On your own, you can do whatever you want – it’s your own little box.
If there are two of you, you take different corners. Standing diagonally across from each other creates the greatest distance.
When a third person enters, you will unconsciously form a triangle (breaking the analogy that some have made with dots on a dice). And when there is a fourth person it’s a square, with someone in every corner. A fifth person is probably going to have to stand in the middle.
Now we are in uncharted territory. New entrants to the lift will need to size up the situation when the doors slide open and then act decisively. Once in, for most people the protocol is simple – look down, or examine your phone.
Most of the political economy and development literature seems to focus on national formal institutions, like courts and constitutions and executive constraints. Obviously important stuff. But less talked about are informal institutions like norms of acceptable behavior, how those come about and get enforced, and how (perhaps more importantly) they get internalized–meaning we punish ourselves if we violate them, rather than fearing what others will do. Internalization is handy because it means society doesn’t have sanction people in costly ways.
My posts on the subject (an ongoing series, mostly mini-case studies) can be found starting here.
One phrase I kept hearing over and over last week in coverage of Egypt’s election results is that the two frontrunners were each “worse than the other.” The two top finishers, Mohamed Morsi and and Ahmed Shafiq, received a combined total of 48.5 percent of the votes. Overall turnout was estimated at about 43 percent. Some commentators took this as bad news, arguing that a mere one-fifth of the eligible voting population had decided which candidates would make it to the run-off round. I take a more neutral view, and see the election results as an illustration of two well-known paradoxes of democracy.
The first paradox, named after the Marquis de Condorcet, is that in an election with three or more candidates, it may be impossible to select a single candidate that satisfies majority preferences. The canonical example is with three candidates and three voters, but let’s use the five candidates who received more than 10 percent of Egyptian votes (of the other eight candidates, none received more than 1.1 percent). Of these five, two will continue to the next round.
Suppose the Egyptian electorate voted according to its true preferences (i.e. not strategically). For simplicity’s sake, let’s round the votes of the top five candidates to 25, 25, 20, 15, and 15 percent. We know that Morsi’s and Shafiq’s supporters each consider their own candidate to be better than the other. Now suppose that the remaining 50 percent are evenly split between the two candidates (they each prefer their own candidate first, but half think Morsi is better than Shafiq and half think the opposite). This means that 50 percent of voters think that Shafiq is the better choice between the two, and the other half think Morsi is better than Shafiq.
We also do not know how much better the voters rank their preferred candidate. It could be that everyone who likes Shafiq best likes Morsi the least, rather than second best. There is no way to accurately deduce the overall preferences of Egyptian’s from the initial results without making strong assumptions like we did here. This is why the runoff is required. But this example does explain how so many people can be unhappy with the results of a democratic election.
The second paradox of voting is known as Arrow’s impossibility theorem. This theorem–one of the few great contributions of game theory to modern social science–says that no election system can accurately translate ranked preferences over three or more candidates into an overall ranking for the community while maintaining three criteria. These three are:
If every voter prefers alternative X over alternative Y, then the group prefers X over Y.
If every voter’s preference between X and Y remains unchanged, then the group’s preference between X and Y will also remain unchanged (even if voters’ preferences between other pairs like X and Z, Y and Z, or Z and W change).
There is no “dictator”: no single voter possesses the power to always determine the group’s preference.
For the first round of elections, the second criterion is most important. Changing the amount by which Shafiq or Morsi is preferred could have changed the results. In the runoff elections, the third principle will be violated. Once you are down to only two candidates, majority rule takes over. Egyptians got rid of one dictator, Mubarak, but exchanged him for the median voter. We will see how satisfied they are with that decision later this month.
From a 2002 paper by Eric Friedman and David Parkes entitled, “Pricing WiFi at Starbucks – Issues in Online Mechanism Design”:
We consider the problem of designing mechanisms for online problems in which agents arrive over time and the mechanism is unaware of the agent until the agent announces her arrival. Problems of this sort are becoming extremely common particularly in a wide variety of problems involving wireless networking.
We show how the standard results of mechanism design can be modified to apply to this setting, provide conditions under which efficient and incentive compatible mechanisms exist and analyze several important online models including wireless networks and web serving.
The authors basically boil the problem of socially optimal wifi pricing down to four variables: users’ valuations of wifi, the capacity of the wifi network, and the number of potential users at a given time (which is a ratio of the arrival and departure rates). If game theory does not interest you, you can stop reading now.
It’s an interesting paper and very readable for someone who was introduced to mechanism design only recently. However, there are three weak points in the paper. The first is their (lack of a) proof for Theorem 1. The theorem states that,
For both dominant-strategy and Bayesian equilibrium, if a social choice function can be implemented online, then it can be truthfully implemented by a direct-revelation online mechanism. (p. 7)
As proof, they cite the standard proof of the Revelation Principle in the Mas-Colell et. al. (1995) textbook. Given that this result is so important to their paper, it seems that they should have spent more time explaining it. The way it is, they rely on an offline result for an online mechanism, and are thus assuming–rather than demonstrating–that there is no qualitative difference. Note that I am not arguing a difference exists, but that it should be shown to exist or not.
