Content uploaded by Melanie Koch
Author content
All content in this area was uploaded by Melanie Koch on Jan 12, 2024
Content may be subject to copyright.
Discussion
Papers
The Non-linear Impact of
Risk Tolerance on Entrepreneurial
Profit and Business Survival
Melanie Koch, Lukas Menkhoff
2067
Deutsches Institut für Wirtschaftsforschung 2024
Electronic copy available at: https://ssrn.com/abstract=4692520
Opinions expressed in this paper are those of the author(s) and do not necessarily reflect views of the institute.
IMPRESSUM
DIW Berlin, 2024
DIW Berlin
German Institute for Economic Research
Mohrenstr. 58
10117 Berlin
Tel. +49 (30) 897 89-0
Fax +49 (30) 897 89-200
https://www.diw.de
ISSN electronic edition 1619-4535
Papers can be downloaded free of charge from the DIW Berlin website:
https://www.diw.de/discussionpapers
Discussion Papers of DIW Berlin are indexed in RePEc and SSRN:
https://ideas.repec.org/s/diw/diwwpp.html
https://www.ssrn.com/link/DIW-Berlin-German-Inst-Econ-Res.html
Electronic copy available at: https://ssrn.com/abstract=4692520
The non-linear impact of risk tolerance on
entrepreneurial profit and business survival∗
Melanie Koch†Lukas Menkhoff‡
January 2024
Abstract
Entrepreneurs tend to be risk tolerant but is more risk tolerance always better? In
a sample of about 2,100 small businesses, we find an inverted U-shaped relation
between risk tolerance and profitability. This relationship holds in a simple bilat-
eral regression and also when we control for a large set of individual and business
characteristics. Apparently, one major transmission goes from risk tolerance via
investments to profits. This is quite robust as it applies for past investments as
well as planned investments. Considering business survival, we show, first, that less
profitable businesses leave the market while moderately risk tolerant entrepreneurs
survive more often. Second, the high risk-low profit part of the U-shaped relation
seems to disappear among businesses being four years and older, indicating that such
inferior risk-profit combinations disappear over time. These findings are important
for the concept of business readiness trainings as the motivation (and ability) to take
risks should potentially be accompanied by some warning that too much risk taking
can be detrimental to long-term business success.
Keywords: risk tolerance; entrepreneurs; profits; investments
JEL: D22; D81; L26; M21
∗We thank Filder Aryemo, Daniel Graeber, Jana Hamdan, Tim Kaiser, Alexander Kritikos, Kilian
Mazurek, Helke Seitz, and Yuanwei Xu for very helpful comments, and some of them also for excellent
field work. This study was partially funded by the German Research Foundation through the Research
Training Group (RTG) 1723 “Globalization and Development.” This paper is coauthored by Melanie
Koch in her personal capacity. The opinions expressed do not necessarily reflect the official viewpoint of
the Oesterreichische Nationalbank or the Eurosystem. Declarations of interest: none.
†Oesterreichische Nationalbank (OeNB), Vienna, Austria;
Email: melanie.koch@oenb.at
‡Corresponding author. Humboldt-Universität zu Berlin, German Institute for Economic
Research (DIW Berlin), and IfW Kiel, Germany;
Email: lmenkhoff@diw.de
Electronic copy available at: https://ssrn.com/abstract=4692520
1 Introduction
Entrepreneurs leading small businesses have a strong impact on how their business oper-
ates. Personal characteristics of the owner and manager shape the course of the business.
A core characteristic of interest is the individual risk tolerance of such entrepreneurs. It is
often found that entrepreneurs have a higher risk tolerance than the average adult. Run-
ning a business is inherently riskier than being employed as it comes with more volatile
returns, including the failure of operations (e.g., Caliendo et al.,2009,2014;Kerr et al.,
2019;Chanda and Unel,2021). However, personality characteristics that motivate one
to become an entrepreneur may be not necessarily the same that make one a successful
entrepreneur (Hamilton et al.,2019).
In many economic models on decision-making, profit is maximized under risk neutral
preferences but not under risk averse or risk seeking preferences. In particular, for decision
making under risk, the risk neutral agent is the one who maximizes expected returns.
