WPS8236


Policy Research Working Paper                        8236




  Small Firm Death in Developing Countries
                               David McKenzie
                             Anna Luisa Paffhausen




Development Research Group
Finance and Private Sector Development Team
November 2017
Policy Research Working Paper 8236


  Abstract
  Small firms are an important source of income for the poor                         half of all firms observed to be operating at a given point
  in developing countries, and the target of many interven-                          in time are dead within 6 years. Death rates are higher for
  tions designed to help them grow. But there is no systematic                       small firms in richer countries, younger firms, retail firms,
  information on the failure or death of such firms. The paper                       less productive and less profitable firms, and those whose
  puts together 16 panel surveys from 12 different developing                        owners are female and not middle-aged. The paper proposes
  countries to develop stylized facts from over 14,000 firms                         three theories of why small firms die: firm competition and
  on how much firm death there is; on which types of these                           firm shocks, occupational choice, and non-separability from
  firms are most likely to die; and on why they die, paying                          the household. It finds the cause of firm death to be het-
  careful attention to issues of measurement and attrition. The                      erogeneous, with different subgroups of firms more likely
  authors find small firms die at an average rate of 8.3 percent                     to die for reasons consistent with each of these theories.
  per year over the first five years of following them, so that




  This paper is a product of the Finance and Private Sector Development Team, Development Research Group. It is part of
  a larger effort by the World Bank to provide open access to its research and make a contribution to development policy
  discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org.
  The authors may be contacted at dmckenzie@worldbank.org.




          The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
          issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
          names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
          of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
          its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.


                                                        Produced by the Research Support Team
                           Small Firm Death in Developing Countries#

               David McKenzie, Development Research Group, World Bank
Anna Luisa Paffhausen, University of Passau and Development Research Group, World Bank




Keywords: firm death; survival; microenterprise dynamics
JEL codes: O12, O17, D22, L26




#
  We gratefully acknowledge funding for this work from the Knowledge for Change Program (KCP) and Strategic
Research Program (SRP) of the World Bank. We thank our collaborators and funders on the many different surveys
which this work draws on; and Lorenzo Casaburi, Markus Goldstein, and participants at the 1st TDC/LSE/CEPR
workshop in Development Economics for helpful comments.
    1. Introduction
Twenty-seven percent of the non-agricultural labor force in developing countries consists of
self-employed business owners with no employees (Gindling and Newhouse, 2014), and 99
percent of the firms in many poor countries have 10 workers or fewer (McKenzie, 2017). These
firms are an important source of income for the poor, and many policy interventions are
designed to help people start and grow such firms. Yet there has been much less attention
devoted to the death of such firms, with no systematic evidence available as to the rate of small
firm death, which firms are more likely to die, and why they die. Almost two-thirds of published
randomized experiments testing policy interventions for small firms in developing countries
ignore firm death completely, neither reporting the death rate nor examining it as an outcome.1


Understanding the prevalence, characteristics, and causes of firm death is important for poverty,
productivity, and policy. Since self-employment is an important source of income for the poor,
firms shutting down could mean a large loss in income for firm owners. A growing body of
literature (e.g. Hsieh and Klenow, 2009) has noted large misallocation of resources in
developing countries. Firm death can improve aggregate productivity if less productive firms
die and reallocate resources and customers to more efficient competitors. The optimal policy
response to firm death then depends on whether it involves large income losses for the poor,
and whether it is efficiently selecting out the least productive firms or not.


This paper provides systematic evidence on firm death in small firms in developing countries
by collating data on more than 14,000 small firms from 16 firm panel surveys in 12 countries,
enabling estimation of the rate of firm death over horizons as short as 3 months, and as long as
17 years. This overcomes many of the limitations of the existing literature on firm death in
developing countries, which have had to rely on a small number of time periods, usually just in
a single country, often with small samples of micro firms, and high rates of attrition.2 Moreover,
we included detailed questions in nine of these surveys to measure cause of death, and most of
our surveys also continue to track the firm owner after their firm has died to provide data on
what they do once their firm dies.




1
  Table A.1 provides details from 35 published randomized experiments testing different interventions for small
firms in developing countries. Out of these 35 experiments, only 12 report the firm death rate and 13 examine firm
death as an outcome.
2
  We discuss how firm death has been measured in the existing literature in the next section, and in Appendix
Figure A.2. and Table A.3, compare our estimates of death rates to those in existing studies.
                                                                                                                2
We begin by estimating the prevalence of firm death. Existing estimates of the rate at which
small firms die in developing countries range from as low as 3 percent per year (Frazer, 2005;
Davies and Kerr, 2015) to over 30 percent annually (Fajnzylber et al, 2006; Nagler and Naudé,
2017). We find the rate of firm death is approximately linear over time horizons up to five
years, averaging 8.3 percent per year, but then declines over longer time horizons, so that firm
death rates average less than 5 percent per year when looking over intervals of 10 years or
longer. The result is that half the stock of firms operating at a given point in time will die within
the next six years.


We then examine which firms are more likely to die, through examining the extent to which
firm and owner characteristics predict firm death. Younger firms are more likely to die, with
the first year particularly failure-prone: a firm in its first year is twice as likely to die in the next
year as a five-year old firm. Retail firms and less profitable firms are more likely to die. Death
rates are higher for younger and older entrepreneurs than for middle-aged owners, and higher
for female-owners than males. Firm death rates are higher in richer developing countries than
poorer countries.


Finally, we turn to why firms die. We consider three separate theories of firm death. The first
views firm death as arising from firm competition and firm-level shocks causing firms to make
losses and exit. Here firm death is involuntary, tends to cull less productive firms, and lowers
the income of the owner. A second theory predicts firm death as arising from occupational
choice decisions of the owner. Firm death can be voluntary here, with owners choosing to shut
down when better outside opportunities arise, with less impact on poverty and less selectivity
on productivity. A final theory is that firm death results from non-separability of business and
household decisions due to imperfect markets. The result is that illness and shocks in the
household can cause the business to have to close, resulting in income loss for the owner and
not necessarily the least productive firms closing.


We bring these three theories to our data by examining the extent to which firm death occurs
for more, or less, productive firms, measuring the change in total labor earnings of the owner
after death, and using cause of death data. On average, firms that die are less productive to
begin with, and their owners suffer a fall in labor earnings. This is most consistent with the firm
competition theory, and least consistent with firm death being a voluntary occupational choice.
However, we find considerable heterogeneity in the cause of death, with subsets of firms


                                                                                                       3
appearing to die in ways consistent with each of these three theories. We characterize which
types of firms are most likely to die for each of these three reasons.

The remainder of the paper is structured as follows: Section 2 discusses the different approaches
to measuring firm death that have been used in the existing literature; Section 3 outlines our
approach to firm death and the sample we have put together; Section 4 examines the prevalence
of firm death; Section 5 which firms are more likely to die; Section 6 why firms die; and Section
7 concludes.

2. Measurement of Firm Death in the Existing Literature
At least four different approaches have been used to measure firm death in the existing
literature, each with particular strengths and drawbacks.3 A first approach is seen in some of
the earliest empirical work on the topic, undertaken by Carl Liedholm and Donald Mead in the
early 1990s in several sub-Saharan African countries (Liedholm and Mead, 1995; Mead and
Liedholm, 1998). They carried out cross-sectional “closed firm” surveys by asking a random
sample of households about enterprises they previously ran, but no longer operate, in addition
to those they currently run. Based on this data, they report an average closure rate of 12.9
percent per year; although using the same data McPherson (1995) reports annual hazard rates
of firm death of 3 to 4 percent per year. This approach has the advantage of allowing for large
samples, and for representativeness of the areas in which sampling occurs. However, as the
authors themselves acknowledge, it is likely to be much less accurate than panel surveys which
track businesses over time, since people may forget or not want to talk about businesses that
failed, and there is likely to be substantial recall error as to exactly when the business closed.
Moreover, it will not capture people who close their businesses and then migrate out of the area.


A second approach has been to use manufacturing censuses or company registers. Examples
include Roberts and Tybout (1996) who provide annual firm exit rates of 10.8 percent in Chile,
11.1 percent in Colombia, and 6.0 percent in Morocco; Bartelsman et al. (2009) who graphically
show annual exit rates of approximately 4 to 5 percent in 10 Latin American and Eastern
European economies; Klapper and Richmond (2011) who use a firm census of formal firms in
Côte d’Ivoire and find annual exit rates of around 10 percent; and Shiferaw (2009) who reports


3
  A fifth, related, approach has been to use panel labor force surveys to track movements into and out of self-
employment status (e.g. Demirgüc–Kunt et al. 2011; Fajnzylber et al, 2006). This does not enable examination of
whether a particular firm survives, since individuals can shut down one firm and start another one, whilst remaining
self-employed. Furthermore, labor force surveys typically do not contain information on the firm such as firm age,
number of employees, or profitability.
                                                                                                                  4
annual exit rates of 16 percent for privately-owned Ethiopian manufacturing firms in the census.
These censuses offer the advantage of providing population rather than sample data on the
dynamics of larger firms. However, they are unsuited for examining the death of small firms
for several reasons. First, many of the censuses are restricted to firms with 10 or more workers,
and/or those which are formally registered, and they are extremely unlikely to capture non-
visible businesses operating within households. Second, Shiferaw (2009) notes that it is not
possible to distinguish firms that have closed from those that have switched sector out of
manufacturing, or which have fallen below the cut-off point for inclusion. Third, they do not
include firms in the retail and service sectors, which comprise the majority of small businesses.


A third approach is to use multi-purpose household living standards panel surveys such as the
World Bank’s Living Standards Measurement Study (LSMS). This approach was used in
Vietnam by Vijverberg and Haughton (2004) who report 61 percent of firms dying over a five-
year period; and McCaig and Pavcnik (2016) who find 30 to 35 percent death rates over two
year periods. Kraftt (2016) uses the Egypt Labor Market Panel Survey to find 51 percent of
firms died over an 8-year period, and 61 percent over a subsequent 6-year period. The
advantages of these datasets are that the samples can be large and representative, and they
capture micro and small businesses operating within households. However, these authors note
several challenges to the use of these household surveys to track enterprises. Since the surveys
do not identify specific enterprises by either name or code, authors need to match firms over
time based on characteristics such as the age of the firm, identity of which household member
runs the firm, and its sector. This can result in measurement errors if firms switch their sector
of business or have trouble recalling the age of the firm. Further downsides are that the surveys
are often conducted at irregular intervals, limiting the periods over which death rates can be
measured, and often contain relatively little information on firm characteristics to enable
exploration of which firms die and why.


The final approach is then to use dedicated firm panel surveys to track a sample of firms over
time. The main challenge for this approach has been the dearth of firm panel surveys in
developing countries, small sample sizes, and the difficulty of tracking firms with relatively
low attrition. This approach was first tried by Mead and Liedholm (1998), who report an annual
closure rate for micro and small enterprises in the Dominican Republic of 29 percent in 1992
and 22 percent in 1993, based on a sample of unspecified size with an unspecified attrition rate.
They note that they also tried the approach in Zimbabwe, but were unable to relocate 42 percent


                                                                                               5
of firms. Panel surveys of manufacturing firms in several sub-Saharan African countries were
collected as part of the Regional Program on Enterprise Development (RPED) by Oxford
University and the World Bank, and have been used in several papers on firm dynamics (Frazer
2005; Söderbom et al, 2006; Sandefur, 2010). These surveys typically include 100-200
manufacturing firms per country, and do not include household enterprises. They report firm
death rates of 6.3 percent over two years (Frazer, 2005) and 19 to 44 percent over 5 years
(Söderbom et al, 2006), although they do not report attrition rates. Davies and Kerr (2015)
conducted a 10-year follow-up survey of a random sample of 1000 firms surveyed in the 2003
Ghanaian manufacturing census. They could not find what had happened to 29 percent of the
386 firms with under 10 workers in their sample, while 25 percent had died.


This review of the existing literature shows that there has been relatively little evidence on the
rates of firm death in the most prevalent types of enterprises found in developing countries:
micro and small household enterprises that include retail and service firms in addition to
manufacturing. Few of these existing studies measure attrition rates, nor do any provide bounds
on what the rate of firm death will be when this is taken into account. A number of the studies
provide descriptive information on what types of firms are more likely to die, and we will
compare our findings on these correlates in Section 5. However, the fact that firm death is only
a fraction of the sample, and that single country samples have often been relatively small, has
meant that studies typically have relatively few deaths with which to examine the characteristics
of who dies. For example, Frazer (2005) pools four rounds of firm surveys from Ghana, and
still has only 30 deaths versus 479 survival episodes to compare; while Davies and Kerr (2015)
have 95 deaths of firms with 0 to 9 workers.


3. Our Approach to Measurement and Our Sample
Our conceptual unit of analysis is the firm, rather than the entrepreneur, and our focus is on
non-farm micro and small enterprises in developing countries. The vast majority of such firms
have fewer than ten workers, with the modal firm in many developing countries consisting of
just the owner with no paid workers (McKenzie, 2017). We define firm death as having
occurred if a firm is open at one point in time, and then is reported as having shut down by the
owner in a subsequent survey round. By shut down, we mean that the owner of the firm has
decided to stop operating the firm, and no one else is operating it. It is not intended to include
temporary closures of a few days or weeks that may occur when the owner is ill or away.



                                                                                                6
3.1 Putting Together a Large Panel of Micro and Small Firms
We draw on information on firms from both multi-purpose household and firm panel surveys
in order to construct a large panel with which to examine death rates. Our resulting dataset
combines data from 16 surveys in 12 developing countries, and includes a total of 14,4264 firms
at baseline, with a longitudinal dimension that is able to cover the dynamics of these firms in
the short, medium and long term. Table 1 lists the surveys underlying the dataset, appendix A.1
discusses more details of each survey and how firm death was measured in each case.


Our inclusion criteria for using multi-purpose household panel surveys was to include
nationally representative surveys that included a detailed non-farm business module, had low
rates of attrition, a relatively large sample size, and enabled us to measure firm survival and
death rates over at least three or more waves. This yielded four surveys. The first two are the
Mexican Family Life Survey (MxFLS) and the Indonesian Family Life Survey (IFLS), both of
which provide large samples of households. These surveys occur at low frequency, enabling us
to measure firm death over horizons of three, four and a half, and eight years in Mexico, and
seven, eight, and fifteen years in Indonesia. We examined the set of LSMS surveys to determine
the ease at which they could be used to construct household firm panels. This led to the inclusion
of the Nigeria General Household Survey (NGGHS), which allows measurement of firm death
over 0.5, 1.5, 2, and 2.5 year horizons. Finally, we also include data from the Townsend Thai
Project. This is a smaller sample, but resamples households annually, over an extended period
of time, enabling firm death to be measured over periods from one to seventeen years.


These household panel surveys were not designed to track businesses over time, and do not link
the businesses reported from one survey wave to the next. We therefore consider households
that only operate one enterprise, and follow Vijverberg and Haughton (2004), McCaig and
Pavcnik (2016) and Kraftt (2016) in using information on firm age, business ownership, and
sector to attempt to link firms over time.5 Since these household panels also contain data on
individuals who are not operating firms in the first period, they enable us to also capture
information on new firms that households start between one survey round and the next. We
then pool together multiple survey rounds to get a larger sample, Appendix A.1. details these
procedures in more detail.


4
  This number refers to the total number of unique firms we observe.
5
  The majority of households who operate a business operate only one business. In the 2000-round of the IFLS
these are 80 percent, in the 2002-round of the MxFLS 93 percent, in the 2010-round of the NGGHS 58 percent,
and in the 1997-round of the Townsend Thai Project 83 percent.
                                                                                                               7
The second form of data comes from specialized firm panel surveys that have been collected as
part of randomized controlled trials to assess different interventions designed to help micro and
small firms in different countries. In each case we only use the control group data, in order to
be able to assess firm death in the absence of any intervention. We searched publicly available
datasets for impact evaluations targeted at microenterprises, but found most data made available
only provided sufficient detail for replicating particular papers, and were not suitable for
assessing survival dynamics. We use 11 panel surveys carried out in randomized trials
conducted by one of the authors.6 These include microenterprise surveys of informal firms
used for formalization experiments in Benin, Malawi and Sri Lanka; surveys of
microenterprises used in business training interventions in Kenya, Sri Lanka, and Togo;
microenterprises that were part of cash grant interventions in Ghana and Sri Lanka;
microfinance clients that were part of an insurance intervention in Egypt; business plan
competition applicants in Nigeria; and microenterprises that were part of an intervention that
gave wage subsidies, savings help, and business training in Sri Lanka. In addition, the one
additional publicly available dataset which we were able to include was data from a program
for female small businesses in Uganda from Blattman et al. (2016).7


Many of these specialized firm surveys are representative of urban microenterprises of
particular size cutoffs or informality status, having been found through door-to-door listing
surveys. Some of the samples also screen on the gender of the owner. The main exceptions are
the business plan competition applicants in Nigeria, the microfinance clients in Egypt, and the
Uganda study which was targeted at women in marginalized villages in Northern Uganda.


These specialized firm surveys collect data at relatively high frequencies, with a number
collecting data at quarterly or semi-annual frequencies. However, few of them follow firms for
more than three and a half years. The exceptions are three surveys from Sri Lanka, which
tracked firms at 5 to 6 years, and at 10 to 11 years in the case of the Sri Lanka Microenterprise
Survey (SLMS). Table A.4. provides an overview of the follow-up periods provided by each
survey.




6
  We excluded data from surveys with high attrition rates for which death rates could not be accurately calculated,
including data from a grants experiment in Mexico and a formalization experiment in Brazil.
7
  The Uganda study targeted women, rather than firms, as the unit of observation. We take the sample of control
group participants who started a firm and use this sub-sample to track firm dynamics.
                                                                                                                 8
3.2 Dealing with Attrition
A key threat to our ability to measure firm death rates comes from survey attrition, in which
owners of businesses could not be found or refused to be re-interviewed in the follow-up
survey. The studies we draw on mitigate this threat in two main ways. The first has been to
conduct multiple follow-up attempts and track individuals who move, thereby reducing overall
levels of attrition. Second, the specialized firm surveys usually record whether the business is
still in operation or not, even if the owner refuses to be re-interviewed or is away from the
business when survey attempts are made. Table A.5. summarizes the attrition rates in our
combined dataset by period of measurement. Attrition rates are relatively low for this
combined sample, at 4.8 percent for one year, 10 percent for two years, and 9.5 percent over
three years, and are around 20 percent in many periods beyond 5 years, with some exceptions
of lower attrition rates.


We then form bounds for death rates to account for attrition. A lower bound for firm death is
obtained by assuming that all enterprises which were operating at baseline, and then have
missing values for firm death status at follow-up survived. An upper bound for firm death is
obtained by assuming that all these firms died.


3.3 Summary Statistics on Firm Characteristics
Table 2 provides baseline summary statistics of our sample of firms in order to provide a
description of the types of firms for which we are able to examine firm death rates. The sample
consists of small enterprises run by individuals with relatively low levels of education (only 6
percent have any tertiary schooling), and who have an average age of 43 years. 56 percent of
the owners are female. The firms have been in business for a mean of 9 and median of 6 years,
with just over half (52 percent) in retail, 31 percent in services, and 14 percent in manufacturing.
Only 29 percent of firms have any paid workers, with a mean of 0.75 workers per firm.
Conditional on having paid workers, the median firm has only 1 worker and only 10 percent
have 5 or more workers. Mean (median) monthly profits are US$129 (US$63). The median
firm is thus earning approximately $2.65 per day it operates.