The second weak point is closely parallel to the first, and can be found in Theorem 2:
An online VCG mechanism is strategyproof if the online choice rule is offline optimal, in the sense that it maximizes the utilitarian value. (p. 8)
Again, they declare that the proof “is analogous to standard proofs for the strategyproofness of VCG mechanisms” and move on.
Third and finally, from a more general economic standpoint it is troubling that they ignore the fact that coffee and wifi may be complementary goods. More specifically, they state that “all customers depart at rate μ, not only the ones in service [of the wifi].” (p. 12) This is problematic because it seems that the purpose of café wifi in the first place is at least partially to get customers to stay longer and purchase more coffee.
Overall, despite these weak points the paper does a good job of showing that optimal pricing is a computationally hard problem. They suggest that a simple mechanism would be a series of (m+1)st price auctions, where m is the wifi capacity, but that this is not implementable when users are uninterruptible (even in the non-strict sense where interrupting their wifi means they might throw hot coffee on you). This explains why the optimal mechanism is not observed in the real world, which is one of the major problems for mechanism design as a field. A good paper for beginners like myself; comments welcome.
Thanks to Josh Cutler and Shahryar Minhas for their comments on a draft of this post.
EJ Marey's French Train Schedule, c. 1880 in Tufte (2001) via Marlena Compton
Like many other universities, Duke has a lack of parking in close proximity to its class buildings, and so it operates a free shuttle service to take students and employees from remote lots to their desired locations. Today I parked my car as the shuttle was pulling up to the lot. The driver did not wait. It would have been easy for me to selfishly have become angry, but I didn’t. Why not? Certainly it is not because I am an unusually patient or moral person. It is because I appreciate institutional constraints.
What was the institutional constraint in this case? The bus schedule: the driver was not refusing to pick me up, he was sticking to his schedule. He went on to pick up other people who were already waiting at their stops, rather than making them wait an extra minute for me. If there was even one person waiting, this balanced out on net. If there was more than one person waiting, everyone benefited from him leaving me behind. Everyone? Even me? Yes, because I knew that he would be back less than 10 minutes later, and so I was able to wait. Everyone’s expectations were clear, and everything worked out fine.
How can we generalize this to other cases? Through the development of a theory about institutions. In general, I would suggest that an institution does the following:
Clarifies priorities. The priorities of a bus schedule are regular and timely service. By making these explicit and clear, both the bus driver and I knew what to expect from our interaction. He did not have to feel guilty about leaving me behind, because he would be back in a few minutes and there were other stops on other routes that I could go to instead. I did not have to get frustrated, because I knew those same two facts.
Delineates roles. The bus schedule makes it clear who is doing what and when. It is not my job to drive the bus, but to be on time. This clear delineation of roles made it clear who or what was responsible for each contingency. If I miss the bus, it is my fault. If I make it onto the bus, it is because I was on time. But if I make the bus late by causing it to wait for me, other passengers will blame the bus driver and not me. The delineation of roles then, helps with accountability and fairness.
Depersonalizes interactions. You may have noticed that I keep using the phrase “bus driver.” Is it rude of me not to know his name? I don’t think so. I encounter at least a dozen different bus drivers a week, so while it might be polite of me to say “please” and “thank you” (as I do), it requires a lot of memory to get to know all of their names. They know this, and do not expect it. No matter which driver it is, though, I expect to get to my destination on schedule and they expect a modicum of politeness. They neither get certain people there faster due to VIP status, nor refuse service to individuals because they are, say, preoccupied Ph.D. students.
One important caveat to the last suggestion is that the same person performing a number of different roles over time can re-personalize them. If the bus driver does only that, I do not get mad at him for sticking to the schedule. But if I encounter him next week as a server at a restaurant, and next year as a student, my attitude toward him will depend more on the quality of my interactions with him than the institutional constraints. That is, I am more likely to get mad at him if he is rude to me every time than to attribute his rudeness to an institutional constraint. This is how Robert Gates went from being a virtually anonymous CIA employee to a very recognizable, political personality by the time he retired.
How general and how valid do my thoughts about institutions seem to you?
In the previous post I gave a very simple–not to say uncontroversial or entirely accurate–definition of an institution: an actor who codifies constraints upon his/her/its own behavior. This post provides two related examples of what I mean, taken from games.
One clear example of an actor placing explicit and clear constraints upon himself is the dealer at a blackjack table. While the rules can vary, the basic version is that the dealer has two cards, one of which is face down. Of the many potential variations, one simple example to consider is whether or not the dealer hits on 17. According to Wikipedia, a dealer who hits on 17 decreases the house edge by about 0.2 percent. Over the long run this is not a trivial sum when you consider the average revenue of a casino. Thus, it is more common for the dealer to stand when his cards total 17.
Either way, though, the dealer’s rules are literally written on the table. His behavior is codified there for all players to see. This can influence there calculus not only of whether to enter the game, but of what decisions to make once they are playing; a dealer who stands on 17 encourages players to make riskier bets when their own total is less than 17. It is this marginal increase in risk that leads to the casino’s higher profit under the stand-on-17 rule. The key thing to notice here is that the institutions restraints on its own behavior affect the behavior of other players.
