Fiscal policy in the age of COVID-19 : does it "get in all of the cracks"?

We study the effects of fiscal policy in response to the COVID-19 pandemic at the firm, sector, country and global levels. First, we estimate the impact of COVID-19 and policy responses on small and medium-sized enterprise (SME) business failures. We combine firm-level financial data from 50 sectors in 27 countries, a detailed I-O network, real-time data on lockdown policies and mobility patterns, and a rich model of firm behavior that allows for several dimensions of heterogeneity. We find that absent government support, the failure rate of SMEs would have increased by 9 percentage points, significantly more so in emerging-market economies (EMs). With policy support it increased by only 4.3 percentage points, and even decreased in advanced economies (AEs). We also find that fiscal policy was poorly targeted: most of the funds disbursed went to firms that did not need it. Nevertheless, we find little evidence of the policy merely postponing mass business failures or creating many "zombie" firms: failure rates rise only slightly in 2021 once policy support is removed. Next, we build a tractable global intertemporal general equilibrium I-O model with fiscal policy. We calibrate the model to 64 countries and 36 sectors. We find that a sizeable share of the global economy is demand-constrained under COVID-19, especially so in EMs. Globally, fiscal policy helped offset about 8% of the downturn in COVID-19, with a low "traditional" fiscal multiplier. Yet it significantly reduced the share of demand-constrained sectors, preserving employment in these sectors. Fiscal policy exerted small and negative spillovers to output in other countries but positive spillovers on employment. A two-speed recovery would put significant upward pressure on global interest rates which imposes an additional headwind on the EM recovery. Corporate and sovereign spreads rise when global rates increase, suggesting that EMs may face challenging external funding conditions as AEs normalize.