Possible Insight

Most people believe humans evolved intelligence because using tools was an advantage.  However, I believe tool use was secondary.  Group cooperation was the primary advantage conferred by intelligence.  You see, cooperation is fundamentally difficult.

This insight coalesced when I was reading about Mark Satterthwaite, an economist at Northwestern’s Kellogg School of Management.  He’s famous for two important impossibility theorems: (1) the Myerson-Satterthwaite Theorem and (2) the Gibbard-Satterthwaite Theorem.

Informally, (1) says that there is no bargaining mechanism that can guarantee a buyer and seller will trade if there are potential gains from trade, while  (2) says that there is no voting mechanism for determining a single winner that can induce people to vote their true preferences.  In both cases, the reason for the impossibility is that people have incentives to hide their actual values to achieve a strategic advantage.

Add these to the Prisoner’s Dilemma and Arrow’s Impossibility Theorem on the list of fundamental barriers to cooperation (Holmstrom’s Theorem is another good one; it explains why you can’t get everyone in a firm to exert maxium effort).  By “fundamental”, I mean there is no general solution.  So the evolutionary process cannot just discover a mechanism that guarantees cooperation when it is efficient.  There will always be the opportunity for individuals to subvert the cooperative process to promote themselves, thus creating selection pressure against the cooperation mechanism.

(Note that there is a hack: make sure each individual has the same genes.  This is how multicellular and hive organisms get around the problem.  But the existence of cancer in the former case and the reduced genetic diversity in the latter case make them limited solutions.)

To achieve extensive cooperation in large groups, individuals need the ability to model the strategic situation, estimate the payoffs to various group members, and continuously assess what strategies other members may be playing. On top of that, there’s an arms race between deceiving and detecting deception.  It’s the old, “I know that you know that I know…” schtick.  The smarter you are, the further you can compute this series.

Bottom line: the impossibility theorems mean the only way to achieve cooperation is to have the machinery in place to make detailed case-by-case determinations.  We’ve talked about the Dunbar Number before: the maximum size of primate groups is determined in large part by a species’ average neocortical volume.   I claim you need to be smarter to process more complex strategic configurations and maintain models of more individuals’ goals.

If I’m right, there are two interesting implications.  First, politics will be with us forever.  No magical technology or philosophical enlightenment will eliminate it.  Second, if we ever encounter intelligent aliens, they’ll have politics too.  Nothing else about them may be recognizable, but they’ll have analogs of haggling over price and building political coalitions.

My wife has an Art History degree from Princeton.  However, she often has excellent insights into economics. So I think either (a) one of the reasons she fell in love with me is that she’s a latent economics geek or (b) she loves me so much that she actually pays enough attention to my economics ramblings that some of it rubs off.

About a year ago, she came up with a solution to the “employee union problem”.  With the recent public employee union showdown in Wisconsin, I thought I should share it with you.  What’s particularly ironic is that she’s out in front with some pretty serious economists in thinking about issue.  For example, check out this Wall Street Journal article by Robert Barro, a Harvard professor and author of my undergraduate macroeconomics textbook.

Most people think that unions are simply workers exercising their rights to band together.  Actually, this isn’t the case.  As Barro points out, unions are a monopolies that are specially exempted from anti-trust law. Moreover, the government actually enforces their monopolies.

The key law here is the National Labor Relations Act (NLRA) as administered by the National Labor Relations Board (NLRB).  The NLRB protect the rights of employees to collectively “improve working terms and conditions” regardless of whether they are part of a union.  In addition, the NLRB can force all employees at a private firm to join a union, or at least pay union dues.  This process is called “certification”.  Once a union is certified, it has the presumptive right to negotiate on behalf of all employees at the firm, and collect dues from them, for one year.

