Using Infectious Disease Modelling To Understand Public Policy Spread
In January I attended a course on infectious disease modelling as part of the AMSI Summer School. The rapid spread of COVID-19 made this surprisingly timely. By the middle of the course only 500 infected individuals had been identified, all in mainland China. As of 1 Mar 2020, this has risen to more than 86,000 cases in over 60 nations. Sustained outbreaks have now occurred in South Korea, Iran, and Italy, and markets have reacted sharply - both the NASDAQ and Dow Jones having fallen by more than 12% across a single week.
I'm not an infectious disease modeller, so this post won't be about the spread of the novel coronavirus. But what I want to highlight is that the same models which describe the spread of diseases can also be used to understand how policies, social, and legal changes can propagate between different jurisdictions and countries.
This is easiest seen by viewing the world as a network: a collection of nodes (individuals, countries, etc) and edges (showing the connection between each node). We can also consider each node having attributes (e.g. whether a node is 'infected'), and each edge having a weight showing connection strength. These are powerful and generic concepts, and networks have been used to describe everything from the structure of economies, friend groups, sexual relationships, diseases, memes, gossip, and much much more.
When looking at how attributes spread across a network there a few key assumptions: firstly, each node in contact with an infected individual has a chance of becoming infected; secondly, this chance of becoming infected rises with the number of infected nodes it is connected to and the strength of those connections. There may also be individual characteristics of nodes that make them more of less susceptible to infection.
The game Plague Inc is built off this process: you take control of a virus, and by manipulating its characteristics and vectors attempt to infect the world's population. As network structures are generic, the game also has versions showing spreads of misinformation, political movements, and insurgency.
This brings us to the key point - by looking at network structure we can explain how policy positions and legal reforms become adopted between different jurisdictions. The case of same-sex marriage legislation illustrates this perfectly. The first nation to provide for same-sex marriage was the Netherlands in 2001. By the start of last year, when I first conducted some rudimentary analysis on the issue in response to Brunei's reprehensible introduction of the death penalty for same-sex relations, over 25 nations had implemented same-sex marriage legislation (map source). Transmission followed clear lines across close cultural, linguistic, geographic, and economic neighbours and was often preceded by civil union legislation at the state, provincial, or city level.*
Using the dynamics of network structure to understand how policies, norms, and legislation spread between different jurisdictions is intuitively appealing. We can readily find examples of policy innovations where the spread is both striking and interesting - my favourite examples include universal suffrage (a long evolution from the first French introduction of adult male in 1791, and the New Zealand introduction of adult female voting in 1893), and the back-and-forth of prohibition for alcohol and marijuana (where both the prohibition and reversal can be viewed as separate spreading processes).
Beyond this we can also use networks as a framework for predicting how a particular issue may evolve. Infectious disease modellers often use passenger flight volumes to forecast the likelihood that a particular location will experience an outbreak (see here for an example produced early in the COVID-19 spread). In principle, similar models can be used to develop predictions for how policy implementations can spread by looking at connections weighted by factors including trade volume, linguistic and cultural links, strategic alliances, shared media and social media, governance, and political party correlations between nations.
Developing these models would be a significant undertaking - and would almost certainly miss the complexity of individual interactions and characteristics which determine spread in the real world - but that's the eternal modeller's dilemma: "All models are wrong, but some are useful." Analysts will find there's a lot of new use to be found in epidemiological models.
* When I first analysed this data I predicted that these trends indicated that Japan was likely to enact same-sex marriage within the next couple of years. At the time 24 jurisdictions in Japan had implemented registered partnerships. This has since increased to 32 municipalities and two prefectures, with an additional 12 jurisdictions to be implemented soon.