Regime Changes: Tour de France vs the U.S. Deficit

The essay begins lower down. First, however, some updates on prior notes. And, as usual, selected readings are at the end.

Updates and Erasures: Low Growth and AI Data Centers

Life Under Two Percent

I recently wrote about how lower growth leads to zero-sum thinking. Various developments continue to remind me of the piece, like this one from The Economist. It is about some recent Harvard experimental work on scarcity and zero-sum thinking.

Here is a key paragraph on where zero-sum thinking comes from:

People who have done better than their parents, or whose families have experienced upward mobility over time, are less likely to think in zero-sum terms. This also explains why younger generations in America are more zero-sum: they have grown up during times of slower economic growth and lower mobility. Other rich countries exhibit a similar generational pattern. In contrast, in many poorer countries where younger generations have experienced more growth than earlier ones, the pattern flips. [Emphasis added]

It can be highly generational and perception-driven. If you think the future will be worse than the past you are more likely to embrace zero-sum thinking—and that is increasingly the case in richer countries, like the U.S. If you think the reverse, that the future looks better, as is the case in developing or poorer countries, then the tendency to zero-sum thinking can dissolve. You think there will be enough for everyone.

And so, absent explosive growth, which seems unlikely, which policies reduce zero-sum thinking? Back to that Economist piece:

Some policies are especially likely to create win-win outcomes, particularly over the long run. These include policies that expand opportunity, such as strong public education, access to health care and support for poorer families; investments in innovation to expand the overall economic pie; and policies to mitigate climate change, protect the environment and conserve natural resources to reduce the sense of scarcity that fuels zero-sum thinking. [Emphasis added]

Sadly, those are not the policies being pursued in most countries as life under two percent becomes locked in. Granted, most countries, other than the U.S., already have solid access to healthcare and strong public education, so there is not materially more to be done on those fronts. Nevertheless, you can see how things are likely to get worse and more zero-sum before they get better if they get better.

SPVs, Credit, and AI Datacenters

I continue to mull the inevitable consequences of the securitizations of AI data center assets for financial system stability. In particular, I worried aloud about the inevitable arrival of securitization in this picture, given massive capital spending and the perception of annuity income with low risk. (Btw, since last writing about this I saw that Grant's Interest Rate Observer is onto this topic with similar skepticism, so I'm in good company.)

First, let's consider this piece on Wall Street working up some too-clever-by-half ideas about bespoke securitization of data center leases.

Goldman Sachs has floated one alternative: that they sell off their stakes to other investors once long-term tenants are locked in and buildings are “stabilized.” 

That would spread out the risks and free up more of that initial capital for further data center development and the utilities to power them, according to a recent Goldman Sachs white paper, “Powering the AI Era.”

This should sound familiar, of course. It is a form of securitization of these data center assets, at least in the context of the income from them, which I have written about. But given technological change and replacement cost, this is arguably less stable than simply owning mortgages on real estate, so .. for sophisticated investors only?

Nah:

The investment bank said insurance funds and 401(k)s looking for long-term, stable returns would be good candidates to buy the stakes off initial investors. Jason Tofsky, global head of digital infrastructure banking at Goldman Sachs, called it the next turn of the capital “flywheel” in an interview.

It's all still bespoke, of course, not standardized, so prospective issuers are still trying to come up with rules of the road, including locking people into data center lease contracts for longer. What could go wrong, given the rapid pace of technological change?

To attract insurance and retirement funds, the industry would need to introduce more certainty and less risk into data center lease contracts. That includes nixing early termination clauses for tenants and extending leases to 17 to 20 years, from 10 to 15 years, according to the white paper. Whether those stricter terms become market standards remains to be seen. 

So these aren't yet collateralized AI obligations, which I wrote about in my prior piece. They're still single asset and closer to sale-leaseback structures, not structured issues, but ... they're headed in that general direction. Will it be the next phase? It would surprise me more if it wasn't.

Rough Notes: Tour de France vs the U.S. Deficit

The world's largest in-person sporting event started this week: the Tour de France cycling race. Nothing else is close in terms of size, with more than 10 million spectators likely along its route. And attendance is free.

Given its long history—the first edition was in 1903—the Tour is a trove of interesting data. For example, it used to be even longer than it is at present, with the Tour previously having had single-day stages as long as 500 km. It also used to be raced sometimes without gears, with very little training, on unpaved roads, and more.

The modern Tour is a much slicker and more professional thing. It has a caravan of hundreds of vehicles, worldwide television coverage, professional athletes sometimes paid millions of dollars to pedal a bike, roadside giveaways, regular scandals, and live pee breaks. It is an ... event.

Given its scandals, however, one of the first thoughts many people have in thinking about the Tour is the use of performance-enhancing drugs, like in the Lance Armstrong era. Granted, Tour athletes have never really hidden that you can't race 3,000 km on clean living and daily pasta.

Here is former professional cyclist Jacques Anquetil puffing out his Gallic cheeks and chiding critics on the topic:

"On ne gagne pas le Tour de France à l’eau minérale." ("You can't win the Tour de France on mineral water.")
— Jacques Anquetil [c. 1965]

Indeed. Given tens of thousands of vertical feet of climbing, high daily speeds, and caloric needs of 6,000-8,000 daily, going to over 10,000 on the most difficult days it pushes up against the limits of human performance.

And that, of course, is worth considering. What are the limits of human performance in this context? And how would we know?

