Capitalism and neoliberalism have been severely questioned lately. The rise of inequality in many countries (even though globally may be balancing due to China’s ascent) is a source of social and political tension. And productivity stagnation also raises doubts about the whole economic system. COVID-19 pandemic has only exacerbated the doubts and the debate, and not surprisingly a lot of new and not so new but marginal ideas are back on the scene. For example, the opportunity and potential benefits of a universal basic income.
Not surprisingly, in countries like my own, Spain, where there is a widespread lack of knowledge about technical and economic issues among the political class (not necessarily some people behind them who nobody hears) , and with governments controlled by populist parties and political value propositions, the debate is clumsy, to say the least.
Therefore it is refreshing and encouraging to see that there are people contributing with new ideas to this old and tough debate. It is more necessary than ever. One of this promising ideas is the use of simulation techniques to analyse and design better economic policies. Economics is all about difficult trade-offs, and testing in the real economy is neither feasible not desirable. For example, finding a tax policy that optimizes equality along with productivity is an unsolved problem.
High income inequality can negatively impact economic growth and economic opportunity. Taxes can help reduce inequality, but it is hard to find the optimal tax policy. Economic theory cannot fully model the complexities of the real world. Instead, tax theory relies on simplifying assumptions that are hard to validate, for example, about the effect of taxes on how much people work. Moreover, real-world experimentation with taxes is almost impossible.
A team of scientists at Salesforce, a technology company, are using reinforcement learning —the same sort of technique behind DeepMind’s AlphaGo and AlphaZero— to optimize dynamic tax policies. They have created AI Economist to identify optimal tax policies for a simulated economy. Admittedly, it is still a very simple economy, but it is a promising first step toward evaluating policies in an entirely different way.
(As they work for a large sucessful company, they can do nice things, like this video:)
Here is the abstract of the paper they have published in arXiv:
Tackling real-world socio-economic challenges requires designing and testing economic policies. However, this is hard in practice, due to a lack of appropriate (micro-level) economic data and limited opportunity to experiment. In this work, we train social planners that discover tax policies in dynamic economies that can effectively trade-off economic equality and productivity. We propose a two-level deep reinforcement learning approach to learn dynamic tax policies, based on economic simulations in which both agents and a government learn and adapt. Our data-driven approach does not make use of economic modeling assumptions, and learns from observational data alone. We make four main contributions. First, we present an economic simulation environment that features competitive pressures and market dynamics. We validate the simulation by showing that baseline tax systems perform in a way that is consistent with economic theory, including in regard to learned agent behaviors and specializations. Second, we show that AI-driven tax policies improve the trade-off between equality and productivity by 16% over baseline policies, including the prominent Saez tax framework. Third, we showcase several emergent features: AI-driven tax policies are qualitatively different from baselines, setting a higher top tax rate and higher net subsidies for low incomes. Moreover, AI-driven tax policies perform strongly in the face of emergent tax-gaming strategies learned by AI agents. Lastly, AI-driven tax policies are also effective when used in experiments with human participants. In experiments conducted on MTurk, an AI tax policy provides an equality-productivity trade-off that is similar to that provided by the Saez framework along with higher inverse-income weighted social welfare.Zheng, Stephan, Alexander Trott, Sunil Srinivasa, Nikhil Naik, Melvin Gruesbeck, David C. Parkes, y Richard Socher. «The AI Economist: Improving Equality and Productivity with AI-Driven Tax Policies». arXiv:2004.13332 [cs, econ, q-fin, stat], 28 de abril de 2020. http://arxiv.org/abs/2004.13332.
I can only share with the authors their wish that “It would be amazing to make tax policy less political and more data driven,”
Now, the question is: who can help explain the average politician what we are talking about? The noise of “caceroladas” today make it very difficult to have any sort of smart conversation… and even if they are not very knowledgeable, they are not completely stupid. I’m sure they can also see the robots coming. This time for their jobs…