Meditations On Barbells

On Socio-Moses parting socioeconomic distributions, barbells, dinosaurs, and how competition hollows out the middle

I initially used the image of the barbell to describe a dual attitude of playing it safe in some areas (robust to negative Black Swans) and taking a lot of small risks in others (open to positive Black Swans), hence achieving antifragility. That is extreme risk aversion on one side and extreme risk loving on the other, rather than just the “medium” or the beastly “moderate” risk attitude that in fact is a sucker game…

A barbell can be any dual strategy composed of extremes, without the corruption of the middle — somehow they all result in favorable asymmetries.

So just as Stoicism is the domestication, not the elimination, of emotions, so is the barbell a domestication, not the elimination, of uncertainty.

Nassim Taleb

[Epistemic status: trepidatious]


There's an observation about the world that's rather curious. Across a wide variety of settings, from business to biology to politics to online discourse, we see a phenomenon where the middle seems to be disappearing!

Much like Moses parting the sea, we see the ends and are left asking why things seem to tend towards extremes so often.

Taleb's view, to start with, is that since the world is so unpredictable, you'd rather be hyperconservative or hyperaggressive to survive and then succeed. We can take it as a phenomenon of the world if we'd like, but it is descriptive rather than explanatory. The answer is that barbells exist everywhere because they protect against bad black swans and help bring good black swans. But the very mechanism by which such a black swan comes about is in itself interesting, and understudied.

Its not as if there's an actual Moses who waves his hands to move actual sociological phenomena towards their extremes following the invisible hand of Moloch. If there were we could take the fight to him, but barring that we're left with the world as it is.

So we're left asking ourselves what is that process we can anthropomorphise as socio-Moses and understand it.

To begin with, the world of business. The biggest news of the first wave of retail was how Walmart destroyed communities. Before Walmart, there were mom and pop shops everywhere selling everything from toothpaste to t-shirts. After Walmart, there was the behemoth who sold most things and the small specialty shops that sold town trinkets that Walmart didn't care about.

The Walmart Effect is the effect that Walmart has been known to have on the communities in which it builds locations. The presence of a Walmart store can hurt the business of smaller companies and lower wages for local workers.

Same for Starbucks. Before they came cafes were idiosyncratic and colourful and plentiful (and more expensive). But afterwards, not so much.

Same with Amazon. Go ask any VC or any consumer if they'd like to fund a generic ecommerce company and they'll say the same thing - the only survivors will be the behemoths like Amazon and the truly unique/ bespoke/ differentiated ones where the largest one might be Etsy.

This isn't an ethical point or some sort of an indictment. The impact I'm talking about here is on the supply side. It's just interesting.

When you have competition along a commonly seen dimension, it bifurcates and polarises that dimension to the point that the distribution becomes bimodal. The middle gets decimated.

All the three companies mentioned above excelled in standardisation, in bringing efficiency and some level of cost control. They reduce the dimensionality of the problem in their efforts at efficiency.

But it's not just companies. Search engines change the way we consume information. I'm not even talking about human curation the way a librarian might have done for you. Instead of serendipity we have the problem of plenty. What was a multidimensional search pattern to find interesting things breaks down into similarity patterns to what you've searched before (or others like you have searched).

Matthew effect gets stronger and the well hyperlinked get even more traffic. Even if you hold that this has been fine, as I do, the existence of SEO as a category belies the fact that this is a clear point of differentiation in how information reaches us.

What we compete on defines us and rearranges the world accordingly. It subsumes the multidimensionality of real life to a smaller number of tractable realms.

Increasing legibility of the exact dimensions you're competing on helps Moses hollow out the middle.


Now, off from business and onto ecology, let's have a look at what Moses has been upto in speciation.

Most ecology work talks about how speciation comes about as a result of geographic separation. If there are multiple ecological niches then that will result in multiple speciation events. If the niches get joined, then the species count falls as a few dominate the newly available large domain. The competitive exclusion principle says this.

… famous example of the competitive exclusion principle is shown in the figure below, which features two types of single-celled microorganisms, Paramecium aurelia and Paramecium caudatum. When grown individually in the lab, both species thrive. But when they are grown in the same test tube (habitat) with a fixed amount of nutrients, both grow more poorly and P. aurelia eventually outcompetes P. caudatum for food, leading to P. caudatum's extinction.

Now this is pretty rare in nature, what instead happens is something called resource partitioning.

Competitive exclusion may be avoided if one or both of the competing species evolves to use a different resource, occupy a different area of the habitat, or feed during a different time of day. The result of this kind of evolution is that two similar species use largely non-overlapping resources and thus have different niches.

By finding a new avenue of differentiation the speciation resumes apace. Darwin's finches provide an interesting observation here, as in so many cases. This is character displacement.

Darwin's finches can be found alone or together on the Galapagos Islands. Both species' populations actually have more individuals with intermediate-sized beaks when they live on islands without the other species present.

Increased competition along a dimension makes the population seem bimodal. Intermediates die out. That's also what we saw with Walmart and Amazon above, the death of the middle.


In the world of business Moses' actions can seem kind of banal. Since the competition is so often on price, convenience and selection, this has become internalised to the point its commonplace to conclude, for instance, that Amazon's success was inevitable.