4. How Much Firm Death is There?
The first question that we wish to use this data to answer is just what the rate of firm death is
for small firms in the economy. That is, starting with a sample of existing firms today, how
many of them will have closed a year from now, or five years from now? The answer to this


                                                                                                  9
question is useful for understanding the rate of churn in these small firms, as well as for
policymakers and researchers planning interventions to help these firms.


4.1 The Fitted Relationship Between Death Rates and Time
The surveys vary in their frequency, so that we cannot measure firm death for the same interval
(e.g. one year) for all the different surveys. We therefore begin by calculating firm death rates
for each of the 79 survey-time interval combinations in our data, and report these rates in
Appendix Table A.2. along with an upper and lower bound that accounts for survey attrition.
We then plot this data in Figure 1, using a point to indicate the death rate for a survey-period
combination, and bands around the point to indicate the bounds on this rate once attrition is
allowed for. We fit a quadratic to these points, to give the predicted relationship:


Firm death rate = 0.015 + 0.092*Years Elapsed -0.00215*Years Elapsed2                   R2 =0.789
                   [0.024] [0.013]                  [0.00079]


where robust standard errors are shown in parentheses. This relationship suggests firms die at
the rate of approximately 9 percentage points per year over the first three years of following
them, with this rate then declining over time, to around 7 percentage points per year at five
years and 4.6 percentage points a year at ten years.


The majority of our data (58 out of 79 survey-time intervals) comes from horizons of up to 5
years. Figure 2 focuses on this sub-sample to better visualize the data within this interval. We
cannot reject linearity of the relationship between death rates and time over this range (p=0.837
on the quadratic term), so we fit the following linear relationship for periods of up to five years:


Firm death rate = 0.024 + 0.083*Years Elapsed                         R2 = 0.409
                   [0.027] [0.016]


That is, firms die at an average of 8.3 percentage points per year over the first five years. Fitting
the same slope through the lower bounds of the intervals gives a slope of 6.9 percentage points
per year, while fitting it through the upper bounds of the intervals gives a slope of 8.7 percentage
points per year.




                                                                                                  10
In appendix Table A.3. we collate together estimates of firm death in developing countries from
other literature. Figure A.1 then plots these alongside with our data and fitted quadratic. The
firm death rates from other studies are all relatively close to the fitted quadratic, showing that
results are generally consistent across a number of contexts. The one exception comes from the
ten year death rates of Ghanaian manufacturing firms reported in Davies and Kerr (2015), which
are much lower than our data would predict. Their death rates would look more similar if we
assume that the more than 20 percent attrition comes from firms that die.


It may also be of interest to some readers to benchmark these death rates against those in the
United States. The main source of firm dynamics data in the U.S. comes from data from the
Bureau of Labor Statistics and U.S. Census Bureau for firms with at least one paid employee.
Among firms with 1 to 4 paid workers, the one year death rate over 1995 to 2016 averaged 19.7
percent. It was highest for younger firms, averaging 28.4 percent for firms aged 1 to 4 years,
20.8 percent for firms aged 5 to 9 years, and 16.6 percent for firms aged 10 or more years. Death
rates were lower for firms with more workers, averaging 4.6 percent for firms with 5 to 9
workers and 4.0 percent for firms with 10 to 19 workers8. U.S. data also shows high exit rates
from self-employment, with 28 percent of the incorporated self-employed exiting over a year
(Schweitzer and Shane, 2016). This evidence suggests that firm death rates for small firms are
higher in the U.S. than the averages seen in our sample.


4.2 The Half-Life of Firms and Variability
Using these fitted relationships, we can then estimate the half-life of firms that are observed to
be in existence at a given point in time. From the quadratic fit, 50 percent of firms are predicted
to die within 6.2 years, while from the linear fit, 50 percent are predicted to die within 5.7 years.
By way of comparison, Klapper and Richmond (2011) find the median lifespan of formally
registered firms in Côte d’Ivoire is five years.


Since the death rates are approximately linear over the first five years, we can annualize them
to get a death rate per year. This enables examination of the variability in death rates across the
different samples. The mean annualized death rate over the 58 survey-time intervals is 10.4
percent, with a standard deviation of 7.9 percent. The [10,90] range is [4.0%, 22.7%]. So there



8
 Source: Bureau of Labor Statistics: https://www.bls.gov/bdm/business-employment-dynamics-data-by-age-
and-size.htm


                                                                                                         11
is considerable variability in the death rates found. We explore determinants of this variability
in the next section.


4.3 Firm Death Need Not Mean Permanent Exit From Self-Employment
As noted, our focus is on the firm. However, using our panel surveys we can also track what
happens to the owner once the firm has closed. Of all owners that we observe closing a firm
within a year, 20 percent are already operating a new firm again within the same year. Figure
A.2. looks at these re-opening rates for the five surveys for which we have the largest numbers
of closed firms. We see re-opening rates increase with time, with approximately 40 percent of
owners of closed firms opening a different firm within 3 years, although this rate is lower in the
Townsend Thai data than in the specialized firm surveys. We will return to this issue when
examining the causes of firm death, but note the importance of being able to track specific firms
rather than just self-employment status in order to accurately measure firm lifespans.


5. Which firms are more likely to die?
We next examine how firm death varies with several key firm and owner characteristics. The
aim in this section is to provide evidence that helps answer descriptive questions about which
types of firms are more likely to die. The next section then provides some theoretical structure
to help better understand why certain firms are more likely to die.


We begin with examination of univariate associations, and then move to multivariate analysis.
Since the previous section shows that unconditional death rates are approximately linear over a
five-year period, we summarize some of the information graphically by calculating
standardized annual death rates. Graphically, we focus on time horizons of up to 3 years, since
this is the range over which we observe firm death rates in a wide range of surveys and
countries. Standardized annual death rates are then calculated by dividing the death rate by the
period of measurement, so that a death rate of 18 percent over two years will be shown
graphically as an annualized rate of 9 percent.


We then use probit and proportional hazard models to complement this graphical analysis and
to examine the multivariate associations of firm and owner characteristics with firm death. We
estimate a probit regression specification of the following form:


                             (     ℎ    = 1	|	 ) = (        + )                              (1)


                                                                                               12
with       ℎ   a dummy indicating whether firm i dies over observation period T,         includes
the firm and owner characteristics of interest, as well as the horizon T over which the firm is
observed, and survey dummies, and          being the error term. To account for the fact that the
same firm enters the regression more than once if it is observed over multiple time horizons,
we cluster the standard errors at the firm level. We restrict this analysis to time periods T of
less than or equal to five years. This accounts for the majority of our observations and enables
us to describe results in terms of annualized death rates since death rates are linear with time
over this interval (see Figure 2).


The Cox proportional hazard specification we estimate has the following form:


                              ℎ ( 	|	 ) = ℎ ( )      (    )                                  (2)


where ℎ ( ) is the hazard at time t, i.e. the likelihood that firm i dies at time t, ℎ ( ) is the
baseline hazard, and     includes firm and owner characteristics and survey dummies. Again,
standard errors are clustered at the firm level.


The choice of firm and owner characteristics Xi here was determined by a combination of
commonly asked descriptive questions about which firms are more likely to die, insights from
the existing literature, and data availability. While the existing literature has examined some of
these correlations, the number of firm deaths observed in any single study is typically small,
limiting their ability to examine multiple predictors of firm death.


5.1 Univariate Firm-Level Correlates of Firm Death
We begin with an exploration of univariate associations between firm and owner characteristics
and firm death, to establish stylized facts. Figure 3 graphically displays the univariate
associations between firm characteristics and firm death. Each point is the mean death rate or
annualized death rate for all firms with a given characteristic for a survey-year combination.
The first five columns of Table 3 then provide the marginal effects from probit estimation, after
controlling for survey fixed effects and a quadratic in time.


Are younger firms more likely to die? Perhaps the most common stylized fact about firm death
is that younger firms are more likely to die than older firms. Theoretically this can arise from

                                                                                               13
new entrepreneurs learning that they are inefficient and exiting as in Jovanovic (1982), from
starting with a low level of investment and not being able to keep up with the advances of
competitors as in Ericson and Pakes (1995), or from taking on high levels of risk and failing
early as in Cressy (2006). Empirically younger firms have been found to be more likely to die
in the United States (e.g. Bernard and Jensen, 2002; Dunne, Roberts and Samuelson, 1989;
Evans, 1987) and in developing countries (e.g. Vijverberg and Haughton, 2004; Frazer, 2005;
Shiferaw, 2009) although Söderbom et al. (2006) found no significant relationship between
firm age and death rates, and Davies and Kerr (2016) find that firms in the middle age category
(5-14 years old) are less likely to die than younger firms (less than 5 years old), while older
firms (older than 14 years) do not have higher likelihoods of dying than younger firms.


Figure 3a shows the relationship between annualized death rates and firm age in our data. There
is a negative relationship, with older firms less likely to die on average than younger firms.
However, we also see considerable heterogeneity in death rates across the different surveys.
Among firms that are less than 2 years old, we see annual death rates ranging from less than 10
percent to more than 40 percent. Column 1 of Table 3 shows the marginal effect of age on the
likelihood of firm death in our probit estimation when a linear term in age is included. An
additional year of firm age is associated with a 0.6 percentage point lower likelihood of dying.
Column 2 allows for firm age to have a non-linear effect, adding a quadratic term in firm age,
and also a dummy variable for being in the first year of the business. Both these additional
terms are significant, and show that an additional year of age matters more when firms are very
young. The predicted probability of dying in the next year is then 26.6 percent for firms in their
first year, 17.1 percent for firms in their second year, 13.7 percent for firms that are five years
old, and 11.5 percent for firms that are ten years old.


Are smaller firms more likely to die? Firm size is closely linked to firm age in many theoretical
models, with firms starting small and either growing or dying. Lower death rates for larger
firms do appear to be the case in studies which compare medium or large-sized firms to small
firms. For example, Davies and Kerr (2015) find Ghanaian firms with 75 or more workers are
5 percentage points more likely to survive over ten years than firms with 0 to 9 workers.
Likewise Söderbom et al. (2006) estimate that a firm with 50 workers is 7 percentage points
more likely to survive than a firm with 10 workers in their data from Ghana, Kenya and
Tanzania, and Shiferaw (2009) finds Ethiopian firms with more than 100 employees are more
likely to survive than those with 10 to 30 workers. The relationship has been less strong when


                                                                                                14
examining smaller firms and household enterprises. Vijverberg and Haughton (2004) find
larger firms less likely to die in Vietnam, whereas McPherson (1995) finds no significant
relationship with firm size in Swaziland and Botswana, and even finds larger firms being more
likely to die in Zimbabwe.


Figures 3b and 3c examine the relationship between firm death and firm size, considering two
measures of firm size: employment, and profitability. Since the modal firm in our sample has
no paid workers, and 99.4 percent have 10 workers or fewer, we compare firms with no workers
to those with 1 to 10 workers in Figure 3b. There is no strong size effect apparent in this figure.
Column 3 of Table 3 confirms this lack of employment size effect. Figure 3c shows firms
earning more profits are slightly less likely to die than the smallest firms. Column 4 of Table 3
confirms this, with log profits being a significant negative predictor of firm death.


Are firms in certain sectors more likely to die? While many studies have just used data from
manufacturing firms, studies which have included other sectors have found death rates to vary
with sector. However, they come to different conclusions about which sectors have higher death
rates. McPherson (1995) and Mead and Liedholm (1998) find that firms in the retail sector are
more likely to close than firms in the services or manufacturing sector; Klapper and Richmond
(2011) find manufacturing firms to face a higher risk of closing compared to firms in the
services sector and Vijverberg and Haughton (2004) find manufacturing and services firms to
be significantly more likely to close than those in the retail sector. Figure 3d shows that in our
data, retail firms have the highest death rates, and manufacturing firms the lowest, with services
in the middle. However, column 5 of Table 3 shows that these differences are not statistically
significant.


5.2 Univariate Individual-Level Correlates of Firm Death
Figure 4 and columns 6 through 8 of Table 3 examine how the characteristics of the owner are
correlated with firm death. These have been less studied in much of the literature, since many
surveys of firms do not provide personal characteristics of the owner.


Are firm death rates higher for younger or older entrepreneurs? Younger owners have had
less time to accumulate skills and capital, and are at an age where job mobility is also higher
even in wage work. But older entrepreneurs may close as they near retirement, or see less time
to recover from shocks. The result might be that survival probabilities are highest for the middle


                                                                                                15
aged. Fajnzylber et al. (2006) find this in the context of the likelihood of staying self-employed
(not necessarily in the same business) over a year in Mexico, where the maximum is found for
ages 36 to 50. Likewise, Vijverberg and Haughton (2004) find firm death rates to be lower for
26 to 45 year olds than for older or younger firm owners. Figure 4a suggests this inverse-U
shaped pattern, which is confirmed in column 6 of Table 3. Firm death rates are highest for
youth, fall until age 50, and then start rising again. The predicted annualized death rate averages
18.0% for 20 to 24 year olds, compared to 9.6% for 45 to 49 year olds.


Are male or female owners more likely to have their firms die? Female-owned firms are, on
average, smaller and less profitable than male-owned firms in developing countries. Both these
factors would suggest their firms would be more susceptible to failure. However, the existing
literature has found mixed associations between gender and firm death. McPherson (1995) finds
female-owned firms are more likely to close in two of the four countries he studies, but finds
no significant relationship in the other two. In contrast, Shiferaw (2009) and Vijverberg and
Haughton (2004) find female-owned firms to be less likely to die than male-owned firms in
Ethiopia and Vietnam respectively. In contrast, female-owned firms are more likely to die in
our data. Figure 4b shows that female-owned firms have higher death rates across the range of
time intervals we consider. Column 7 of Table 3 shows this difference is statistically significant,
with a 2.7 percentage point higher death rate for female-owned firms.


Are firms run by more educated owners less likely to die? To the extent that higher education
also reflects higher ability, we might expect firms run by more educated owners to be more
likely to survive. However, the existing literature has either found no relation between firm
death and education (Vijverberg and Haughton (2004); Krafft (2016) in one time period), or
that firms with more educated owners are actually more likely to die (Fajnzylber et al. (2006)
for self-employment; Krafft (2016) in a second time period). Figure 4c shows that firm death
rates appear higher for the more educated in our data. However, this appears to largely reflect
differences across surveys. After controlling for survey-year fixed effects, column 8 of Table 3
shows that the firm death rate for tertiary educated owners is not statistically significant
(p=0.366).


5.3 Multivariate analysis of the correlates of firm death
Column 9 of Table 3 then includes all these variables together to examine how they jointly
predict firm death with the probit. Column 10 gives odds ratios from the analogous proportional


                                                                                                16
hazard model. The results are largely similar to the univariate results: firm death is higher for
younger firms, especially those in their first year, lower for firms with higher baseline profits,
lower for middle-aged owners compared to youth or older owners, and higher for female
owners. Business sector, which was not statistically significant in the univariate specification,
is significant in both the probit and hazard multivariate models. Manufacturing firms are 2.7
percentage points less likely to die than retail firms, while service firms are 1.7 percentage
points less likely to die compared to retail firms. The number of employees also becomes
significant at the 10 percent level in the hazard model, suggesting that larger firms are less
likely to die. However, the odds ratio of 0.98 shows this effect is small in magnitude.


5.4 Do death rates vary with level of development?
All of the above analysis controls for survey fixed effects, which captures country-level
differences across the samples. But it is still of interest to see descriptively whether firm death
rates vary with level of development. On one hand, we might expect businesses in less
developed countries to face more constraints such as less access to credit, and so be less likely
to survive negative shocks. However, there may be fewer other options for employment in
poorer economies, suggesting fewer voluntary exits. Figure 5 plots the observed correlation
between annualized firm death rates and per-capita GDP. We see a positive relationship
(correlation 0.41), with firm death rates higher in richer developing countries. Each log point
increase in per capita GDP is associated with a 5.2 percentage point higher annualized firm
death rate (p=0.004). This is also consistent with the evidence presented earlier in this paper
suggesting that small firm death rates are higher in the U.S. than in our sample.


6.   Why do firms die?
The previous section shows that firm death does not simply occur at random, but is more likely
to occur for certain types of firms and firm owners than others. We consider three competing
theories of why micro and small firms may close down. These theories differ in the extent to
which firm closure is a voluntary choice for the owner, and in terms of the relationship that can
be expected between firm productivity and firm death. As a result, they lead to different
conclusions about how firm death will affect aggregate productivity and individual welfare. We
then relate these theories back to our data on which firms die, and use data on cause of death to
help understand why firms die.




                                                                                                17
6.1 Three theories of why firms die
Firm competition and firm-level shocks: A first view of firm death takes the firm as the
decision-making unit, and comes from firm competition and industrial organization models of
firm dynamics (e.g. Hopenhayn, 1992; Ericson and Pakes, 1995). Firm exit occurs for two main
reasons in these models. The first is the entry of more productive competitors, which take away
market share from less productive incumbents, ultimately leading them to close down.
Secondly, firms face exogeneous shocks, such as shocks in the demand for the firm’s products.
Firm closure is thus involuntary in this model, and is expected to lower the welfare and earnings
of the firm owner since it arises from deteriorating firm conditions. In the absence of distortions
and adjustment frictions, it will be the least productive firms which close down, resulting in an
increase in aggregate productivity. However, distortions can allow some unproductive firms to
remain in business and cause some more productive firms to exit (Bartelsman et al, 2013),
weakening the extent to which firm death selects on productivity.


Occupational choice and positive shocks to the entrepreneur’s outside options: the first view
completely ignores the role of the firm owner or entrepreneur in decision-making. This may be
appropriate for examining the death of larger firms with many owners, but for micro and small
firms, the firm is tightly linked to the occupation of the owner, and it is rare for the firm to
continue when the owner leaves.9 This leads to examining firm death through theories of
occupational choice (e.g. Evans and Jovanovic, 1989; Banerjee and Newman, 1993). In these
theories individuals initially make their choice to be self-employed in a particular firm given
their endowments of financial assets and human capital, and the prevailing returns to their skills
and assets in alternative occupations. Business closure then occurs when relative returns change
so that working in a wage job or starting a different firm becomes more attractive. This could
occur through negative shocks to the existing firm, in which case we are back in the first theory,
or through positive shocks to outside options such as an attractive wage job offer or idea for
starting a different business. Firm closure is then voluntary, and will improve, or at least not
lower, welfare and earnings of the firm owner. The extent to which firm death selects on
productivity will depend on which types of entrepreneurs are most likely to receive these
outside offers. Low productivity entrepreneurs may be close to the margin of preferring wage


9
  For instance, in the SLMS, only 1.5 percent of owners who stopped operating their business from round to round
had sold it, while only 2.9 percent said the firm was being operated by another family member. In the Lomé
Informal Enterprise Survey, 1.2 percent of owners who were not running their business anymore had sold it, and
1.8 percent had passed it to a family member. In the Ghana Microenterprises Survey, 7.5% of the businesses that
were not being operated anymore by the original owner were being run by another family member, while no
business had been sold.
                                                                                                             18
work, so that small improvements in the wage market may induce them to quit. Conversely,
high productivity individuals may receive more outside wage offers and also be more likely to
think of ideas for alternative businesses they could run.


