Due to the global coronavirus (COVID-19) pandemic, 2020 is set to be a tragic year for many businesses. Startups are likely to be affected particularly strongly as they find themselves in a fragile state, being sensitive to disruptions in demand, supply, or credit conditions. This is already showing in the statistics. In the last week of March, new business applications were down 40% compared to the same week one year earlier, a contraction that is even sharper than that during the Great Recession (Haltiwanger 2020).
These developments are likely to have important macroeconomic implications that may last well beyond the pandemic itself. Seemingly small changes to startups can create persistent and increasingly strong ripple effects on the macroeconomy as cohorts of new firms age and grow into larger businesses. Therefore, startups deserve special attention in this situation.
The Startup Calculator
We share an empirical perspective on what the disruption of startup activity might imply for the US economy, in terms of the severity and persistence of employment losses. To this end, we developed a simple Startup Calculator, available on our website, which allows anyone to easily compute employment losses under various scenarios.
The calculator allows one to vary three key margins that pertain to entry and exit of young businesses. These are not easily reversed and may have important effects on the macroeconomy in the medium and long runs.
The first margin is the number of startups. A fall in this number directly reduces the number of new jobs created by startups. This ‘lost generation’ of firms then creates a persistent dent in aggregate employment as subsequent years are characterised by a lower number of young firms (e.g. Gourio et al. 2016, Sedláček 2019).
The second margin is the growth potential of startups. Sedláček and Sterk (2017) show that firms born during recessions not only start smaller but also tend to stay smaller in future years even when the aggregate economy recovers. These differences in growth potential are attributed to changes in the composition of the type of startups. In the current situation, it seems particularly difficult to start highly scalable businesses, since supply chains are heavily distorted, credit conditions are poor, and customer demand is difficult to acquire during a lockdown.1
The third and final margin we consider is the survival rate of young businesses. Startups and young firms in general have much higher exit rates than older firms (see e.g. Haltiwanger et al. 2013), and during downturns, these exit rates tend to increase.2
The Startup Calculator uses publicly available data from the US Business Dynamics Statistics. We take a conservative stance and only consider changes to firms younger than 10 years of age. In other words, we leave 40% of all businesses unaffected in our calculations and, as such, the results may be taken as lower bounds. Details underlying the calculations can be found on our website.
What is so special about startups?
There are two main reasons why we focus on startups and, in turn, young firms. First, new and young businesses are the dominant creators of new jobs. To get out of the current labour market contraction, hiring by firms will be key (see also Merkl and Weber 2020). In the US, an average of 16.3 million jobs are created and about 14.9 million jobs are destroyed every year. This means that, annually, about a third of all jobs in the US are either new or get destroyed. Strikingly, startups create a net amount of 2.9 million jobs per year. These values suggest that startups are the only business category characterised by positive net job creation and existing firms only shed jobs on average.
It is also true, however, that young firms exhibit a higher rate of exit, suggesting that not all jobs created by startups are long-lasting. Nevertheless, the data show that surviving young firms tend to grow faster than the average incumbent (see e.g. Haltiwanger et al. 2013). These patterns of high rates of exit and growth among young firms have been dubbed ‘up-or-out dynamics’.
The second reason to focus on startups relates precisely to the up-or-out dynamics. The high rate of labour-market churn associated with startups has been linked to measures of productivity and profitability growth (see e.g. Bartelsman and Doms 2000 or Foster et al. 2001). Therefore, the data suggest that surviving young businesses are the ones crucial for aggregate productivity growth.
Finally, these findings are supported by new evidence on young high-growth firms, or ‘gazelles’. Haltiwanger et al. (2017) document that this small group of startups with exceptional growth potential accounts for about 40% of aggregate growth in total factor productivity, 50% of aggregate output growth, and 60% of aggregate employment growth.
Startup activity since the COVID-19 pandemic
It is still too early to tell exactly how hard startups will be hit by the COVID-19 crisis. The available data, however, suggest that the situation is severe. Figure 1 plots state-level data on COVID-19 deaths versus the number of (high-propensity) business applications, a strong early indicator of startup activity (see Bayard et al. 2020).3 Haltiwanger (2020) shows that in late March 2020, business applications in the US declined strongly, about as much as during the Great Recession (although it is unclear how long the decline will last this time).
