At the University of Sao Paulo, I am building a policy simulator for the Brazilian education system. Brazil's inequality in educational service provision is extremely high. To find a way to mitigate the differences in educational outcomes, I analyse the problem using a computational model of the money transfers between governmental entities. The model incorporates about 11,000 entities, representing all the municipalities, school districts, states, and redistribution agencies in Brazil.
In the baseline version, all agents’ behaviour is driven by the actual laws, while their parameters, such as tax income and school/student numbers and types, are taken from the actual data.
With the model, I am able to show that quantitative changes in the current financing rules cannot decrease inequality. An alternative system with redistribution at the federal level is instead necessary. This new system has been designed and tested with this model. As this model can be run from a web application, it can be used by policymakers to test policies and to train their intuition and understanding.
At MIT's Systems Engineering Division we developed my first web-enabled public policy simulation.
The team I led worked in close collaboration with the Saudi Arabian Ministry of Labor to develop a data-driven model of the Saudi Arabian labour market. Saudi Arabia’s principal problem is unemployment of Saudi Nationals, which is, at least partially, caused by the fact that seven million expatriates reside in the country, and they hold the majority of the available jobs.
The purpose of our model was to evaluate Saudi Arabia’s policy options to increase employment of Saudi nationals: taxes on expatriates, quotas, and minimum wages (in general and for expatriates). We show that the best policy answer depends on the sector. The labor market model that we developed was calibrated using time-series and firm-level data. This model was delivered not only as part of an academic paper but also as a web-enabled policy simulator. Beyond testing and designing policy, the model could also be used to train policymakers’ intuition and understanding of the labour market.
Unfortunately only the agent-based simulation is publicly available, while the knowledge base is confidential property of the Kingdom of Saudi Arabia's Ministry of Labor.
Studies of the economic impact and mitigation of climate change usually use computable general equilibrium models (CGE). Equilibrium models, as the name suggests, model the economy as in equilibrium, the transitions to the equilibrium are ignored. In the time spent outside equilibrium, the economy produces different quantities of goods and pollution as predicted by the equilibrium model. If the economy in this time outside of the equilibrium produces more climate gasses the predictions are dangerously wrong.
We present in this paper a computational generalization of the Arrow-Debreu general equilibrium model, which is not in equilibrium during the transitions, but converges to the same equilibrium as a CGE model with the same data and assumption. We call this new class of models Computational Complete Economy models.
Computational Complete Economy models have other interesting applications for example in international trade, tax policy and macroeconomics.
ABCE is a python-based modelling platform for economics. ABCE allows to express economic models in a concise manner. It automatises data collection and visualisation and makes it easy to create a web-app out of an agent-based model.