Non-separability of business and household decisions: in both of the first two theories, business
decisions are made separately from household time allocation and consumption decisions of
the business owner. However, the classic agricultural household model (e.g. Singh, Squire and
Strauss, 1986; Bardhan and Udry, 1999) notes that production and consumption decisions may
not be separable in the presence of missing markets. Adapting this idea to an urban context,
with imperfect labor markets, illness of household members may require the firm owner to shut
down the business in order to look after sick family members as they cannot hire someone else
to do the job. Likewise, illness of the owner may lead to the business shutting down.10 Firm
closure is involuntary, and will be associated with lower earnings and welfare for the owner.
The impact on aggregate productivity will depend on which types of businesses are most likely
to shut down from household shocks. Gender norms may make women more likely to shut their
business to look after other family members, and since female-owned businesses tend to be less
profitable and less productive on average, this would lead to less productive businesses being
more likely to die. Secondly, less profitable and less productive businesses may be closer to a
viability margin below which it no longer is profitable to run the business. As a result, a
temporary reduction in sales during a period of illness and the possibility of having to sell
business assets to cover household expenses may result in closure for these less profitable firms.


6.2 Which theories are our data on correlates of firm death most consistent with?
The patterns found in section 5 for which firms are most likely to die can generally be consistent
with any one of these three theories. For example, consider the finding that younger (and
smaller, less profitable) firms are more likely to die. This is a direct prediction of the firm
competition and firm-shocks literature, as young firms learn that they are inefficient or are
unable to keep up with competition. But if we think of firm owners having heterogeneous
likelihoods of receiving positive wage job offers or of experiencing negative health shocks,
then death rates will be high for young firms as the highest risk firms exit, and then older firms
will be those with lower risks of experiencing these shocks. Thus, the other two theories would
also give rise to older firms being less likely to die. Likewise, the higher death rate for female-


10
 People with higher likelihoods of illness may also self-select into running household enterprises rather than
more brawn-based activities such as farming (Adhvaryu and Nyshadham, 2017).
                                                                                                                 19
owned firms is a direct prediction from the non-separability theory, but could also arise as a
prediction from firm competition theory since female-owned firms are less productive on
average and so may find it harder to compete with new entrants; and could also occur in the
occupational choice theory since females may be on average less attached to the labor market,
and so may need less of a change in outside options to lead them to shut down their firm.


The U-shaped relationship between firm death and age of the owner, and the finding that firm
death rates are higher in richer developing countries are most consistent with the occupational
choice model. Young workers are likely to experience a higher arrival rate of alternative job
offers than middle-aged workers, while older workers may be choosing between continuing to
run the business and retirement or at least working less. Richer countries may offer a wider
range of employment opportunities beyond subsistence self-employment, pulling people out of
jobs.


The theories differ most in their predictions for the effect of firm exit on individual labor
earnings, and in their predictions of the link between productivity and firm death. We therefore
estimate the following regression for the outcome (Y) of log total labor earnings (the sum of
wage earnings plus profits+1, in last month):
 ,   =   +    ∗           ℎ, +∑        1(   , = )+       ,                                   (3)
Where FirmDeath takes value one if firm i is dead at time t, and 1(            , = ) are a set of
dummy variables for survey j measuring follow-up in period s. The standard errors are clustered
at the firm level. The coefficient b then measures the earnings loss or gain associated with the
firm dying.


To test whether it is the more or less productive, or profitable, firms who die, we instead use
baseline labor productivity (log of sales per worker) and log profitability, and estimate the
regression:
     =   +    ∗          +∑     1(   )+                                              (4)
Where EverDie takes value one if firm i is dead by the last survey round (within 5 years) of a
survey. The coefficient b then measures whether those firms who go on to die were lower or
higher productivity and profitability to begin with than those that survive.


Panel A of Table 4 shows the results of estimating equations (3) and (4). In column 1, we see
that firm death is associated with a sizeable fall in total labor earnings. The coefficient suggests

                                                                                                   20
an average 84 percent fall in earnings of the owner when the firm dies. Columns 2 and 3 show
that the firms which die are, on average, 17 percent less productive and 19 percent less
profitable in the baseline period than those that go on to survive.


These results are not in line with the occupational choice theory in which firm owners are
voluntarily closing down to pursue better opportunities. They are more in line with the firm
competition and firm shocks theory, or non-separability theory. However, Figure 6 shows that
this average hides considerable overlap in the productivity distributions of firms that die with
those that survive (Figure A.3. shows this same pattern within surveys). There is not a sharp
productivity cutoff below which death occurs, but rather many deaths from high productivity
firms as well as many low productivity firms surviving. Similarly, while Figure 7 shows
considerable mass in the distribution of labor income near zero for owners whose firms die,
there are also firm owners who earn relatively high incomes after firm death. These results
suggest that there may be some firms that die for reasons related to these alternate theories,
even if the average is due to firm competition and firm shocks.


6.3 Cause of Death Data to Examine Heterogeneity in Reasons for Firm Death
To further explore the heterogeneity in reasons why firms die, and to see whether each of the
theories seems applicable for different subsets of firms, we use cause of death data. This comes
from direct questions asked of firm owners whose businesses have closed, and is asked in 9 of
our 16 surveys. Table 5 summarizes what the owners report as the main cause of death of their
firm, and what their main activity is after the firm dies. We see that the most common cause of
death is in line with the first theory of being driven out by competition and firm shocks: 41
percent say they closed the firm because it was making a loss. The second most common cause
is in line with the non-separability theory, with 26 percent closing for sickness or family
reasons, while 11 percent give responses in line with occupational mobility theory, closing to
pursue better wage opportunities or another idea for a firm.11 Appendix Table A.6 shows that
the cause of death varies by gender. Among male enterprise owners, 49 percent say that they
closed their firm because it was making a loss, while among women 35.5 percent give this
reason. For women, sickness (16 percent) or family reasons (18 percent) taken together are
almost as common as making a loss, and only 9 percent of female owners give reasons related
to the occupational mobility theory. We observe a different pattern for male enterprise owners,


11
  22 percent give “other” as cause of death. This includes some firms in Sri Lanka, where the data for what “other”
refers to was never entered, as well as reasons like the government shutting down the firm, disasters causing assets
to be destroyed, and international migration.
                                                                                                                21
of which 12 percent give reasons in line with the non-separability theory – and among which
only 2 percent say they closed their firm because of family reasons. Instead, closing for better
opportunities is the second most common reason of closure given by male enterprise owners
(16 percent). Appendix Table A.7 strengthens the suggestion that non-separability is especially
relevant for female business owners.


One might question whether these self-administered firm autopsies give reliable information
on cause of death. A first check is then to see whether differences in cause of death are
accompanied by differences in what the owner is doing after the firm dies. Table 5 shows this
to be the case. Those whose firms died because they sought better opportunities are most likely
to be in wage work or operating new firms; those whose firms died for family reasons or
sickness are most likely to be doing housework and family care; while those whose firms died
because the firm was making a loss are spread across a range of activities, including having the
highest rate of still looking for a job of any of the causes of death.


We can then test whether different causes of death result in differences in selection on
productivity and profitability, and in changes in labor earnings, in the way predicted by the
three theories. Panel B of Table 4 re-runs the regressions in equations (3) and (4) separately by
cause of death, while Figures 8 and 9 show the distributions of initial productivity and follow-
up labor earnings in the month preceding a given survey round by cause of death. We see that
the mean selection on labor productivity is similar, and negative, across all three causes of
death. However, Figure 9 shows the variance is much higher for those closing for reasons of
better opportunity – some very high productivity firms close for this reason, as well as some
very low productivity firms. In contrast, the different reasons for closure are associated with
very different changes in labor earnings. Those closing for illness or family reasons have a large
mass earning zero, and an average 97 percent drop in labor earnings. Those closing because the
business was making a loss also experience a large earnings fall, averaging 77 percent. In
contrast, the earnings change associated with closing for better opportunities is not statistically
different from zero. We see in Figure 8 that this group has a bimodal earnings distribution, with
many earning higher incomes than survivors, while a larger proportion also have zero income
compared to the survivors. This suggests that the better opportunities do not always work out,
but that when they do, they do result in higher incomes.




                                                                                                22
6.4 What determines cause of death?
These results suggest that firms do close for all three of the theoretical reasons described above,
but with closing for firm competition and firm shocks the most common. This then leads to the
question of which types of firms are more likely to die for each of the theoretical reasons. To
examine this, we estimate a multinomial logit model, where the outcome is firm operating
status. The omitted (base) case is surviving, and the coefficients are then odds ratios for the
likelihood of the firm dying for each of the three reasons (business loss, family and illness,
better opportunities) relative to survival. We include the same firm and owner characteristics
we used in Table 3 to enable comparability.


Table 6 presents the results. We see that firms are more likely to close because they are making
a loss if they have lower baseline profits to start with, are in retail rather than manufacturing or
the services sector, and if they are run by women. The chance of closing for this reason
decreases with firm age, but at a decreasing rate. Firms are more likely to close for family
reasons and illness if they have fewer employees and lower profits, and especially if they are
run by female owners, with the chance decreasing with firm age, at a decreasing rate. In
contrast, the strongest predictor of closing for reasons of better opportunity is if the owner has
tertiary education, and if the firm is in the first year of business, with no significant size effect,
nor a significant effect of firm age after the first year. There is no significant impact of gender
on closing for this reason, and firms in manufacturing are less likely to close for reasons of
better opportunity than retail firms.


Taken together with the results of the previous subsection, this highlights that the firm death
experience differs by type of firm. One group of less productive and profitable firms, which are
more commonly run by women, tend to die because they are making a loss, and suffer a loss in
labor income for the owner when they do. Another group of less productive and profitable firms,
with fewer employees, and even more likely to be run by women, tend to die because of the
family reasons or illness of the owner, and result in an even larger drop in labor earnings for
the owner. In contrast, a smaller group of firms, that tend to be young and run by highly
educated owners, die because better opportunities come about for the owner, and there is not a
large fall in owner labor earnings as a result.

7. Conclusions

Firm death is extremely common among small firms, with half of the current stock of small
firms in a developing country likely to die within the next six years. Yet because panel surveys

                                                                                                   23
of such firms are still relatively rare, often do not track firms for long periods, and often lose
firms which die to attrition, systematic data on the rates, correlates, and causes of firm death
have not been available. This paper overcomes these issues by pooling together data from 16
panel surveys in 12 countries, in which special efforts were made to limit attrition and ascertain
the status of firms which closed down. Using this, we are able to provide stylized facts on the
rate of firm death over time, provide descriptive evidence on which types of firms are most
likely to die, and test three competing theories for why firms die. Our evidence suggests that
the most common reason for firm death is that less profitable and less productive firms end up
making losses and closing. However, other small firms, particularly those run by women, close
because of illness and family reasons, suggesting non-separability between the household and
firm, while a minority of firms, largely run by more educated owners, close because better
opportunities arise for the owner.


We see several potential uses for this research. The first is as a benchmarking tool. There is
currently little evidence for researchers or policymakers tracking the failure rates of firms in
their study or intervention to compare to, to know whether the rates they see in their data are
high or low. This paper provides guidance as to what death rate can be expected at different
time horizons. The second potential use is for targeting of programs. Programs that want to
work actively with businesses for several years may wish to avoid small firms at greatest risk
of closing down, while other programs that are designed to help firms at risk of failure to survive
may wish to use the information provided here to decide which firms to offer the program to.
Finally, the results may be of interest in deciding whether interventions are needed, and what
types of intervention. To the extent that firms are exiting because they are low productivity and
competition is reallocating production to more efficient firms, policymakers may not want to
intervene. Similarly, if firm death arises because of voluntary exit in pursuit of better
opportunities, again no intervention may be warranted. In contrast, if firms die because of a
lack of separability with household shocks, or because of firm shocks that were not able to be
insured, policymakers may want to experiment with policies to insure these types of shocks.
Our results suggest that there is a subset of smaller firms, run by women, for whom such policies
may be of particular importance.




                                                                                                24
References
Adhvaryu, Achyuta and Anant Nyshadham (2017) “Health, enterprise, and labor
complementarity in the household”, Journal of Development Economics 126: 91-111.

Banerjee, Abhijit and Andrew Newman (1993) “Occupational choice and the process of
development”, Journal of Political Economy 101(2): 274-98.

Bardhan, Pranab and Christopher Udry (1999) Development Microeconomics. Oxford
University Press, Oxford.

Bartelsman, Eric, John Haltiwanger and Stefano Scarpetta (2013) “Cross-country differences
in productivity: The role of allocation and selection”, American Economic Review 103(1): 305-
34.

Bartelsman, Eric, John Haltiwanger and Stefano Scarpetta (2009) “Measuring and Analyzing
Cross-country Differences in Firm Dynamics”, pp. 15-76 in Timothy Dunne, J. Bradford
Jensen, and Mark J. Roberts (editors) Producer Dynamics: New Evidence from Micro Data,
NBER, Cambridge, MA.

Benhassine, Najy, David McKenzie, Victor Pouliquen and Massimiliano Santini (2016) “Can
enhancing the benefits of formalization induce informal firms to become formal? Experimental
evidence from Benin “, World Bank Policy Research Working Paper 7900.

Bernard, Andrew B. and J. Bradford Jensen (2002) “The deaths of manufacturing plants”,
NBER Working Paper 9026.

Blattman, Christoper, Jeannie Annan, Eric P. Green and Julian Jamison (2016) “The returns to
microenterprise support among the Ultrapoor: A field experiment in Postwar Uganda",
American Economic Journal: Applied Economics 8(2): 35-64.

Campos, Francisco, Michael Frese, Markus Goldstein, Leonardo Iacovone, Hillary Johnson,
David McKenzie and Mona Mensmann (2017) “Teaching personal initiative beats traditional
business training in boosting small business in West Africa”, Science, forthcoming.

Campos, Francisco, Markus Goldstein and David McKenzie (2015) “Short-term impacts of
Formalization Assistance and a Bank Information Session on Business Registration and Access
to Finance in Malawi”, World Bank Policy Research Working Paper no. 7183

Cressy, Robert (2006) “Why do most firms die young?”, Small Business Economics 26: 103-
116.

Davies, Elwyn and Andrew Kerr (2015) “Firm Survival and Change in Ghana, 2003-2013”,
CSAE Working Paper 2015-06.

De Mel, Suresh, David McKenzie, and Christopher Woodruff (2016) “Labor drops:
Experimental evidence on the return to additional labor in microenterprises”, World Bank
Policy Research Working Paper 7924.

De Mel, Suresh, David McKenzie, and Christopher Woodruff (2014) “Business training and
female enterprise start-up, growth, and dynamics: Experimental evidence from Sri Lanka”,
Journal of Development Economics 106: 199-210.



                                                                                          25
De Mel, Suresh, David McKenzie, and Christopher Woodruff (2013) “The demand for, and
consequences of, formalization among informal firms in Sri Lanka”, American Economic
Journal: Applied Economics 5(2): 122-150.

De Mel, Suresh, David McKenzie, and Christopher Woodruff (2012) “One-time transfers of
cash or capital have long-lasting effects on microenterprises in Sri Lanka”, Science 335(6071):
962-966.

De Mel, Suresh, David McKenzie, and Christopher Woodruff (2010) “Wage subsidies for
microenterprises”, The American Economic Review Papers and Proceedings 100(2): 614-618.

De Mel, Suresh, David McKenzie, and Christopher Woodruff (2009) “Are women more credit
constrained? Experimental evidence on gender and microenterprise returns”, American
Economic Journal: Applied Economics 1(3): 1-32.

De Mel, Suresh, David McKenzie, and Christopher Woodruff (2008) “Returns to capital in
microenterprises: Evidence from a field experiment”, The Quarterly Journal of Economics
123(4): 1329-1372.

Demirgüc-Kunt, Asli, Leora F. Klapper, and Georgios A. Panos (2006) “Entrepreneurship in
post-conflict transition. The role of informality and access to finance”, Economics of Transition
19(1): 27-78.Dunne, Timothy, Mark J. Roberts, and Larry Samuelson (1989) "The growth and
failure of US manufacturing plants", The Quarterly Journal of Economics 104(4): 671-698.

Ericson, Richard and Ariel Pakes (1995) “Markov-Perfect Industry Dynamics: A Framework
for Empirical Work.” Review of Economic Studies 62 (January): 53–82.

Evans, David S., and Boyan Jovanovic (1989) ‘‘An Estimated Model of Entrepreneurial Choice
under Liquidity Constraints.’’ Journal of Political Economy 97(4):808–27.

Evans, David S. (1987) "The relationship between firm growth, size, and age: Estimates for
100 manufacturing industries." The Journal of Industrial Economics 35(4): 567-581.

Fafchamps, Marcel, David McKenzie, Simon Quinn and Christopher Woodruff (2014)
“Microenterprise growth and the flypaper effect: Evidence from a randomized experiment in
Ghana”, Journal of Development Economics 106: 211-226.

Fajnzylber, Pablo, William Maloney, and Gabriel Montes Rojas (2006) “Microenterprise
Dynamics in Developing Countries: How Similar are They to Those in the Industrialized
World? Evidence from Mexico”, World Bank Economic Review 20(3): 389-419.

Frazer, Garth (2005) “Which Firms Die? A Look at Manufacturing Firm Exit in Ghana”,
Economic Development and Cultural Change 53(3): 585-617.

Gindling, T.H. and David Newhouse (2014) “Self-employment in the Developing World”,
World Development 56: 313-31.

Groh, Matthew and David McKenzie (2016) “Macroinsurance for microenterprises: A
randomized experiment in post-revolution Egypt”, Journal of Development Economics 118:
13-25.

Hopenhayn, Hugo (1992) “Entry, Exit, and Firm Dynamics in Long-Run Equilibrium.”
Econometrica 60:1127–50.

                                                                                              26
Hsieh, Chang-Tai and Peter Klenow (2009) “Misallocation and Manufacturing TFP in China
and India”, Quarterly Journal of Economics 124(4): 1403-48.

Jovanovic, Boyan (1982) “Selection and the Evolution of Industry”, Econometrica 50(3): 649-
70.

Klapper, Leora and Christine Richmond (2011) “Patterns of business creation, survival and
growth: Evidence from Africa”, Labour Economics 18(1): S32-S44.

Krafft, Caroline (2016) “Understanding the Dynamics of Household Enterprises: Birth, Death,
Growth, and Transformation”, Economic Research Forum Working Paper Series 983.

Liedholm, Carl and Donald Mead (1995) “GEMINI Action Research Program I: Final Report”
– The Dynamic Role of Micro and Small Enterprises in the Development Process, Gemini,
Bethesda, MA.

McCaig Brian and Pavcnik Nina (2016) “Out with the old and unproductive, in with the new
and similarly unproductive: Microenterprise dynamics in a growing low-income economy”,
Mimeo.

McKenzie, David (2017) „Identifying and Spurring High-Growth Entrepreneurship:
Experimental Evidence from a Business Plan Competition“, American Economic Review,
107(8): 2278-2307.

McKenzie, David and Susana Puerto (2017) “Growing markets through business training for
female entrepreneurs: A market-level randomized experiment in Kenya” World Bank Policy
Research Working Paper 7993.

McPherson, Michael A. (1995) "The hazards of small firms in Southern Africa.", The Journal
of Development Studies 32(1): 31-54.

Mead, Donald and Carl Liedholm (1998) “The Dynamics of Micro and Small Enterprises in
Developing Countries”, World Development 26(1): 61-74.

Nagler, Paula and Wim Naudé (2017) “Non-farm entrepreneurship in rural sub-Saharan Africa:
New empirical evidences”, Food Policy 67: 175-191.

National Bureau of Statistics of the Federal Republic of Nigeria (2014) “Nigeria General
Household       Survey    –      Panel    Wave      2    2012/2013”,     Available    at
http://microdata.worldbank.org/index.php/catalog/1952.