One way to approach the problem is to consider average Tour de France speeds. And one media organization tried that recently, producing this graph:

It should be obvious that there are multiple issues with this graphic. First, it is a short time series, only two decades. Second, there isn't much variance, with speeds only about +/- 2% of the mean over the period. While you could draw a trend line up and to the right through this data, you could equally draw a fairly flat line.

Does it change if you bring in more data? Here is a scattergraph of all Tour average speeds from the first edition up to the present.

This is much more interesting. To some degree, you could argue that recent performance improvements are actually below trend, that we are only returning to historical norms.

While true, the problem is that it's not entirely clear what "normal" is when it comes to Tour performances. After all, the post-1980 period was tainted by performance-enhancing drugs and blood doping, from Marco Pantani to Lance Armstrong, with a stop-off for Bjarne Riis.

In a sense, the Tour has gone through various eras. There was the early period, raced on gravel roads and with fixed-gear bikes. And then growing professionalism before WWII, a hiatus, and more professionalism, and so on.

In the following graph, I try to make this clearer via a segmented graph through the scatterplot points:

While not incontestable, this is more satisfying. You can see the pre-war period, the hiatus during WWII becomes clear, the rapid post-war professionalization, and then a long interregnum during which average speeds didn't improve much, if at all.

And then it all changed in the late 1980s. Not coincidentally, that was when Amgen's anti-anemia drug Epogen (EPO) hit the market. Its main function was to stimulate the production of red blood cells (RBCs) in the bone marrow. And RBCs are the primary carriers of oxygen to stressed muscles in endurance athletes. It took approximately zero time for EPO to begin being used by endurance athletes looking for an edge.

But after a halcyon doping decade, it all ended in the early 2000s with endless legal suits, better testing protocols, and so-called biological passports that made clear what your "normal" red blood cell count was. And wasn't.

More recently, however, as you see in the segmented graph above, speeds have begun trending upward again. Speeds are now higher than they were at the peak of the blood doping era 20 years ago. And the pace of improvement is on par with the EPO era.

Much journalist time is spent fussing over this and trying to parcel out the likely (non-drug) contributors, from better nutrition, to more aerodynamic bicycles, to wider tires, and more. There are entire online forums devoted to the cause of trying to figure out which drugs cyclists are taking, from testosterone, to human growth hormone, and more.

We simply don't know. What we do know, however, is there has been a recent regime change in cycling, and by that I don't mean the rise of a specific cyclist, like Tadej Pogacar. I mean it in the technical sense: a bifurcation where the behavior of the system changes qualitatively, as it does at various points above. The past of cycling is a different place.

But thinking in terms of piecewise regressions and regime changes is highly useful, whether you are considering Tour de France average speeds, or the changing nature of risk in capital markets. Things are the same until they aren't, from credit spreads, to interest rates, to price-earnings multiple, to deficits, and trying to map the new dynamical system regime onto the old one is generally an error.

It is about as sensible as trying to understand current Tour de France average speeds without acknowledging we are in a new era, even if we don't know how it happened, or why.

Regime Changes and U.S. Deficits

While fiscal deficits are not as exciting as Tour de France races, the crashes can be worse. Let's turn the analytical lens from cycling races to the U.S. deficit and consider the regime changes therein, and their consequence.

Economic series are just as amenable to this piecewise approach as cycling times. Consider the following graphic showing regime changes in the US federal deficit as a percentage of GDP, by year, since WW II.

You can see that there are similar patterns as we saw previously with average Tour de France speeds, in particular, pre- and post-war effects. But in more recent decades, that pattern has broken down.

We see an administration-specific pattern of high early deficits, followed by subsequent-year improvements. And that pattern persisted until the global financial crisis, after which the regimes became more jagged and fragmented.

The modest recovery in deficits after the financial crisis bailouts is followed immediately by 2016 tax cuts, COVID spending, and higher post-COVID spending on various new initiatives. The result is an inversion of the usual pattern of high early deficits and lower later ones. Instead, we get high early deficits, and larger ones later, a pattern that looks likely to be replicated in the current budget cycle.

This is a classic regime change. We see changes in level, slope, and duration, all at once, and that sort of tripartite break with the past is very hard for financial markets to discount, given that it is alien to their recent experience, and thus to pattern matching.

Just as the Tour de France reveals distinct performance eras through shifts in speed and doping patterns, the U.S. fiscal deficit shows clear regime changes marked by breakpoints in policy, crisis response, and spending behavior. Both systems—athletic and economic—exhibit piecewise dynamics, where past patterns give way to new structures. Recognizing these transitions is essential: applying old models to new regimes blinds us to underlying shifts. Whether analyzing cyclists or capital, the regime matters, perhaps even more than the trend.

Rougher Notes:

Some Chinese millennials are competing at indebtedness ... University endowments face diminishing returns as illiquid assets strain liquidity and spending flexibility... Popular beverages in France are found to contain microplastic contamination, raising health and regulatory concerns... Extreme convective rain is intensifying at alarming rates, with ~14% more rainfall per °C of warming making prediction difficult... UTIMCO increases private equity exposure while exploring the secondaries market to improve portfolio liquidity... Some CEOs now openly admit that AI adoption will wipe out white-collar jobs... Observers mistake intelligence for experience in AI, a form of semantic pareidolia linked to anthropomorphic bias...Investment workflows are being reshaped as AI systems take over core analytical processes ... Big Food retrenches as snack consumption declines and Americans turn away from sugary treats... Young college graduates enter an employment crisis as job markets shrink and ROI on degrees weakens... In Japan, biodiversity shifts under rural depopulation reveal complex ecosystem responses... A rare behind-the-scenes look at SNL showcases the precision of live boom mic work ... The end of the essay and its consequences ...