Developing Jen

A very simple game called the ultimatum game has become an excellent tool for studying social preferences (Guth, Schmittberger and Schwarze 1982, Thaler 1988).  In this game, one player (call him the “Proposer”) makes the first move and offers a split of a certain amount, say $10.  The second player (“Responder”) decides whether to accept or reject this split. If it is accepted, P and R get the proposed split. If rejected, then both get zero.  This game has been run in many, many countries and for stakes that range from a few dollars in the US to a few months of income in many countries. Yet the pattern of findings is relatively constant.  First, responders often reject unfair offers (i.e. away from 50-50 splits).  Second, proposers often make very fair offers, close to 50-50 or 60- 40.  Moreover, proposers’ fair offers are not just driven by fear of rejection. They tend to make offers larger than what a simple (risk-neutral) fear of rejection implies. The ultimatum game illustrates two facts about interpersonal preferences: (i) people care about others and are willing to give up resources to help others and (ii) people react negatively to perceived unfair behavior and are willing to give up resources to punish it.  The second fact illustrates part of the “dark side” of interpersonal preferences. In simple altruistic models, interpersonal preferences are only a good thing: having one person care in a positive way about another only makes it easier to deal with externalities and so on. The Responder’s possible “punishment” behavior shows, however, the way in which interpersonal preferences could potentially cause inefficiencies and conflicts.  

This possibility is clearest in a classic experiment by Messick and Sentis (1979).  They ask subjects to imagine they have completed a job with a partner. They are asked to decide what “fair” pay for their work is.  They divide the subjects into two groups however. One group is told to imagine that they had worked 7 hours on the task while the partner had worked 10. The other group is told to imagine that they had worked 10 hours while the partner had worked 7. Both groups were told that the person who had worked 7 hours had been paid $25 and were asked what the 10 hour person should be paid.  Those who were told that they had worked 7 hours (and paid $25) tended to feel that the 10-hour subject should be paid $30.29.  In contrast, those who were told that they had worked 10 hours felt they should be paid $35.24.  The source of the bias can be seen in the bimodality of the distribution of perceived “fair” wages. One mode was at equal pay($25 for both) while the other mode was at equal hourly wage (so the 10 hour worker gets paid approximately $35.70).  Interestingly, the difference between the two treatments was mainly in the proportion at each mode.  Those who had worked 7 hours showed more subjects at the equal pay level mode while those who had been told they’d worked 10 hours showed more subjects at the equal hourly pay mode.  In other words, both groups recognized two compelling norms: equal pay for equal work and equal pay for equal output.  Yet their roles determined (in part) which norm they picked.  Such conflicts could easily arise even if there’s disagreement about measuring input levels (which often are not fully observed). More broadly, when there is not universal agreement about what is fair division, individuals trying to act “fair” may produce even more conflict than individuals acting in a self-interested manner.

Let us return to the case of teacher absenteeism. The PROBE report surveyed teachers extensively in many areas of India and noted high absenteeism levels. Its in-depth interviews are illuminating about their attitudes and highlights how teachers often feel unmotivated. Some of this discouragement can be viewed as a perceived failure of reciprocity. As noted earlier, individuals strongly adhere to the norm of reciprocity.  Failures of reciprocity (or perceived failures) can result in punitive or self-interested behavior in response.  Teachers may feel a strong social preference early on and be motivated to teach and give much more than they need to.  After all, from a pure self- interest motive, they know they can get away with very little teaching.  Yet they may be initially motivated to do more, to come to school, to struggle with tougher students and so on.  They may view these contributions as a "gift" in large part due to the initial framing of the job (as a “plum job, with good salaries, secure employment and plenty of other time for other activities”).  Thus, a young teacher may think, “I am giving a lot to the school”. As with any giving, however, the teacher may expect strong reciprocity and see (perhaps in a self-interested way) many outcomes as a lack of reciprocity. For example, the Probe report notes that many schools have terrible infrastructure; accordingly, teachers may feel that the government is not reciprocating their "gifts". This may be especially exaggerated by the transfer system in India, which moves teachers to various areas, disrupting the lives of teachers. Thus both the benign neglect of schooling and the active transfers could easily drive teachers to feel that the government does not reciprocate their efforts. They may also come to feel similarly vis a vis parents, who they may feel do not care about their children's education. 

