Inflation, Corporate Profits, and Artificial Intelligence
Shrugging corporate profits aside as a partial explanation of inflation is dumb; but there may be something new afoot
Catherine Rampell of WaPo thinks the solution to inflation is faster immigration and removing import tariffs. She claims that only ‘conspiracy theorists’ believe corporate profit margins play a significant role.
Odd. Adam Smith (among many others) believed markets were ALWAYS threatened by collusion among market participants. His proposed solution developed into a theory of ‘free market’ competition. But given market concentration, the default of ‘unfree market collusion’ must prevail. Maybe Adam Smith didn’t know what he was talking about, or maybe his successors believe he was stupid in fretting about collusion (and how the government could make that worse).
The Economic Policy Institute debunks both the ‘immigration’ and the ‘tariff’ planks in Rampell’s theory, and their arguments are persuasive. However, perhaps there’s more to the story than meets the eye here?
Perhaps the ‘collusion’ in the markets is not human beings trying to abuse the law for their own profit — but rather, delegating that function to artificial intelligence and then shrugging aside any legal or other liability?
Would expediting immigration reduce inflation?
No, and this appears to be a pretty stupid argument. Immigration would only reduce inflation if (1) wage growth is causing inflation, and (2) immigrants would stop wage growth. Perhaps Rampell is making a joke, or perhaps the proposition is intended to gain access to some other story from some other sponsor who actually believes this claim.
EPI puts it this way:
Putting further downward pressure on wage growth that is already lagging far behind inflation would just increase the burden of adjustment to more normal inflation that will be borne by workers.
Josh Bivens, Economic Policy Institute, Working Economics Blog
Concur.
There’s even an argument to be made that ‘expediting immigration’ would actually INCREASE inflation — dramatically — rather than reduce it.
Rampell is probably only referring to certain categories of employment visas. Currently, the US gives out about 140,000 or so ‘employment’ visas each fiscal year. In a country of 330 million people, that’s not exactly a major component in the labor force. But it would be quite useful for a handful of companies with global workforces that would want certain work (particularly patentable research and development) to occur exclusively in the U.S.A.
But there are between 10 and 14 million undocumented persons in America. (Pew and the Department of Homeland Security estimate about 11 million as of 2014; other researchers estimated the total population to be between 16 and 29 million — the total number is hard to know with precision since undocumented persons try to avoid becoming ‘persons facing deportation.’)
So while some people fret about 140,000 visas, ‘expediting immigration’ that referred to a group 100 times larger than that small chunk of people would have far greater effects than either Rampell or the EPI consider.
What happens if millions of agriculture, janitorial, health services, and other workers suddenly acquires legal status and the capacity to demand minimum wages, or change employment to a different minimum wage job elsewhere?
Rampell doesn’t care. She’s probably focused on ‘expediting immigration’ for people like this:
EPI probably does care, but that’s too big a question to be addressed in such a blog post. But perhaps there’s a reason to consider such claims anyway that are bigger than meets the eye, and which have nothing to do with ridiculous claims about immigration reducing inflation?
Consider Artificial Intelligence. Any major police/DHS dragnet to investigate and capture a large number of undocumented people would leave traces. But what sort of traces?
Say there are 100 ICE officials conducting ‘major’ investigations in a territory away from the border, like Mississippi (I have no idea how many officials they use for such purposes, only that it’s a finite number)(. One doesn’t necessarily know who these people are. However, if they carry a personal cell phone with them when they travel, you could plot their movements through apps that they installed on that phone (not their government phone) — they wouldn’t even realize who was tracking them and why.
Say that 80% of those agents had installed one of many apps on their personal phones, so that all their movements might be tracked. Say 1 of them visits Jackson, Mississippi every month. Say 2 fly out from Washington to Jackson, all three are tracked. Say that local police and governor staff also have their own personal cell phones around with them, separate from their ‘official’ phone.
AI, looking merely at basic geographical tags and travel movements, might deduce a raid like this one coming weeks in advance — then after tracking those movements, tracking who they met with and where, interpret that data, and even respond to it as follows:
…investigation underway in Mississippi…major raid imminent…poultry most likely target…raid may disrupt poultry availability for 2–5 weeks…stockpile poultry recommended…orders placed and arranged for XXX tons of poultry…shall we execute the orders? …
All of this is basic ‘signals intelligence.’ The sort of stuff that the US Navy was doing 40 years ago to track Soviet fleets. Didn’t even need AI back then — just needed a large enough set of computers + a vast trove of well-trained sailors to evaluate the data and use it.
These days, the same computation can be done with tiny computers, possibly even cell phones. The role of AI matters in investigating thousands/millions of other possible explanations, checking on them, and rejecting them one by one in milliseconds. It’s not that the AI is so ‘smart’ — but that it can run math and assess probabilities for more possibilities than humans can, and do so much, much faster. Like playing chess. Only with a much broader set of possibilities than those available on a game board.
