Dunbar's Number and Startup Companies, Part 2

More on the Dunbar number transition in startups from Ribbonfarm.

Many people have noted and describe the culture shift that occurs at startup companies - curiously, usually at a headcount right around Dunbar's number. Of course, differences occur over the causes.

One explanation from personal experience as a consultant at multiple companies (full post here) is that the influx during this growth phase consists mostly of people who are middle- or lower-tier contributors, whose strategy is primarily that of seeking a stable institution and burrowing in. This is a rational strategy for these folks, because as the company grows, the individual's contribution (especially the lower-level, later hires) is less important to the company's health and therefore feeds back less to the individual's fortune. There's no reason to knock yourself out, if you know it won't make much difference to your future well-being, and you don't have confidence in your talents anyway.

(When aggressive young managers talk to these people, it sometimes feels like you're playing two different games on the same field. Manager is rewarded by scoring. S/he wants to perform, and wants others to perform, so risk-taking and occasional annoying of others with non-company-aligned-goals is worth it. Late-comer wants to keep a low profile and not rock the boat and often seems frankly confused at any suggestion that they should take a risk to accomplish something, but of course s/he can't say this out loud. Or sometimes, they actually do. I once had an employee, a recent college grad, who looked at me with surprise at the start of such a meeting and said, "This isn't me, this job isn't what I really do." I was grateful for this honesty because it saved time. Not in the way that I think this employee was expecting.)

Netflix has a set of slides that's been making the rounds in the Valley recently, with a different (though not necessarily conflicting) explanation. This explains the dilution of high performers as an efflux of talent (driven out by bureaucratic process sclerosis) rather than the influx of low performers. This certainly occurs as well though in my experience, it's a much slower process, as people tend to stay at a company they believe in with colleagues they enjoy even as administrative annoyances accumulate.

Culture from Reed Hastings

I also think that they're being too kind about the motivations behind the proliferation of bureaucratic rules, which they ascribe to well-meaning management trying to control the chaos at a company that's become too large and complex to run informally. I think there's a lot more conscious justifying-of-one's-own-existence processes that new administrators put in place, although I concede this is easier to do and probably happens more in my regulation-heavy world of biotechnology than in the tech world that the Netflix people know.

Loyalty, Optimization and Corporate Empires

Peter Turchin's War and Peace and War analyzes patterns of empires in history, extending the work of Gibbon's Roman history and especially the Arab historian Ibn Khaldun, in particular his idea of asabiya, or solidarity. Turchin argues that everyone's recent favorite whipping boy, the rationalist concept of Homo economicus, cannot explain how states can fluorish or fall; that there has to be a social glue which is not simply aggregate self-optimization. This (Khaldun and Tuchin state) is asabiya. In a recent article he argues that this can also explain the decline and fall of corporate empires like Microsoft.

There are some problems with Turchin's very interesting book, and it's interesting to apply asabiya to corporations, but I think it's not needed to explain the free-rider problems of big companies. It's certainly true as Turchin says that small-ish companies, as they are making the transition to larger ones, often has plenty of knaves weaseling their way in. If you've ever been present during this transition you've no doubt seen this first-hand.

The problem here is that you can still explain this in terms of the behavior of Homo economicus. The founders and the few people who come after them understand what is needed for the company's success, act in plain view of all the other (few) employees with the risk of termination and a damaged reputation, and understand that their own behavior has a big impact on that success. That is to say, when even a rational low-asabiya individual comes into a small company, they know that if they free-ride, it will hurt the company badly and soon, and everyone will know it. As the company gets bigger, it gets easier for free-riders to hide, and their free-riding hurts the company less and over a longer-term.

It's also worth asking how there could be asabiya in an organization that just appeared? The closest thing to this is the camaraderie that emerges at small companies (which is not a bad thing) but that camaraderie is less about some abstract idea of the company and more about a social connection between the founding few individuals. And asabiya to the founders in place of the company as a whole can and does end up damaging the quality of decisions and therefore the fortunes of the company as a whole.