Even an initially motivated teacher may very quickly feel justified in their growing apathy.  They gave it their best and think that their efforts were not reciprocated. Are these inferences justified?  Perhaps not. As in the Messick and Sentis study teachers may very well be making such inferences in a self-interested way. The failure of the context may very well be in it allowing teachers to make such biased attributions of fairness. Alternatively, teachers may very well be justified in these attributions. We simply cannot tell. 

At a deeper level, these studies of fairness suggest that the problem of corruption may have interesting social preference wrinkles.  People may be more willing to avoid taxes if they feel they are not "fair".  This judgment of unfairness could be the result of getting very few government services or having to bribe corrupt middleman in order to procure government services. Economic models of corruption, by assuming blatant self-interest, ignore the tension corruption generates. If most people, as the evidence suggests, have strong social preferences, then corrupt acts will require self-justification.  This resembles anecdotal evidence on corruption experiences. It is very rare that an official simply asks for a bribe. The request is often couched with an explanation for the reason for the bribe. Even though the bribe is clearly a violation of the law, there is usually a story that serves to justify it in the context of the law.  For example, a customs officer may point to improper packaging as a reason for an extra payment. These kinds of insights may one day help to better understand the nature of corruption and anti-corruption policies.  

Of concern, in the two years since Tanzania’s backbone policy was first put in place, the market has shifted rapidly.  Currently the private sector is eager to invest in national backbone infrastructure on a cost effective basis, which implies public investment at the scale of the existing policy is no longer required.  Additionally, the current policy carries substantial risks that remain unmitigated in implementation to date: 1) the government’s ownership of a single national backbone may bias policy decisions; 2) a public monopoly limits incentives for efficiency and current costs for the national backbone are higher than industry norm; 3) the government’s participation in the retail market through the government-owned Tanzania Telecommunications Company Limited (TTCL) compromises neutral management of the backbone and exacerbates the concerns outlined above.  As a consequence of these issues, the current policy risks undermining the very objectives it was formulated to achieve.

Recommendations for policy improvements must consider the substantial government investment to date.  Furthermore, the government must take care to foster private investment and competition without creating market inefficiencies.  By selling conditional indefeasible rights of use (IRUs) and dark fiber across the government-owned national backbone, Tanzania can enable competition in the wholesale broadband market alongside a consolidated infrastructure.  Offering IRUs and dark fiber will shift capital away from duplicative investments in new networks and allow the government to quickly recover a substantial portion of its investment.  This approach will achieve scale in the national backbone and foster competition in the last mile, updating the current policy in light of the changes that have occurred in the market since its inception.  Furthermore, these policies do not carry the significant implementation risks currently observed.

This is of course impossible at the macro level.  We obviously can’t randomly assign policies to countries and even if we could there aren’t enough of them to be sure of our conclusions.  Instead at the macro level we use creative econometrics to do our best to isolate the relationship between policies and outcomes.  And even then we run into issues of reverse causality.  This gives macro level studies an issue with internal validity.  It’s messy econometrics.

Opponents of randomized experiments argue that their results can’t be generalized to a broader context (i.e. they aren’t externally valid).  They may be useful for academic research, but they are of limited policy value.  Messy econometrics, on the other hand, incorporates many places across many time periods, which makes a stronger case for their results being broadly applicable.

Do you see where I’m going with this?  Could it be that development economics has to live with a minimum level of uncertainty with its experimental outcomes, trading off internal and external validity?

Dani Rodrik’s “We Shall Experiment, But Shall We Learn?” elaborates (and inspired this post).  I highly recommend it, he even uses one of our favorite examples, bed nets for malaria, to explain the merits of various experimental approaches.  The real aha moment hit me on page 4, talking about Cohen and Dupas' random control trial:

“But do the results extend to other settings in Africa as well? One can certainly make the case that it does, but the arguments one would need to deploy are perforce informal ones and they are convincing to varying degrees.  In fact such arguments are not too different in kind from those that researchers may offer in defense of a set of instrumental variables employed in a conventional econometrics study with weaker internal validity.” 

I knew my physics degree would come in handy some day.