Figure 1 shows the decline in high-propensity business applications during late March and April 2020 by state, relative to the same weeks in 2019. These data are plotted versus the number of COVID-19 deaths by state. The figure shows that, not only have business applications declined strongly in many states, there is also a clear relation with the severity of the pandemic. Particularly striking is New York State (NY), which suffered both the largest number of deaths and the strongest declines in business applications.
Figure 1 Decline in startups and COVID-19 deaths by state
Notes: Horizontal axis: Change in high-propensity business applications during weeks 12–15 of 2020, relative to the same weeks in 2019, by US state. Source: Business Formation Statistics. Vertical axis: Total number of COVID-19-related deaths in the US. Source: Centers for Disease Control and Prevention.
At this point, we do not know whether the current contraction will be short-lived or develop into a full-blown recession. Therefore, we take a scenario-based approach. Based on the early indicator discussed above, we select as a baseline scenario a strong but short-lived contraction. Specifically, we assume that the startup rate, the growth potential, and the survival rate all drop to their lowest levels since 1977 (the beginning of our data sample). These values are in fact closely linked to the Great Recession, which was the worst period for startup activity since the start of the sample.4 However, we let the contraction last for just one year, based on the observation that several countries seem to have moved past the peak of the pandemic within several months, and assuming a relatively swift recovery of overall macroeconomic conditions. Figure 2 plots the effects on aggregate employment.
Figure 2 Baseline scenario in the calculator
Two key features stand out. First, the decline in startup activity has sizeable aggregate effects. In the first year, about 1.5 million jobs are lost, relative to a scenario without the pandemic. This loss is about 6% of the employment of firms aged below 10 years, and 1.1% of aggregate employment.
Second, the macroeconomic effects are very persistent, even though the shock itself lasts for only one year. Cumulated from 2020 until 2030, the job losses are about 10.6 million. Moreover, each of the three margins plays a substantial role. The decline in the number of startups accounts for about 4.6 million of the cumulated job losses, the decline in growth potential for about 2 million, and the decline in survival for about 3.5 million. The remaining 0.5 loss is due to interactions between the three margins.
Quite possibly, however, the shock will last longer than one year. Based on the calculator, we find that the cumulative employment loss is roughly proportional to the duration of the shock. If the crisis lasts for two years, it will result in roughly 20 million jobs lost between 2020 and 2030.
Alternatively, it is possible that the shock will be followed by a ‘bounceback’, which is also allowed for in the calculator. Figure 3 shows a scenario in which one year after the pandemic, all three margins reach the highest levels observed in our data sample. In this case, aggregate employment losses are much shorter lived, but nonetheless, some effects persist. Not only is the cumulative job loss until 2030 about 2 million, but it is only around 2028 that aggregate employment finally catches up to its initial trajectory. In other words, even a short-lived crisis with a strong bounceback will have a sizeable negative impact on the aggregate economy for the next decade.5
Figure 3 Bounceback scenario in the calculator
How likely is such a reversal scenario? This question is difficult to answer. Historically, however, strong bouncebacks have been uncommon, as in the data, all three margins show strong and positive autocorrelations over time. Another possibility is that older firms will hire more, compensating for the employment losses from startups. To fully offset the startup job losses in the baseline scenario, older firms would need to create an additional 1.5 million jobs in 2020. For comparison, net job creation by firms older than 10 years was only about 0.6 million. From this perspective, creating the 1.5 million extra jobs needed appears to be a large challenge.6
While the outlook for startups may be gloomy, there are also some glimmers of hope. First, the high sensitivity of startups to economic conditions implies that they may also respond positively to policies that aim to support them. Given that startups can be identified fairly easily, such policies might be relatively cost effective. Second, the change in our daily lives might inspire entrepreneurs to come up with new ideas and new ways of running businesses, which could foster growth in the long run.
