Geographical distribution of start-up layoffs in the US
To analyze the effect of COVID-19 on layoffs in US startups, we work on a sample of 297 startups from the "Layoffs.fyi Tracker" database covering the period from 3/31/2020 to 3/9/2021. For this purpose, the following industries were included in the study, depending on data availability: Media, Infrastructure, Data, Consumer, Food, Transportation, Retail, Recruiting, Finance, Education, Security, Travel, Healthcare, Real Estate, Marketing, Construction, HR, Product, Logistics, Support and Other. The number of start-ups that had made layoffs of 15% or less of their workforce was 102, of which 57 were in California, 26 in New York, 7 in Massachusetts, etc. In addition, 141 start-ups had layoffs between 15 and 50% of their workforce, of which 69 were in California, 29 in New York, 14 in Massachusetts, 9 in Washington, etc. Finally, 44 start-ups had layoffs between 50 and 100% of their workforce, of which 22 were in California, 8 in New York, 6 in Washington, etc. The purpose of this section is not to analyze the spatial distribution of layoffs as in the literature (Batty, 2020; Antipova, 2021). We wish to enrich the existing literature with an original approach that incorporates membership in a specific branch of activity. Thus, we calculate the probability of a worker becoming unemployed according to his or her branch of activity.
The objective is to predict the layoff variable (LAIDOFF) defined in with two modalities for the variable LAIDOFF using a binary logistic regression. We note LAIDOFF (branch), the value taken by LAIDOFF in a given branch of activity. Let be the J descriptors. The vector of values for a given branch of activity is then written :
Thus, the probability for LAIDOFF to come from a given branch of activity is noted:
Therefore, the estimated logistic model is the following:
With , , , ..., the parameters to be estimated. The table below gives the results of four estimations. If we denote L layoffs, in model 1, modality 1 corresponds to L≤15% and 0 the rest. In model 2, modality 1 corresponds to 15%<L≤50% and 0 the rest. In model 3, modality 1 corresponds to 50%<L≤75% and 0 the rest. Finally, in model 4, modality 1 corresponds to 75%<L≤100% and 0 the rest.
Table 1: Estimation results
|
Model 1
|
Model 2
|
Model 3
|
Model 4
|
L
|
15% L
|
50%
|
|
Laid Off
|
Odds Ratio
|
Odds Ratio
|
Odds Ratio
|
Odds Ratio
|
Media
|
0,45**
(3,46)
|
0,208***
(0,153)
|
|
0,5*
(5.24)
|
Infrastructure
|
0,1***
(9,247)
|
|
|
0,15***
(16.8)
|
Data
|
0,337*
(2,608)
|
0,5
(0,36)
|
|
|
Consumer
|
0,692
(0,65)
|
1,031
(0,778)
|
0,115*
(1,667)
|
0,214
(2.7)
|
Food
|
0,257
(2,148)
|
0,313
(0,241)
|
|
0,3
(3.81)
|
Transportation
|
0,415***
(2,835)
|
0,295*
(0,184)
|
|
0,261
(2.686)
|
Retail
|
0,128
(1,252)
|
0,3
(0,247)
|
|
0,15***
(15.1)
|
Recruiting
|
0,204
(1,579)
|
0,375
(0,259)
|
0,108*
(1,5549)
|
0,428
(4.47)
|
Finance
|
0,236
(1,603)
|
0,35*
(0,212)
|
0,5785
(8.267)
|
0,48*
(4.31)
|
Education
|
0,45
(5.141)
|
|
0,54***
(8,259)
|
0,1*
(1,36)
|
Security
|
|
0,75
(0,986)
|
|
0,15**
(21.3)
|
Travel
|
0.281
(0.328)
|
0,9
(0,644)
|
0,101*
(1,456)
|
0,642**
(6.17)
|
Healthcare
|
0,321*
(2,588)
|
0,375
(0,281)
|
|
0,2727
(3.45)
|
Real Estate
|
1
(0,781)
|
0,541
(0,349)
|
0,36***
(4,123)
|
0,1428
(1.78)
|
Marketing
|
0,421***
(2,781)
|
0,309**
(0,185)
|
0,54
(7,71)
|
0,1
(1.24)
|
Construction
|
0,135***
(1,728)
|
0,125*
(0,156)
|
|
|
HR
|
0,169
(1,474)
|
0,1
(0,829)
|
|
|
Product
|
0,45
(5,141)
|
0,125*
(0,156)
|
|
0,1*
(1,36)
|
Logistics
|
0,45*
(4,038)
|
0.375
(0.32)
|
|
|
Support
|
0,45
(5,141)
|
0,125*
(0,156)
|
|
0,1*
(1,36)
|
Other
|
0,135
(1,16)
|
0,437
(0,321)
|
|
0,1***
(8.77)
|
Cst
|
0,222***
(0,122)
|
0,2666***