Roberts, Mark J., and James R. Tybout (1996) “A Preview of the Country Studies” pp. 188-
199 in Roberts, Mark J., and James R. Tybout, (editors) Industrial Evolution in Developing
Countries: Micro Patterns of Turnover, Productivity, and Market Structure, Oxford University
Press, New York.

Rubalcava, Luis and Graciela Teruel (2013) “Mexican Family Life Survey, Third Round”,
Available at: www.ennvih-mxfls.org.

Sandefur, Justin (2010) “On the Evolution of the Firm Size Distribution in an African
Economy”, CSAE Working Paper no. 2010-05.




                                                                                         27
Schweitzer, Mark E., and Scott Shane (2016). “The Ins and Outs of Self-Employment: An
Estimate of Business Cycle and Trend Effects,” Federal Reserve Bank of Cleveland Working
Paper, no. 16-21.

Shiferaw, Admasu, (2009) “Survival of private sector manufacturing establishments in Africa:
The role of productivity and ownership”, World Development 37(3), 572–584.

Singh, Inderjit J., Lyn Squire, and John Strauss (1986) Agricultural Household Models:
Extensions, Applications and Policy. John Hopkins University Press, Baltimore.

Söderbom, Måns, Francis Teal, and Alan Harding (2006) “The Determinants of Survival
Among African Manufacturing Firms”, Economic Development and Cultural Change 54(3):
533-55.

Strauss, John, Firman Witoelar, and Bondan Sikoki (2016) “The Fifth Wave of the Indonesia
Family Life Survey (IFLS5): Overview and Field Report”, WR-1143/1-NIA/NICHD.

Townsend, Robert M. (2016) “Townsend Thai Project Household Annual Resurvey, 2014
(Urban)”, doi:10.7910/DVN/UOOUAZ, Harvard Dataverse, V1.

Van Biesebroeck, Johannes (2005) “Firm Size Matters: Growth and Productivity Growth in
African Manufacturing”, Economic Development and Cultural Change 53(3): 545-83.

Vijverberg, Wim and Jonathan Haughton (2004) “Household Enterprises in Vietnam: Survival,
Growth, and Living Standards”, pp. 95-132 in David Dollar and Paul Glewwe (eds.) Economic
Growth, Poverty, and Household Welfare: Policy Lessons from Vietnam. The World Bank,
Washington D.C.




                                                                                         28
Figure 1. Firm death rates over different time horizons




Figure 2. Firm death rates over different time horizons – up to five years




                                                                             29
Figure 3. Death rates by firm characteristics
(a) Standardized death rates by firm age        (b) Death rates by firm size (employment)




(c) Standardized death rates by daily profits   (d) Death rates by firm sector




Figure 4. Death rates by owner characteristics
(a) Standardized death rates by age of owner    (b) Death rates by gender of owner




(c) Death rates by education of owner




                                                                                            30
Figure 5. Death rates by per capita GDP




Figure 6. Comparison of labor productivity at baseline for firms that closed vs. survivors




Includes data from BENINFORM, GHANAFLYP, IFLS, KENYAGETAHEAD, MALAWIFORM,
MXFLS, NGLSMS-ISA, NGYOUWIN, SLKFEMBUSTRAINING, SLKINFORMALITY, SLLSE, SLMS,
TOGOINF, TTHAI.




                                                                                             31
Figure 7. Comparison of monthly labor income at follow-up of owners of firms that closed
vs. survivors




Figure 8. Comparison of labor productivity at baseline for firms that closed vs. survivors, by
reason for closure




                                                                                           32
Figure 9. Comparison of monthly labor income for owners of firms that closed vs.
survivors, by reason for closure




                                                                                   33
Table 1. Overview of included surveys
                                                                                     Sample
                          Survey                                   Rounds              Size
A. General Household Surveys:
Indonesian Family Life Survey (IFLS)                      2000,2007/2008,2014/2       2,462
                                                          015
Mexican Family Life Survey (MxFLS)                        2002,2005/2006,2009/2         723
                                                          010/2011/2012
Nigeria General Household Survey (NGGHS)                  2010/2011 - 2012/2013       1,804
                                                          (post-planting and post-
                                                          harvest each)
Townsend Thai Survey (TTHAI)                              1997 - 2014 (Annual)         1,122
B. Specialized Firm Impact Evaluation Surveys
Cotonou Informal Enterprise Survey (BJINFORMALITY)           2014 - 2016               1,197
Egypt Macroinsurance for Microenterprises Survey             2012-03 - 2012-11         1,441
(EGMACROINSURANCE)
Ghana Microenterprises Survey (GHMS)                         2008 - 2010                 397
Kenya GET Ahead Business Training Program Impact             2013 - 2016               2,365
Evaluation Survey (KEGETAHEAD)
Malawi Business Registration Impact Evaluation Survey        2011/2012 - 2013/2014       757
(MWBRIE)
Nigeria YouWiN! National Business Plan Competition           2011 - 2016                 280
Impact Evaluation Survey (NGYOUWIN)
Sri Lanka Microenterprise Survey (SLMS)                      2000 - 2016                 269
Sri Lanka Impact Evaluation Survey of Business Training for 2009 - 2015                  260
Women (SLFEMBUSTRAINING)
Sri Lanka Informal Enterprises Survey (SLINFORMALITY)        2008 - 2011                 105
Sri Lankan Longitudinal Survey of Enterprises (SLLSE)        2008 - 2014                 284
Lomé Informal Enterprise Survey (TGINFORMALITY)              2013/2014 - 2015            499
Uganda Women’s Income Generating Support Impact              2009 - 2011                 461
Evaluation Survey (UGWINGS)
Note: Only the control group is used from surveys that are part of randomized experiments.




                                                                                       34
Table 2. Summary Statistics on Firms
                                Number of
                               Observations   Mean    S.D.    10th   Median   90th
Owner Characteristics
Owner is Female                  18,761       0.56     0.50   0         1     1
Age of Owner                     17,961       43.41   13.36   27       42     62
Owner has tertiary education     15,294       0.06     0.24   0         0     0
Firm Characteristics
Age of firm (years)              15,027       9.47     9.50    1       6       22
Monthly sales (USD)              17,159       1053    18175    2      205     1664
Monthly profits (USD)            15,424       129      733    -63     63      417
Any paid worker                  15,737       0.29     0.45    0       0       1
Number of paid workers           15,645       0.75     5.02    0       0       2
Retail firm                      17,457       0.52     0.50    0       1       1
Manufacturing firm               17,449       0.14     0.34    0       0       1
Services firm                    17,449       0.31     0.46    0       0       1




                                                                                     35
Table 3. Univariate and Multivariate Correlates of Firm Death
                                                  (1)              (2)             (3)           (4)            (5)             (6)             (7)            (8)            (9)             (10)
 Years since baseline                         0.0679***        0.0674***       0.0688***      0.0673***     0.0688***       0.0684***       0.0686***      0.0688***      0.0663***
                                              (0.00186)        (0.00186)       (0.00183)      (0.00184)      (0.00183)       (0.00183)       (0.00183)     (0.00183)       (0.00186)
 Age of firm (years)                         -0.00575***      -0.0108***                                                                                                 -0.00929***       0.960***
                                              (0.000601)       (0.00135)                                                                                                   (0.00138)       (0.00632)
 Age of firm squared                                         0.000180***                                                                                                 0.000150***       1.001***
                                                               (3.20e-05)                                                                                                  (3.28e-05)     (0.000152)
 Firm in first year of life                                    0.0774***                                                                                                  0.0666***        1.283***
                                                                (0.0114)                                                                                                    (0.0113)       (0.0620)
 Number of employees                                                           -0.00148                                                                                    -0.00224         0.983*
                                                                               (0.00261)                                                                                   (0.00241)       (0.00889)
 Log Business Profits                                                                        -0.0363***                                                                   -0.0302***       0.892***
 Business sector (omitted category
 retail):                                                                                     (0.00320)                                                                    (0.00314)       (0.0119)
   Manufacturing                                                                                              -0.0183                                                      -0.0269**        0.878**
                                                                                                              (0.0131)                                                      (0.0124)       (0.0544)
   Services                                                                                                  -0.00619                                                      -0.0171*          0.928
                                                                                                              (0.0106)                                                     (0.00994)       (0.0436)
   Other                                                                                                     -0.00831                                                      -0.00466          1.001
                                                                                                              (0.0259)                                                      (0.0264)        (0.120)
 Age of Owner                                                                                                               -0.0159***                                    -0.0111***       0.950***
                                                                                                                             (0.00178)                                     (0.00179)       (0.00698)
 Age of Owner Squared                                                                                                      0.000154***                                   0.000114***       1.001***
                                                                                                                            (1.93e-05)                                     (1.95e-05)     (7.63e-05)
 Owner is Female                                                                                                                            0.0273***                       0.0197*         1.112**
                                                                                                                                             (0.0105)                       (0.0102)       (0.0521)
 Owner has tertiary education                                                                                                                                0.0195         0.0166           1.105
                                                                                                                                                            (0.0216)        (0.0203)       (0.0951)


 Observations                                   40,954            40,954           40,957         40,957         40,957         40,957          40,957        40,957          40,954         40,954
 Standard errors in parentheses, clustered at the firm level. All specifications include survey fixed effects. Columns 1-9 are marginal effects from a probit of firm death; column 10 are odds ratios
 from a proportional hazard model. Estimation is over time horizons of five years or less. *, **, and *** denotes significance at the 10, 5, and 1 percent levels respectively.


                                                                                                                                                                                                      36
Table 4. Changes in Productivity, Profitability, and Labor Earnings Associated with Firm
Death
                                 Log(Labor Earnings + 1)           Log(Labor Productivity)         Log(Profitability+1)
Panel A: Full Sample
Firm Death                             -1.8567***                        -0.1829***                    -0.2095***
                                         (0.0666)                          (0.0313)                      (0.0318)
Observations                              15,187                             7,927                        7,261
Panel B: By Cause of Death
Making a Loss                            -1.5123***                       -0.1774***                     -0.2207***
                                           (0.1109)                         (0.0591)                       (0.0579)
Illness or Family Care                   -3.3645***                        -0.1828**                     -0.1990***
                                           (0.1051)                         (0.0769)                       (0.0635)
Better Opportunities                        -0.1911                         -0.1642                        -0.0739
                                           (0.1650)                         (0.1232)                       (0.1040)
Observations                                13,114                            6,035                         5,661
Note: Robust standard errors, clustered at the firm level, in parentheses. Labor earnings are the sum of wage earnings plus
profits in the month preceding a given survey wave. Labor productivity is measured as business sales in the month preceding
a given survey wave divided by the number of employees +1. Profitability is measured by business profits in the month
preceding a given survey wave.




Table 5. Cause of Firm Death and Main Activity of Owner After Firm Death
Reason for closure       Percent Frequency       Working     Looking  Operating        Housework       Other      Total
                                                for a wage   for a joba different     looking after
                                                                       business         children
Making a loss                40.55        590        30.92       7.96       18.26             17.54     25.32         553
Sickness                     13.54        197         8.13       2.50       10.63             46.25     32.50         160
Care for family              12.37        180         5.29       1.76          4.12           71.18     17.65         170
Better wage job               6.53         95        95.56       1.11          2.22            1.11       0.00            90
Better business
                              4.88         71        10.77       1.54        75.38             6.15       6.15            65
opportunity
Other                         22.13         322        19.51        6.27      18.12           33.10     23.00       287
Total                        100.00       1,455          342          71        228             392       292     1,325
Note: Data for columns (1)-(2) from BENINFORM, KENYAGETAHEAD, MALAWIFORM, NGYOUWIN, SLKFEMBUSTRAINING,
SLKINFORMALITY, SLLSE, SLMS, and TOGOINF. The category “Other” includes 3.09% of businesses, which had to close
because of a catastrophe (e.g. fire or burglary), 1.51% of business which were shut down by government, a remaining
1.72% of owners who discontinued their businesses because either the business was taken from them, they married, or
migrated. The remaining 15.81% of owners who closed their businesses either did not specify this other reason more in
detail or closed because of any other miscellaneous reason, which could not be coded into one of the above categories
(for instance because of a lack of workers or inputs, or a machine broke down).
Data for columns (3)-(8) from BENINFORM, KENYAGETAHEAD, MALAWIFORM, SLKFEMBUSTRAINING, SLLSE, SLMS, and
TOGOINF.




                                                                                                                          37
Table 6. Multinomial logit regression results
                                                                 Reason for closure
                                                         (Base outcome: Survived at time t)
                                              (1)                     (2)                       (3)
                                         Making a loss     Illness and family care    Better opportunities
Years since baseline                      1.1824***               1.3050***                 1.2972***
                                           (0.0514)                (0.0838)                  (0.1104)
Age of firm (years)                       0.8967***               0.9249***                   0.9661
                                           (0.0180)                (0.0195)                  (0.0442)
Age of firm squared                       1.0016***               1.0012***                   0.9997
                                           (0.0004)                (0.0004)                  (0.0014)
Firm in first year of life                 1.3569*                 1.4249*                  2.0584***
                                           (0.2228)                (0.2610)                  (0.5764)
Number of employees                         0.9674                0.7591**                    0.9270
                                           (0.0529)                (0.1044)                  (0.1973)
Log business profits                      0.8272***               0.8608**                    0.9343
                                           (0.0459)                (0.0580)                  (0.0803)
Sector (Omitted category = Retail)
Sector=Manufacturing                       0.6488**                0.6803                   0.4160**
                                           (0.1320)               (0.1999)                   (0.1441)
Sector=Services                            0.7583**                0.8832                     1.0270
                                           (0.1055)               (0.1441)                   (0.2254)
Sector=Other                                0.4181                 0.6776                     0.4866
                                           (0.2410)               (0.4621)                   (0.4990)
Age of owner                                0.9714                 0.9467                     1.2123
                                           (0.0389)               (0.0337)                   (0.1486)
Age of owner squared                        1.0002                1.0008*                    0.9971*
                                           (0.0005)               (0.0005)                   (0.0017)
Owner is female                            1.3811**              2.4835***                    0.8244
                                           (0.2107)               (0.6069)                   (0.2129)
Owner has tertiary education                1.4118                 0.9010                   2.1314**
                                           (0.3170)               (0.2553)                   (0.6714)
Constant                                    0.1851               0.0010***                10799000***
                                           (0.3269)               (0.0018)                 (25950000)

Observations                               16,308                  16,308                    16,308
 Note: Robust standard errors in parentheses, clustered at the firm level. The specification includes survey
 fixed effects and controls for missing values in firm age, number of employees, log business profits,
 sector=other, age of owner, and owner has tertiary education. Estimation is over time horizons of five years
 or less. *** p<0.01, ** p<0.05, * p<0.1.




                                                                                                          38
Online Appendix
A.1 Additional details on surveys
We discuss each of the sixteen surveys used in our study, and how firm death is defined and
calculated in each. We begin with the multi-purpose household surveys, before turning to
specialty firm panel surveys.

Multi-purpose household surveys:

Indonesian Family Life Survey (IFLS)
The Indonesian Family Life Survey (IFLS) is a publicly available multi-purpose panel survey
of the Indonesian population. It surveys over 7,000 households, consisting of over 30,000
individuals, and is representative of about 83 percent of the Indonesian population. Further
details on study design and data can be obtained in Strauss et al. (2016) and online at
http://www.rand.org/labor/FLS/IFLS.html. There are currently five waves of data available and
our sample draws from IFLS3, which was fielded in 2000, IFLS4 (fielded in 2007/2008) and
IFLS-5 (fielded in 2014/2015). Non-farm business modules of IFLS1 and IFLS2 are less
detailed, complicating the identification of the business owner as well as the matching of
businesses with later rounds, which is why we decided not to include them.

We used the non-farm business modules of the different survey waves to identify household
businesses in each of the three waves we included in our sample. A household was coded as
operating a business if there was a householder who had worked in a family-owned, non-farm
business like trade/retailing or been self-employed in a non-farm enterprise during the past 12
months preceding the survey, which was entirely owned by the household, and still operating
at the time of the survey. Owners were identified as those household members who were
indicated in the non-farm business module as being primarily responsible for the reported
business. Up to two members could be reported per business and their individual survey
identification numbers were given, so that information on owner characteristics could be
matched from the other survey modules.

Since the IFLS is not a panel of firms, but of households who are asked in each round if they
operate a business, it allows us to capture information on firms that were not operating in IFLS3,
but in IFLS4. We can thus use two baselines for the analysis of firm dynamics over time, as
each round provides us with a sample of firms operating at a given point in time. Taking IFLS3
as baseline, we observe horizons of 7.5 and 14.5 years to IFLS4 and IFLS-5, respectively. We
observe a horizon of seven years from IFLS4 to IFLS-5 with IFLS4 as the baseline. Since
horizons using IFLS3 and IFLS4 as baseline differ, we do not pool together data for the same
observation period with this approach, but increase the sample with information on different
horizons.

The non-farm business modules of the IFLS did not ask whether the business reported in a
given round is the same as the business reported in the previous round, or what happened to the
one reported earlier. We therefore needed to match reported businesses across rounds in order
to be able to determine whether the business survived or not, and in the latter case, whether a
new firm was started. To simplify this matching exercise, we only kept firms of households
who never operated more than one enterprise in any of the three survey rounds in our sample.

Taking IFLS3 [IFLS4] as our baseline, and IFLS4 [IFLS-5] as follow-up round, we then coded
the business reported at baseline as surviving, if a business was reported in the follow-up round,
and:
                                                                                               39
      start dates (i.e. date when the business was started) were identical across the two waves;
       or
      the field of the business was the same across the two waves, and the start date of the
       business reported in the follow-up round was a date preceding the interview date of the
       baseline wave; or
      at least one of the business owners (person(s) indicated as being responsible for the
       business) was the same across the two waves, and the start date of the business reported
       in the follow-up round was a date preceding the interview date of the baseline wave.

We coded the business reported at baseline as closed, if:
   no business was reported in a subsequent round.

Otherwise, we coded survival/closure as missing.

We coded a new firm start, if:
   the start date of the business reported in the follow-up round was a date after the
      interview date of the baseline wave.

We coded the variable newfirmstart to zero, if:
   the business reported in the baseline round survived; or
   no business was reported in the follow-up round and the business of the baseline round
      was coded as closed; or
   the start date of the business reported in the follow-up round was a date preceding the
      interview date of the baseline wave and no business was reported in the baseline wave.

Otherwise, the variable newfirmstart was coded as missing.

We used these three types of business characteristics (start date / age of the business, ownership,
and business sector) together, because start dates turned out not to be very accurate, and it is
possible that ownership of a household business changes over time. Even so, there remained a
number of inconsistencies and cases which could not be matched, so that in a final step, we
inspected the characteristics of these businesses in all three consecutive survey waves visually
and coded survival on a case-by-case basis, again, taking into account information on firm age,
ownership and the business sector reported in the three survey waves. For these remaining
observations, we decided to exclude firms from the sample if information on firm age was
missing, since this information is important in determining survival and newfirmstart.

Given that the non-farm business module of the IFLS allowed for more than one owner to be
reported, and ownership could change from one round to another, we also had to decide on
which of the owners to include in case we identified multiple owners for a business, as we were
interested in assessing also owner characteristics in our study. For firms for which we had
identified more than one owner, we decided to include the owner, who was reported as the
owner in all of the waves in which we observed the firm. If this was the case for both owners,
we included the owner,
     who had the lowest household roster number (this chose the household head as the
        owner if he or she was among the owners), if no other business was reported for at least
        one of the owners over the whole observation period; or
     who had already been identified as the owner of a different business, observed during
        one or more different waves over the whole observation period;



                                                                                                40
and in the remaining cases we chose the owner who was reported as the owner in most of the
waves in which we observed the firm, or, if this was the case for both owners, who had the
lowest household roster number.

