Do Social Networks Matter More Than Institutions?

Image result for world connection cell phone

I’ve posted before about McKinsey’s findings regarding digital globalization. They reported,

Data flows directly accounted for $2.2 trillion, or nearly one-third, of [globalization’s] effect [in a decade]—more than foreign direct investment. In their indirect role enabling other types of cross-border exchanges, they added $2.8 trillion to the world economy. These combined effects of data flows on GDP exceeded the impact of global trade in goods.

This in turn supported research by economist Andreas Bergh, who found that

the poverty-decreasing effect of globalization is bigger in countries where institutions are worse. The graph below shows how the marginal effect of information flows on poverty varies depending on the level of bureaucratic quality. The slope looks the same for all institutional indicators, suggesting that globalization is especially important for the poor in countries with high corruption levels and inefficient public sectors.

A new Harvard working paper supports these findings, suggesting that communication networks and social interactions are more important than institutions. The authors explain,

Telling institutional versus socio-technological interpretations apart has been challenging. This paper tests these two hypotheses by measuring convergence in income across Colombian municipalities along two distinct geospatial divisions: one institutional, one socio-technological. The institutional explanation would emphasize the role that belonging to a particular departamento, or state, has on the institutional arrangements and the provision of public goods, thus affecting the incentive structure of agents to operate with better technology.

Although Colombia is a unitary republic, not a federation, states have significant autonomy1 . Studies on Colombia, including those that take an institutional perspective such as Acemoglu et al (2015)…utilize state-level data, as do almost all studies of intra-national unconditional convergence worldwide. Under the institutional assumption, a municipality should tend to converge to the income of the state to which it belongs.

The socio-technological explanation would predict that municipal income convergence should occur within the cluster of municipalities that interact intensely with each other, whether or not they belong to the same state. This is due to the need for intensive social interactions for knowhow to diffuse. To form these socio-technological groupings, we utilize a unique dataset of cellphone calls to group municipalities so that most of the phone calls happen within rather than between these clusters. To facilitate comparison with the 32 states of the institutional state aggregation, we group municipalities into 32 communication clusters…Thus, communication clusters are groups of municipalities that are densely connected through phone calls, meaning that they are significantly more likely to call members of the cluster than they are to call other municipalities (pgs. 4-5).

The authors conclude,

To test these two interpretations in a more direct way, we use municipal level data for Colombia, which we aggregate using two different grouping criteria: the departamento or state to capture institutional variation; and the communication cluster to which a municipality belongs, to capture the intensity of social interaction. We use formal wages per capita as our measure of income per capita, as it can be measured at the municipal level. We use cellphone data to group municipalities into communication clusters of intense interaction.

In this setting, we find evidence of absolute convergence in Colombia at the municipal level. We find evidence that the process is accelerated when the municipality belongs to a richer communication cluster. However, we do not find evidence of a positive influence of belonging to a richer state. We interpret these results as evidence in favor of the idea that obstacles to technology diffusion may be related to the fact that the use of technology requires tacit knowledge which tends to move slowly between brains through a protracted process of imitation and repetition as occurs in learning by doing. Within communications clusters, there seems to be accelerated convergence. Obstacles to convergence in developing countries may be related to the paucity of social interactions between citizens of the same country

…From a policy perspective, the findings emphasize the fact that economic convergence requires intense social interaction, not just the presence of institutions of a certain quality. Regions that are formally part of the same nation-state but do not really interact with the more advanced parts of the country cannot expect to share similar development outcomes.(pg. 19).

Fascinating stuff.