Dunbar's Number and Startup Adolescence

Dunbar's Number Part 2 can be found here.

Kevin Simler has written about a connection that I've also noticed, that between Dunbar's Number of 150 and startup organizations. He goes further and analyzes startups in terms of anthropology, which is excellent. Looking at it one way, all business analysis really is just a branch of anthropology that uses a scoring system, because the larger civilization these mini-tribes exist in have an agreement about the scores the mini-tribes have to achieve to continue getting the chief's approval to exist as such. (This is called accounting.)

I was often brought into biotech companies when they had stalled out at this number because of external constraints (failure of product in clinical trials) to try to accelerate the process or jump the track to a new indication for the drug. There are growing pains at and beyond this number that often significantly damage the company's prospects in its transition to being a larger organization.  What I noticed is that by the Dunbar Point:

1) Corporate culture, for better or worse, is set in stone. I suspect this actually occurs much earlier in growth than Dunbar's Number, resulting from the personalities of remaining founders and challenges the company encountered - likely in that order.

2) This seems to be an inflection point at which the free rider problem becomes much worse; non-productive people who want a stable income are hired, and, recognizing rationally that at this size firm or larger their own performance won't impact the future prospects of the institution, they often don't work too hard. They also recognize that in the presence of remaining founders, they're not going to climb the ladder far anyway. And finally, karma is no longer real at this stage. There are too many people for us each to keep score on the rest of the people in the organization. Bad deeds go unpunished and good deeds go down a black hole. Karma is only real when there are people who can remember the deeds that you're in frequent contact with and who will and act on them in the future.

3) Possibly because of #2 above, rules that don't make sense or don't seem to connect to the organization's purpose and means of self-perpetuation multiply and become entrenched. This occurs because the people enforcing them now justify their existence in the organization, and because as an organization grows there are more people between any individual and the customer and investors, i.e. the entities in the outside world that the organization needs to perpetuate itself. Consequently it's more important for an individual concerned about their future prospects to please other people inside the organization, regardless of how irrational the rules are with respect to the future prospects of the organization. You're more likely to be fired (and lose money) if you defy a rule to do what you think is best for the company but it makes your boss look bad, than you are likely to lose your job due to downsizing or company bankruptcy if you follow the irrational rule, thus harming the company over the long-term. In the abstract, what is rational in the near-term for an individual departs from what is good for the company. This also seems to hit an inflection point at about 150. There's a great Dilbert (which of course I couldn't find) where Dilbert asks the boss, who's obsessed with screwing over another department as well as the customer, when they're going to think about beating the competition.

It seems to many observers that something happens as previously successful start-ups grow to make them go off the rails, largely because of internal cultural changes that cause them to self-sabotage. My observations of what happens during the startup transition contains that implicit assumption, although we have to admit it might be false: it might just be that we're seeing statistics, and regression to the mean implies that you should expect to see a bunch of companies doing well for a few years (during which time they will grow) and then their luck runs out. But the following suggestions assume that there really is a true negative effect on corporate culture and success in future profits of growth beyond Dunbar's Number.

To #1 and 3 above: these are intractable and I can offer no remedies. Removing founders often doesn't work; and status is important to humans. See this primate behavior experiment to understand why.

To #2: - Set performance metrics and relate them simply, visually, and frequently to the performance of the organization as a whole and against anonymized members of their teams. Automate it to the extent possible, so they can get frequent feedback (not just once a year or quarter) without consuming too much time generating the report. Make it clear that these are the same metrics their bosses are using. Suddenly you see that fake sick day you took, and how it affected a deadline that cut into profits (or some other metric elsewhere in the organization.) Of course, sometimes, the reality is that people's jobs really won't have a huge impact on the company's fortunes. But if you can't think at all of how to measure someone's performance and connect it to how the company is doing, then you either need to think harder, or get rid of the position. And if you're not willing to think of eliminating the position because Suzy in marketing is a nice lady and she knows Jim the Senior Director of XYZ, then you're already way beyond the startup transition.