Why should this be different for entrepreneurial decision making, where outcomes are
inherently risky or even uncertain? Following this reasoning, we ask whether successful
entrepreneurs are characterized by a moderate level of risk tolerance that is largely in
line with risk neutrality. While there are studies suggesting related arguments (which we
discuss below), our question about a potentially non-linear impact of risk tolerance on
profitability seems to be neglected in existing empirical work.
We conduct a survey study with about 2,100 small entrepreneurs in Western Uganda
comprising rich information about their business and personal characteristics. These
small entrepreneurs operate in trading centers, i.e., groups of several enterprises of various
kinds. All of them are retailers, service providers, or small manufacturers. Two-thirds
of the entrepreneurs work alone in their business, i.e., have no regular employees. The
others typically have between one and three employees. We measure risk tolerance using
a self-assessment of each individual’s “willingness to take financial risk,” as introduced
by Dohmen et al. (2011); a closely related measure has been experimentally validated
by them. We also measure profits via self-assessment of the entrepreneurs, a procedure
1
Electronic copy available at: https://ssrn.com/abstract=4692520
whose usefulness is demonstrated for small enterprises in low-income countries by De Mel
et al. (2009). As profits are intended to capture business success and not just business
size, we prefer to control for size which is approximated by the number of workers. Thus,
the respective main outcome measure is profit per worker.
We regress profit per worker and total profits on financial risk tolerance. Figure 1illus-
trates our main result using a linear and a non-linear (quadratic) approach, respectively.
Starting with the linear approach, the two grey regression lines trend clearly upwards,
confirming the expectation that risk taking and profits go hand in hand. In particular,
we are interested whether there is something like “too much risk tolerance,” i.e., that risk
tolerance above a certain level may no longer be helpful in realizing higher profitability.
Therefore, we also estimate the quadratic model. Indeed, the non-linear relations, plotted
as black lines, initially increase with risk tolerance before declining for high risk tolerance.
100 150 200
Fitted values, profits (per worker) in UGX
0 2 4 6 8 10
Financial risk tolerance (from low to high)
Linear fit, profits/worker Quadratic fit, profits/worker
Linear fit, profits Quadratic fit, profits
Figure 1: Regressing profits (per worker) on risk tolerance, linear and quadratic fit
Note: Fits are obtained via bivariate regressions. The turning points for the quadratic fits are at values
6.7 and 9 respectively. Multivariate linear regressions show a similar picture with turning points
between 5.5-6.5 for the quadratic fits.
2
Electronic copy available at: https://ssrn.com/abstract=4692520
While Figure 1shows the bivariate relation of interest, we further conduct a multi-
variate regression analysis in which we control for individual characteristics of the en-
trepreneurs and characteristics of their businesses. The consideration of these variables
shows that many of them are reasonably related to the profitability of businesses, so that
– for example – better educated entrepreneurs generate higher profits. Reassuringly, how-
ever, if considering them, they do not qualitatively change the significant and quadratic
relationship between risk tolerance and profits.
An issue that can arise in such a regression analysis is the potential endogeneity of
risk tolerance. Existing literature does not raise this concern often because individual
preferences, such as risk tolerance, were regarded as invariant and exogenous. This has
changed. Three influences seem to be systematic, i.e., age, past shocks, and current
emotions (Schildberg-Hörisch,2018). We do not observe emotions during the survey
interview, which increases noise in the large cross-sectional regression, but we can control
for age and past shocks, which confirms our results. Further, our results also hold when
we analyze the risk-profitability relation of those small businesses that still operate more
than one year after measuring risk tolerance.
We provide a possible channel of how risk tolerance leads to the non-linear profitability
pattern we document. Motivated by previous literature, we look at investments as poten-
tial mediator. Indeed, the volume of investments during the preceding year has a strong
influence on profitability while much reducing the influence of risk tolerance in both the
linear and the quadratic terms. Thus, investments serve as a mediator variable. Surpris-
ingly, (past and planned) investments themselves show the inverted U-shaped pattern in
relation to the risk tolerance that we have seen for profitability. This may be interpreted
as suggesting that risk averse entrepreneurs invest too little and that the most risk toler-
ant may underestimate the necessity of investments (and overestimate their success). In
contrast, moderate levels of risk tolerance go along with the highest investments and then
the highest profitability and profits.