Baker, S, N Bloom, S Davis and S Terry (2020), “COVID-induced economic uncertainty and its consequences”, VoxEU.org, 13 April.
Bartelsman, E J, and Mark Doms (2000), “Understanding productivity: Lessons from longitudinal microdata”, Journal of Economic Literature 38(3): 569–94.
Bayard, K, E Dinlersoz, T Dunne, J Haltiwanger, J Miranda and J Stevens (2020), “Early-stage business formation: An analysis of applications for employer identification numbers”, CES Working Paper 18-52.
Faria e Castro, M (2020), “Back-of-the-envelope estimates of next quarter’s US unemployment rate”, VoxEU.org, 11 April.
Coibion O, Y Gorodnichenko and M Weber (2020), “Labour markets during the COVID-19 crisis: A preliminary view”, VoxEU.org, 14 April.
Foster, L, J Haltiwanger and C J Krizan (2001), “Aggregate productivity growth: Lessons from microeconomic evidence”, in C Hulten, E Dean, and M Harper (eds.), New Developments in Productivity Analysis, NBER Book Series Studies in Income and Wealth, Chicago: University of Chicago Press, pp. 303–72.
Haltiwanger, J (2020), “Applications for new businesses contract sharply in recent weeks: A first look at the weekly Business Formation Statistics”, mimeo, April.
Haltiwanger, J, R Jarmin, R Kulick and J Miranda (2017), “High growth firms: Contribution to job, output and productivity growth”, in J Haltiwanger, E Hurst, J Miranda and A Schoar (eds.), Measuring Entrepreneurial Businesses: Current Knowledge and Challenges, NBER Book Series Studies in Income and Wealth, Chicago: University of Chicago Press, pp. 11–62.
Haltiwanger, J, R Jarmin and J Miranda (2013), “Who creates jobs? Small versus large versus young”, The Review of Economics and Statistics 95(2): 347–61.
Gourio, F, T Messner and M Siemer (2016), “Firm entry and macroeconomic dynamics: a State-level analysis”, American Economic Review Papers and Proceedings 116(5): 214–8.
Lee, Y, and T Mukoyama (2015), “Entry and exit of manufacturing plants over the business cycle”, European Economic Review 77: 20–7.
Merkl, C, and E Weber (2020), “Rescuing the labour market in times of COVID-19: Don’t forget new hires!”, VoxEU.org, 4 April.
Sedláček, P (forthcoming), “Lost generations of firms and aggregate labor market dynamics”, Journal of Monetary Economics.
Sedláček, P, and V Sterk (2017), “The growth potential of startups over the business cycle”, American Economics Review 107(10): 3182–210.
1 The growth potential margin only includes cohort-specific effects, i.e. factors that were there at the entry stage. The size of cohorts may also vary due to post-entry macro fluctuations. Given that such effects may be relatively easily reversed, they are less relevant for the medium and long runs and, hence, the calculator abstracts from them. In this sense, the calculator may understate the total impact of startups on aggregate employment.
2 Given our focus on entry and exit of startups, a fourth potential margin would be the size of exiting startups. However, we abstract from this, given that variations in this margin tend to be moderate compared to the other three (see also Lee and Mukoyama 2015).
3 These data are taken from the Business Formation Statistics, a new and experimental data set released by the US Census. Before starting up a company, an entrepreneur must apply for an employer identification number. Business Formation Statistics report the number of applications and flag a subset of them as ‘high propensity’ based on predictive characteristics, see Bayard et al. (2020).
4 This is supported by the observed drops in business applications (see e.g. Haltiwanger 2020), the surge in unemployment claims and their possible underestimation (see e.g. Coibion et al. 2020), the estimated unemployment rate in the second quarter of 2020 (see Faria e Castro 2020) or the estimated increase in uncertainty (see e.g. Baker et al. 2020).
5 The reason for this persistent drag is that, historically, the worst-case scenarios for our three margins have been worse than the best-case scenarios have been good.
6 Net job creation by older firms is low, in part due to job destruction by exiting firms. In 2016, net job creation by continuing firms older than 10 years of age was about 0.9 million.