[论文解读] COVID-Town: An Integrated Economic-Epidemiological Agent-Based Model

Patrick Mellacher|arXiv (Cornell University)|Nov 12, 2020

COVID-19 epidemiological studies被引用 8

一句话总结

本文提出 COVID-Town，一种基于代理的模型，整合了流行病学与宏观经济动态，以模拟COVID-19疫情期间政策干预的影响。结果表明，若将防控措施延迟一周实施，德国的死亡人数将增加180%；而若及早采取行动，死亡人数可减少60%。此外，研究显示，反周期财政政策可在不增加死亡率的情况下实现V型经济复苏。

ABSTRACT

I develop a novel macroeconomic epidemiological agent-based model to study the impact of the COVID-19 pandemic under varying policy scenarios. Agents differ with regard to their profession, family status and age and interact with other agents at home, work or during leisure activities. The model allows to implement and test actually used or counterfactual policies such as closing schools or the leisure industry explicitly in the model in order to explore their impact on the spread of the virus, and their economic consequences. The model is calibrated with German statistical data on time use, demography, households, firm demography, employment, company profits and wages. I set up a baseline scenario based on the German containment policies and fit the epidemiological parameters of the simulation to the observed German death curve and an estimated infection curve of the first COVID-19 wave. My model suggests that by acting one week later, the death toll of the first wave in Germany would have been 180% higher, whereas it would have been 60% lower, if the policies had been enacted a week earlier. I finally discuss two stylized fiscal policy scenarios: procyclical (zero-deficit) and anticyclical fiscal policy. In the zero-deficit scenario a vicious circle emerges, in which the economic recession spreads from the high-interaction leisure industry to the rest of the economy. Even after eliminating the virus and lifting the restrictions, the economic recovery is incomplete. Anticyclical fiscal policy on the other hand limits the economic losses and allows for a V-shaped recovery, but does not increase the number of deaths. These results suggest that an optimal response to the pandemic aiming at containment or holding out for a vaccine combines early introduction of containment measures to keep the number of infected low with expansionary fiscal policy to keep output in lower risk sectors high.

研究动机与目标

• 开发一个综合的基于代理的模型，以捕捉疫情期间的经济与疾病传播动态。
• 评估政策干预的时机与设计对感染传播与经济结果的影响。
• 利用德国的人口、就业与时间使用数据校准模型，以实现现实的模拟。
• 测试反事实政策情景，如学校或休闲产业的关闭。
• 评估在大流行条件下财政政策应对的宏观后果。

提出的方法

• 模型模拟具有不同属性（如年龄、职业与家庭状况）的个体代理，他们在家庭、工作与休闲场所中互动。
• 代理之间的互动由德国的时间使用数据决定，以建模不同部门的接触模式。
• 流行病学动态采用类似SIR的分 compartment 框架建模，包含按年龄与部门划分的传播率。
• 通过部门就业、工资与企业利润建模宏观经济反馈，高接触部门的产出损失在经济中传播。
• 通过匹配德国在首波疫情中的实际死亡曲线与估算的感染曲线对模型进行校准。
• 通过调整政府转移支付来模拟财政政策情景，以维持预算平衡（顺周期）或刺激需求（反周期）。

实验结果

研究问题

• RQ1 若德国的防控政策提前或推迟一周实施，对首波疫情中死亡人数的影响如何？
• RQ2 关闭特定部门（如学校或休闲产业）对感染传播与经济产出的影响如何？
• RQ3 在大流行期间，顺周期财政政策的长期宏观经济后果是什么？
• RQ4 反周期财政政策是否可在不增加感染或死亡率的情况下实现快速经济复苏？
• RQ5 政策干预的时机如何影响公共卫生与经济稳定之间的整体权衡？

主要发现

• 若将防控措施延迟一周，德国首波疫情的死亡人数将增加180%。
• 若提前一周实施政策，死亡人数可减少60%。
• 顺周期财政政策（零赤字）导致经济持续下滑的恶性循环，即使病毒被消除，经济也未能完全恢复。
• 反周期财政政策可在不增加死亡率的情况下实现V型经济复苏。
• 模型表明，早期防控结合扩张性财政政策可最优地平衡健康与经济结果。
• 高接触行业（如休闲产业）的部门溢出效应显著放大了整个经济的损失。