After that, it is possible to “decertify” a union.  However, the union usually has enough time to consolidate its power and then use it to keep the rank and file in line.  When you have monopoly power, you have the opportunity to abuse it.  In fact, there’s a whole government agency devoted to investigating such abuses: the Office of Labor-Management Standards.  UnionFacts.com has helpfully collated the related crime statistics for 2001-2005.  Assuming that only a fraction of abusive behavior faces prosecution, these statistics are pretty sobering.

Of course, federal, state, and local government employees are exempt from the NLRA.  You might think this exemption is a good thing for limiting union power.  However, what it means in practice is that each level of government is free to offer special treatment to their employee’s unions without oversight from the NLRB.  As you can imagine, the politicians and public employee union leaders get nice and cozy.  The politicians give the unions a sweet deal and the unions give politicians their political support.  Everybody wins.  Except ordinary citizens.

Personally, my solution would have been to completely eliminate the government-enforced monopoly of unions.  However, I admit this blanket approach could swing power too far towards management in some industries.  My wife’s solution is better.  She says the unions can get monopolies, but only for a set period of time.  Say 3 or 5 years.

From an economics standpoint, this approach is really insightful.  First, it removes union leaders’ incentive to form a union just to accumulate power.  It will all go away pretty soon.  Second, it prevents originally well meaning union leaders from getting corrupted over time.  Pretty soon they’re ordinary workers again.  Third, it does provide help to those workers who feel management is truly abusing them.  They can form a union and get better treatment.  When the union’s existence terminates, they can still bargain collectively, just not exclusively.  If management tries to screw them again, they will have the example of how to work together.  An economist would call this “moving to a better equilibrium”.

I’ll admit this solution isn’t perfect.  Some management abuses will slip through the cracks.  But I’m pretty confident they’ll be less extensive than the current union abuses.  It’s also probably better than my original thought of banning unions altogether.  And there’s some small chance my wife’s approach would actually be politically feasible.  Nice work, Jane!

Previously, I used my Production Function Space (PFS) hypothesis to illuminate the differences between startups, small businesses, and large companies. Now, I’d like to turn my attention to the implications of PFS on a firm’s demand for labor.

I don’t know about you, but a lot of hiring behavior baffles me. I see companies that appear clearly and consistently understaffed or overstaffed, relative to demand for their offerings.  Then the hiring process itself is strange.  I’ve consistently seen companies burn all kinds of energy and incur all sorts of angst just to come up with a job description. Shouldn’t it be obvious what isn’t getting done? Why do they delegate apparently core aspects of production to contractors? And despite decades of evidence, why do firms insist on using selection procedures like unstructured interviews that aren’t very effective.  Also, there’s the mystery of why incentive-based pay doesn’t work in general despite plenty of evidence that humans respond well to incentives in other circumstances.

Can the PFS hypothesis shed any light?  I think so.  But this hypothesis implies that a firm’s labor decisions are substantially more complicated than we thought, so don’t expect a nice “just-so” story.

I see three big implications if many employees benefit the firm primarily through searching PFS instead of producing goods and services:

1. Uncertainty. The payoffs from searching PFS are uncertain.  In many cases, they’re really uncertain. You could end up with a curious but unmarketable adhesive or you could end up with the bestselling Post-it notes. You could end up with just another search engine or you could combine it with AdWords and end up with Google. A search over a given region is essentially a call option on implementing discovered production functions.
2. Specificity. Economists refer to an asset as “specific” if its usefulness is limited primarily to a certain situation. The classic example is the railroad track leading up to the mouth of a coal mine. I think employees searching PFS are fairly specific. Each firm’s ability to exploit production functions is rather unique. Google and Microsoft can’t do exactly the same things. Moreover, each firm’s strategy for exploring PFS is different. So it takes time for an employee to “tune” himself to searching PFS for a particular firm. All other things being equal, an employee with 3 months on the job is not as effective at searching PFS as an employee with 3 years. And an employee that leaves firm A will not be as effective at searching PFS for firm B for a significant period of time. Think of specificity in searching PFS as a fancy way of justifying the concept of “corporate culture”.
3. Network Effect. The number of people searching a given region seems to matter. I’m not at all sure if it’s a network effect, a threshold effect, or something else. But there seems to be a “critical mass” of people necessary to search a coherent region of PFS. You need a certain collection of skills to evaluate the economic feasibility of a production function. The larger a firm’s production footprint and the larger the search area, the greater the collective skill that is required.