Such ‘artificial intelligence’ could give a major corporation — or even a modest private equity firm — all the head start needed to profit immensely from the information.
To pretend that this isn’t happening already is silly.
Perhaps there’s more to AI’s role in the markets linked to immigration than simply identifying an opportunity and setting things up to exploit it?
What if the AI systems monitor one another, as well as tracking the movements by the Feds? What sort of idiot would fail to monitor business rivals using the same algorithms they are using?
In stock markets, such efforts can create feedback loops — pushing prices up or down drastically, triggering built in controls. But stock markets are simple mechanisms (again, they’ve been around far longer than 100 years). And stock markets can deploy AI themselves to defend against certain forms of abuses, once they identify the possibility.
Say one AI identifies “risk of poultry shortages” based on tracking movements in Mississippi, and another AI identifies “risk of poultry shortages” based on a large trove of purchases arranged by a specific AI. Immediately, it may step in to place orders for poultry as well, without knowing precisely why it should do so.
Thus, ‘supply chain shortages’ manifest. Covid-19 becomes an excuse. What is actually happening is AI interacting with one another — creating shortages. All linked to guesses about enforcement of immigration law.
Finally, assume some insiders have more information than others about imminent ICE operations. Assume that the most senior officials play golf with their buddies and donors in the corporate sector, and tell them MONTHS ahead of time what is likely to happen. Those corporations can then set up their own AI to parse the data and identify how both to profit the most from that data, AND how to disguise their operations and position from other players.
Here’s how their AI might interpret the same ‘immigration’ data:
Poultry shortage imminent due to ICE raids scheduled in 6 months…recommend (1) sell all positions in existing factories, (2) remove personnel, consultants, and all operational links to targeted factories, (3) raise prices by XXX before other AI identifies imminent major raid, (4) if prices do not escalate adequately to cover position, issue coupons or discounts to major wholesalers; otherwise, prepare to raise prices further, citing market volatility caused by COVID…(5) WARNING: avoid any reference to raid in public and purge disposable records.
The basic problem is one of corruption, or rather, ‘inside information.’ But trading on ‘inside information’ that does not involve securities is…not always criminal. Even prominent political people in charge of regulating currencies like Federal Reserve Board members (is there anyone with more importance in the universe of inflation?) could trade on ‘inside information’ about bonds because no criminal law extends the same rules to their particular area of regulation.
ICE would probably require all their personnel not to talk to outsiders with an eye toward contaminating an investigation. None of their own internal mechanisms would necessarily catch someone trading on that information. And who appointed the inspectors general tasked with monitoring all this anyway?
Would lowering tariffs reduce inflation?
As a thought exercise, the above story slips away from immigration toward AI as an illustration of the sorts of mechanisms that could be at work, just beneath the surface. Yes, actual corruption is possible (the possibility of ICE officials ‘selling’ data to corporate buddies). But even that isn’t strictly necessary to create feedback loops that would drive up inflation while also increasing corporate profits.
Tariffs are even more easily manipulated, if the tariffs themselves are implemented in an ad hoc manner.
First, Rampell proposes that reducing tariffs would naturally reduce inflation. Cute. EPI’ concluded after looking into this that —
…eliminating tariffs would offset just 7.2% of the total increase in consumer prices, but provide no buffer to price increases thereafter.
- Robert E. Scott and Adam S. Hersh, Tariff increases did not cause inflation, Economic Policy Institute, January 19, 2022
By EPI’s reasoning, Rampell may actually be about 7% “correct” (which also means 93% wrong). Their numbers appear compelling:
Total U.S. tariff and customs duties collected rose from $36.6 billion in the fourth quarter of 2016 to $85.7 billion in the third quarter of 2021. This increase of $49.1 billion represents only 0.3% of total U.S. personal consumer expenditures of $16.0 trillion (Bureau of Economic Analysis). Rolling back or even eliminating U.S. tariffs would have only a minimal and transitory impact, at best, on price levels and inflation in the United States
- Scott and Hersh
Yes, that would be too little to matter…unless one used leverage to spread these effects far more broadly than the amount of tariff and customs duties actually collected.
AI can trade in leverage. Computational tools have always been used to assess leverage.
Computers themselves were invented largely to assist in complex calculations of leverage — like how much social insurance ought to cost, given a vast trove of assumptions about population, health, etc.
It’s not that this is anything new — just the tools to do this have gotten far more sophisticated than they were 80 or so years ago when they were first deployed for that purpose.
If a tariff system includes a large number of exemptions, exceptions, adjustments, permits, licenses, and other systems that operate in an ad hoc matter depending on country of origin + nature of goods + (insert a dozen other variables), then that will be too much for humans to calculate on the fly. BUT AI could extrapolate based on completely different data what goods will pass when — and then buy/sell to create shortages and then exploit them.
That would probably happen at the ‘private equity’ level — too small for most major corporations (though billions of dollars would be on the table, and if they didn’t pay attention, then private equity firms would eat into THEIR profits).