Don’t these self imposed rules make it easier to cheat? Of course they do. Even when cheating is difficult (some casinos use as many as eight decks of cards to discourage counting), there will be some people who are smart enough, bold enough, and greedy enough to try it. And none of that is to criticize them–I’m not even entirely comfortable calling it cheating, but even so it is entirely rational.
This is where the second element of institutions, at least political ones, comes into play. Once they have imposed constraints on themselves they become predictable. In order to keep from getting screwed by people who find loopholes in the rules, casinos need some way to make cheating less likely. Enforcement often takes the form of brawny guys with brass knuckles. More generally, our government and others around the world maintain police and militaries to impose constraints on domestic and international actors who might try to take advantage of their predictability.
To simplify the casino example and connect it to government, let us consider a very basic and common game: Matching Pennies. Readers not familiar with game theory can find a treatment of it in Osborne (2004: 19-20, 111-112, 136-137) or at Wikipedia, where it is described as the two-strategy equivalent of Rock, Paper, Scissors. There are two players, which we will call Government and Citizen, two actions (heads or tails), and four possible outcomes. If they show the same side of their coins, the government takes the citizen’s penny. If they show different sides of their coins, the citizen takes the government’s penny. These outcomes are shown in the following table:
The Matching Pennies Game
Predictability in this game means losing. Whichever player I am, if I know what you are going to do I get to choose the action that will give me the highest payoff. However, the game is typically played simultaneously. When it is played this way there is no best pure strategy in which you always do one thing or the other. As Osborne will tell you, the best thing to do is just flip your coin and let it land randomly on heads or tails. If both players do this and the coins are fair (50/50 odds of heads/tails), over the long run both will break even.
But say you have two types of governments: autocrats and populists. Autocrats really like the strong personality of Abraham Lincoln, so they always show the head of their coin. Populists feel more strongly about the e pluribus unum message on the tail side of the coin, and always show it. Either one is predictable, and a citizen or foreign government playing against the government can choose the outcome because the government is predictable. This is what I meant by an institution codifying its own behavior.
Say the government is an autocratic institution that always shows heads. How will it be able to continue pursuing its favored course of action and keep from losing its penny in every iteration of the game? It could change the rules, start randomizing its actions, or maybe just poke the other player in the eye. Empirically, we observe all of these outcomes in different scenarios. This extends the argument of the previous post in two ways:
A political institution is an actor who codifies constraints on their own actions in a way that reveals their preferences AND
contrains the other players’ actions in such a way as to lead to the institution’s preferred set of outcomes.
The Duke Political Science Department is not organized around the traditional disciplinary subfields of comparative, international relations, and American. While we do retain political theory, political economy, and political methodology, the three research areas above are re-arranged (with varying levels of correspondence) into security, peace, and conflict studies; political institutions; and political behavior. This has been a positive transformation both for the department and for myself as a new student, but I have had the lingering question, “isn’t it all about behavior?” (And I don’t mean this as a subfield chauvinist–my first field is SPC and the second is methods–but it may reflect my guidance by professors at my previous university.)
Last night in class with Guillermo Trejo we came one step closer to solving this conundrum. We put forth the tentative definition of an institution as “an actor who codifies constraints upon his/her/its own behavior.” This could apply to a state’s constitution, a subnational group’s manifesto, or the UN charter. It may need further refinement to avoid overgeneralization, but for now it seems helpful.
Today I came across this example, which helps point to the role of narrative, theology, and hermeneutics in defining religious institutions:
Goldman grew up in New York City as an orthodox Jew for whom religion was a central focal point of everyday life. He saw religion as a communal force and a public issue, and he has spent a career following those principles.
He teaches through tales. In his courses, he uses the great stories of the Bible and the Quran to illustrate the ways and beliefs of Christians, Jews and Muslims. His master’s thesis compared the story of King Solomon and the Queen of Sheba as it appears in the Quran and the Hebrew Bible.
(The Hebrew version is more about power and international relations, Goldman reports. The Quran presents a more strictly religious version of events.)
A religion’s stories — like Moses, and the Garden of Eden — are good teaching tools because they’re well-told and compelling, and thus, broadly influential, Goldman says.
“These texts govern behavior for many people,” he says. “So the way these stories are told influences behavior.“
I will leave it to readers to reflect upon and argue for or against any of the points I’m making here (either the definition of an institution as self-restrained behavior, or the role of religious rhetoric in defining its own institutions). It does seem to me, however, that Goldman’s interpretation of the Hebrew account of Solomon and Sheba as political and the Quranic account as religious is more reflective of the modern flexibility in interpreting Jewish and Christian testaments that does not yet extend–for various and sundry reasons–to the Quran. For more on this point, see the poorly-titled writings of Ibn Warraq.
On your own, you can do whatever you want – it’s your own little box.
accurate–definition of an institution: an actor who codifies constraints upon his/her/its own behavior. This post provides two related examples of what I mean, taken from games.
Matt Dickenson 