Dinlersoz et al., Business Formation: A Tale of Two Recessions (2021-01)

Summary

CES working paper 21-01, dated January 2021, by Emin Dinlersoz, Timothy Dunne, John Haltiwanger and Veronika Penciakova, comparing business formation in the Great Recession and the COVID-19 Recession. It uses the Business Formation Statistics (BFS) data on business applications together with records from the Longitudinal Business Database (LBD). The paper states that by 2018 the startup rate of new employer businesses remained 24% lower than in 2006. It reports that business applications fell slowly and persistently in the Great Recession but had a sharp V-shaped recovery in the COVID-19 Recession, resulting in a surge in applications during 2020. According to the paper, the composition of applications in 2020 shifted toward those more likely to be nonemployers, which dampens projected transitions to employer businesses.

Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used

Cite this document

“Dinlersoz et al., Business Formation: A Tale of Two Recessions (2021-01).” Original at www2.census.gov. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/dinlersoz-et-al-business-formation-a-tale-of-two-recessions-2021-01-21c1e7bb4a/, SHA-256 5fe6866f7b485c51dca943f8136f3eb2e68327552e9cd34c021a3cf68da2d1c3.

Full text

Business Formation: A Tale of Two Recessions

Emin Dinlersoz
U.S. Census Bureau

Timothy Dunne
University of Notre Dame

John Haltiwanger
University of Maryland

Veronika Penciakova
Federal Reserve Bank of Atlanta

CES 21-01 January 2021

The research program of the Center for Economic Studies (CES) produces a wide range of
economic analyses to improve the statistical programs of the U.S. Census Bureau. Many of
these analyses take the form of CES research papers. The papers have not undergone the
review accorded Census Bureau publications and no endorsement should be inferred. Any
opinions and conclusions expressed herein are those of the author(s) and do not necessarily
represent the views of the U.S. Census Bureau. All results have been reviewed to ensure that no
confidential information is disclosed. Republication in whole or part must be cleared with the
authors.

To obtain information about the series, see www.census.gov/ces or contact Christopher Goetz,
Editor, Discussion Papers, U.S. Census Bureau, Center for Economic Studies 5K038E, 4600 Silver
Hill Road, Washington, DC 20233, CES.Working.Papers@census.gov. To subscribe to the series,
please click here.
Abstract

The trajectory of new business applications and transitions to employer businesses differ
markedly during the Great Recession and COVID-19 Recession. Both applications and
transitions to employer startups decreased slowly but persistently in the post-Lehman crisis
period of the Great Recession. In contrast, during the COVID-19 Recession new applications
initially declined but have since sharply rebounded, resulting in a surge in applications during
2020. Projected transitions to employer businesses also rise but this is dampened by a change
in the composition of applications in 2020 towards applications that are more likely to be
nonemployers.

*
We thank Shawn Klimek and Scott Ohlmacher for helpful comments. Any opinions and conclusions expressed
herein are those of the authors and do not represent the views of the U.S. Census Bureau, the Federal Reserve
System, Board of Governors, or their staff. The Census Bureau’s Disclosure Review Board and Disclosure Avoidance
Officers have reviewed this data product for unauthorized disclosure of confidential information and have
approved the disclosure avoidance practices applied to this release (Approval Number CBDRB-FY21-062). John
Haltiwanger is also a part-time Schedule A employee of the Census Bureau.
Dinlersoz: emin.m.dinlersoz@census.gov
Dunne: University of Notre Dame; tdunne1@nd.edu
Haltiwanger: haltiwan@econ.umd.edu
Penciakova: veronika.penciakova@atl.frb.org
How have recent recessions affected business formation? The startup rate of new employer
businesses declined sharply in the Great Recession and rose slowly in the subsequent recovery.
Indeed by 2018, the startup rate remained 24% lower than in 2006 (see online appendix). There
is evidence that the COVID-19 Recession is leading to widespread business closure (Leland
Crane, Ryan Decker, Aaron Flaaen, Adrian Hamins-Puertolas and Christopher Kurz, 2020),
but less is known about its impact on business formation.
In this paper, we use a new Census Bureau data series on applications for Employer Iden-
tification Numbers (EIN), the Business Formation Statistics (BFS), to describe the path of busi-
ness formation in the Great Recession and the COVID-19 Recession. There are three advan-
tages of the BFS. First, it can shed light on the path of current and future business formation
because new employer businesses must obtain an EIN in order to file payroll taxes. Second,
it is currently published on a weekly basis, enabling a high frequency and timely analysis.
Finally, Kimberly Bayard, Emin Dinlersoz, Timothy Dunne, John Haltiwanger, Javier Miranda
and John Stevens (2018) shows that new business applications can be used to track the fluctu-
ations in actual new employer business formation.
By combining the BFS data on applications with records of employer businesses from the
Longitudinal Business Database (LBD), we can measure both the evolution of business appli-
cations (BA) and the observed or modeled transition of BA to employer businesses. Our find-
ings show marked differences in the path of BA between the Great Recession and the COVID-
19 Recession. The fall in BA is slow and persistent during the Great Recession, whereas BA
experience a sharp V-shaped recovery in the COVID-19 Recession. However, the composition
of BA shifts substantially during the COVID-19 Recession toward those with a lower propen-
sity to transition to employer businesses. This composition shift implies that the surge in new
BA in 2020 is likely to yield a more substantial increase in new non-employer businesses rela-
tive to new employer businesses.