Finally, we excluded firms for which we could not identify at least one owner among the
household members who were in the IFLS sample.

Mexican Family Life Survey (MxFLS)
The Mexican Family Life Survey (MxFLS) is a publicly available multi-purpose panel survey
conducted in Mexico. It surveys 8,400 households (35,000 individuals) and is representative of
the Mexican population at the national, urban, rural, and regional level. Further details on study
design and the data can be obtained in Rubalcava and Teruel (2013). Three rounds of data have
been collected (2002, 2005-2006, and 2009-2012) and we included all three of them in our
sample.

We identified household businesses in the MxFLS using the non-farm business modules of the
different survey waves. A household was coded as operating a business if there was a household
member who had owned or shared a non-agricultural business during the past 12 months
preceding the survey. Owners were identified as those household members who took the most
important decisions regarding the business. Given that only their relationship to the respondent
was reported, as opposed to their survey identification number, owners could only be uniquely
identified, if they were either the respondent or his or her spouse.

Given the household panel structure of the MxFLS, we also extracted information on firms that
were not operating in round 1, but in round 2, and use two baselines for the analysis of firm
dynamics over time. Taking round 1 as baseline, we observe horizons of three and 7.5 years to
rounds 2 and 3, respectively. We observe a horizon of 4.5 years from round 2 to round 3 with
round 2 as the baseline. Since horizons using rounds 1 and 2 as baseline differ, we do not pool
together data for the same observation period, but increase the sample with information on
different horizons with this procedure.

In order to code whether a business observed at baseline survived or not, and in the latter case,
whether a new firm was started, we needed to match reported businesses between rounds based
on a number of business characteristics, as the MxFLS does not link household businesses from
round to round. For simplification, we decided to only include firms of households in our
sample who never operated more than one enterprise in any of the three survey rounds.

Taking round 1 [round 2] as our baseline, and round 2 [round 3] as follow-up rounds, we then
coded the business reported at baseline as surviving, if a business was reported in the follow-
up round, and:
     start dates (i.e. year the business was started) were identical across the two waves; or
     at least one of the business owners (person(s) taking the most important decisions
       regarding the business) was the same across the two waves, and the business reported
       in the follow-up round was not started after 2001 [in or after the year in which the
       baseline survey was conducted].

We coded the business reported at baseline as closed, if:
   no business was reported in a subsequent round.

Otherwise, we coded survival as missing.


                                                                                               41
We coded a new firm start, if:
   the start date of the business reported in the follow-up round was 2002 or later [the year
      in which the baseline survey was conducted or later].

We coded the variable newfirmstart to zero, if:
   the business reported in the baseline round survived; or
   no business was reported in the follow-up round and the business of the baseline round
      was coded as closed; or
   the start date of the business reported in the follow-up round was a date preceding 2002
      [the interview date of the baseline wave] and no business was reported in the baseline
      wave.

Otherwise, the variable newfirmstart was coded as missing.

We coded survival and newfirmstart from round 1 to round 3 as well, based on these same
criteria described above.

After this matching and coding, we inspected the characteristics of the businesses for which
survival and newfirmstart remained unclear or inconsistent on a case-by-case basis, taking into
account information on firm age and ownership reported in the three survey waves. For this,
we decided to exclude firms from the sample if information on firm age was missing, since this
information is important in determining survival and newfirmstart.

For firms for which we had identified more than one owner, we decided to include the owner,
who was reported as the owner in all of the waves in which we observed the firm. If this was
the case for both owners, we included the owner,
     who had the lowest household roster number (this chose the household head as the
        owner if he or she was among the owners), if no other business was reported for at least
        one of the owners over the whole observation period; or
     who had already been identified as the owner of a different business, observed during
        one or more different waves over the whole observation period;
and in the remaining cases we chose the owner who was reported as the owner in most of the
waves in which we observed the firm, or, if this was the case for both owners, who had the
lowest household roster number.

Finally, we excluded firms for which we could not identify at least one owner among the
household members who were in the MxFLS sample.

Nigeria General Household Survey (NGGHS)
The Nigeria General Household Survey (NGGHS) is a cross-sectional multipurpose household
survey of 22,000 households, carried out annually in Nigeria. Under the World Bank’s Living
Standards Measurement Study - Integrated Surveys on Agriculture (LSMS-ISA) project a panel
component was introduced into the multi-purpose NGGHS, which collects longitudinal data for
5,000 households. The sample is representative at the national level and provides reliable
estimates of key socio-economic variables for the six zones in the country. Further details on
study design and the data can be obtained in National Bureau of Statistics of the Federal
Republic of Nigeria (2014).

The panel component was first introduced in 2010/2011 and has since then been implemented
every two years. There are currently three waves available, of which we use the two first ones
(2010/2011, and 2012/2013), given that the third wave was not yet available at the time of our

                                                                                             42
data analysis. Each wave consists of two visits, one in the post-planting period and another in
the post-harvest period, which we consider separately, resulting in a total of four survey waves
from the NGGHS in our sample.

We identified household businesses in the NGGHS panel component using the non-farm
business modules of the different survey waves. A household was coded as operating a business
if there was a household member who had worked for himself, other than on a farm or raising
animals (e.g. operating his or her own business, trade, working as a self-employed professional
or craftsman) during the past 12 months preceding the survey. Owners were identified using
information on who in the household owned the business, who managed it, and who was the
survey respondent for the non-farm business module. Up to two members could be reported as
owners and/or managers per business and their individual survey identification numbers were
given, so that information on owner characteristics could be matched from the other survey
modules. We then identified the business owner as the person who was indicated as the business
owner if the household indicated only one and the same owner over the whole observation
period. For the remaining cases (i.e. households who reported two owners in the same wave,
or different owners in different waves) we identified the business owner(s) in a given round as
the owner(s):
      who was also reported to be the manager and was the respondent for the non-farm
        business module; or
      who was/were also the manager of the business, if the respondent was neither a business
        owner nor a manager of the business; or
      who was also the respondent for the non-farm business module, if no information on
        the manager was available or none of the owners were identified as the manager; or
      who was/were the business owner(s), if no information was provided on manager and
        respondent; or
      who was the household head if no information was given on who was the business
        owner, manager, and respondent; or
      who was also reported to be the manager if one owner was indicated as manager and
        the other was the respondent for the business module.

In contrast to the other multi-purpose household surveys we use, the non-farm business modules
of the NGGHS panel component to check whether the households surveyed reported any non-
farm household enterprise or income generating activity in the previous round. If so, they ask,
whether this activity is still operating, closed permanently, temporarily or seasonally. We coded
enterprises as surviving, if they were still being operated, and as closed if they were either
closed permanently, temporarily, or seasonally. Households were also asked if the firm they
reported was a new or an original one. We coded the variable newfirmstart to one if the reported
firm was a new one, and to zero if it was an original one. For simplification, we only included
firms in our sample if the household never operated more than one firm in any of the four rounds
we considered. Nevertheless, relying on the matching and information on closure and new firm
start provided by the NGGHS was not sufficient. There remained a number of inconsistent and
unclear cases, which we then coded on a case by case basis. For this we considered ownership
and business sector over the horizon over which we observed these unclear cases for a given
household, as well as, if available, information provided on closure and new firm start.

Given that the NGGHS panel component asks households in each survey round, whether they
are operating a business, regardless of whether they were operating one in the previous round,
we also extracted information on firms that were not operating in round 1, but in round 2, or 3
and use three baselines for the analysis of firm dynamics over time. Taking round 1 as baseline,
we observe horizons of six months, 1.5 years, and two years to rounds 2, 3, and 4 respectively.

                                                                                              43
With round 2 as the baseline, we observe horizons of 1.5 and two years to round 3 and 4,
respectively. And with round 3 as baseline, we observe a horizon of six months to round 4. In
the analysis, we pool the data for the same horizons with different baselines together to increase
the sample size for a given horizon. We also use different baselines when horizons differ to
increase the sample with information on different observation periods.

In order to be able to assess owner characteristics in our study, for jointly owned businesses,
and businesses for which ownership changed over the observation horizon, we determined one
owner as the owner for all waves in which we observed the firm. In the case of jointly owned
businesses, we decided to include the owner, who was reported as the owner in all of the waves
in which we observed the firm. If this was the case for both owners, we included the owner,
     who had the lowest household roster number (this chose the household head as the
        owner if he or she was among the owners), if no other business was reported for at least
        one of the owners over the whole observation period; or
     who had already been identified as the owner of a different business, observed during
        one or more different waves over the whole observation period;
and in the remaining cases we chose the owner who was reported as the owner in most of the
waves in which we observed the firm, or, if this was the case for both owners, who had the
lowest household roster number.

Townsend Thai Survey (TTHAI)
The Townsend Thai Project is a data collection project consisting of ongoing annual and
monthly multi-purpose household, institutional, and key informant (village leader) panel
surveys, in addition to collecting environmental data in Thailand. The baseline was conducted
in 1997, collecting data from rural areas in four Thai provinces, two in the Northeast and two
in the Central region. In the early and mid-2000s the sample was extended to include data from
urban areas as well as from the Southern and Northern region of Thailand. Further details on
study design can be obtained in Townsend et al. (2016). We use data from the annual household
surveys conducted between 1997 and 2014, resulting in the inclusion of 18 survey waves in our
sample.

We identified non-farm household businesses using the household business modules of the
different survey waves and coded a household as operating a business if anyone in this
household owned a business, which was entirely owned by the household or this person. Since
our interest was in non-farm enterprises, we did not consider entrepreneurial activities if they
were classified as rice mills, shrimp farms, fish farms, shrimp and fish farms, or any other
agricultural activity, as provided by the description of the type of business operated. The
household business module did not ask who was the owner or manager of the business reported.
We therefore identified business owners using the occupation module, as those household
members whose occupation was reported to be “owner of a business” in case the household
was operating a non-farm household business.

Households in the panel were asked in each survey round if anyone in the household owns a
business. Each round provides us hence with a sample of firms operating at a given point in
time. Taking advantage of this, we pool together data for the same observation period to
increase our sample for a given horizon. For instance, we include baseline information from 17
survey waves (with baselines from 1997 to 2013) to assess survival over a horizon of one year,
from 16 survey waves (with baselines from 1997 to 2012) for an observation period of two
years, and so forth. For the longest observation horizon, 17 years, we can only use information
on baseline firms from one survey round, i.e. from 1997.


                                                                                               44
The annual Townsend Thai Survey (TTHAI) we use, does not link household businesses from
round to round. It does hence not provide information about survival of a given firm that has
been identified in one of the survey rounds to later rounds. In order to code whether a business
observed in a given round survived or not, and in the latter case, whether a new firm was started,
we needed to match reported businesses between rounds based on a number of business
characteristics. For simplification, we decided to only include firms of households in our sample
who never operated more than one enterprise in any of the three survey rounds.

We then coded the business of a household reported in a given survey round as surviving to a
given follow-up round, if a business was reported for the same household in the follow-up
round, and:
    firm ages coincided across the two waves; or
    the sector of the business was the same across the two waves, and the age of the business
       reported in the follow-up round was not less than the time difference between the two
       survey waves; or
    at least one of the household members identified as the business owners was the same
       across the two waves, and the age of the business reported in the follow-up round was
       not less than the time difference between the two survey waves.

We coded the business reported at baseline as closed, if:
   no business was reported in the follow-up round.

Otherwise, we coded survival/closure as missing.

We coded a new firm start, if:
   the age of the business reported in the follow-up round was less than the time difference
      between the two survey waves.

We coded the variable newfirmstart to zero, if:
   the business reported in the baseline round survived; or
   no business was reported in the follow-up round and the business of the baseline round
      was coded as closed; or
   the age of the business reported in the follow-up round was not less than the time
      difference between the two survey waves.

Otherwise, the variable newfirmstart was coded to missing.

For businesses for which survival and newfirmstart remained unclear or inconsistent, we
inspected the characteristics on a case-by-case basis, taking into account information on firm
age, type of activity, and ownership reported in all survey waves in which we observed these
unclear cases. In addition, we used information on whether the household decided to stop
running any business in the past 12 months, and if so, the type of activity they were no longer
running, as well as whether the business reported in a given round had been founded in the past
12 months, which was provided in the household business modules of the follow-up
questionnaires.

Finally, we included one owner per firm in our sample, in case we had identified more than one
owner for a given firm, either because we had identified more than one owner in one or more
waves, or because different owners were reported in different waves. In these cases, we decided
to include the owner, who was reported as the owner in all of the waves in which we observed
the firm. If this was the case for both owners, we included the owner,

                                                                                               45
       who had the lowest household roster number (this chose the household head as the
        owner if he or she was among the owners), if no other business was reported for at least
        one of the owners over the whole observation period; or
    who had already been identified as the owner of a different business, observed during
        one or more different waves over the whole observation period;
and in the remaining cases we chose the owner who was reported as the owner in most of the
waves in which we observed the firm, or, if this was the case for both owners, who had the
lowest household roster number.


Specialized firm impact evaluation surveys:
We use data from the control group only in a number of panel data surveys collected to
measure the impacts of firm interventions in randomized experiments.

Cotonou Informal Enterprise Survey (BJINFORMALITY)
The sample surveyed for the Cotonou Informal Enterprise Survey (BJINFORMALITY)
consists of 3,596 informal businesses and is representative of informal businesses operating in
Cotonou, Benin. The survey was designed to evaluate the effect of a simple free business
registration process on formalization, and test the effectiveness of supplementary efforts to
enhance the presumed benefits of formalization by facilitating its links to government training
programs, support to open bank accounts, and tax mediation services. Details on sampling and
the interventions tested can be found in Benhassine et al., (2016). Survey participants were
randomly allocated into three treatment groups and one control group, with the control group
consisting of 1,197 informal firms. The baseline survey was conducted between March and
April 2014. Two follow-up surveys followed, the first was conducted between April and June
2015, and the second between May and June 2016. Attrition rates at first and second follow-up
were 11.8 percent and 15.9 respectively and were not correlated with treatment status.

We coded survival to one if:
   the owner reported to be still operating in the same sector as in the previous wave; or
   the business was said by a person knowledgeable on this matter to be still operating, if
      the owner refused to be re-interviewed or could not be interviewed when survey
      attempts were made.

We coded survival to zero if:
   the owner reported not to be operating anymore in the same sector as in the previous
      wave because of a change in activity; or
   the survey was not completed because the business was found to have been closed
      definitely; or
   the business was said by a person knowledgeable on this matter not to be operating
      anymore, if the owner refused to be re-interviewed or could not be interviewed when
      survey attempts were made.

We coded newfirmstart to one if the owner reported not to be operating anymore in the same
sector as in the previous wave because of a change in activity. For the endline wave, we
additionally coded new firm start to one if the enterprise had closed definitely and the owner
was now operating a different enterprises.

Newfirmstart was coded to zero if:
    an enterprise survived; or


                                                                                             46
      the survey was not completed because the business was found to have been closed
       definitely (and in the endline: the owner was not operating a different enterprise); or
       the business was said by a person knowledgeable on this matter not to be operating
       anymore, if the owner refused to be re-interviewed or could not be interviewed when
       survey attempts were made.

With this procedure, we coded survival and newfirmstart from round to round. Given that we
were interested in following only the baseline businesses, and not businesses started in the
midline, we recoded survival to be capturing survival of the baseline firm. This implied,
recoding survival to zero in the endline wave, if the business had been coded as closed and a
new firm had been started in the midline wave.

Egypt Macroinsurance for Microenterprises Survey (EGMACROINSURANCE)
The survey was conducted to evaluate the impact of an insurance to protect microenterprise
owners against macroeconomic and political shocks in Egypt, described in detail in Groh and
McKenzie (2016). The baseline survey was conducted in March 2012 among a sample of clients
from Egypt’s largest microfinance institution at particular loan sizes, who were deciding about
the renewal of their loans. The sample consisted of 2,980 individuals, with equal numbers
assigned to treatment and control. In November 2012, a follow-up survey was conducted, which
was able to re-interview 98.8 percent of respondents in the control group and 98.9 percent in
the treatment group. We use these two survey waves and are hence able to examine survival
over a horizon of approximately 6 months. Using only data from the control group yields 1,441
business owners at baseline, which we include in our sample.

We coded survival to one, if:
   the respondent said that he or she had not closed or sold the business in which he or she
      worked at the time of the baseline survey in March 2012; or
   the business was said by a person knowledgeable on this matter, to be still being
      operated by either the business owner, an immediate family member, or by some other
      person, if the respondent refused to participate in the survey or could not be located.

We coded survival to zero, if:
   the respondent said that he or she had closed or sold the business in which he or she
      worked at the time of the baseline survey in March 2012; or
   the business was said by a person knowledgeable on this matter, to be closed, if the
      respondent refused to participate in the survey or could not be located.

If the respondent said that he or she had closed or sold the business in which he or she had
worked at the time of the baseline survey in March 2012, we coded newfirmstart to one, if the
respondent said that he or she was operating a different business after closing or selling the
business previously operated, and had opened this business since March 2012. Otherwise, we
coded newfirmstart to zero.

Ghana Microenterprises Survey (GHMS)
The Ghana Microenterprises Survey (GHMS) was conducted to evaluate the impact of cash and
in-kind grants to microentrepreneurs in urban Ghana. Details of the intervention and study
design are described in Fafchamps et al. (2014). The baseline survey was conducted in October
and November 2009, surveying a representative sample of 907 microenterprise owners in urban
Ghana, in particular in the cities of Accra and Tema. A second pre-treatment survey of these
firms was conducted in February 2009 to eliminate firms most likely to attrit, yielding a final
baseline sample of 793 firms for the experiment, of which 397 were assigned to the control

                                                                                            47
group. Four additional quarterly follow-up surveys followed in May 2009, August 2009,
November 2009, and February 2010. In March 2012 a final, longer-term follow-up survey was
conducted. We include the 397 baseline control group enterprises in our sample and use data
from all six survey waves.

Using information from round to round, we coded survival in each round to one if:
    the owner said that he or she was still owning and operating a business, and the line of
       the business had not changed; or
    the line of the business had changed, but when asked what happened to the business
       previously ran, the owner said that he or she was either operating the same business, or
       that the business was now being run by another family member; or
    the owner or a person knowledgeable of the matter told the enumerator that the owner
       was still operating the business, if the respondent refused to participate in the survey or
       could not be located; or
    the owner or a person knowledgeable of the matter told the enumerator that he owner
       was not operating the business anymore but that someone else was operating it, if the
       respondent refused to participate in the survey or could not be located.

We coded survival to zero if:
   the line of the business had changed, and when asked what happened to the business
      previously run, the owner said that the business had closed down, had been modified to
      become the current business, or gave any other answer that implied the closure of the
      original business; or
   the owner or a person knowledgeable of the matter told the enumerator that the business
      had closed down, if the respondent refused to participate in the survey or could not be
      located.

We coded newfirmstart to one, if the line of the business had changed from one round to another
and the respondent said that his or her main activity was then operating a different business.
Otherwise, we coded newfirmstart to zero.

To be able to assess survival of firms operating at baseline, we recoded survival in a final step,
by recoding it to zero for all subsequent rounds, if it had been coded as zero in a given round
and a new firm start had been coded in the same round or later.