But his effects are vast.

For instance, even when we look at corporate organisation sizes, we see that "the middle" seems to be disappearing everywhere. In this analysis from a McKinsey paper (disclosure: I was one of the authors), we see that the later years lets us see high ROIC (return on invested capital) and high variance in ROIC, indicating that there's a divergence happening between the largest firms and the smallest firms. The top 10% of firms account for 80% of all profits.

And mostly, the size difference between the large and small companies is still growing!

The 5 largest holdings in the S&P 500 are larger than at any point in the past few decades. That concentration is increasing!

The S&P 500 being dominated by a few companies isn’t new, but the concentration is more extreme. Over the last 25 years, the five largest stocks represented about 13% of the index, on average. In 1999, during the dot-com bubble, the number rose as high as 16%. It didn’t get that high again until 2018.

There's more, such as the belief that the US economy is leaving midsized companies behind. From HBR.

Data suggests that a perfect storm has been brewing for midsize companies over the past 50 years. In every successive decade since 1970–79, the annual growth rates of assets, sales, and profits have been declining for midsize companies, which are increasingly struggling to earn profits.

In an ecological sense, when you look at how companies are distributed, this suggests that the natural tendency for systems is to tend towards producing a barbell. Just as Moses' actions would have caused.


Now all of this is interesting, but did you know that Moses also got his hands on dinosaurs? It's true. This is why there were barely any medium sized carnivorous dinosaurs around in the Cretaceous. There were mostly either giant carnivores, weighing north of 1000 kilograms, or small ones, weighing less than a 100.

If you're not interested in dinosaurs, feel free to just look at the pictures and skip this section (though I'll look at you in askance now and forever). From a wonderful Science paper exploring this:

We found that megatheropods (more than 1000 kg) such as tyrannosaurs had specific effects on dinosaur community structure. Although herbivores spanned the body size range, communities with megatheropods lacked carnivores weighing 100 to 1000 kg.

The answer that the paper cites is that the juveniles of the giant theropods, like the overly rambunctious teenage Tyrannosauruses, occupied the niche that would otherwise be occupied by a traditional medium sized dinosaur.

Needless to say this is weird. If you look around at the world today, there are carnivores in all shapes and sizes. There's no precocious baby lion cub that's making sure that hyenas don't exist. There's a clear continuous line from the tiny shrew to the giant tiger. Particularly enlightening is this analysis of current breakdown of carnivores today vs that which existed for the dinosaurs.

At least with the dinosaurs what seems to have happened is that as Jurassic slowly (very slowly) moved to the Cretaceous period, the overall size gap in carnivores increased! The theory is that there was increased competition amongst the carnivores during this time as the prey for the dinosaurs reduced in size.

The splitting up of Gondwanaland meant Jurassic was a time of multiple niches being created. The hot, humid climate meant plants grew plentiful, which meant herbivores grew big and plentiful, which made the carnivores pretty happy in turn.

The Jurassic period therefore had many large allosauroids and medium sized ceratosaurs which meant a smaller size gap. The young Allosaurs were similar in size to the fully grown ceratosaurs, which meant there were fewer prey niches to much upon. If the larger Allosaurs ended up eating the Sauropods, that meant the medium sized ceratosaurs would've gone and eaten fish or other smaller prey.

However when the Cretaceous period came about the diversity of the sauropods and stegosaurs decreased. This meant that the tyrannosaurs and the abelisaurs had added competitive pressure on the prey base, which was now truncated, which in turn changed their relative size distribution.

We suggest that competition for a limited prey source by both large and small carnivores, and the broadening of megatheropod niches, resulted in a widening of the carnivore gap.

The “grow fast, die young” approach of megatheropods resulted in a predominance of juveniles in communities, filling the morphological and functional role of mesocarnivores, which as a result are absent from the fossil record as individual species, artificially deflating diversity indices of dinosaurs as a whole.

The carnivores here got a barbell-shaped population distribution. There were the very largest who survived and thrived, and then there were the tiny ones, who also thrived. For all those in the middle, they were left out in the cold.

Only times when there was the availability of multiple enormous prey species which reduced interspecific competition, allowed the coexistence of an unusually diverse assortment of carnivores.

The polarisation of outcomes here is fascinating. If you're a measly 15 kg dino baby just after hatching, and you need to grow 100x to be a 1500 kg carnivore king, then you got to find a lot of food to eat, and pretty quickly. So if a fast growing juvenile tyrannosaurus starts snacking on the available food sources, they will quickly outcompete a "normal" medium sized carnivore who would've been occupying that niche otherwise.

Which means the body size distribution of dinosaurs, there's two interesting observations:

  1. Periods of higher interspecies competition results in a gap of middle sized dinosaurs, and the size of this gap is correlated with the size of the largest dinosaur

  2. A larger species that have to grow across body sizes means that their young ones go through the different size distributions


So clearly Moses has been busy, but is that all? Walmarts and Allosaurs? Well, no. This holds across all swathes of human experience where legibility has engendered competition. The idea that increased hunt for efficiency increases focus and competition on one axis while making the distribution barbell shaped.