- Focus on loyalty to small teams of up to 10-12 people. This is the military's strategy of enforcing emotional loyalty within a unit; it's not possible to be emotionally loyal to a large abstract organization unless you were raised in it as a child (like a state or religion.) Not only will this create more feedback for people - easier to to feel rewarded for doing something for a co-worker you respect than the organization, and know there will be real karma as a result (and these bonds of respect are created by the company). This also helps break down other kinds of alliances that form within the company (but not in the company's interests).

- When the founders aren't going anywhere, the new blood you get knows it's not going to the top any time soon, and you can't groom everyone to replace them. That's okay, as long as you're networked with the rest of your industry/space, and the hard work and reputation that they build while they're with you will count elsewhere in the network. Companies that don't expect young blood to leave are stupid, and some have actual functions in HR to address this. People in high-turnover, high-human-capital industries are saying "duh" to this one, but this should be something that's addressed explicitly right from the beginning of their association with the company.

Dunbar's Number Part 2 can be found here.

Intrade Manipulation and the Epistemic Rationality of the Pundits

Karl Rove on FOX, Election Night 2012.  Delusional, or crazy like a...

Intrade was still only in the high 60s for Obama until polls actually started closing on the American East Coast. Then it only went into the 70s. This, despite other prediction markets having much higher odds for him for many days before that, and of course the various statistics witches like Nate Silver, who was in the 80s two days before. Math whizzes can look at numbers if people decide this question is important enough to merit a more rigorous look, but for now I'll have to go on these facts for my argument:

a) Intrade users think it was manipulated.

b) For a campaign that wants to appear as if it will win, Intrade is a good candidate to manipulate because it's better represented in American media than other predictions markets.

c) There was a single buyer making many of the Romney purchases.

d) It was in the Romney campaign's interest to try to appear as if it had a chance going into the election, because no one gives money to a doomed loser.

The question does remain as to whether it was deliberate manipulation, or purchases by people (or one person) who genuinely believed Romney would win. There are lots of people who believed Romney would win, and if reports are to be believed, Romney himself was one of them. And of course prediction markets do fail when enough people with bad/non-updateable beliefs are investing in them. "Fail" in the sense of not converging on the actual outcome significantly prior to the event, of course. If you had money to put into this contract and you bet on Obama though, I bet you didn't think it failed you. If it was the result of deliberate manipulation, you should be thanking Romney campaign donors who redistributed their wealth into your pocket continually for several weeks so you could bet on Obama. The twenty-first century is an interesting place.

In all the post-election laughing/anger at delusional or propagandist pundits, it is interesting that even moderately well-informed news junkies saw it coming.  But not everyone:  the pre-election denial pieces were fascinating and sounded like a combination of hippies, religious fanatics and alternative medicine nutbars: Peggy Noonan's "All the vibrations are right" is my personal favorite.

But it's also worth pointing out that from one standpoint, Karl Rove's famous meltdown was rational - for next time around. And game theorists know that games are very different when there are more (or indefinite) rounds. Other GOP higher-ups will remember that this was the guy that just would not give up. He's invested. I bet those higher-ups don't care if Jon Stewart makes fun of Rove. So neither does Rove.  Neither do any of the other pundits.  Really, what would have been the benefits to a Dick Morris of calling it like it really was, even if he actually knew how it really was? (That said, Pundit Tracker is the best site ever.)

Of course, Rove may really have believed Romney was going to win, which makes him epistemically irrational. But this false belief may lead to future GOP campaign consulting gigs as described above, which makes him every bit as epistemically rational, regardless of whether the FOX meltdown was theater or not.