1 Measuring Business Formation
The analysis uses the BFS, which tracks the weekly number of BA by state. The BFS is derived
from administrative data from the IRS on EIN applications. All employer businesses in the
United States are required to have an EIN to file payroll taxes. New non-employer businesses
also file for an EIN if forming a partnership or an incorporated business. Even new sole pro-
prietor non-employers often file for an EIN to facilitate their business activity (e.g., working
with other businesses or opening a business bank account). The application form includes the
name and address of the applicant and business, business start date, type of business entity,
principal industry, and planned date of initial wage payments. The filing date and business lo-
cation information are used to aggregate individual applications to the state level at a weekly
frequency. The main data series analyzed, BA, represents the number of EIN applications

1
received in a state in a specific week of the year.
We complement the confidential microdata underlying the weekly BFS with the LBD. The
LBD contains firm and establishment-level information on age, location, industry, number of
employees, quarterly payroll, and EIN for the near-universe of employer businesses in the
United States. Using the EIN, the BFS is matched to the LBD to identify the incidence and
timing of transitions to employer businesses. In particular, we focus on transitions that occur
within the first four quarters of the application date, which are formally referred to in the
BFS as BF4Q. Bayard et al. (2018) find that the majority of the transitions from application to
employer business occur within the first year of the EIN filing.
The matched data are employed in two ways to obtain the number of applications that
transition, or are likely to transition, to employer businesses. For the Great Recession, the
count of realized transitions is aggregated to the state-week level, as this is directly observable
in the matched data. For the COVID-19 Recession, transitions to employer business are not
yet measured in the LBD, which is currently available through 2018. Instead, we project the
number of transitions during 2019 and 2020 using a linear probability model (LPM) following
the approach discussed in Bayard et al. (2018). The model generates a predicted probability of
transitioning to an employer business for each application, including the out-of-sample period
from 2019 through 2020. The probabilities are aggregated to yield an estimate of the number
of applications that will transition to employer status at the state-week level. This approach
allows for the number of estimated transitions to vary by both the number of applications
received and the composition of those applications (see online appendix).
We use a regression framework to evaluate whether the Great Recession and COVID-19
Recession significantly affected the trajectory of BA and transitions. In doing so, we first con-
struct measures of cumulative applications and transitions for crisis and normal time bench-
mark (reference period), forwards and backwards relative to a week 0 defined below. This
approach smooths out seasonality observed in the weekly flow of these data. Second, we con-
struct our outcome variables as the growth rate in cumulative applications (or transitions) dur-
ing the crisis period relative to the reference period. Finally, we estimate application-weighted
(or transition-weighted) regressions of this growth rate on distance from week 0 dummies and
state fixed effects to capture how deviations from normal times evolve within states during
each crisis. Note that each distance dummy contains two weeks, with the week 0 dummy
containing weeks zero and one.