Let’s start with hiring and firing decisions.  As you can see, firms face a really complex optimization problem when choosing how many people to employ and with what skills.  Suppose demand for a firm’s products suddenly declines.  What’s the optimal response?  Due to the the network effect, firing x% of the workforce reduces the ability to search PFS by more than x%.  Due to specificity, this reduction in capability will last much longer than the time it takes to rehire the same number of people.  Thus, waiting to see if the drop in demand is temporary or permanent provides substantial option value.  Of course, a small firm doesn’t have much cushion, so may have to lay off people anyway.

Thus I predict a sudden drop in demand will result in disproportionately low or significantly delayed layoffs, and the disproportion or delay will be positively correlated with firm size.  Moreover, firms will tend to concentrate layoffs among direct production workers to minimize the effect on searching PFS.  This tendency may explain why they delegate some apparently core functions.  Being able to flexibly adjust those direct costs preserves the ability to search PFS. This hypothesis implies that the more volatile the demand for a firms’ products, the more they will outsource direct production.

Conversely, what should a firm do if demand suddenly increases?  Based on the PFS hypothesis, I have three predictions: the firm will (1) delay hiring to see if the demand increase is sustained, (2) “over hire” relative to the size of the demand increase, and (3) hire a disproportionate number of people outside of core production.  The reason is simple, diversification.  Due to uncertainty, the best way for a firm to ensure its long-term survival is to have a large portfolio of ongoing PFS searches. Extra dollars should therefore be allocated to PFS searching labor rather than capital or direct production labor.  However, because a firm knows that it will be reluctant to fire in the future, it will initially be conservative in deciding to hire.

It seems like these predictions should be testable.  I wish I had a research assistant to go through the available data and crank through some econometric analysis.  I’m thinking the next step is to work through the implications of PFS searching on employee behavior.  Unless anyone has other thoughts.

One of the reasons I like my Production Function Space (PFS) hypothesis, is that it clarifies a lot of issues that have puzzled me for a while.  For example, as part of my work on seed-stage startup investing with RSCM, I have struggled with two questions: (1) what’s the difference between a startup  and a small business and (2) why do some large companies have initiatives like venture groups and startup incubators?

To answer these questions, I’m going to have to get a little mathematical.  Don’t worry.  No derivatives or integrals, but I need to introduce some notation to keep the story straight.  First, let’s define production footprint and search area.

A production footprint is the surface that encloses a set of points in PFS.*  If A is a production function for basketballs, B is a production function for baseballs, and C is a production function for golfballs, then pf(ABC) looks like this in two dimensions:

We can also talk about the production footprint of a company like Google.  pf(Google) is the surface that encloses all the production functions that Google currently uses.

A search area is the surface defined by extending the production footprint outward by the search radius.** Imagine that we wanted to see if making basketballs, baseballs, and golfballs would enable a company to make footballs using production function F, we might want to compute sa(r, ABC).  It looks like this in two dimensions:

We can also talk about search areas for a company.  sa(r,Google) is the surface enclosing all the points in PFS within r units of Google’s current production footprint.  If we write just sa(Google), we mean search area using Google’s actual search radius.