Kenya GET Ahead Business Training Program Impact Evaluation Survey (KEGETAHEAD)
The survey was part of an evaluation of a business training program for women in four counties
of Kenya: Kakamega and Kisii in the Western region, and Embu and Kitui in the Eastern region.
In each county a census was taken of all female-owned businesses in market centers deemed as
medium-sized or large, and then a screening was done to select a sample in which the business
did not have more than 3 employees; the business had profits in the past week between 0 and
4000 Kenyan Shillings (KSH) (1 US Dollar averaged approximately 85 KSH over the survey
period); sales in the past week less than or equal to 50,000 KSH; and the individual had at least
one year of schooling. The survey is therefore representative of microenterprises of this size
run by women in these four counties.

The baseline survey took place between June and November 2013, and included 3,537
individuals. These individuals were then assigned to treatment and control for the business
training intervention in a two-stage process, with markets being assigned to treatment or control
in the first stage, and then, within markets that had been assigned to treatment, individuals being
assigned to treatment or control. We include all control group observations, i.e. individuals in

                                                                                                48
pure control markets and individuals assigned to the control group in treatment markets, totaling
2,365 firms, in our sample. One year after the baseline survey, between June and October 2014,
the first follow-up survey was conducted, which was supplemented by a rapid short survey that
was conducted between November 2014 and February 2015, and which we mainly used to
obtain information on firms not found in the first follow-up. The second follow-up survey took
place between February 2016 and July 2016. Further details on the survey, intervention, and
study design can be found in McKenzie and Puerto (2017). We used all four survey waves,
joining information from the first follow-up survey and the subsequent short follow-up survey
and treating them as one round, hence including three rounds of data from this survey in our
sample.

Using information from round to round, we coded survival in each round to one if:
    the owner answered that she was still operating the same business as in the previous
       round; or
    the owner was said by a proxy respondent to still be operating a business, if the
       respondent refused to participate in the survey or could not be located.

We coded survival in each round to zero if:
   the owner answered that she had closed the business and was now operating a different
      business; or
   the owner answered that she was no longer operating a business; or
   the owner said that she had never operated a business; or
   the owner said or was said by a proxy respondent, not to be in business anymore or to
      have left the country, if the respondent refused to participate in the survey or could not
      be located; or
   the owner had died.

We coded newfirmstart to one, if the owner answered that she had closed the business and was
now operating a different business. If the owner answered either that she was still operating the
same business as in the previous round, or that she was no longer operating a business, we
coded newfirmstart to one. In the remaining cases newfirmstart was coded to missing.

To transform the coding into survival to a given round for firms that were operating at baseline,
and to deal with some inconsistencies that were due to the combination of data from the first
follow-up and the subsequent short follow-up, we recoded survival in a final step. If survival to
round 2, as well as newfirmstart to round 2 had been coded to one, we recoded survival to round
2 to zero, assuming that these were either cases in which the original business had survived to
the first follow-up, but not to the subsequent short follow-up survey, and the owner was
operating a new firm by the time of the short follow-up survey, or cases in which the owner had
closed and opened new firm in the first follow-up and then the business was coded as surviving
in the short subsequent follow-up survey, because this new business continued to be open. If a
business had been coded to be closed by round 2, but survived to round 3 and a new firm had
been started by round 2, we recoded survival to round 3 to zero.

Malawi Business Registration Impact Evaluation Survey (MWBRIE)
The Malawi Business Registration Impact Evaluation Survey (MWBRIE) was designed to
evaluate the impact of a randomized experiment designed to increase business formalization in
Malawi. The survey was conducted on a sample of 3,002 informal firms in urban Lilongwe and
Blantyre, which had been identified through a listing of over 100 business centers (i.e.
concentrations of firms including industrial parks, markets, streets with shops, set of
workshops, etc.). As a result the sample does not include household-based enterprises.

                                                                                              49
Moreover, firms in the sample comply with one of the following criteria: (i) they had at least
one worker contracted outside of family members and business owners, (ii) they were operating
in a fixed location with more than one person working in the business, (iii) they were at the 25
percentile of revenues or above. More details are provided in Campos et al. (2015).

The baseline survey was conducted between December 2011 and April 2012. After that, the
sample was randomly assigned to either one of three treatment arms or the control group, which
consisted of 757 firms. Details on the intervention and experimental design can be found in
Campos et al. (2015). A first follow-up survey took place between November 2012 and March
2013. Attrition was 5.7 percent, and uncorrelated with treatment status. The second follow-up
survey was implemented between November 2013 and March 2014. Attrition for this second
follow-up survey was 9.4 percent, and also uncorrelated with treatment status. Between
November 2014 and April 2015, a third follow-up survey took place, and a fourth follow-up
survey was conducted between July 2015 and January 2016. Including data from all five survey
rounds into our sample, we are able to observe firms over horizons of up to 3.5 years.

We coded survival to one if owners said that they were still operating the business that they had
when they were first interviewed, i.e. around December 2011 to April 2012. If they said they
were not operating this business anymore, we coded survival to zero. Newfirmstart was coded
to one if the business had closed and the owner had started a new business, and to zero if the
business had closed but the owner was engaged in an activity different than operating another
business.

Nigeria YouWiN! National Business Plan Competition Impact Evaluation Survey
(NGYOUWIN)
The survey sample consists of applicants to a nationwide business plan competition launched
in late 2011 and described in McKenzie (2017). Applicants submitted their business plans in
January 2012, together with baseline information on firm and owner characteristics collected
via a baseline data sheet. Business plans were then narrowed down to 2,312 semi-finalists, of
which 1,200 received awards averaging US$50,000 each, and among which 729 winners had
been randomly selected from a group of 1,841 semi-finalists. This left 1,112 control group
individuals. Three follow-up surveys were taken, approximately at yearly intervals after
individuals had applied for the program, with the first follow-up conducted between November
2012 and May 2013, the second between October 2013 and February 2014, and the third
between September 2014 and February 2015. A fourth follow-up survey took place between
July and November 2016.

Given that at baseline, there were not only already existing firms in the sample, but also
individuals who intended to start a new business with the grant, and that, in addition, limited
firm data collected at baseline, we decided to include only the four follow-up surveys in our
sample, taking businesses that were existing by follow-up 1 (either because they survived from
the baseline to this round, or because they had been created in between these two rounds) as the
baseline sample. In addition, we only kept firms, which were not in the business of agriculture,
forestry or fishing, whose owners never operated more than one business in any of the rounds
in which we observed them, and which had been assigned to the control group. This resulted in
the inclusion of 280 firms from the Nigeria YouWiN! National Business Plan Competition
Impact Evaluation Survey (NGYOUWIN) in our sample at baseline.

We coded survival to one, if:



                                                                                              50
        the respondent was currently operating a business in the follow-up round and start dates
         of the business were the same across the follow-up round and the baseline round (i.e.
         follow-up 1); or
        the respondent was currently operating a business in the follow-up round and the start
         date of the business was missing in the follow-up round12; or
        a person knowledgeable on the matter answered that the respondent was currently
         owning a business in the follow-up round, if the owner could not be interviewed after
         repeated attempts.

We coded survival to zero, if:
   the respondent was not operating a business in the follow-up round; or
   the respondent was currently operating a business in the follow-up round, but start dates
      of the businesses reported in the follow-up and baseline round (i.e. follow-up 1) did not
      coincide; or
   a person knowledgeable on the matter answered that the respondent was not owning a
      business in the follow-up round, if the owner could not be interviewed after repeated
      attempts.

We coded newfirmstart to one, if the respondent was currently operating a business in the
follow-up round, but start dates of the businesses reported in the follow-up and baseline round
(i.e. follow-up 1) did not coincide. Given that firms were only asked for their start dates in
follow-up rounds 2 to 4, if they had been established in the survey year, this should have picked
up new firm start.

We coded newfirmstart to zero, if the respondent was not operating a business, or survival had
been coded to one.

In a final step, we then recoded survival for firms that had been started in a given round and
had then been coded as surviving in a later round, because they had appeared with missing start
dates in this round.

Sri Lanka Microenterprise Survey (SLMS)
The Sri Lanka Microenterprise Survey (SLMS) was designed to evaluate an experiment to
measure the impact of providing capital grants to microenterprises. It took place in three
Southern and South-Western districts of Sri Lanka: Kalutara, Galle and Matara. One-time
grants of between US$100 and US$200 were given in cash or in-kind (as business equipment)
to male and female-owned enterprises, some of whom had been affected by the December 2004
tsunami. A door-to-door screening survey of households was used to identify enterprises with
invested capital of 100,000 rupees (about US$1000) or less, excluding investments in land and
buildings, resulting in a baseline sample of 617 enterprises in retail trade and manufacturing,
operated by owners 22 to 65 years old, and with no paid employees.


12
   Businesses were only asked their start dates in follow-up rounds 2 to 4 if they had been established in the survey
year. We hence assumed that all firms which had missing values for start dates in these rounds, had been
established earlier. We corrected for potential erroneous coding as a result of this rule in a final step. Initially we
had tried to use ISIC codes of the businesses reported in different rounds to code survival, coding survival to one
if a firm reported in a follow-up round was operating in the same sector as in in the baseline (i.e. follow-up 1)
round and as closed, if they were in different sectors. This led however to extremely high closure rates, and a visual
inspection of the sectors in two consecutive survey waves revealed that sectors of firms coded as closed with this
rule, were very similar, although having different ISIC codes. As a result, we decided not to use ISIC codes for
the coding of survival.


                                                                                                                   51
The baseline survey was conducted in April 2005. Firms in the sample were then randomly
allocated into one of two treatment groups, which received US $100 or $200 either after the
first or the third survey round and either in cash or in-kind. Details on the experimental design
and intervention can be found in de Mel et al. (2008, 2009, and 2012). Follow-up surveys were
conducted at quarterly intervals, from July 2005 to April 2007, and semi-annually in October
2007 and April 2008. Further follow-up surveys were carried out in June and December 2010,
in September 2015, and in March 2016. We included all 15 waves of data collection in our
sample, which allows us to observe firms for horizons of up to eleven years. Using firms
assigned to the control group, we include 269 firms at baseline from the SLMS in our sample

For rounds 2 to 11, we coded survival from round to round to one, if the owner was still working
in the same line of business, he or she had been working in, in the previous round., and to zero
if the line of business had changed. Newfirmstart was coded to one for rounds 2 to 11 if the line
of business had changed and the owner stated that his or her main activity was to operate a new
business. If the line of business had changed but the main activity was a different one (for
instance, working for a wage, looking for work, or housework or looking after children), we
coded newfirmstart to zero.

In round 12, questions on business continuation referred to the original baseline business. We
coded survival to one, if the owner was still working in the same line of business, he or she had
been working in, during the baseline survey in April 2005, and to zero if the line of business
had changed. Newfirmstart was coded to one if either the line of the original business had
changed or the owner had started a new line of business while continuing to operate the original
business, and the owner stated that his or her main activity after closing, selling or changing the
business was either to operate a different business or to ooperate a new business along with the
original one. If the line of business had changed but the main activity was different to operating
another business, we coded newfirmstart to zero. For round 13, we coded survival and
newfirmstart based on the same criteria, only that the questions asked here referred to the
business being operated in round 12 and survival and newfirmstart were hence being coded
from round to round.

In rounds 14 and 15 we coded survival from round to round to one, if:
     the owner stated that he or she was currently owning a business, and had not changed
       the industry of the business since the previous round or closed down the business he or
       she had been running and started a new business; or
     a person knowledgeable of the matter answered that either the enterprise owner was still
       running the business, or the business was being run by an immediate family member,
       or the business was being run by some other person, if the respondent refused to
       participate in the survey or could not be located.

We coded survival from round to round to zero in these rounds, if:
   the owner stated that he or she was currently not owning a business; or
   the owner said that he or she had changed the industry of the business since the previous
      round, or that he or she had closed down the business he or she had been running and
      started a new business; or
   a person knowledgeable of the matter answered that the enterprise was closed or not
      functioning, if the respondent refused to participate in the survey or could not be located.

Newfirmstart was coded in rounds 14 and 15 to one, if:


                                                                                                52
        the business being owned in the given round had been started after the date of the
         previous survey round; or
        the owner stated that he or she had either changed the industry in which the business
         was operating, or started another business while still operating the original business, or
         closed down the previously run business and started a new one.

Newfirmstart was coded to zero in these rounds, if:
    the owner stated that he or she was currently not owning a business.

Except for round 12, we then recoded survival in a way that it referred to survival of firms
operating at baseline by recoding it to zero for all subsequent rounds, if it had been coded as
zero in a given round and a new firm start had been coded in the same round or later.

Sri Lanka Impact Evaluation Survey of Business Training for Women
(SLFEMBUSTRAINING)
The Sri Lanka Impact Evaluation Survey of Business Training for Women
(SLFEMBUSTRAINING) was conducted to evaluate a business training intervention aimed at
increasing female labor force participation rates and to raise the income levels of low-earning
women already in business. It surveys two different groups of females: (i) a representative
sample of 628 current female business owners operating in greater Colombo and greater Kandy
areas, who had monthly profits of 5,000 Rs or less (US$43), and (ii) a representative sample of
628 women between 25 and 45 years of age in greater Colombo and greater Kandy areas, who
were out of the labor force, but who were likely to start working within the next year, called
potential business owners. These two groups were first interviewed in January 2009, after which
400 current and 400 potential business owners were allocated to treatment. Four rounds of
follow-up surveys were then conducted in September 2009, January 2010, September 2010,
and June 2011. A fifth and long-term follow-up survey was conducted in June 2015. Details on
the survey and study design are described in de Mel et al. (2014).

Since our interest lies in existing businesses, we use follow-up 1 to 5, including current business
owners who were operating a business by the first follow-up round and potential business
owners who had started a business by then, and who had been assigned to the control group, as
our baseline sample.13 This resulted in the inclusion of 260 firms at baseline in our sample.

We coded survival from round 1 to round 2 ( i.e. from follow-up 1 to follow-up2) to one, if the
owner reported that she was still working in the same line of business as the one she was
working in in September 2009. If the owner had changed the line of business of the original
business, or was not operating a business anymore, we coded survival from round 1 to 2 to zero.
Newfirmstart was coded to one from round 1 to round 2, if the owner answered that she was
either operating a different business or operating a new business along with the old one, after
closing, changing or selling the business, and to zero, if she was engaged in a different activity
then.

In rounds 3 and 4 (the fourth and fifth follow-up) separate questionnaires were given to females
owning only one business and those owning more than one business. For single business owners
in rounds 3 and 4, we coded survival from round to round to one, if the owner reported that she
was still working in the same line of business as the one she was working in in the previous
wave. If the owner had changed the line of business of the original business, or was not

13
  For current business owners we also use round 1 to extract information on time-invariant characteristics
that were not asked in round 2.
                                                                                                             53
operating a business anymore, we coded survival from round to round to zero. In round 4, we
additionally used information from the attrition section to code survival, coding a firm of a
single business owner as surviving if a person knowledgeable on the matter said that the owner
was still running the business, the business was being run by an immediate family member, or
the business was being run by some other person, in case the respondent refused to participate
in the survey or could not be located. If the proxy respondent answered that the business was
closed, we coded survival to zero. We coded newfirmstart to one for single business owners in
rounds 3 and 4, if the owner answered that she was either operating a different business or
operating a new business along with the old one after closing, changing or selling the business,
and to zero, if she was engaged in an activity different than operating another business, after
closing, changing or selling the business.

For owners of multiple businesses in rounds 3 and 4, we coded survival from round to round to
one, if:
    none of the businesses operated in the previous round had been closed and business
         lines for all businesses had not changed since the previous round; or
    after closing, changing or selling one of the businesses operated in the previous round,
         the owner was either operating a new business along with one previous business, or
         operating only one of the previous businesses.

We coded survival from round to round to zero, for owners of multiple businesses in rounds 3
and 4, if, after closing, changing or selling the businesses operated in the previous round, the
owner was either working for a wage, looking for work, operating a different business, or doing
housework or looking after children.

In round 4, we additionally used information from the attrition section to code survival, coding
a firm of a multiple business owner as surviving if a proxy respondent answered that:
      the owner was still running all businesses; or
      the owner was not running all businesses, but at least one; or
      the owner had closed down at least one of the businesses, but not all; or
      at least one of the businesses was being run by an immediate family member; or
      at least one of the businesses was being run by some other person,
if the respondent refused to participate in the survey or could not be located, and coding survival
to zero if the proxy respondent said that all businesses had closed down.

For owners of multiple businesses in rounds 3 and 4, we coded newfirmstart to one, if, after
closing, changing, or selling at least one of the businesses, or starting another business, the
owner was either operating a different business, or operating a new business along with one or
both previous businesses, and to zero, if after closing, changing, or selling the business, the
owner was either engaged in wage work, looking for work, doing housework or looking after
children, or she was operating only one of the previous businesses, or none of the businesses
operated in the previous round had been closed and business lines for both businesses had not
changed since the previous round.

In the final round, we coded survival to one, if:
     the respondent was currently owning a business and, since January 2009, she had not
        changed the industry in which the business was operating, or closed down a business
        she was operating and then started a new business; or
     a proxy respondent reported that the owner was still running the business, or the
        business was being run by an immediate family member, or the business was being run


                                                                                                54
       by some other person, if the respondent refused to participate in the survey or could not
       be located.

We coded survival to zero in the final round, if:
   the respondent was not owning a business; or
   the respondent said that, since January 2009, she had changed the industry in which the
      business was operating or she had closed down a business she was operating and then
      started a new business; or
   a proxy respondent said that the business was closed, if the respondent refused to
      participate in the survey or could not be located.

We coded newfirmstart in round 5 (the final round) to one, if the business had been started after
the previous survey round, or, since January 2009, the owner had either changed the industry
in which the business was operating, started another business while still operating the original
one, or closed down the original business and then started a new business. If the respondent
was not operating a business in the final round or the business she was operating had been
started before the previous survey round, we coded newfirmstart to zero.

According to the criteria described above, survival had been coded since baseline only in the
second and final round. We hence recoded survival in rounds 3 and 4, in order to reflect survival
since baseline, by recoding it to zero for all subsequent rounds, if it had been coded as zero in
a given round and a new firm start had been coded in the same round or later.

Sri Lanka Informal Enterprises Survey (SLINFORMALITY)
The Sri Lanka Informal Enterprises Survey (SLINFORMALITY) was part of a field experiment
that provided informal firms incentives to formalize. A sample of 520 informal firms owned
either exclusively by men or jointly owned by husband and wife, with one to 14 paid employees
in the two largest Sri Lankan cities – Colombo and Kandy – were randomly divided into four
treatment groups and a control group. Details on sample selection, experimental and survey
design can be found in de Mel et al. (2013).

The baseline survey was conducted in December 2008. Treatment was then assigned randomly,
with 105 businesses being assigned to the control group. Three follow-up surveys were carried
out in August 2010, March 2011 and December 2011, and we include data on the control firms
from all four survey rounds in our sample.

We coded survival from round to round to one, if, in a given wave, the owner was still engaged
in the same line of business as in the previous survey round, and to zero, if the line of business
had changed. In the last round, we were able to use information obtained from proxy
respondents to code survival, if the owner had refused to participate in the survey or could not
be located. If the proxy respondent answered that the owner was still running the business, or
the business was being run by an immediate family member, or the business was being run by
some other person, we coded survival to one. We coded survival to zero if the proxy respondent
said that the business was closed.

For rounds 3 and 4 we then recoded survival, so that it reflected survival since baseline, by
recoding it to zero for all subsequent rounds, if it had been coded as zero in a given round and
a new firm start had been coded in the same round or later.