2 Evolution of Business Applications and Transitions
We begin by comparing the paths of BA and transitions for the two recessions. Figure 1a
depicts the growth rates of cumulative applications and transitions during the crisis period
relative to the reference period of the Great Recession. All statistics should be interpreted as

2
deviations relative to the two-week period containing week 0, which is normalized to zero.1
Focusing on deviations relative to week 0 allows us to highlight high frequency cyclical vari-
ation, abstracting from lower frequency trends. The bands represent 95 percent confidence
intervals. Application growth rates (the black line) decline after week 0 and remain persis-
tently lower 52 weeks after the Lehman collapse. The second line, depicting transition growth
rates, exhibits a sharp initial decline and a subsequent flattening out.
Figure 1b focuses on the COVID-19 Recession.2 In March and April 2020, the growth rate
of applications drops sharply, but subsequently experiences a strong rebound. Indeed within
18 weeks, applications have recovered and in subsequent weeks rise sharply above the refer-
ence week level of activity. The differences in projected transition growth rates fall even more
sharply than applications and recover more slowly. As will become clear below, to ascertain
the quantitative implications for projected transitions in 2020, we need to consider patterns
without normalizing week 0 differences to zero.
Figures 2a and 2b highlight the differences in the evolution of transitions by depicting the
implied differences in transition rates from Figure 1. They represent the difference between the
crisis and reference period in the transition rate of cumulative applications. Figure 2a focuses
on the Great Recession, and shows a decline post-Lehman, followed by a recovery in transition
rates within 30 weeks.3 During COVID-19, depicted in Figure 2b, the projected transition rate
shows a sharp decline at the onset of the crisis and no recovery 30 weeks later.
The contrast between the Great Recession and the COVID-19 Recession is stark. Applica-
tions dropped and remained low in the Great Recession, but after an initial decline, transition
rates recovered. While actual transitions fell, their decline was driven by the fall in applica-
tions, rather than a persistent decline in transition rates. In contrast, during the COVID-19
Recession applications initially fell sharply and rebounded several weeks later, while the tran-
sition rate also fell but remain depressed through week 30. Differences between the two re-
cessions in the evolution of transition rates reflect differences in the compositional changes in
applications.

1 Week 0 is the two-week period ending September 20, 2008 in the crisis period, which coincides with the

Lehman Brothers collapse, and the equivalent week in 2006 for the reference period.
2 Week 0 is the two-week period ending March 14, 2020, and the equivalent week in 2019 for the reference

period. The growth rate of applications during the crisis relative to the reference period are shown for weeks
0 through 40 using the public domain BFS. The growth rate of transitions during the crisis relative to reference
period is reported for weeks 0 through 30.
3 The differences are normalized to zero in the two-week period containing week 0. Some caution is required

in interpreting the path of the transition rates post-Lehman. As shown in the appendix there is a decline in
transition rates from 2006 to 2008 pre-Lehman – much of this occurs in 2007, before the Great Recession.

3
DHS: cumulative crisis relative to reference DHS: cumulative crisis relative to reference

−.44 −.34 −.24 −.14 −.04 .06 .16 .26 −.44 −.34 −.24 −.14 −.04 .06 .16 .26

−25 −20 −15 −10 −5 0 5 10 15 20 25 30 35 40 45 50 55 −25 −20 −15 −10 −5 0 5 10 15 20 25 30 35 40
Time since start of crisis (2−week periods) Time since start of crisis (2−week periods)

Business Applications Transitions (actual) Business Applications Transitions (projected)
Figure 1a: Great Recession −− Applications and Transitions Figure 1b: COVID −− Applications and Transitions

Figure 1: Business Applications and Transitions

Difference: cumulative crisis relative to reference Difference: cumulative crisis relative to reference

−.02 −.01 0 .01 .02 −.02 −.01 0 .01 .02

−25 −20 −15 −10 −5 0 5 10 15 20 25 30 35 40 45 50 55 −25 −20 −15 −10 −5 0 5 10 15 20 25 30 35 40
Time since start of crisis (2−week periods) Time since start of crisis (2−week periods)
Figure 2a: Great Recession −− Transition Rate (actual) Figure 2b: COVID −− Transition Rate (projected)