Finally, we analyze small business survival by surveying a second time 18 to 24
3
Electronic copy available at: https://ssrn.com/abstract=4692520
months after compiling the baseline data. Surviving businesses differ from closed busi-
nesses mainly by their higher average profitability and higher investments at baseline (see
Manso,2016), in line with the consequence of conventional market forces. Interestingly,
entrepreneurs still in business are slightly more risk tolerant at baseline. We confirm these
patterns when “looking backwards,” i.e., by comparing businesses being in operation for a
different length of time at baseline. Here, we find for businesses in operation for four years
or longer that the tentatively inefficient combination of high risk tolerance and limited
profitability seems to disappear; this would also indicate the working of market forces.
Results of this study cannot be automatically generalized as with other case studies
dominating the literature. This holds, in particular, when comparing a poor rural area
with high-income countries. However, studies such as ours compare entrepreneurs with
different risk tolerance within a given institutional setting, so that for example hindrances
due to limited financial development apply in principle to all entrepreneurs operating in
rural Uganda. Thus, it may be reassuring that we find a very similar set of determinants
explaining profits or survival as has been found in related studies before.
Theoretical considerations. In microeconomic models of the firm, optimal decision-
making is typically derived under risk neutrality. This is because risk neutrality usually
maximizes expected returns (not utility); for example, under expected utility theory. Risk
neutrality translates into a real world where managers maximize firm owners’ value and
firm owners, who bear the risk of these decisions, are well diversified. However, the case
is different for entrepreneurs who hold much, if not most, of their wealth in their own
business. Here, risky entrepreneurial decisions may affect personal consumption possi-
bilities such that the degree of individual risk tolerance will impact business decisions.
With personal consumption at stake, small-business owners might rather resort to less
risky firm decisions. In this sense, risk averse decisions lead to lower investments in gen-
eral (Panousi and Papanikolaou,2012) or lower innovation-related investments (Caggese,
2012); consequently, lower growth and/or lower profitability follows.
Arguably, the concept of Knightian risk used in standard models might not be appro-
4
Electronic copy available at: https://ssrn.com/abstract=4692520
priate for real-life decision making. Under Knightian risk, objective outcome probabili-
ties are always given. In entrepreneurial decision-making, this is rather unlikely; instead,
outcome probabilities are uncertain. The point that research should focus more on the
concept of uncertainty instead of risk when thinking about entrepreneurship was already
made in Knight (1921). While acknowledging this fact, we still focus on risk in line with
the literature on which our work is based. Moreover, the self-reported measure of risk is a
compromise in this respect: even though economists differentiate between Knightian risk
and Knightian uncertainty, this is not necessarily true for the general population or small
entrepreneurs. When asking them how willing they are to take risks, they well might
have uncertainty in mind instead of risk. We nevertheless continue to use the term risk
throughout the paper.
While the positive relation between risk tolerance and business success is well estab-
lished, there is less evidence about an effect of highly risk tolerant, i.e., risk seeking,
entrepreneurs on business success. The theoretical response may be the analogy to risk
averse entrepreneurs, implying that risk seeking provides utility that compensates for
lower returns. Thus, the resulting outcomes may be due to a fully rational decision, as
modeled for some entrepreneurs in Vereshchagina and Hopenhayn (2009). This impact
of risk-seeking may be similar in its consequences to overconfidence, which has a long-
standing tradition in entrepreneurship research (e.g., Moskowitz and Vissing-Jørgensen,
2002). Overconfidence can lead to too much optimism about business outcomes, tending
to reduce business success (Koellinger et al.,2007). With respect to investment be-
havior, over- (under)confidence is shown to be related to over- (under)investment (e.g.
Malmendier and Tate,2005;Pikulina et al.,2017).