Some Problems in Turchin's War and Peace and War

Peter Turchin aims to be the real-life Hari Seldon, so it's admirable he gets the inevitable comparison out of the way early in the work. (If you're not familiar with Turchin and/or Seldon, go here and here respectively.) And the goal is an admirable one: model history mathematically. His arguments center on the consequences of interactions at culture frontiers between empires and their neighbors, on the mechanisms which cause state-decay from within, and on the crucial effect of metaethnic solidarity, which his macrohistorical predecessor Ibn Khaldun called asabiya.

Indeed one of the strongest values of the book for me was the synthesis of the two most famous works of macrohistory, namely Gibbon's Decline and Fall of the Roman Empire and Ibn Khaldun's Prolegomenon.  Turchin summarizes Khaldun as follows:  empires decline, former barbarians on the borders who were forced to organize are more vigorous, and either supplant the empire through conquest or infiltration, or surround and surpass them. Turchin's work can be seen as largely an expansion of Ibn Khaldun's.  Of course viewing the book's main value as an executive summary of a classic isn't necessarily a compliment, and even in Turchin's goals themselves there are issues. First, the fixation with the history of empires (which, granted, is interesting) is different than a focus on human flourishing, which is much more important. To be fair Turchin never implies that this is what he's doing but if you're going to model history, it seems a more worthy goal is a direct focus on happiness than grand sweeping armies.

Second, and I think far more damaging to the value of the book, the point of a model is to predict. Yes, it's interesting if a model that you built based on the Romans and Byzantines fits the Arabs, but the proof of the pudding (and its true value) is in forecasting the future. To this end he makes only two predictions: one, that if the Islamist Chechen incursion into the Russian Caucasus expands, this means the withering of the Russian state and the growth of a new imperial state will grow from the nucleus of the Chechen insurgency; and second, that China and the United States will be competitors in the twenty-first century. I don't think you need to be a macrohistorian to arrive at the second one. As to the first, while I don't take issue with a growing Chechen insurgency damaging the Russian state's legitimacy, I do find it dubious that an Islamic successor state in the Caucasus would be an "empire" or even a functioning state by any means. I refer to Fukuyama's argument that the Islamist movements of the world are desperate rear-guard measures that will have no lasting impact. To be a real state today you need a real economy, with a real, complex, technological industry. That is to say, the Taliban were never in any danger of building an empire in Afghanistan.

This brings up a point which Turchin half-makes, which is that Things Have Changed. He dedicates a chapter to the effect of cell phones, but I think international economic organizations, a world full of clearly defined borders and terrain, widely available media, and wealthy democratic countries dependent on value-adding labor and human capital might behave differently than the island-civilizations he's writing about. Today, there's just no equivalent to Rome and the Germans on the frontier, or Russians and the Mongols. The world is neatly divided up, everybody knows what everyone else is doing, and plunder does not equate to wealth. The Mongols were the last hurrah of the nomads, the last wave of horsemen emerging from central Eurasia, but since the fourteenth century the balance of strength has favored the sedentary since improving technology enables them to accumulate capital (and defenses) more than adequate to repel would-be marauders. (I'm suggesting here in contradiction to Turchin that it wasn't just the Russian frontierspeople's asabiya that helped them drive out the Mongols, it was improved farming and metalworking.) And yes, today there are certainly cultural fault lines revealed by the end of the Cold War, as Turchin alludes to by referencing Huntingdon, but there are no frontiers. This isn't Turchin's fault but the reader can't help but wonder if his theory still applies. If not, then I can continue predicting a dim future for his hypothetical Chechen Caliphate. 

