375% (median)
6 / 20.507812%
7 / 11.718750%
8 / 4.394531%
9 / 0.976562%
10 / 0.097656%
Now, for any given fixed ratio strategy, it’s possible to figure out the 11 different outcomes, then apply the probabilities above and get both a mean value and a median value. Kelly’s Criterion optimizes the median, however just betting the full pot every round does optimize the Average (mean) value outcome. Here’s a series of approx results for different strategies:
Bet ratio: 10% Median: $104.06 Average: $105.12
Bet ratio: 15% Median: $105.36 Average: $107.76
Bet ratio: 20% Median: $106.15 Average: $110.43
Bet ratio: 25% Median: $106.41 Average: $113.19
Bet ratio: 30% Median: $106.15 Average: $115.97
Bet ratio: 35% Median: $105.36 Average: $118.97
Bet ratio: 40% Median: $104.06 Average: $121.96
Bet ratio: 50% Median: $100.00 Average: $127.98
Bet ratio: 90% Median: $68.82 Average: $154.84
Bet ratio: 100% Median: $59.05 Average: $162.89
You can see that the median starts going down sharply, as the average goes up. That’s because the winnings are skewed into a smaller and smaller probability outcome (but not zero probability). Thus, you need to decide whether you want a small probability of being fabulously wealthy (with a good overall average) or else if you want a reasonable probability of doing OK (with lower average but still acceptable).
That’s where “risk tolerance” comes into the picture … and ultimately there’s no unique mathematical answer to that. By adjusting the bet ratio, you can trade off the chance of reliable and ordinary gains, against the chance of risky and extraordinary gains … but if you did have a million lifetimes to repeat the same thing over again, then on average you are always better off betting more.
Paul Samuelson was right and the LSE was wrong … at least that outcome should make us all comfortable that the world is once again set to rights.
Let me point out, that in this game, even if you do bet everything and “let it ride” each round for 10 rounds, you still have a 37.65% chance of at least making some kind of profit … that’s way better than a lottery ticket, and plenty of people do buy lottery tickets. That’s just to put it in context.
What about more complex strategies? I intuitively think you will always be balancing off the average against the median … but possibly there are ways to nuance it if anyone can come up with some kind of overall “utility” function which merges both greed and risk tolerance. There would be families of such functions I suppose. I can see why investment advisors recommend different strategies as people get older … thus trying to capture extra gains in early rounds and then reverting over to low risk as they get older. Also, no investment advisor wants to recommend something which will produce more unhappy customers than happy … for obvious reasons!