Figure 2: Transition Rates

3 Changing Composition of Business Applications
The decline in observed and projected transition rates during both crises is indicative of com-
positional changes in applications playing a role in business formation during recessions. To
explore the relevance of these compositional changes, we first disaggregate applications into
two groups – business applications with planned wages (WBA) and non-wage business appli-
cations (non-WBA). WBA are submissions that indicate the intent to pay wages. Bayard et al.
(2018) report the transition rate for WBA to be substantially higher than non-WBA.
Figures 3a and 3b contrast the differences in WBA and non-WBA growth rates in cumula-
tive differences for the two recessions. In the Great Recession, WBA initially fell more rapidly
than non-WBA. However, after about 20 weeks, the gap between the decline in WBA and
non-WBA stabilized.
During the COVID-19 Recession, both types of applications initially declined sharply. Af-
ter around 4 weeks, non-WBA began to rebound, while WBA continued to decline. Though
WBA started to rebound after about 8 weeks, by week 40, the gap between the cumulative dif-
ferences in the two types of applications remains substantial. This pattern is consistent with
the fall in the projected transition rate.

4
Other compositional changes are also important for understanding the decline in projected
transition rates during the COVID-19 Recession. A special projects release from the weekly
BFS reports that BA among non-store retailers (NAICS 454) account for the vast majority of
the observed surge in BA in 2020. This is not surprising, given the pandemic has provided
strong incentives for shifting retail activity to online transactions. Yet, in 2018 over 90% of
non-store retailers are non-employer businesses (see combined NES-CBP data). Therefore,
this concentration of applications among non-store retailers also helps explain the declining
transition rate in Figure 2b.
To further highlight the importance of the changing composition of applications during the
COVID-19 Recession, we evaluate the number of new businesses that would be formed during
this period if applications filed during the COVID-19 Recession had the same composition as
those filed during 2019. The projected number of transitions would be 17.2 percent higher
during COVID-19 if the composition of applications remained unchanged from the reference
period.
While the compositional change implies a reduction in the transition rate, our findings still
imply an increase in employer startups in 2020, relative to 2019, from the increase in applica-
tions. To understand this implication, it is instructive to consider the patterns of transitions
in the COVID-19 Recession without normalizing the week 0 difference to zero. In online ap-
pendix Figure A1, we find that by week 30, cumulative transitions are about 8% higher in 2020
relative to 2019. Consistent with this finding, Figure A4 in the online appendix shows that ap-
plications with a high propensity to transition to employer business (HBA) are 17% higher by
week 40 in 2020 relative to 2019. At the national level, this implies about 200 thousand more
HBA by week 40 in 2020 relative to 2019. These findings suggest that the surge in applications
in 2020 will translate into an increase in actual startups in 2020 relative to 2019 – it just will not
be as large as implied by the increase in overall applications.
The compositional changes imply an even larger surge in non-employer businesses in 2020.
We find that the cumulative increase in applications for likely non-employers (Figure A.4, non-
HBA) is 30% in week 40 of 2020 relative to 2019. At the national level, this implies about 600
thousand more non-HBA by week 40 in 2020 relative to 2019. Non-employer businesses are
an important source of business activity. Counting each non-employer establishment as one
job and combining with County Business Patterns data, non-employers accounted for about
17 percent of jobs in 2018. Many non=employers generate a small amount of revenue, but EIN
non=employers are three times as large in terms of revenue as SSN non=employers (see online
appendix for more discussion).

