“Experimenting with Social Norms” in Small-Scale Societies

Social norms, institutions, and economic development. (A companion post to “Where do pro-social institutions come from?”)

Although the main focus of “cultural evolution” research seems to be on the big picture of how humans, culturally, went from the “Lascaux caves to Goldman Sachs”, it may have some relevance for explaining the wealth and poverty of nations today. And behavioural experiments are a crucial part of the cultural-evolutionists’ ambition to discover the psychological primitives behind “human ultra-sociality” and social institutions.

The vast majority of such experiments to date have been conducted at university labs with students in developed countries, with the results having a pronounced WEIRD bias. In order to take a close, micro-level look at the global diversity of cooperative norms, researchers led by Joe Henrich have been conducting experiments in many small-scale societies and comparing them with reference populations in developed countries.

Ensminger & Henrich 2014, which replicates and extends findings from Henrich et al. 2004, document the variation in ‘fairness’ norms within diverse societies around the world. As Ensminger & Henrich put it,

“The sample of societies from which we draw the data for this project is virtually unique in that it runs the gamut from almost pure hunter-gatherers (absent most traces of modern development and material possessions) through numerous horticultural and nomadic herding societies (some equally remote from modern markets), to cash-cropping farmers, urban African workers [Accra], and small-town residents in rural America.”

The three games conducted in the field with substantial monetary stakes were the Dictator Game (DG), the Ultimatum Game (UG), and the Third Party Punishment Game (TPG). All of these simulate, in controlled settings, bargaining between strangers intended to tease out and measure the existence of ‘fairness’ norms.

In DG, the ‘proposer’ is given a certain amount of money and is permitted to choose whether to keep all for himself, or share some part of it with a passive second party. In UG, the second party or ‘responder’ can engage in costly punishment by rejecting the proposer’s offer, if it is deemed unfair. Rejection would mean neither gets any money. In TPG, a third party is permitted, if he chooses, to punish the proposer’s choice of allocation to the second party, but only by paying a cost (real money) to do so.

The big general finding is that people in larger, more complex societies are more likely to split pots of money with strangers on a 50-50 basis. They are also more likely to have norms which enforce that outcome.

Ensminger & Henrich find there is a lot of variation in behaviour across the sample societies, but purely selfish behaviour, on average, is not found in any of them. (An exception which comes closest may be the Hadza, a small hunter-gatherer group in Tanzania, whose modal offer in DG was zero). Every society shows a positive mean offer in DG (as opposed to zero). And in all societies sampled, there is some offer that is perceived as just too low and gets rejected (in UG) or punished (in TPG), with both rejections and punishments declining as the offer approaches 50% of the total. However, the “minimum acceptable offers” do vary substantially from society to society.

The X-axis = the percentage of the money offered by the ‘dictator’. Grey bars = the mean offer for each population. The size of the bubbles = the fraction of the sample making those offers. Also see the chart of distribution of results for UG and TPG.

What accounts for the variability in these behaviours ? After a series of regressions, Ensminger & Henrich come to the following conclusions:

1. The average amount of money offered by first parties in all three games increases with the degree of their community’s market integration (as measured by % of calories obtained in the market).
2.  ‘Fairness’ also increases with individuals’ participation in a major world religion, as represented in the sample by Christianity and Islam (as opposed to a ‘traditional’ religion lacking “high moralising gods“).
3. “People from larger communities punish more” in TPG. This is consistent with claims that more anonymous societies depend on “altruistic punishment”, the supposed tendency of some people to violate norm-violators and free-riders even when they are not direct victims.
4. The amount of third-party punishment meted out in TPG is a robust predictor of offers made by the proposer in the DG (a game with no punishment option) and of offers made by the proposer in the two-party UG. This may be evidence for the internalisation of fairness norms.
5. “[L]arger communities with greater punishment and fairness…are strongly associated with larger ethnic groups”, which is taken as consistent with their theoretical prediction that “[fairness] norms can spread as a consequence of their group-beneficial effects”.

The statistical tests show the two variables — market integration and world religion — are decent predictors of the prevalence of fairness norms in all the experiments. (In the Third Party Punishment, community size is also important.)

But the amount of variation in the data explained by the variables of interest is fairly small. “Low R-squared” is usually not important, but in this case it matters. There are real trends in the noise, but obviously much more is going on in the data to differentiate these societies than Ensminger & Henrich explore.