The frontierless world we now occupy does raise the question of what mechanisms of competition, if any, apply to governments. Nation states are effectively a cartel, enabled by force and their monopoly on a finitely divisible resource (land) - this is a staple observation by many libertarian theorists who point out that such arrangements are much less voluntary than most of us would like. That I know of, this problem has not been discussed in the context of a recently frontierless world where there is no "correction" mechanism for weak, badly run states. Before, complacent despots at least feared finding themselves facing an invading horde that overran their badly-guarded borders. I don't favor a return to the system of state-level survival of the fittest - and in any event an enduring state is not necessary a pleasant one to live in - but one can't help but worry that a fickle and irrational electorate is enough to keep a large and complex modern state running well, when said modern state has, literally, nothing to worry about in terms of its continuing survival.  No barbarians are coming, and it can keep muddling along.

One of Turchin's objections to other theories of history and economics is the failure of rationally self-interested models to explain the mass behavior of humans with respect to their meta-ethnic groupings.  His claim is that asabiya would not exist in a rationally optimizing animal. This is flawed on several levels.

The first flaw is that pre-committed solidarity is actually a very powerful and therefore rational strategy - just not over "single-round games" for a single individual. In fact we see it in organisms besides humans (e.g. ants). These strategies rely on a) a pre-committed signal that one is a member of a group and b) behaving differently toward members of said group. (Richard Dawkins referred to this in hypothetical genetic terms as the green beard gene.) The emphasis on pre-commitment in humans is that our behavior is much more plastic. An ant is pre-programmed to attack those with a different scent; humans have more degrees of freedom. Consequently we rely on early life experiences (family and ethnic identification) that we can't choose to opt out of, because there's an emotional response built into us that remains with us for life, like it or not. Individually and in single round games, this behavior seems irrational, but in the aggregate and over indefinite rounds (in the massively multiplayer game of real-life) it's very powerful. Indeed a lot of human behavior that seems irrational in isolation is actually pre-programmed game-theoretic heuristics that make sense over multiple rounds; e.g. that we are willing to punish wrongdoers even when such punishment costs us and doesn't reverse the crime. (The point? Stop future wrongdoers!) The more generalized solution for human behavior is that we maximize utility, which is partly calculated with money, and partly with feeling different degrees of good or bad about helping or hurting human beings that we have varying connections to.

The second flaw is that it's difficult to imagine the motivations of the peasants that Turchin describes as "patriotic" in any sense that we would recognize.  Patriotism is a relatively recent development. A Chinese emperor didn't much care if peasants were patriotic Chinese - and why would an absolute monarch, as most rulers were throughout history, give a damn? What would the peasants do about it?  There are also considerable direct costs to not risking one's life to go to war - censure from neighbors (or outright violence), plus throughout history people were making these decisions based on poor information, plus in the desperate poverty that existed throughout most of history, the chance at plunder or a few dollars from a royal coffer may have been a better bet than staying at home barely subsisting on the feudal lord's land.  All this is to say that the medieval Russian farmers that he focuses on never had a choice to say, "I could peacefully trade with other Russian farmers, or Mongol merchants. I shall choose the Russian farmers out of ethnic solidarity." It was economic reality that drove them. Since Turchin is more realistic about the economic forces that drove change in thirteenth-fourteenth century Europe (climate shfit and plague), you would think he would be less prone to romanticizing. That said, culture does matter - it profoundly affects institutions and delay of gratification, and therefore wealth and happiness. The two options on this question, both unpalatable to the politically correct, are that either culture is pointless noise and mere window-dressing, or it has an impact on material well-being. Russian peasants chose to accumulate agricultural wealth and defend their land while Mongols continued raiding. The farmers' successor state today is much bigger and more successful than the Mongols'. An accumulation of coincidences, or a real trend?