(1,276)
|
0,006***
(0,006)
|
0.033
(0.023)
|
Observations
|
297
|
297
|
297
|
297
|
Source: authors
Regarding model 1, the probability of layoffs that are less than or equal to 15% of the start-up's workforce is 45% if it belongs to the "Media" branch, 10% if it is the "Infrastructure" branch, 33.7% if it is the "Data" branch, 41.5% if it is the "Transportation" branch, 32.1% if it is the "Healthcare" branch, 42.1% if it is the "Marketing" branch, 13.5% if it is the "Construction" branch and 45% if it is the "Logistics" branch.
As for model 2, the probability of making layoffs whose number is strictly greater than 15% of the Start-up's workforce, but less than or equal to 50% of the workforce is 20.8% if it belongs to the "Media" branch, 29.5% if it belongs to the "Transportation" branch, 35% if it belongs to the "Finance" branch, 30.9% if it belongs to the "Marketing" branch, 12.5% if it belongs to the "Construction" branch, 12.5% if it belongs to the "Product" branch and 12.5% if it belongs to the "Support" branch.
For model 3, the probability of layoffs that are strictly greater than 50% of the start-up's workforce, but less than or equal to 75% of the workforce is 11.5% for the "Consumer" branch, 10.8% for the "Recruiting" branch, 54% for the "Education" branch, 10.1% for the "Travel" branch and 36% for the "Real Estate" branch.
Finally, for model 4, the probability of making layoffs whose number is strictly greater than 75% of the Start-up's workforce, but less than or equal to 100% of the workforce is 50% for the "Media" branch, 15% for the "Infrastructure" branch, 15% for the "Retail" branch, 48% for the "Finance" branch, 10% for the "Education" branch, 15% for the "Security" branch, 64.2% for the "Travel" branch, 10% for the "Product" branch, 10% for the "Support" branch, and 10% for the other branches.
Overall, the results show that the "Media" industry is significant in all three models where it was tested, with a 50% probability of being in the highest layoffs. In the two models in which it was tested, the "Infrastructure" industry remains significant. The same is true for the "Construction" branch. The other branches most affected are: "Transportation" , "Product", "Support" and "Education" with two significant models out of three, and "Finance", "Travel" and "Marketing" with two significant models out of four. Apart from "HR" and "Food" which are not significant in all the models tested, all the other branches are significant in at least one of the models tested. Thus, we can see that the COVID-19 health shock did not impact all the branchs of activity in the same way and to the same extent in the US.
These results suggest that, in addition to the factors identified in previous work, the branch of activity in which an individual is employed determines his or her level of exposure to layoff. Thus, our results are in line with those of Brinca et al. (2020). The insignificance of the Food industry may be explained by the resilience of this industry during the health crisis. In general, the food industry maintained its activities and therefore its jobs. It has even undergone changes, notably with home delivery services. This is not the case, for example, in the "Construction", "Education", "Transportation" and "Travel" sectors, which were strongly affected by the total or partial containment measures, leading to waves of layoffs.
Therefore, the implementation of policies to support activity should take into account these disparities by targeting the most affected sectors. It is also important to analyze the factors that make certain industries more vulnerable than others, in order to provide better support and understanding of their difficulties.