5
DHS: cumulative crisis relative to reference DHS: cumulative crisis relative to reference

−.44 −.34 −.24 −.14 −.04 .06 .16 .26 .36 −.44 −.34 −.24 −.14 −.04 .06 .16 .26 .36

WBA Non−WBA WBA Non−WBA
Figure 3a: Great Recession −− WBA vs. non−WBA Figure 3b: COVID −− WBA vs. non−WBA

Figure 3: WBA versus non-WBA

4 Conclusion
Our findings show that the trajectory of business applications and transitions to employer
businesses differ markedly during the Great Recession and COVID-19 Recession. The persis-
tent reduction in actual employer business formation during the post-Lehman crisis period
of the Great Recession is primarily accounted for by a decrease in applications, as transition
rates to employer businesses recover after an initial decline. In contrast, during the COVID-19
Recession new applications initially declined but have since sharply rebounded, resulting in a
surge in applications during 2020. Although our analysis predicts a rise in employer startups
in 2020 relative to 2019, this rise is dampened by a reduction in the projected transition rate
of applications to employer businesses. Compositional changes in applications underlie this
reduction, including a decrease in the share of applications with intent to pay wages and an
increase in the share of applications in industries dominated by non-employers (e.g., non-store
retailers). These compositional changes suggests some caution regarding the observed surge
in new applications in 2020 in terms of their implications for new employer businesses. On the
other hand, the surge in applications of likely non-employers in 2020 has been quite dramatic.

6
A Appendix
Applications and Transitions Data: The data for all business application series (BA, HBA,
WBA) are obtained from the public domain Business Formation Statistics (BFS). The BFS
publishes quarterly data on business applications, as well as realized (BF4Q) and projected
(PBF4Q) transitions. Starting soon after the onset of the COVID-19 Recession, the BFS began
releasing weekly data on business applications at the national, regional, and state level. In our
analysis, we use the public domain weekly BFS business application data. Since actual and
projected transitions are only available at a quarterly frequency in the public domain BFS, the
actual and projected transitions and transition rates reported in this paper are based on the
integration of the BFS and Longitudinal Business Database (LBD) micro data, as described in
the main text. In addition, we used the newly redesigned LBD in this process which enabled
us to project transitions for 2019 and 2020 based on actual transitions through 2018.
Projecting Transitions: The empirical model used to predict transitions is described in
detail in Bayard et al. (2018). Briefly, the projections are based on a linear probability model
(LPM), which relates observed transitions of business applications to employer businesses
in the LBD to a rich set of predictors from the EIN application form. The variables include
indicators of the business start date, type of entity, industry, limited liability status, reason for
applying, and wage date. The model also controls for location (state), week of application
submission within the year, presence of prior EIN, and whether the application indicates a
trade name, an executor’s name, or a distinct business address. The covariates of the empirical
model also include a rich set of interactions between industry, wage date, type of entity, and
reason for applying. Figure 6 of Bayard et al. (2018) shows that projected transitions track
actual transitions closely, both within and out of sample.
Crisis and Reference Periods: For the reference period (or normal time benchmark) for the
Great Recession, we use 2006 to avoid any overlap between the crisis and reference periods
between weeks 0 and 52. For the COVID-19 Recession, at the time the projection model was
estimated, BFS data were available through the week ending October 3, 2020 (which is 30
weeks after the start of the crisis period). Since the BFS currently releases new data weekly, in
the main text and appendix figures we use applications through the week ending December
19, 2020 in our analysis. The estimates in the figures for 40 weeks after the crisis period reflect
the two week period that ends December 19 , 2020.
Growth Rate Calculation: For the analysis, we compute growth rates for cumulative ap-
plications and transitions using the DHS (Steven Davis, John Haltiwanger and Scott Schuh
(1996)) growth rate, defined as ( Xcw − Xrw )/(0.5 ∗ ( Xcw + Xrw )) for variable X, where r is the
reference period, c is the crisis period and w is the week through which X has been calcu-
lated. The DHS growth rate is a second order approximation of the log first difference and
accommodates zeroes.