                                                                                               55
Sri Lankan Longitudinal Survey of Enterprises (SLLSE)
The Sri Lankan Longitudinal Survey of Enterprises (SLLSE) is a long-term panel survey of
microenterprises designed to assess the impact of a wage subsidy, plus a savings and a business
training intervention as supplementary interventions (see de Mel et al, 2010, 2016 for details
on the study and survey design). The baseline survey was intended to be a representative survey
of urban microenterprises with two or fewer paid employees, owned by males aged 20 to 45,
and operating in non-agricultural sectors, in the Colombo, Kandy, and Galle-Matara areas of
Sri Lanka. It was conducted between April and October 2008. The first phase occurred in April
2008 as part of a larger panel survey and also included other urban areas in Sri Lanka. In
October 2008, a booster listing exercise and survey took place in order to attain a larger sample
for the intervention. The original sample was also re-interviewed during the October 2008
survey, resulting in a total sample of 1,533 firms. These firms were then randomly allocated
into treatment and control groups, the latter of which consisted of 286 enterprises. After these
two survey rounds, the sample was re-interviewed in 10 additional follow-up rounds, consisting
of six-monthly surveys every April and October from 2009 through 2012, and two additional
surveys in April 2013 and April 2014.

We included firms allocated to the control group that were either part of the booster sample, or
surveyed in April 2008 and reported no change in the line of business in October 2008, taking
the survey conducted in October 2008 as the baseline. Our baseline sample then comprises 284
firms, which we follow for a total of eleven rounds (baseline plus ten follow-ups).

We coded survival from round to round to one, if:
     the respondent reported that he was still working in the same line of business as in the
       previous round; or
     the business was now being operated by another household member; or
     the respondent had sold the business to someone else to run.
Otherwise, survival was coded to zero.

Starting in round 6, we additionally used information from the attrition section to code survival,
coding it to one if the respondent could not be located or refused to participate in the survey,
and the proxy respondent said that the owner was either still operating the business, or the
business was being run by another family member or by someone else. If the proxy respondent
said that the business had closed down, we coded survival to zero.

Newfirmstart was coded to one, if the respondent was operating a different business, after
closing, changing or selling the business, and to zero otherwise.

In a final step, we recoded survival for rounds 3 to 11, so that it reflected survival since baseline,
by recoding it to zero for all subsequent rounds if it had been coded as zero in a given round
and a new firm start had been coded in the same round or later.

Lomé Informal Enterprise Survey (TGINFORMALITY)
The Lomé Informal Enterprise Survey (TGINFORMALITY) was used for an impact evaluation
of two training interventions for informal businesses in Lomé. The sample was drawn from
applicants to the government project under which these trainings were conducted and consists
of 1,500 microenterprises. Applicants had to be in business for at least 12 months, have fewer
than 50 employees, operate outside of agriculture, and not be a formally registered company.
Further details on the intervention, sample selection and survey can be found in Campos et al.
(2017). The baseline survey was conducted between October 2013 and January 2014. Firms
were then randomly assigned into one of two treatment groups and one control group, each of

                                                                                                   56
which included 500 firms. The first follow-up survey took place in September 2014, followed
by a second in January 2015, a third in September 2015 and a fourth in September 2016. We
included data for control group firms from all five survey rounds and excluded firms that were
in the agricultural sector, resulting in the inclusion of 499 firms at baseline into our sample.

In each of the follow-up rounds, we coded survival to one, it the owner reported to still be
working for the enterprise he or she operated in the previous round, and to zero if the owner
said he or she was not working for this enterprise anymore, either because it was closed down
or because of another reason. In round five, we additionally used information for owners, who
refused to participate in the survey or could not be located, to code survival. If the proxy
respondent said that the owner was still working for the enterprise he or she operated in the
previous round, we coded survival to one, and if the proxy respondent said that the owner was
not working for this enterprise anymore, we coded survival to zero. Newfirmstart was coded to
one if the owner was operating a business and was not working for the previous business
anymore, and to zero, if the owner was engaged in a different activity than operating a business
and not working for the previous business anymore. For rounds 3 to 5, in a final step, we then
recoded survival to zero, if it had been coded as zero in a given round and a new firm start had
been coded in the same round or later, so that it reflected survival since baseline.

Uganda Women’s Income Generating Support Impact Evaluation Survey (UGWINGS)
The Uganda Women’s Income Generating Support Impact Evaluation Survey (UGWINGS)
was conducted to evaluate individual business training, a one-time unconditional cash grant of
US$150, supervision, and business advising, provided to very poor Ugandans, mostly women.
Details of the intervention and experimental design are described in Blattmann et al. (2016).
The baseline survey was conducted on a sample of 1,800 individuals from April to June 2009
who had been selected for the intervention. Of these, 896 were randomized to receive either
training, or cash and supervision (phase 1 treatment group), and 904 were randomized to
waitlist treatments (phase 2). The first follow-up survey was conducted between November
2010 and February 2011. Phase 2 participants then all received business training and cash and
in addition supervision and advise, which was randomized among them. They received a short
follow-up survey between September and October 2011 (roughly a month after receiving the
grant), and a further, more detailed, follow-up survey between June and August 2012.

Individuals were not asked whether they were already operating a small business at baseline,
but in the follow-up surveys conducted between November 2010 and February 2011, and
between June and August 2012, respondents were asked whether they had started a small
business between the last time they had been surveyed and the current survey. In particular they
were asked how many small businesses they had started or had tried to start within this period.
They were then asked how many of these were still being operated. We used this information
to construct the variables newfirmstart and survival. In particular, we constructed them in the
following way:
      We coded newfirmstart to 1 if the individual reported to have started or tried to start at
        least one new small business since the last time he or she was surveyed.
      We coded survival to 1 if the individual reported to have started or tried to start at least
        one new small business since the last time he or she was surveyed and at least one of
        these businesses was still active. If none of these businesses was active anymore, we
        coded survival to zero.

In the endline for phase-2 individuals (June and August 2012), individuals were not asked if
they had started a business between the baseline and the midline survey (conducted between
November 2010 and February 2011), so it was not possible to follow the businesses they had
                                                                                                57
started then until the final round. They were only asked whether they started a new small
business between the last time they were surveyed and the current
survey. Furthermore, at that time they had already received treatment. We hence included in
our sample data from 461 phase-2 businesses which had been started between round 1 and
round 2, and observe them until round 2, which is a maximum horizon of 18 months.




                                                                                        58
Table A.1. Many Published RCTs of Small Firm Interventions do not discuss Firm Death
       Author(s)               Type of        Firm Size     Country      Sample Size       Time Frame         Attrition           Firm        Death Rate in   Note
                            Intervention                                                  (approximate)         Rate         Death/Survival   Control Group
                                                                                                                             is an Outcome
Angelucci et al. (2015)     microcredit        micro        Mexico      Total: 2,912     20 months              37%                yes            n.a.         a.
Atkin et al. (2017)       export promotion     micro         Egypt      Total: 219       2.5 years (S1)      28% (S1)             no                           b.
                                                                        Control: 145     15 months (S2)       7% (S2)

Beaman et al. (2014)            other          micro        Kenya       Total: 508       n.a. / n.r.           2-3%               no                           c.
Berge et al. (2015)       business training    micro       Tanzania     Total: 644       3 years               2.6%               no
                           and cash grant                               Control: 199
Blattman and Dercon       business training    micro       Ethiopia     Total: 643       1 year                 14%               no                           d.
(forthcoming)              and cash grant                               Control: 358
Blattman et al. (2014)       cash grants        micro       Uganda      Total: 2,677     4 years                18%               no
Bruhn et al.                management          micro,      Mexico      Total: 432       1 year                 12%               no              3.3%         e.
(forthcoming)                consulting       small and                 Control: 282
                                               medium
Bruhn and Zia (2013)      business training     micro     Bosnia and    Total: 445       1 year                 11%               yes             36%
                                                          Herzegovina   Control: 148
De Andrade et al.           formalization      micro         Brazil     Total: 1,348     1 year                36.8%              yes            14-16%        f.
(2016)                        incentives                                Control: 689
De Giorgi and Rahman        formalization     small and   Bangladesh    Total: 3,000     2 years                n.r.              no
(2013)                        incentives       medium                   Control: 1,500
De Mel et al. (2014)      business training     micro      Sri Lanka    Total: 624       2.5 years             7.9%               yes             17%          g.
                           and cash grant                               Control: 224
De Mel et al. (2013)        formalization       micro      Sri Lanka    Total: 520       31 months              11%               yes             n.r.
                              incentives      and small                 Control: 105
De Mel et al. (2012)      cash and in-kind      micro      Sri Lanka    Total: 387       5.25 years       10% (5.25 years)        yes          29% (males)     h.
                                grants                                                   5.75 years        8% (5.75 years)                    26% (females)
De Mel et al. (2009)      cash and in-kind     micro       Sri Lanka    Total: 387       3 years               13.8%              no
                                grants
De Mel et al. (2008)      cash and in-kind     micro       Sri Lanka    Total: 408       2 years               9.6%               no
                                grants                                  Control: 280
Drexler et al. (2014)     business training    micro      Dominican     Total: 1,193     1.5 years              13%               no
                                                           Republic     Control: 387



                                                                                                                                                                59
Dupas and Robinson       formal savings       micro        Kenya       Total: 392         6-18 months         36%          no
(2013)                      accounts                                   Control: 197
Fafchamps and Quinn           other            small      Ethiopia     Total: 126 (E),    1 year              n.r.         no                      i.
(forthcoming)                                             Tanzania     134 (T), 85 (Z)
                                                           Zambia      Control: 40 (E),
                                                                       44 (T), 22 (Z)
Fafchamps and           business training      small       Ghana       Total: 140         2 years              0%          yes        n.r.         j.
Woodruff (2016)                                                        Control:70
Fafchamps et al.        cash and in-kind      micro        Ghana       Total: 793         3 years             14%          no
(2014)                       grants                                    Control: 396
Field et al. (2013)       microcredit         micro         India      Total: 845         3 years              9%          yes       38.6%
                                                                       Control: 425
Field et al. (2010)     business training     micro        India       Total: 636         4 months            5.3%         no
                                                                       Control: 212
Glaub et al. (2014)     business training      micro      Uganda       Total: 100         1 year               0%          yes        9.4%         k.
                                             and small                 Control: 53
Groh and McKenzie       insurance against      micro       Egypt       Total: 2,961       20 months           1.2%         no
(2016)                   macroeconomic                                 Control: 1,481
                           uncertainty
Higuchi et al. (2015)    business training     small      Vietnam      Total: 312         2.5 years          12.2%         no
                                                                       Control: 115
Karlan et al. (2015)      management          micro        Ghana       Total: 160         2 years             6.9%         no                      l.
                           consulting                                  Control: 45
                        services and cash
                              grants
Karlan and Valdivia     business training     micro         Peru       Total: 4,591       1-2 years           24%          no
(2011)
Karlan and Zinman          microcredit        micro      Philippines   Total: 1,601       11-22 months        30%          no
(2011)                                                                 Control: 329
Mano et al. (2012)      business training      micro       Ghana       Total: 167         3.5 years          10.8%         yes        9.3%         k.
                                             and small                 Control: 107
Martínez et al.         business training      micro        Chile      Total: 1,948       45 months           23%          yes       75.1%         m.
(forthcoming)              and in-kind                                 Control: 566
                             transfer
McKenzie (2017)         large cash grants      micro      Nigeria      Total: 541         3 years        28.1% (1 year)    yes    13% (1 year)     n.
                                             and small                 Control: 263                      12.8% (2 years)         15.6% (2 years)
                                                                                                         19.5% (3 years)         24.1% (3 years)


                                                                                                                                                    60
McKenzie et al. (2017)     matching grants     small and       Yemen        Total: 416        15-17 months              51%                 no                 1%          k.
                                                medium                      Control: 200
McKenzie and              cash and in-kind       micro         Mexico       Total: 207        1 year                   33.8%                no
Woodruff (2008)                grants                                       Control: 77
Premand et al. (2016)     business training      micro         Tunisia      Total; 1,702      1.5 years                 7.2%                no
                                                                            Control: 851
Valdivia (2015)           business training      micro          Peru        Total: 1,979      1 year (FU1)          16.2% (FU1)             yes           15.7% (FU1)      o.
                           and technical                                    Control: 565      2 years (FU2)         19.8% (FU2)                            20% (FU2)
                             assistance
Notes:
n.r. denotes not reported. n.a. denotes not applicable.
Where available attrition rate refers to the control group, otherwise attrition rate refers to the total sample.
a. Only part of the full sample has a panel structure. Sample size, time frame, and attrition reported here refer only to the panel sample. One of the outcomes the
     authors assess is whether the respondent no longer has a business. The timeframe is indefinite and the outcome is measured as a binary variable equal to 1 if
     the respondent used to have a business but no longer has one.
b. The authors use different samples in the analysis. S1 denotes sample 1 and S2 denotes sample 2. Sample sizes refer to the combined sample, which is the
     analysis sample.
c. The intervention consisted of different methods to increase the salience of lost sales due to changeouts (running out of change). Firms were surveyed weekly
     and the experimental design relied on random variation in the introduction into the study. Hence the number of survey waves and length of follow up vary.
d. There was an additional treatment arm, to which 304 study participants were assigned. We report only on the entrepreneurship program treatment arm and
     the control group.
e. Data comes from two sources: (1) a one-year follow-up and (7) social security data, covering a horizon of up to 7 years. We do not consider the longer term
     horizon since the data is not able to assess survival. The death rate of the control group is provided in a footnote in the context of attrition.
f. Sample sizes and attrition only refer to the panel sample.
g. The authors study two groups of women: women who were already operating a business at baseline (current business owners) and women out of the labor
     force but interested in starting a business (potential business owners). Survival is only assessed for current business owners, and so we focus only on this sample
     here.
h. Death rates are based on the pooled information from the two long-term follow-up surveys.
i. The intervention randomly created social networks among firm owners.
j. The business training was part of a larger business plan competition. Training was randomly assigned among enterprise owners who presented their project
     bevor the competition panel. We report only on the sample that presented before the panel.
k. We calculated the death rate in the control group, based on information provided in the article.
l. The authors note that no firm permanently exited the tailoring business during the course of the study, but that some closed their businesses temporarily for
     health or other reasons.
m. Survival is measured as 1 if the individual was self-employed in 2011 and self-employed in 2013.



                                                                                                                                                                            61
n.   Survival is only measured for firms that already existed at the start of the intervention. Part of the full intervention sample consisted of entrepreneurs interested
     in starting a business. We report sample sizes only for the existing businesses sample.
o.   Death rates are computed asking business owners whether they closed a business in the last two years.




                                                                                                                                                                             62
References for Appendix Table A.1.
Angelucci, Manuela, Dean Karlan, and Jonathan Zinman (2015) “Microcredit impacts:
Evidence from a randomized microcredit program placement experiment by Compartamos
Banco”, American Economic Journal: Applied Economics 7(1):151-182.

Atkin, David, Amit K. Khandelwal, and Adam Osman (2017) “Exporting and firm performance:
Evidence from a randomized experiment”, The Quarterly Journal of Economics 132(2): 551-
615.

Beaman, Lori, Jeremy Magruder, and Jonathan Robinson (2014) “Minding small change among
small firms in Kenya”, Journal of Development Economics 108: 69-86.

Berge, Lars Ivar Oppedal, Kjetil Bjorvatn, and Bertil Tungodden (2015) “Human and financial
capital for microenterprise development: Evidence from a field and lab
experiment”, Management Science 61(4): 707-722.

Blattman, Christopher and Stefan Dercon (forthcoming) “The impacts of industrial and
entrepreneurial work on income and health: Experimental evidence from Ethiopia”, American
Economic Journal: Applied Economics.

Blattman, Christopher, Nathan Fiala, and Sebastian Martinez (2014) “Generating skilled self-
employment in developing countries: Experimental evidence from Uganda”, The Quarterly
Journal of Economics 129(2): 697-752.

Bruhn, Miriam, Dean Karlan, and Antoinette Schoar (forthcoming) “The impact of consulting
services on small and medium enterprises: Evidence from a randomized trial in Mexico”,
Journal of Political Economy.

Bruhn, Miriam and Bilal Zia (2013) „Stimulating managerial capital in emerging markets: The
impact of business training for young entrepreneurs. Journal of Development Effectiveness,
5(2): 232-266.

De Andrade, Gustavo Henrique, Miriam Bruhn, and David McKenzie (2016) “A helping hand or
the long arm of the law? Experimental evidence on what governments can do to formalize
firms”, The World Bank Economic Review 31(1): 24-54.

De Giorgi, Giacomo and Aminur Rahman (2013) “SME’s registration: evidence from an RCT in
Bangladesh”, Economics Letters 120(3): 573-578.

De Mel, Suresh, David McKenzie, and Christopher Woodruff (2014) “Business training and
female enterprise start-up, growth and dynamics: Experimental evidence from Sri Lanka”,
Journal of Development Economics 106: 199-210.

De Mel, Suresh, David McKenzie, and Christopher Woodruff (2013) “The demand for, and
consequences of, formalization among informal firms in Sri Lanka”, American Economic
Journal: Applied Economics, 5(2): 122-150.

De Mel, Suresh, David McKenzie, and Christopher Woodruff (2012) “One-time transfers of
cash or capital have long-lasting effects on microenterprises in Sri Lanka”, Science 335(6071):
962-966.


                                                                                            63
De Mel, Suresh, David McKenzie, and Christopher Woodruff (2009) “Are women more credit
constrained? Experimental evidence on gender and microenterprise returns”, American
Economic Journal: Applied Economics 1(3): 1–32.

De Mel, Suresh, David McKenzie, and Christopher Woodruff (2008) “Returns to capital in
microenterprises: Evidence from a field experiment”, The Quarterly Journal of Economics,
123(4): 1329-72.

Drexler, Alejandro, Greg Fischer and Antoinette Schoar (2014) “Keeping it simple: Financial
literacy and rules of thumb”, American Economic Journal: Applied Economics, 6(2): 1-31.

Dupas, Pascaline and Jonathan Robinson (2013) “Savings constraints and microenterprise
development: Evidence form a field experiment in Kenya”, American Economic Journal:
Applied Economics, 5(1): 163-192.

Fafchamps, Marcel and Simon Quinn (forthcoming) “Networks and manufacturing firms in
Africa: Results from a randomized field experiment”, The World Bank Economic Review.

Fafchamps, Marcel and Christopher Woodruff (forthcoming) “Identifying gazelles: Expert
panels vs. surveys as a means to identify firms with rapid growth potential”, The World Bank
Economic Review.

Fafchamps, Marcel, David McKenzie, Simon Quinn and Christopher Woodruff (2014)
“Microenterprise growth and the flypaper effect: Evidence from a randomized experiment in
Ghana”, Journal of Development Economics 106: 211-226.

Field, Erica, Rohini Pande, John Papp and Natalia Rigol (2013) “Does the classic microfinance
model discourage entrepreneurship among the poor? Experimental evidence from India”,
American Economic Review 103(6): 2196-2226.

Field, Erica, Seema Jayachandran and Rohini Pande (2010) “Do traditional institutions
constrain female entrepreneurship? A field experiment on business training in India”,
American Economic Review: Papers and Proceedings 100(2): 125-129.

Glaub, Matthias E., Michael Frese, Sebastian Fischer and Maria Hoppe (2014) “Increasing
personal initiative in small business managers or owners leads to entrepreneurial success: A
theory-based controlled randomized field intervention for evidence-based management”,
Academy of Management Learning & Education 13(3): 354-379.

