Sustainability with Vontobel: our report 2019
Find out what has made Vontobel a pioneer in sustainable investing – for over 30 years.
In 2015, the member states of the United Nations signed a resolution with a total of 17 goals for sustainable development of the earth to be achieved by 2030. The UN has now published its new patent recipe for achieving these goals: a computer simulation called Policy Priority Inference (PPI).
With the simulation software PPI (Policy Priority Inference) the UN wants to save the world – by better harmonizing budget allocations with its goals for sustainable development (SDG). © Getty
staff editor at WIRED.
We feature his report here as part of our Publishing Partnership with Wired UK.
Policy Priority Inference (PPI) is a budgeting software – it simulates a government and its bureaucrats as they allocate money on projects that might move a country closer to an SDG. The areas a government invests money in are fundamental to understanding a government's priorities, according to Luis Felipe López-Calva, assistant secretary-general at the UN. “Our view is that the structure of the budget is really the manifestation of your policy choices,” he says. “The intention is that this becomes a tool to enhance the discussion around the political processes that leads to a certain allocation of budget.”
This advance would theoretically lead to greater transparency of a government’s reasoning behind a policy, and make it easier for the public to monitor progress on these pledges. López-Calva explains:
“The simulation would also shed light on the relationships between different SDG goals.”
For example, it might show how poverty interacts with educational inequality, or how investment in certain forms of energy might damage the environment.
PPI was co-created by Omar Guerrero, an economist at University College London and a fellow at the Turing Institute, and his research partner, professor Gonzalo Castañeda of the Center for Research and Teaching in Economics in Mexico. “Some years ago, we started to think about how to create this development plan for state governments in Mexico,” says Castañeda. “Because they were doing a very lousy job, and still are.” From this observation grew the first iteration of PPI. After Luis Felipe López-Calva, assistant secretary-general at the UN, read the paper outlining how PPI would work, he thought that adapting the technology might help the UN achieve its SDGs.
Governments haven't used the technology yet, but the team is in discussion with officials from Mexico, Colombia and Uruguay. In Mexico, there have been workshops run with the finance ministry, the national institute of statistics, representatives of six of Mexico's 32 states and NGOs. The idea is that an incoming government would use the tool's insights to better understand how to reach an SDG. “For instance, in Mexico, it is estimated that by next year there will be between five and ten million new poor people, which means that many indicators regarding fighting poverty will go down,” says Guerrero. “So, we can again try to estimate the size of the delays.”
One of the tool’s main innovations, and one that caught López-Calva’s attention, is that it uses a technique called “agent-based modelling”. In its simulated world, individuals and institutions are represented directly by agents – in this case, bureaucrats and governments. “Most of economics is based on what I often call a representative agent, essentially an average person,” says Nigel Gilbert, a professor in computational social science at the University of Surrey. “But you don't have to have average people if you’re using an agent model.”
Traditional neoclassical economics assumes that agents are rational in the sense that they will try and maximise their utility. Problem is – real people aren’t like this, explains Gilbert – not even bureaucrats. “Though they might like to think they are perfectly rational, they are typically ‘boundedly rational’,” he says. “In other words, it’s not that they do silly things and make decisions completely at random, but that they don’t spend huge amounts of time doing infinitely long calculations to work out what to do.”
Agent-based modelling, in this sense, provides a more realistic representation of how a government works. It is particularly useful for understanding the way we are affected by long-standing social norms, says López-Calva. “For example, women not participating in the labour force should not be seen as something that is embedded in the preferences of society, but rather is about constraints as a persistent social norm,” he says.
At the moment, PPI is just a command to be executed in the programming language Python. The team are currently working to generate a friendlier interface and more efficient code. Next year, Guerrero hopes to create a website where you can drag and drop data on indicators and networks, in order to get some statistics, visualisations and run simple scenarios.
How effective PPI proves will to a large extent come down to the quality of the data governments provide. “Some countries don’t have a good quality [government expenditure data],” says Castañeda. “There are some countries – in particular Mexico and Norway – they have good data in that respect. The model becomes more robust and reliable.”
Another potential issue is whether PPI, which is built on its own assumptions on economics and behavioural science, is a valid model for a particular government’s problems. “There are lots of models,” says Gilbert. “The question is which one is the right one giving the right answers?”
That’s why PPI is intended to be advisory. “We don’t want to say this is the model that everyone should adopt,” says Guerrero. “We’re trying to say this is a philosophy, the philosophy of how to think about understanding policy priorities and getting advice.”
PPI shouldn’t be considered a replacement for government, nor fully determine its decisions. “The tool just informs the decision-maker and doesn't determine the budget allocation,” says López-Calva. “Every tool has its strengths and downsides. So it should be used with a grain of salt, in a sense – it isn't a straight jacket.” It’s not government-by-algorithm just yet.