7
Weighting: In the main text, we report application-weighted (or transition-weighted) re-
gression results, where the weights are the relevant state-level DHS denominator. In unre-
ported results, we estimated both population-weighted and unweighted regressions. These
regressions yield results very similar to those reported in the main text.
Startup and Non-employer Statistics: The discussion in the main text of the decline in
actual new employer startups draws from the Business Dynamic Statistics. Further analysis of
the decline in employer startups during the Great Recession can be found in Ryan Decker, John
Haltiwanger, Ron Jarmin and Javier Miranda (2014) and Steven Davis and John Haltiwanger
(2019). The discussion in the main text on the importance of non-employers draws from Steven
Davis, John Haltiwanger, Ron Jarmin, C.J. Krizan, Javier Miranda, Alfred Nucci and Kristin
Sandusky (2009) and the 2018 Non-employer Statistics. Davis et al. (2009) find that more than
ten percent of EIN non-employers ultimately transition to employer businesses. Most of these
transitions occur at least one year after the non-employer business started. Consequently, such
transitions are not included in the application transition rates discussed in the main text.
Additional Figures: Figures A1-A3 are analogous to Figures 1-3, but without normaliz-
ing the week 0 difference to zero.4 Figure A4 depicts the growth rate of cumulative high-
propensity business applications (HBA) and non high-propensity business applications (non-
HBA) in the crisis relative to the reference period, without normalizing week 0 to zero. As
discussed in Bayard et al. (2018), non-HBA have a very low probability of transiting to em-
ployer businesses in the next 4 quarters (less than 3%), while HBA have a 27% transition rate.
HBA are those applications that have either (i) planned wages, (ii) are an application for a
corporation, or (iii) are in small number of industries that have notably higher transition rates.
These characteristics are among those used in the much richer LPM model to project tran-
sitions of applications to employer businesses. HBA are intended as a transparent way to
classify applications on core characteristics.
Figure A1 shows that there is a much larger gap at week 0 between applications and transi-
tions in the Great Recession (Figure A1a) compared to the COVID-19 Recession (Figure A1b).
Indeed, in the Great Recession, the gap is large and negative, while in the COVID-19 Recession
the gap is smaller and positive. For the Great Recession, this means that the growth rate in ap-
plications from reference week 0 to crisis week 0 is close to zero, while the analogous growth
rate for transitions is very negative. This large negative gap, observed during the Great Reces-
sion, is consistent with the large negative decline in the transition rate in Figure A2a in week
0. Importantly, these patterns indicate that the transition rate declined before the Lehman col-
lapse (between 2006 and 2008). Consequently, our inference of the decline in transitions during
the Great Recession being mostly due to the decline in applications refers specifically to the

4 In practice, we run the same regressions, but rather than reporting the coefficients and 95% confidence

interval of the distance from week 0 dummies, we report the predictive margins and their 95% confidence interval
evaluated at the distance from week 0 dummies.

8
post-Lehman period. To help understand the pre-Lehman decline in the transition rate, a few
additional findings are relevant. First, from the BFS, the composition of applications changed
significantly in 2007 even before the beginning of the recession in December 2007. In January
2007, the share of applications that are HBA is 62% and WBA is 38%. In November 2007, these
shares are 51% and 25% respectively. These compositional changes imply a significant decline
in the transition rate pre Great Recession which is consistent with the patterns in Figure 6 of
Bayard et al. (2018). The latter shows both actual and projected transitions declined substan-
tially in 2007 pre Great Recession. In the main text we intentionally abstract from these pre
Great Recession changes in applications and transitions as this reflects lower frequency factors
underlying the well-known negative trend in employer startups in the U.S. that accelerated in
the post-2000 period. As discussed in the main text, we abstract from these trend effects by
focusing on deviations from week 0.
In contrast, in the COVID-19 Recession, the transition rate at week 0 in 2020 is slightly
higher than in week 0 in 2019. There are not sizable pre-recession trend effects in 2019 com-
pared to early 2020. Figure A3 shows similar patterns to Figure A1 for the WBA and non-
WBA during both recessions. Importantly, during the COVID-19 Recession, both WBA and
non-WBA show marked increases by week 40 in 2020, relative to 2019.
For the COVID-19 Recession, Figure A4 shows that cumulative HBA are 17% higher by
week 40 in 2020, relative to 2019. It is instructive to compare this increase of 17% in HBA to the
25% increase in overall applications in Figure A1. This difference is consistent with our finding
that during the COVID-19 Recession, the transition rate has declined as the share of HBA in
overall applications has declined. However, HBA are increasing substantially in 2020, relative
to 2019, which implies that employer startups are likely to be higher in 2020 than in 2019. At
the same time, as discussed in the main text, the increase in employer startups is likely to be
smaller than suggested by the overall increase in business applications.