Interestingly Ensminger & Henrich reject possible genetic causes for the variation, even though they themselves report that the heritability of rejection behaviour in the Ultimatum Game is >40% in a sample of Swedish twins. E & H appeal to the standard logical argument that the within-population heritability of a trait does not imply anything about the between-population heritability of the same trait. That’s true, but it doesn’t foreclose the possibility that between-population variability is also heritable, either.

Their argument is that as societies grow larger and more complex (e.g., more market integrated), the social norms which are required to function in society adjust slowly to match the societal evolution. Of course that’s an inference from a cross-section, not capturing change in norms over time within societies. But they do have a theoretical model for a positive feedback between norms and societal scale with which their data are consistent.

This was recently echoed in a paper by Richerson & Henrich 2012, “Tribal Social Instincts and the Cultural Evolution of Institutions to Solve Collective Action Problems”, which appeared in a special issue devoted to nation-building in failed states in the journal Cliodynamics: The Journal of Quantitative History and Cultural Evolution. It recapitulates the main tenets of cultural evolution and multi-level selection theory, but specifically on the question of the massive cross-country variation in collective action capacity, it ends like this:

“Why does the scale vary so much among human societies, with some societies lacking much collective actions beyond the extended family while others organize millions in modern nation-states?

People began to domesticate plants and animals only about 11,000 years ago. Agriculture and the many arts that grew up with it created the potential for dense societies in favorable locations. Hence villages, towns, and cities began to grow. The pace of evolution varied from region to region, probably for many reasons (Richerson and Boyd 2001), even in the most favorable areas. Today, the world is a mosaic caused by differences in history and ecology. Tropical forest cultivators living a low density in family hamlets have virtually no institutions that operate outside the extended family (Johnson 2003). Densely populated urban cores of societies rich in agricultural, industrial, and human capital resources support modern nation-states. In some places with intermediate productivity or a historically slow trajectory of development tribal-scale institutions are still very strong. Sub-Saharan Africa and parts of the Middle East, most notably the Pashtun parts of Pakistan and Afghanistan are examples (see Turchin’s article in this Special Issue). It is important to note that the time scales of cultural change ranges from generations to millennia. If an institution is destabilized it may change rapidly until a new, usually nearby, equilibrium is established. Unless destabilized, institutions are very resistant to change. Policy makers are fated to be frustrated by the slow and hard-to control nature of cultural evolution.”

Basically, the reasoning is very similar to what you find in modernisation theory. The pace of cultural change is linked to the pace of economic development, which is treated as exogenous; or at least the trigger for the positive feedback cycle of economic development and cultural change is exogenous. So if development is slow or non-existent, then traditional social institutions are also slow to change. Even maladaptive norms can be stuck in a trap.

But a shock can put a society onto a path of evolving pro-social institutions. As social systems based on kin networks get disrupted by economic development, people grow more dependent on impartial norms. This is echoed in numerous papers in the “cultural evolution” literature, e.g., Hruschka et al., Hruschka & Henrich, Newson & Richerson.

Also, Boyd & Richerson’s reasoning is rather similar to that in biological evolution. If the palaeontological or archaeological record shows it took X million years for binocular vision to evolve, then, that’s how long it took and the job of your model is to spell out the mechanism within the dominant theoretical paradigm. In the case of “cultural evolution”, the assumption seems (to me anyway) that how ever long it takes is how ever long it takes, and you should model it as cultural evolution.

§  §  §  §  §

An Indonesian whaling village displays fairness norms in the Ultimatum game which exceed the ‘perfect’ 50-50 (proposers tend to offer more than 50%), presumably because their whale-hunt economy is completely dependent on team effort and strong social cohesion. But is that all it takes? Maybe these one-shot trust games don’t really give us fine enough information about what’s required to evolve into a modern complex market society. I stress I am not questioning the external validity of one-shot games. But they don’t really seem to generate enough variation beyond a certain societal scale, for explaining the cross-country heterogeneity in social institutions. Or so it seems.

The repeated versions of games like the prisoner’s dilemma or the stag hunt or the public goods game are probably more informative about the complex interactions that exist in the most advanced societies. (The 2004 volume Foundations of Human Sociality does report findings from public goods games conducted in a couple of small-scale societies, which I will blog about separately.)

I should note that the economics literature on “culture & institutions” mirrors the argument by cultural evolutionists to some extent: culture and institutions mutually reinforce one another and co-evolve. That is, cultural differences are reinforced by institutions which are themselves path-dependent products of earlier cultural differences. I don’t find that a very satisfying literature on the largest questions of political and economic development. But I will be commenting more on the economics literature on culture in the near future.