When I wrote about how we're being greedy algorithms in our interactions with politicians or CEOs it hinted at this. It forces us to compete in one dimension, for attention and efficiency, which reduces the dimensionality of the phenomenon.

A few other examples.

Value of time

Byrne Hobart writes about the fact that the value of time is either seen as zero or infinity in the information economy. The contention is that the best software either tries to save you every tiny slice of time available, implicitly assuming your time to be worth a lot so that saving you thirty microseconds is valuable, or it tries to treat your time as worthless, by providing endlessly entertaining and addictive content to be consumer in any fraction of time you choose.

Across the information supply chain, the value of time either rounds up to approximately infinite or rounds down to roughly zero. And sometimes, this variance happens within a single app. When Facebook sends a push notification informing you of a new interaction, they're treating your time as priceless—don't be even a second late for this notification! But once you've surmounted the barrier of actually opening the app, future interactions implicitly assume that your goal is to fill as much time as possible.

Similar to the business mix as above, those who fall in the middle don't seem to get much shrift. Its as if there's some force that pulls at the value of time in these instance to either extreme, with the middle being effectively outcompeted and decimated. Once the value of time becomes legible, it gets bifurcated into either being treated as infinitely valuable or completely worthless.

Graduate salaries

If you're a young computer science graduate, say, it seems to exist for you as well. Since software developers are still paid pretty well, there are some fatter spots in the middle. But even so, if you add the H1B workers into the mix it will get even more lopsided.

For instance, this analysis of developer graduate salaries shows the average joe developers earning decent sums and a few superstars earning a ton. Note this is from the before days where the data scientists were earning $1m+ weren’t represented.

If there truly are 10x or 100x or 1000x developers, then it only makes sense to overpay a few by 2x or 3x in order to place the correct bet to capture their talent. Once talent measurement here is legible and you can compete on it, the salaries bifurcate.

What about some other white collar professions?

Here's an analysis of lawyer salaries by Dan Luu which shows the rise of the superstar lawyer. (parenthetically, doesn't this coincide with the rise of TV lawyer dramas?)

We now live in a more bimodal world. Its unclear if law's more legible today than it was in 1991, but I have to imagine perhaps that law being made more legible so superstars can stand out. We can blame automation, or offshoring and outsourcing, and they're probably true too. We can clearly see the result that the middle is being hollowed out. The superstars do pretty well, while the rest separate out into the poorly paid mountain to the left of the chart.


What about the wonderful and illiquid world of venture capital? It seems to also be the case for VCs. As Delian tweeted recently.


The actions of the mythical socioeconomic-Moses is all around us. Whether you believe in an antifragile portfolio or not, it does seem that barring intervention in the ecosystem the natural tendency does seem to be for Moses to act freely. And much as it is attractive to think of this as a peculiar sort of failing in our society it might just be the inevitable result of multiple rounds of selection over specific dimensions made legible.

And if the fact is that selection on the basis of legibility of particular dimensions results in some version of a bimodality, where the middle disappears, that's also indicative of the fact that the way out is to figure out different dimensions of legibility. Maybe not quite getting to illegibility, but at least with interest towards gaining a fuller picture.

It might not just be because of something nefarious going on, but rather it's the outcome of a system that generates polarisation as its output again and again. Barring a regular shakeup of the ecology in which they operate, we're resigning ourselves to a more static field of megatherapods, those it tolerates, and it's prey.

A potential implication of this thought process is that the 'normal' status for the economic (and ecosystem) variables seems eventual stasis, once all the niches are filled. So while we like explosions of creativity like with the semiconductors, with computing, with software, with the internet and with mobiles, these new technologies breaking through to mainstream is essential to maintaining the growth trajectory.

And the way to break through that stasis, and to see a further Cambrian explosion filling all available niches, requires an external event, a catalyst, to shake things up. Usually these are technological (chipsets, mobile, AI), or often cultural (globalisation, move against factory farming). Without new technological vistas, the turnover of species gets lower and existing animals get larger. That's the innovators dilemma.

From the first part of the ecological relationship to markets:

Unleashing a new breakthrough innovation clearly requires a shift in the ecosystem. Now this catastrophe can come from technological shifts (rise of the smartphone, 4G, changing biotech, CRISPR, availability of new materials) or they can come from business environment shifts (globalisation, contract manufacturing several seas away) or they can come from regulatory changes (anti trust). These are all ways in which we regularly see the business world go topsy turvy and a new paradigm emerge and a new suite of kings and queens emerge.

However the parallel potentially also goes beyond the mere existence. It also tells us that the absence of one of these ecosystem wide shifts, internal or external, results in a more static environment. Over time it ossifies into a structure that can potentially endure for long periods of time without meaningful change.

Socio-Moses is busy parting the seas all around us. But the good news is that he's pretty myopic. The only way to stop him is not by fiat, nor by decree, but by constantly distracting him and experimenting with pathways anew, to make the ecology itself more diverse.

Rather than solely hunting for more efficiency, perhaps we can hunt for more diversity? Create more niches? The blame isn’t on the megatherapods, but on us for not experimenting with the ecosystem enough.