A final shortcoming is that Turchin makes a common mistake by correlating two high-level characteristics of societies by cherry-picking; in this case, that the more central a government a region or state had over time, the less overlapping levels of status and trust there will be and therefore the worse the institutions and economic situation will be. (There are other lines of thought converging on the conclusion that multiple overlapping status hierarchies are good.) Talk about cherry-picking; Turchin refers to a specific book about Italy for this argument. True, people have applied similar arguments to the success of Japan and England (multiple feudal states kept decentralized and weak by a remote-from-the-people monarch intentionally keeping the mid-level nobility weak), and to China vs. Europe - yet is England really that much better off than France as a result? As a "dissident" libertarian, I frequently point out that for markets to work you need some commons, and for that you need some centralized functioning government. Otherwise, Nepal and Somalia would be veritable Galt's Gulch-paradises of innovation and venture capital, and guess what? They're not. What about far eastern Siberia for that matter? People who make the kind of argument Turchin is making (about anything, not just economics) usually respond by saying that you need "just the right amount" of centralization. Which is to say, the amount that England and Japan and Northern Italy have is the right amount because they're all successful, except a) that turns it into a completely atheoretic, circular, empirical argument and b) their centralization is in fact all different.

A final observation: despite taking issue elsewhere, I was very happy to see a historian observing that we should be circumspect in our weighting of Machiavelli's the Prince as either an accurate historical document or an effective political manual, since Machiavelli was after all writing it in exile. Why should we listen to this guy? He was a screw-up!

The EAC: East African Community

Burundi, Rwanda, Tanzania, Kenya, and Uganda are turning into a single-visa zone for foreigners. They're already in the process of eliminating work visa requirements for each others' citizens, and ultimately they're aiming for a common currency. This will hopefully stabilize these countries economically and politically (less ethnic violence when all five governments have an interest in each other, and much less probable bad unilateral economic policy decisions.) We'll also have data on another experiment in political and economic cooperation. There is already a West and Central African Franc, although I'm not aware of any studies showing whether this has been good or bad for that currency zone.

Nationalization could be seen as the opposite of this process, in that it makes economics and the security of assets in the country less predictable. It would be interesting to compare the fortunes of eager-to-nationalize states like Zimbabwe, Bolivia, and Venezuela, based on Peruvian economist Hernando de Soto's argument that economic take-off of developing economies is based on transparent property-ownership conventions that encourage value creation. Are there studies correlating nationalization with foreign capital inflows and growth rates? Granted, Zimbabwe's ruination was also caused by a monetary approach that should have gone out with the Weimar government, but looking across all nations, we could still see an effect. You could even make the argument that internal nationalization would have the same effect on private capital within the country. Late-in-the-game price controls on new pharmaceuticals are the classic game theory example of sunk-cost gotcha's.

No doubt these governments and many progressives in the developed world would argue that they have the right to nationalize assets in their country, but unfortunately, in the world of trade and investment this is beside the point. If your assets aren't safe in a certain country, you're not going to put money there. Maybe these governments (or more accurately, their insulated leaders) are willing to endure the pain of international investment pull-back in order to achieve independence from foreign influence, but there is a large trade-off between absolute independence and growth rate requiring foreign capital. How much is absolute economic independence worth?

Productivity Paradox?

An article by Shaun Hendy about New Zealand's poor recent productivity growth points to a paper by McCann. As trade barriers with New Zealand's larger continental neighbor have fallen, New Zealand's productivity growth has been anemic. This is the productivity paradox. Hendy points out that the same thing is going on between Australian states, and links to a paper on the poor performance of South Australia's economy relative to the rest of the country. What's going on?

Hendy argues that as barriers to movement of capital and talent fall, capital and talent will tend to enucleate around the same locations and they will concentrate geographically, in what could be considered a big coordination game. How many Australians are moving to New Zealand to further their careers or find investors? How many people from Sydney are moving to Adelaide?

It's worth asking this same question for the United States. Anecdotally, I have known this recession to have forced multiple families to move to the coasts, and if this is representative, it's going to accelerate an internal U.S. productivity paradox. While Richard Florida's vision of vibrant, tolerant, creative-class-oriented cities is certainly a positive one, not everybody can be San Francisco and Seattle - and we may see a runaway productivity paradox within the U.S. The only productivity growth map I could find (below) does not show a clear separation by region, but it predates this recession.

From Steve Cochrane at Economy.com