Groh, Matthew and David McKenzie (2016) “Macroinsurance for microenterprises: A
randomized experiment in post-revolution Egypt”, Journal of Development Economics 118: 13-
25.

Higuchi, Yuki, Vu Hoang Nam and Tetsushi Sonobe (2015) “Sustained impacts of Kaizen
training”, Journal of Economic Behavior and Organization 120: 189-206.

Karlan, Dean, Ryan Knight and Christopher Udry (2015) “Consulting and capital experiments
with microenterprise tailors in Ghana”, Journal of Economic Behavior and Organization 118:
281-302.



                                                                                          64
Karlan, Dean and Martín Valdivia (2011) “Teaching entrepreneurship: Impact of business
training on microfinance clients and institutions”, The Review of Economics and Statistics
93(2): 510-527.

Karlan, Dean and Jonathan Zinman (2011) “Microcredit in theory and practice: Using
randomized credit scoring for impact evaluation”, Science 332(6035): 1278-1284.

Mano, Yukichi, Alhassan Iddrisu, Yukuta Yoshino and Tetsushi Sonobe (2012) “How can micro
and small enterprises in Sub-Saharan Africa become more productive? The impacts of
experimental basic managerial training”, World Development 40(3): 458-468.

Martínez, Claudia, Esteban Puentes and Jaime Ruiz-Tagle (forthcoming) “The effects of micro
entrepreneurship programs on labor market performance: Experimental evidence from
Chile”, American Economic Journal: Applied Economics.

McKenzie, David (2017) “Identifying and Spurring High-Growth Entrepreneurship:
Experimental Evidence from a Business Plan Competition”, American Economic Review 107(8):
2278-2307.

McKenzie, David, Nabila Assaf and Ana Paula Cusolito (2017) “The additionality impact of a
matching grant programme for small firms: Experimental evidence from Yemen”, Journal of
Development Effectiveness 9(1): 1-14.

McKenzie, David and Christopher Woodruff (2008) “Experimental Evidence on Returns to
Capital and Access to Finance in Mexico”, The World Bank Economic Review 22(3): 457-482.

Premand, Patrick, Stefanie Brodmann, Rita Almeida, Rebekka Grun and Mahdi Barouni (2016)
“Entrepreneurship education and entry into self-employment among university graduates”,
World Development 77: 311-327.

Valdivia, Martín (2015) “Business training plus for female entrepreneurship? Short and
medium-term experimental evidence from Peru”, Journal of Development Economics 113: 33-
51.




                                                                                        65
Table A.2. Firm death rates over different time horizons, pooled sample and individual surveys




                                                                                                                                                                                  SLKFEMBUSTRAINING
 Years since baseline




                                                                                                     KENYAGETAHEAD
                                                          EGYPTMACROINS




                                                                                                                                                                                                         SLKINFORMALITY
                                                                                                                        MALAWIFORM
                           Pooled sample




                                              BENINFORM




                                                                          GHANAFLYP




                                                                                                                                                                    NGYOUWIN
                                                                                                                                                     NGLSMSISA




                                                                                                                                                                                                                                                                             UGWINGS
                                                                                                                                                                                                                                                   TOGOINF
                                                                                                                                        MXFLS




                                                                                                                                                                                                                                                                 TTHAI
                                                                                                                                                                                                                             SLLSE




                                                                                                                                                                                                                                        SLMS
                                                                                         IFLS
                            4%                                 1%                                                                                                                 10%                                                   4%
0.25
                          [4%,5%]                            [1%,1%]                                                                                                           [10%,10%]                                              [4%,8%]
                            9%                        8%       3%                                                                                   12%*                                                                                2%
0.5
                         [8%,18%]                   [8%,8%] [3%,5%]                                                                              [10%,27%]*                                                                          [2%,10%]
                            4%                                 5%                                                                                                                                                                       4%        3%
0.75
                         [3%,10%]                            [5%,8%]                                                                                                                                                                 [3%,14%]   [2%,8%]
                            13%               8%               5%                                    16%       12%                                                  35%           16%                                        5%         1%                      23%*
1
                        [13%,18%]          [7%,15%]         [5%,15%]                              [16%,17%] [11%,18%]                                            [32%,39%]     [16%,18%]                                  [5%,11%]   [1%,14%]                [21%,28%]*
                            5%                                 8%                                                                                                                                                                       3%         2%
1.25
                         [4%,14%]                           [8%,13%]                                                                                                                                                                 [3%,16%]   [2%,14%]
                            20%                                                                                                                     21%                                                                      7%         3%                                   34%
1.5
                        [17%,34%]                                                                                                                [16%,40%]                                                                [7%,17%]   [3%,17%]                             [32%,39%]
                            17%                                                                                         23%                                         40%           25%                    8%                             4%         3%
1.75
                        [15%,28%]                                                                                    [21%,30%]                                   [28%,59%]     [25%,25%]              [7%,20%]                       [3%,20%]   [2%,13%]
                            20%               9%                                                                                                    25%*                                                           13%                  4%                      35%*
2
                        [17%,33%]          [8%,19%]                                                                                              [17%,49%]*                                                     [11%,24%]            [3%,20%]                [32%,40%]*
                            23%                                                                      24%                                             28%                                                 10%       20%                  5%
2.5
                        [19%,36%]                                                                 [23%,29%]                                       [17%,56%]                                            [9%,23%] [19%,21%]            [4%,23%]
                            30%                                                                                         21%       69%                                                                    12%       23%                  4%         8%       44%*
3
                        [26%,38%]                                                                                    [18%,31%] [54%,76%]                                                              [11%,17%] [23%,25%]            [3%,23%]   [8%,12%] [39%,50%]*
                            25%                                                                                         22%                                         40%                                            23%
3.5
                        [22%,35%]                                                                                    [19%,31%]                                   [29%,56%]                                      [22%,26%]
                            36%                                                                                                                                                                                    26%                                          53%*
4
                        [32%,43%]                                                                                                                                                                               [25%,30%]                                    [45%,59%]*
                            51%                                                                                                         67%                                                                        27%
4.5
                        [44%,58%]                                                                                                    [54%,73%]                                                                  [26%,31%]
                            41%                                                                                                                                                                                                        21%                      59%*
5
                        [35%,49%]                                                                                                                                                                                                   [17%,32%]                [50%,65%]*
                            27%                                                                                                                                                                                              29%       25%
5.5
                        [25%,34%]                                                                                                                                                                                         [27%,32%] [22%,35%]
                            50%                                                                                                                                                   35%                                                                           66%*
6
                        [43%,57%]                                                                                                                                              [35%,37%]                                                                     [55%,72%]*
                            58%                                                          53%                                                                                                                                                                    74%*
7
                        [42%,69%]                                                     [37%,67%]                                                                                                                                                              [61%,78%]*
                            65%                                                          56%                                            78%                                                                                                                     79%*
8
                        [47%,74%]                                                     [38%,71%]                                      [64%,82%]                                                                                                               [66%,83%]*




                                                                                                                                                                                                                                                                                       66
                                                                                                                                                            SLKFEMBUSTRAINING
 Years since baseline




                                                                                                KENYAGETAHEAD
                                                        EGYPTMACROINS




                                                                                                                                                                                SLKINFORMALITY
                                                                                                                MALAWIFORM
                            Pooled sample




                                            BENINFORM




                                                                        GHANAFLYP




                                                                                                                                                 NGYOUWIN
                                                                                                                                     NGLSMSISA




                                                                                                                                                                                                                                            UGWINGS
                                                                                                                                                                                                                     TOGOINF
                                                                                                                             MXFLS




                                                                                                                                                                                                                                   TTHAI
                                                                                                                                                                                                 SLLSE




                                                                                                                                                                                                            SLMS
                                                                                       IFLS
                            83%                                                                                                                                                                                                   83%*
9
                         [67%,86%]                                                                                                                                                                                             [68%,86%]*
                            66%                                                                                                                                                                             39%                   87%*
10
                         [58%,70%]                                                                                                                                                                       [37%,43%]             [73%,89%]*
                            65%                                                                                                                                                                             39%                   90%*
11
                         [57%,70%]                                                                                                                                                                       [36%,43%]             [75%,92%]*
                            90%                                                                                                                                                                                                   90%*
12
                         [74%,91%]                                                                                                                                                                                             [75%,92%]*
                            91%                                                                                                                                                                                                   92%*
13
                         [75%,93%]                                                                                                                                                                                             [75%,93%]*
                            92%                                                                                                                                                                                                   93%*
14
                         [76%,93%]                                                                                                                                                                                             [76%,94%]*
                            77%                                                        76%                                                                                                                                        94%*
15
                         [53%,84%]                                                  [52%,84%]                                                                                                                                  [78%,95%]*
                            96%                                                                                                                                                                                                   97%*
16
                         [80%,96%]                                                                                                                                                                                             [80%,97%]*
                           100%                                                                                                                                                                                                   100%
17
                        [80%,100%]                                                                                                                                                                                             [80%,100%]
Notes:
Death rate bounds accounting for missing information on survival/death, including attrition, are given in parentheses.
* indicates that multiple rounds are available for this combination of survey and time horizon, so that death rates, lower and upper bounds shown are averages of all their
respective round-specific values available for this particular horizon.
Observed drops in death rates for a given survey are possible, since some firm closure is temporary (in the sample of unique firms, we observe 6.22 percent as closed at one
point in time and open in a later round), and as a result of attrition (1.1 percent of unique firms in our sample are either closed or open in t and t+2, and not observed in t+1).
For the MXFLS, the baselines for the horizons over which we observe firm death are different. We use the 2002-round for the 3-year and 8-year horizon and the 2005-round for
the 4.5 year horizon.




                                                                                                                                                                                                                                                      67
Table A.3. Estimates of Firm Death Rate in Existing Literature
                                                                                     Sample            Death    Annualized   Attrition
 Study                        Country                        Type of firm              size   Period    rate       Rate        rate      Note
 Cabal (1995)                Dominican      micro and small enterprises                n.r.    n.r.     n.r.     22%-29%       n.r.        e.
                              Republic
 Daniels (1995)              Zimbabwe       micro and small enterprises                n.r.    n.r.     n.r.      11.5%        42%         e.
 Davies and Kerr (2015)        Ghana        All Manufacturing                         1000      10      21%       2.1%         22%
                                                                                              years
                                            0-9 worker manuf.                         386       10     24.7%      2.5%        28.5%
                                                                                              years
 Fajnzylber et al. (2006)     Mexico        self-employed individuals (own-account    n.r.       2      n.r.      30.1%        n.r.
                                            workers or employers)                             years
 Fisscha (1994)               Jamaica       micro and small enterprises               n.r.     n.r.      n.r.     4.1%         n.r.        f.
 Frazer (2005)                 Ghana        selected manufacturing                    200        2      6.3%      3.2%         n.r.        a.
                                                                                              years
 Haddad et al. (1996)        Morocco        Manufacturing firms                       n.r.    1 year   4.6%-    4.6%-7.1%      n.r.
                                                                                                       7.1%
 Kilby (1994)                 Nigeria       micro and small enterprises                n.r.    n.r.     n.r.      1.3%         n.r.        f.
 Klapper and Richmond       Cote d'Ivoire   Formal firms                              5941     20       n.r.      17%          n.r.        b.
 (2011)                                                                                       years
 Krafft (2016)                 Egypt        Household non-farm                        828       8      51.4%      6.4%         n.r.
                                                                                              years
                                            Household non-farm                        1584      6      60.9%      10.2%        n.r.
                                                                                              years
 McCaig and Pavcnik           Vietnam       Household enterprises                     n.r.      2      30-35%    15-17%        n.r.
 (2016)                                                                                       years
 Mead and Liedholm           Botswana       micro and small enterprises               n.r.     n.r.     n.r.      6.0%         n.r.
 (1998)
                               Kenya        micro and small enterprises               n.r.     n.r.     n.r.      15.9%        n.r.
                               Malawi       micro and small enterprises               n.r.     n.r.     n.r.      15.0%        n.r.
                             Swaziland      micro and small enterprises               n.r.     n.r.     n.r.      10.5%        n.r.
                             Zimbabwe       micro and small enterprises               n.r.     n.r.     n.r.      7.0%         n.r.



                                                                                                                                          68
Nagler and Naudé              Uganda         Household non-farm                                   1849      1 year    30.8%     30.8%      n.r.
(2017)
                                                                                                  1829      1 year    26.2%     26.2%      n.r.
                                                                                                  1707         4      30.5%     7.6%       n.r.
                                                                                                            years
                               Nigeria       Household non-farm                                   4448        0.5     13.2%     26.3%      n.r.
                                                                                                            years
                                                                                                  5367        0.5      1.9%     3.8%       n.r.
                                                                                                            years
                                                                                                  4516        1.5      4.5%     3.0%       n.r.
                                                                                                            years
Roberts (1996)                Colombia       Manufacturing firms                                   n.r.     1 year    8.3%-   8.3%-12.9%   n.r.
                                                                                                                      12.9%
Sandefur (2010)                Ghana         formal and informal manuf. firms with less            n.r.      n.r.      n.r.     5.1%       n.r.
                                             than 10 employees
Shiferwaw (2009)              Ethiopia       Manufacturing with 10+ workers                     471-686     1 year     16%      16%        n.r.   c.
Söderbom et al. (2006)         Ghana         Manufacturing firms                                  154          5       19%      3.8%       n.r.
                                                                                                            years
                               Kenya         Manufacturing firms                                  133          5       39%      7.8%       n.r.
                                                                                                            years
                              Tanzania       Manufacturing firms                                  102          5       44%      8.8%       n.r.
                                                                                                            years
Tybout (1996)                   Chile        Manufacturing firms                                   n.r.     1 year    7.1%-   7.1%-18.2%   n.r.
                                                                                                                      18.2%
Vijverberg and                Vietnam         Household enterprises                                2484         5      61%      12.2%      9.6%   d.
Haughton (2004)                                                                                               years
Mean Annualized Rate                                                                                                            11.5%
Median Annualized
Rate                                                                                                                            9.5%
Notes:
n.r. denotes not reported.
Annualized rate is expressed relative to initial stock, and is death rate divided by period over which death is measured.
Sample size reported for experimental studies is number in the control group.


                                                                                                                                                  69
a. Author doesn't report attrition, but reports sample of 800 with only 479 observations on survival status.
b. Death is measured from filings at register of companies, so authors cannot distinguish death from becoming informal/not reporting.
c. Death is measured from manufacturing census, cannot distinguish between dropping below size threshold from death.
d. Attrition rate does not include firms in households that were re-interviewed, but where authors were not able to match the firm over time.
e. Cited in Mead and Liedholm (1998)
f. Cited in Liedholm and Mead (1995). Death rates are based on tracer surveys which are subject to severe selectivity biases that potentially understate these
   death rates.




                                                                                                                                                             70
Table A.4. Overview of the follow-up periods covered, by survey




                                                                                                                                   SLKFEMBUSTRAINING
 Years since baseline




                                                                       KENYAGETAHEAD
                                    EGYPTMACROINS




                                                                                                                                                       SLKINFORMALITY
                                                                                       MALAWIFORM
                        BENINFORM


                                                    GHANAFLYP




                                                                                                                        NGYOUWIN
                                                                                                            NGLSMSISA




                                                                                                                                                                                                         UGWINGS
                                                                                                                                                                                       TOGOINF
                                                                                                    MXFLS




                                                                                                                                                                                                 TTHAI
                                                                                                                                                                        SLLSE
                                                                                                                                                                                SLMS
                                                                IFLS
0.25                                                 x                                                                                x                                         x
0.5                                   x              x                                                       x                                                                  x
0.75                                                 x                                                                                                                          x       x
1                        x                           x                   x              x                                x            x                                 x       x                x
1.25                                                 x                                                                                                                          x       x
1.5                                                                                                          x                                                          x       x                         x
1.75                                                                                    x                                x            x                  x                      x       x
2                        x                                                                                   x                                                          x       x                x
2.5                                                                      x                                   x                                           x              x       x
3                                                                                       x           x                                                    x              x       x       x        x
3.5                                                                                     x                                x                                              x
4                                                                                                                                                                       x                        x
4.5                                                                                                 x                                                                   x
5                                                                                                                                                                               x                x
5.5                                                                                                                                                                     x       x
6                                                                                                                                     x                                                          x
7                                                               x                                                                                                                                x
8                                                               x                                   x                                                                                            x
9                                                                                                                                                                                                x
10                                                                                                                                                                              x                x
11                                                                                                                                                                              x                x
12                                                                                                                                                                                               x
13                                                                                                                                                                                               x
14                                                                                                                                                                                               x
15                                                              x                                                                                                                                x
16                                                                                                                                                                                               x
17                                                                                                                                                                                               x




                                                                                                                                                                                                                   71
Table A.5. Attrition and missing information on survival by follow-up horizon
  Years     Percent Percent Percent
  since     attrited missing attritted +
 baseline                           missing
 0.25             1,1%      0,1%         1,2%
 0.5              2,4%      7,1%         9,5%
 0.75             6,1%      0,0%         6,1%
 1                4,8%      0,0%         4,8%
 1.25             9,2%      0,7%         9,9%
 1.5              6,3%      0,4%         6,7%
 1.75            10,0%      0,6%       10,6%
 2               10,0%     12,0%       21,9%
 2.5              7,0%      9,5%       16,5%
 3                9,5%      3,6%       13,1%
 3.5             11,6%      0,0%       11,6%
 4                5,7%      0,0%         5,7%
 4.5              5,6%      0,0%         5,6%
 5               15,4%      0,0%       15,4%
 5.5              9,0%      0,0%         9,0%
 6                4,1%      0,0%         4,1%
 7               20,0%      0,0%       20,0%
 8               15,5%     13,8%       29,3%
 9               20,0%      0,0%       20,0%
 10               7,4%      0,0%         7,4%
 11               8,0%      0,0%         8,0%
 12              20,0%      0,0%       20,0%
 13              20,0%      0,0%       20,0%
 14              20,0%      0,0%       20,0%
 15              19,3%     12,4%       31,6%
 16              20,0%      0,0%       20,0%
 17              20,0%      0,0%       20,0%



Table A.6. Reason for closure by gender
 Reason for closure      Male       Female            Total
 Making a loss            49.34%      35.50%           590
 Sickness                  9.79%      15.69%           197
 Care for family           2.45%      18.07%           180
 Better wage job           8.10%        5.63%           95
 Better business           7.91%        3.14%           71
 opportunity
 Other                    22.41%      21.97%           322
 Total                       531          924        1,455




                                                                                72
Table A.7. Gender of owner by reason of closure
 Reason for closure     Male       Female         Total
 Making a loss          44.41%       55.59%           590
 Sickness               26.40%       73.60%           197
 Care for family         7.22%       92.78%           180
 Better wage job        45.26%       54.74%             95
 Better business                                        71
 opportunity            59.15%       40.85%
 Other                  36.96%       63.04%           322
 Total                  36.49%       63.51%         1,455




                                                             73
Figure A.1. Comparison of our Firm Death Rates to Those Reported in Existing Literature




Note: Black circles are death rates from existing literature in Table A.4. Open diamonds are
data points in our study. Curve shown is fitted quadratic in Figure 1. The two outliers in the
existing literature at 10 years are from Davies and Kerr (2015).


Figure A.2. Reopening rate over different horizons for firms that closed within 1 year




Notes: only surveys in which 85 or more firms died within a year and subsequent re-opening can then be
observed are shown.


                                                                                                         74
Figure A.3. Comparison of labor productivity at baseline for firms that closed vs. survivors,
by survey




                                                                                           75