Internet measurement & central banks

An analogy

Believe it or not, the Internet is regulated.

For example, Comcast isn't allowed to offer a low-cost plan that only includes Facebook, then charge you more to access the rest of the web [1]. That's thanks to a principle called “net neutrality.” Meanwhile, regulators in the West are increasingly concerned with privacy. They've enacted rules like GDPR and CCPA.

But, globally, regulators have instantiated a variety of different rules. For example, Russia has "Internet sovereignty" laws, which allowed them to disconnect themselves from the global Internet in 2019 as part of a preparedness drill.

The principle here is this: regulators in different countries have different values. They optimize for these values in part by leveraging their legal jurisdiction to manage the Internet. But, as we've shown empirically, Internet regulations have externalities. They correlate with trade; they correlate with military alliance.

Currently, there is no good way for regulators to assess how their decisions about the Internet will impact the world, or their place in it.

Here's a concrete example: recent Australian legislation mandates that Google pay news organizations who show up in search results. Now, Google is threatening to leave the Australian internet.

This kind of spectacular regulatory “misfire” would never be tolerated at the level of, for example, a central bank that overshoots inflation targets. But Internet legislators deal with something fundamental to the economy, too.

A point of comparison: central banks

Let's turn more closely to the concept of a central bank as a point of comparison. The US Federal Reserve, to take an example, has a dual mandate: to promote full employment, consistent with a stable and steady rate of inflation. Monetary policy uses various tools to seek an impact on those two goals, affecting economic activity in direct and indirect ways. While the relationship between the intervention and its effect is delineated by theory, and has been validated by evidence over time, the relationship is complex and, in some ways, unpredictable. [2]

Nevertheless, central banks create these goals and enact their particular policies in response to economic data, and guided by theory. Data motivate interventions. Banks track the effects of these interventions by monitoring data.

Central banks’ decisions are hugely consequential, and the tools they have to achieve these goals are imperfect at best. But at least there are quantitative models to guide their use.

Imagine a world…

Imagine a world in which Internet regulators had well-defined levers to pull—levers that allowed them to optimize for things that they cared about domestically and internationally. Imagine these regulators had good data to track the effect their interventions had.

Drawing out the analogy a bit further: in both central banks and Internet regulators, policy happens “on top of” a very dynamic private market. Google, Facebook, and many small firms manage the domestic Internet, just as private employers manage the domestic economy. Who is above that domestic market, making sure its substrate (the Internet) remains stable and manageable long-term? That kind of body doesn’t exist in any coherent way yet—but it could.

When I think about the long-term vision of our Internet fragmentation work, I think about how our data could create better levers for managing the Internet.

Nations need better levers, and a better idea of what happens when you pull them. Just look at Australia.

Update Feb 5, 2021: See Steve Weber’s reply to this post.

In other news

  • Lab member Lily Bhattacharjee released polisci-vis, a package for visualizing relationships between countries. She developed it while building the interactive visualizations for our First Monday paper. Way to go, Lily!

  • I wrote a post for CLTC Bulletin about our work on educational resources to help students identify bias in machine learning algorithms. Who are these resources for—and who else might make use of them?

Footnotes

[1] Much to the chagrin of Ajit Pai.

[2] As Steve Weber rightly comments, “Monetary policy is just one variable in the equation (fiscal policy, Keynesian animal spirits, bond markets, and so on all respond, sometimes strategically). Rational behavior equilibria depend on time horizons, discount rates, and expectations of future behaviors, all of which are complex social and emotional as well as cognitive variables. The complexity of the system and the time lag between cause and effect makes it very difficult to assess the overall relationship in precise terms.”