Ch. 12: Chance, Action, Innovation: Mechanisms of institutional change
Aaron Swartz
When Cortes visited Mexico in 1519, he described a society much like
any other: markets in the center of the city, a courthouse with a
dozen judges, police enforcing the use of honest weights and measures,
chiefs, beggars, temples, and orderly government. Talcott Parsons
suggested that such basic societal elements naturally evolved many
times because of their utility, the same way many different species
evolved so they could see. Others focus on the differences. In modern
Mexico, villages with high homicide rates sit side-by-side with
villages with low homicide rates, with the nonviolent villages having
traditions of mediation and rotation of village offices.
So how do institutions change? Darwinians argue that institutions can
shift through a form of "genetic drift" -- a combination of people all
randomly happen to do something individually suboptimal at the same
time and the whole society shifts into another equilibrium. Marxists
point to collective action from the people who get the short end of
existing conventions to shift society into a more favorable
equilibrium for themselves. Yet both theories predict things get
better over time, while in fact inefficient, inegalitarian
institutions seem quite stable.
OK, so let's start with a simple game. We have two teams, Capital and
Labor. Each round, each capitalist pairs up with a laborer and both
propose a strategy. There are two possible strategies: equal, in which
the two get equal benefits, and unequal, in which capital gets more
than labor. If they both propose the same strategy, the benefits are
so divided, otherwise neither of them gets anything. Each round they
update their strategy based on what worked last time.
The result is, of course, two possible equilibria, equal and unequal,
and where you end up depends simply on whichever you start out closest
to. But now introduce some randomness such that players occasionally
take actions that are suboptimal for them. With sufficient randomness,
the society can move all over the board. ("The population is
perpetually in motion, or at least susceptible to movement, and its
state is path dependent: where it was in the recent past influences
where it will most likely be at any moment. History matters, and it
never ends.")
In general, though, the two main conventions are _robust_: easy to get
to, hard to leave. It takes a great deal of randomness to dislodge a
society from a stable convention but not much to end up there.
Now imagine that instead the conventions simply being about who gets
more of the spoils, one convention is a Pareto improvement: both
groups are better off by it. We can show that Pareto-inefficient
conventions are not robust. In fact, we can show that the stable state
is the one that maximizes the relative payoffs of the group with the
lowest relative payoff -- a kind of egalitarian inevitability!
Let you get 3 and they get 2 be represented as {3,2}. Efficient
contracts do well because if you're choosing between {3,2} and {3,1},
you should offer the first since it doesn't hurt you and is more
likely to be accepted. Egalitarian contracts do well because if most
people are offering {3,0} but even a couple are offering {3,1}, it
makes sense to hold out for the second offer. ("This is the
evolutionary game theorist's rendition of Marx's rhetoric about the
working class having 'nothing to lose but their chains.' Thus the
unequal convention becomes less persistent as it becomes more
unequal.")
[Speculation: And since the logical conclusion of that direction isn't
stable, the whole area is unstable? Anyone understand this?]
Such robust states aren't strictly necessary, but they are very
likely, providing an analog to Parsons' "evolutionary universals."
Now let us bring in Marx by allowing our players to look forward as
well as backward -- thus the poor can envision that they will do
better if they stage a revolution. Also let us assume that in each
period some number are "organized" and can decide collectively whether
some other offer would be better for them. In the inegalitarian
condition, the laborers can use it to go "on strike" and all demand a
more equal contract, but the capitalists can do the same and go "on
strike" to prevent equal contracts. (Both sides can also use it to
prevent defectors.) If the populations are equal-sized, labor strikes
are more powerful than capital strikes and society moves quickly to
the efficient, egalitarian pole.
But as soon as the poor outnumber the rich, things become very
different. Since it's easier for the few rich to tip things, they can
ensure society stays at conventions that benefit them, even if those
conventions are very inefficient.
Let's make our model a little more realistic. Striking has two
effects: first, there is a cost on those who don't conform (the cost
of being a scab); second, there is the net joy of participating, which
is probably proportional to the potential gains (it's a lot more fun
to strike for a big raise than for a small one). Note that the unequal
equilibrium is even more sticky in such cases -- the purely random
players would at least rarely all decide to randomly defect, whereas
more rational strikers will never do so, seeing that the possible
benefits are outweighed by the probable costs. But technological
changes either making the egalitarian convention more efficient or
cultural changes making strikes more fun can change that.
This "is far from the first proposed marriage of Darwin and Marx."
Marx himself wote Engels that "Although it is developed in the crude
English style, this is the book [_The Origin of Species_] which
contains the basis in natural history for our viewpoint." In his
eulogy for Marx, Engels noted "Just as Darwin discovered the law of
evolution in organic nature, so Marx discovered the law of evolution
in human society."
They have shown us three reasons why inegalitarian, inefficient
institutions are durable. First, moderate inequality may not be enough
to motivate collective action by the poor. Second, it's easier for a
small group of capitalists to go on strike than a large group of
laborers. Third, extreme inequality provides a strong incentive for
the rich to fight (endure the costs of labor's strike) to preserve the
status quo. Thus, if inequality is small, the poor will not be
motivated to end it, while if inequality is big, the rich will be very
motivated to protect it.
Of course, all this evolution works over very long time periods --
probably much longer than the time periods experienced by real human
societies. But making the simulations more realistic can accelerate
the process. Small groups that coordinate internally, migration and
emulation, chance events, additional (closer) conventions, conformism,
and stable latent organizations can all accelerate the process.
Next week: What happens when we add in between-group effects?