We use survey data on an opt-in panel of around 2,500 US small businesses to assess the impact of COVID-19. We find a significant negative sales impact that peaked in Quarter 2 of 2020, with an average loss of 29% in sales. The large negative impact masks significant heterogeneity, with over 40% of firms reporting zero or a positive impact, while almost a quarter report losses of more than 50%. These impacts also appear to be persistent, with firms reporting the largest sales drops in mid-2020 still forecasting large sales losses a year later in mid-2021. In terms of business types, we find that the smallest offline firms experienced sales drops of over 40% compared to less than 10% for the largest online firms. Finally, in terms of owners, we find female and black owners reported significantly larger drops in sales. Owners with a humanities degree also experienced far larger losses, while those with a STEM degree saw the least impact.
We run randomized controlled trials on a panel of 7,300 small U.S. firms to test if we can improve their sales forecasting. At baseline, only 17.4% of entrepreneurs can forecast their firm's sales in the next three months within 10% of the realized value, with 1% of the mean squared error attributable to bias and the remaining 99% attributable to noise. Our first intervention rewards entrepreneurs up to $400 for accurate forecasts, our second requires respondents to review historical sales data, and our third provides forecasting training. Increased reward payments significantly reduce bias but have no effect on noise, despite successfully making entrepreneurs spend more time thinking about their prediction. The historical sales data intervention has no effect on bias but significantly reduces noise. Since bias is only a minor part of overall forecasting errors, we find that the reward payments have small effects on mean squared error, while the historical data intervention reduces it by \DashImpact\%. The training intervention has negligible effects on bias, noise, and ultimately mean squared error. Our results suggest that while offering financial incentives to exert more efforts make forecasts more realistic, firms may not fully realize the benefits of having easy access to past performance data.
The importance of social networks in job search and migration have been well documented. However, spreading information too widely throughout networks when opportunities arise can easily lead to the tragedy of the commons - too many people depleting a limited opportunity can mean no one benefits in the end. Hence, despite the generally positive value of large social networks, we should expect the strategic sharing of information within networks. To better understand this, we study the co-migration decisions of social connections through the movements of gold miners in Colombia. In this setting, we document three facts that are easily interpretable with a model of referrals and scarce resources. First, while working with close social connections is associated with higher production, having too many miners present is ultimately associated with lower production. Second, in line with the first result, we find that more productive miners, for whom depletion of resources is a greater concern, invite fewer social connections. Finally, the connections that miners are willing to invite are heavily selected; miners tend to invite productive over non-productive peers.