But they have AI of their own. That’s not just Google, Amazon, and other trillion dollar ‘tech’ companies — every major corporation can simply hire consulting firms able to crunch these numbers…and then as each interacts with others, prices surge — corporations profit — and none of them need to ‘collude’ with one another to ‘abuse’ their market position (at least, not in anyway that humans can detect without mastering the intricacies of the AI systems at work).
With the magical power of leverage at work, and with extreme uncertainty playing behind the scenes, could a tariff that by itself accounts for ‘merely’ 7% of inflation actually account for significantly more?
I suppose it would depend on exactly what sort of leverage was available (and from what sources). Bitcoin? Real estate? Debt? Inflated equities that they expect to collapse and are positioning themselves for a ‘correction’?
In an inflationary environment, the value of debt changes immensely…could major debtors themselves orchestrate inflation by gaming a cluster of AIs?
There are infinite ways AI might “collude” with one another to enable market manipulations nobody has ever seen before (let alone tried to enforce routine market rules about).
Just a little bit of advance knowledge about ANYTHING would be quickly monetized profitably. Failure to strike first would enable other, less well-connected but equally astute players to eat your lunch.
Some possibilities:
- Trucking. Datasets provide extremely high certainty about truck drivers, including responses to Facebook questionnaires. Those datasets also show which drivers are most likely to respond favorably to conspiracy theories. If it is helpful to create a shortage in some place, and one wanted to arrange for a bunch of truck drivers to protest in one area, a group masterminding such a protest would do so (probably in some other country), and then use AI to parse which truck drivers needed to be recruited to do it, and how best to do so (data sets they’d look into would include credit card transactions for diesel fuel, number of friends who are also truckers, number of friends who are also truckers who respond to their posts, etc.).
- Warehousing. Say AI parsed the ‘contractors’ hired as additional work crews at a warehouse, then watched to see how many of them showed up for work, and when, cross-checking that against goods shipped into that warehouse and competing groups of ‘contractors.’ Hire the essential crews nearby for a small gig exactly at the time they were needed for a big gig, and you can lock up the warehouse.
- Counterfeiting. Say AI parsed shipments of goods and the exact value of Amazon rankings for any specific goods. By injecting counterfeit or inferior items into a set of shipments, or a certain number of ‘wrong size’ or other complaints, AI could which suppliers could be intercepted, extorted, shut down, shaken down — which suppliers could be vulnerable if they lost their sole source supplier of certain crucial components or materials…
How would corporations respond to all of these sorts of threats posed by much smaller entities that were much more dangerous than they had been a decade ago?
Most likely, raising prices would be their first line of defense. Then hiring some of the contractors as their own employees (at least temporarily), paying them enough to verify that they were not part of the problem while investigations were pending, then disposing of them as needed (which would also be identified by…well, something a bit like AI).
Conventional wisdom holds that ‘ignorant farmers’ caused (or contributed to causing) the Dust Bowl in the Midwest that contributed so much suffering to so many thousands of people during the Great Depression. That’s erroneous.
Farmers experienced a massive surge in productivity in the 1920s — and took on debt so they could acquire machinery to compete with other farmers, making that glut even worse. When the Great Depression hit, those farmers struggled to pay off their debts by working even the land they knew they needed to leave fallow to recover — but couldn’t because that would mean missing payments and losing their farms.
Then some droughts hit, one after another. Every vulnerable farm got wiped out by the first one. Then even more robust farms started getting wiped out with the second one.
The farmers did not conspire with one another to cause the Dust Bowl. Nor did the banks, at least, not necessarily (certain bankers might have tried to help certain farmers more than others…). At the end of the day, weather/climate + realities of agriculture + realities of banking interacted to create a catastrophe of global significance.
Not the first time. Not the last time.
Every farmer relied in part on algorithms to determine how much crops they would need to produce each year to pay off the interest on their debts, how much it would take to grow a crop, and gambled on whether their land could handle another more year. So did each bank. So did many other players involved in the system. The error was not a result of bad algorithms, but of interacting systems behaving unexpectedly.
Inflation may be a similar structure — multiple systems performing multiple computations, faster than ever, interacting unexpectedly with one another, and driving something unprecedented. Because I believe that possibility might be true, I share Bivens’ pessimism:
I’m pretty pessimistic that policymakers can move nimbly enough to put a cap on excess profits that will restrain inflation in the near term.
Bivens, “Ignoring the role of profits makes inflation analyses a lot weaker”
Indeed. The only way to move nimbly enough would be to figure out exactly how the mechanisms of inflation can be created by powerful players behind the scenes (and if not by them, then any other player with enough money to invest and pay for carefully crunched computations).
I do not know what to do about this possibility, or even if this thought experiment accurately reflects business operations and judgments in 2022. But even if I am incorrect, a day will come (if it isn’t here already) where humans will use AI that we build to achieve things that humans have done ever since the first humans engaged in trade and discovered how to profit by ‘adjusting’ scales to their own advantage.