DHS: cumulative crisis relative to reference DHS: cumulative crisis relative to reference

−.34 −.24 −.14 −.04 .06 .16 .26 −.34 −.24 −.14 −.04 .06 .16 .26

Business Applications Transitions (actual) Business Applications Transitions (projected)
Figure A.1a: Great Recession −− Applications and Transitions (non−normalized) Figure A.1b: COVID −− Applications and Transitions (non−normalized)

Figure A.1: Business Applications and Transitions (non-normalized)

9
Difference: cumulative crisis relative to reference Difference: cumulative crisis relative to reference

−.04 −.03 −.02 −.01 0 .01 −.04 −.03 −.02 −.01 0 .01

−25 −20 −15 −10 −5 0 5 10 15 20 25 30 35 40 45 50 55 −25 −20 −15 −10 −5 0 5 10 15 20 25 30 35 40
Time since start of crisis (2−week periods) Time since start of crisis (2−week periods)
Figure A.2a: Great Recession −− Transition Rate (actual, non−normalized) Figure A.2b: COVID −− Transition Rate (projected, non−normalized)

Figure A.2: Transition Rates (non-normalized)

DHS: cumulative crisis relative to reference DHS: cumulative crisis relative to reference

−.48 −.38 −.28 −.18 −.08 .02 .12 .22 .32 −.48 −.38 −.28 −.18 −.08 .02 .12 .22 .32

WBA Non−WBA WBA Non−WBA
Figure A.3a: Great Recession −− WBA vs. non−WBA (non−normalized) Figure A.3b: COVID −− WBA vs. non−WBA (non−normalized)

Figure A.3: WBA versus non-WBA (non-normalized)

DHS: cumulative crisis relative to reference DHS: cumulative crisis relative to reference

−.28 −.18 −.08 .02 .12 .22 .32 −.28 −.18 −.08 .02 .12 .22 .32

HBA Non−HBA HBA Non−HBA
Figure A.4a: Great Recession −− HBA vs. non−HBA (non−normalized) Figure A.4b: COVID −− HBA vs. non−HBA (non−normalized)

Figure A.4: HBA versus non-HBA (non-normalized)

10
References
Bayard, Kimberly, Emin Dinlersoz, Timothy Dunne, John Haltiwanger, Javier Miranda,
and John Stevens. 2018. “Early Stage Business Formation: An Analysis of Applications for
Employer Identification Numbers.” NBER Working Paper No. 24364.

Crane, Leland, Ryan Decker, Aaron Flaaen, Adrian Hamins-Puertolas, and Christopher
Kurz. 2020. “Business Exit during the COVID-19 Pandemic: Non-Traditional Measures in
Historical Context.” FEDS Working Paper No. 2020-089.

Davis, Steven, and John Haltiwanger. 2019. “Dynamism Diminished: The Role of Housing
Markets and Credit Conditions.” NBER Working Paper No. 25466.

Davis, Steven, John Haltiwanger, and Scott Schuh. 1996. Job Creation and Destruction. MIT
Press.

Davis, Steven, John Haltiwanger, Ron Jarmin, C.J. Krizan, Javier Miranda, Alfred Nucci,
and Kristin Sandusky. 2009. “Measuring the Dynamics of Young and Small Businesses:
Integrating the Employer and Nonemployer Universes.” In Producer Dynamics: New Evidence
from Micro Data. 329–366. NBER and University of Chicago Press.

Decker, Ryan, John Haltiwanger, Ron Jarmin, and Javier Miranda. 2014. “The Role of En-
trepreurship in U.S. Job Creation and Economic Dynamism.” Journal of Economic Perspectives,
24(3): 3–24.

File and source

File: dinlersoz-dunne-haltiwanger-penciakova-2021-ces-wp-21-01-business-formation-two-recessions.pdf
Size: 376,229 bytes
SHA-256: 5fe6866f7b485c51dca943f8136f3eb2e68327552e9cd34c021a3cf68da2d1c3
Our copy: dinlersoz-dunne-haltiwanger-penciakova-2021-ces-wp-21-01-business-formation-two-recessions.pdf
Original: www2.census.gov