With these two basic concepts in place, I can now easily answer questions (1) and (2).  SU stands for startup, SB stands for small business, and LC stands for large company.  So what’s the difference between a startup and a small business?  Well, when they are founded, a startup doesn’t have any production footprint at all and a small business does.  To a first order, pf(SU)=0 and pf(SB)>0.  A startup doesn’t know with any precision how it’s going to make stuff.  A small business does.  Whether it’s a dry cleaner, law office, or liquor distributor, the founders know pretty precisely what they’re going to do and how they’re going to do it.  However, startups have a much larger search radius than small businesses.   r(SU)>r(SB).  Assuming that we can define a search area on a set of production functions the startup could currently implement (but hasn’t yet), I content that also sa(SU)>sa(SB).

This realization was an epiphany for me.  Even though the average person thinks of startups and small businesses as similar, they are actually polar opposites.  They may both have a few employees working in a small office, but one is widely focused on exploring a huge region of PFS while the other is narrowly focused on implementing  production functions within a tiny region of PFS.  I also realized that you need to evaluate two things in a startup: (a) its ability to search PFS and (b) the ability to implement a production function once it locates a promising region of PFS.  But the magnitude of impact for (a) is at least as big as (b) in the very early stages.

Now on to the issue of large companies.  The problem here is search costs.  Remember that, in three-dimensional space, volume increases as the cube of distance.  In PFS, volume increase as an exponent of distance equal to the dimensionality of PFS.  I posit that PFS is high-dimension, so this volume increases very quickly indeed.  Now try to visualize the production footprint and search area of a large company.  A large company has a lot of production functions in play so pf(LC) is large.  But sa(r,LC) increases exponentially from this large volume by a large exponent.

In three dimensions, imagine that pf(LC) is like a hot air balloon.  Extending just 10 feet out from the hot air balloon’s surface encompasses a huge volume of additional air.  But in high-dimension PFS, the effect is… well… exponentially greater.  So a large company has a problem.  On the one hand, increasing its search radius is enormously costly.  On the other hand, we know that Black Swan shifts in the fabric of PFS will occasionally render a huge volume dramatically less profitable, probably killing companies limited to that volume.  So it’s only a matter of time before sa(r,LC) is hit by one of these shifts, for any value of r.

Obviously, there will be some equilibrium value of r where the cost balances the risk, but that also implies there’s an equilibrium value for the expected time-to-live of large companies.  Yikes!  This explains why the average life expectancy of a Fortune 500 company is measured in decades. Another epiphany.

Internal venture groups and incubators represent a hack that attempts to circumvent this cold, hard calculation. The problem is that it’s difficult to explore a region of PFS without actually trying to implement a production function in that region.  Sure, paper analysis and small experiments buy you some visibility, but not very much.  Also, in most cases, you don’t get very good information from other firms on their explorations of PFS, unless you observe a massive success or failure, at which point it’s too late to do much about your position.  That’s why search costs are so darn high.  Enter corporate venture groups and startup incubators.

These initiatives require some capital investment by the large company.  But this investment is then multiplied by the monetary capital of other investors as well as the human capital of the entrepreneurs.  With careful management, a large company can get almost as much insight into explorations of PFS by these startups as it would from its own direct efforts.  Moreover, because startups are willing to explore PFS farther from existing production footprints, the large company actually gets better search coverage of PFS.

This framework answers a key question about such initiatives.  To what extent should corporate venture groups and startup incubators restrict the startups they back to those with a “strategic” fit”?  If you believe in my PFS hypothesis, the answer is close to zero or perhaps even less than zero (look for startups in areas outside your company’s area of expertise).  Otherwise, they’re biasing their searches to the region of PFS that’s close to the region they’re already searching.  It doesn’t increase the ability to survive that much.  As far as I know, Google is the only company that adopts this approach.  I think they’re right and now I think I understand why.  Epiphany number three.

* There are some mathematical details here that need to be fleshed out to define this surface.  But I don’t think they add to the discussion and my topology is really rusty.

** More mathematical details omitted.  The only important one is that the extension outward doesn’t have to be by a constant radius.  It can be a function of the point on the pf. In that case, r is a global scaling parameter for the function.