In summary, theoretical considerations about entrepreneurial decision making show
that risk neutral decision makers may realize the highest degree of business success, i.e.,
here measured via profitability. By contrast, an increasing level of risk aversion will lead to
less investments and less business success. Toward the other extreme of risk tolerance, i.e.,
risk seeking behavior, theory indicates that this will not be compensated by higher returns.
5
Electronic copy available at: https://ssrn.com/abstract=4692520
In total, these patterns suggest an inverse U-shaped relation between risk tolerance and
profitability that is mediated by investment behavior.
Literature. Our study contributes to the literature examining the role of risk toler-
ance on entrepreneurial behavior. The decision to run a business, even a very small one,
requires some degree of risk tolerance, also in low-income countries (e.g., Willebrands
et al.,2012;Sharma and Tarp,2018). While this is intensively studied, there is much
less empirical work on the potentially non-linear relation of risk tolerance to measures of
business success (Kritikos,2022). An early study by Begley and Boyd (1987) covering
147 entrepreneurs (and 92 other CEOs) finds that, among entrepreneurs, risk aversion is
related to return on assets in the form of an inverted U-shape. Caliendo et al. (2010), ana-
lyzing the fact of remaining self-employed in a large representative sample of the German
population, the German Socio-Economic Panel (SOEP), find a U-shaped relation between
risk tolerance and exit from being self-employed, i.e., an inverted U-shape for remaining
in the market. Nieß and Biemann (2014) extend and confirm the Caliendo et al. (2010)
study by considering further waves of the SOEP-data. Only the study by Kreiser et al.
(2013), analyzing a heterogeneous sample of 1,600 firms from ten countries, finds that
moderate risk aversion comes with low sales growth. Further studies do not apply an
explicit non-linear approach but conclude that a moderate degree of risk tolerance may
facilitate success; these include Hvide and Panos (2014) on Norway and Willebrands et
al. (2012) for Nigerian small entrepreneurs’ revenues.
While these studies tend to support an inverted U-shaped relation between risk tol-
erance and return on assets or survival, respectively, they do not consider profitability,
which seems to be a core criterion of success. Moreover, these earlier studies do not
always analyze a transmission mechanism through which risk tolerance may impact busi-
ness success. Finally, we analyze entrepreneurs over time and find market forces at work:
those exiting the market are primarily characterized by lower profitability and low risk
tolerance, those surviving are more similar over time in their profitability.
From a methodological perspective, revealing an impact of risk tolerance on profitabil-
6
Electronic copy available at: https://ssrn.com/abstract=4692520
ity requires controlling for individual and business characteristics (or a quasi-experiment
as in De Blasio et al.,2021). In particular, these controls include the age and education
of the entrepreneur and the age of the business (see,e.g., McKenzie and Woodruff,2017;
Bernstein et al.,2022). Further, as risk tolerance depends on systematic influences, these
should be considered. Schildberg-Hörisch (2018) states a pattern over the life-cycle (e.g.,
Dohmen et al.,2017), an impact of shocks in life (e.g., Malmendier and Nagel,2011), and
a temporary impact from all kinds of emotions (Meier,2022). Unlike most other studies,
we can control for age and shocks (albeit imperfectly). Regarding emotions, there are no
survey items available, so that we have to accept the resulting “noise,” from which we do
not expect a systematic influence on the risk-profit-relation.
Putting the case of a potentially non-linear effect of risk tolerance on profits into a
broader perspective, there are other studies discussing related issues. Puri and Robinson
(2007), for example, find that some degree of optimism is useful for firm success but
that a very high degree of optimism leads to suboptimal results. De Meza et al. (2019)
show that more optimistic small entrepreneurs earn less, where optimism is measured
before the entry decision. Malmendier and Tate (2015) review findings on the impact of
overconfidence on managers’ behavior and outcomes; as the word overconfidence indicates,
its impact is tentatively negative, while confidence is definitely important for running a
firm. This distinction between the starting and running phase of new businesses in also
found in the meta-analysis of Kraft et al. (2022). Interestingly, there may be a few
instances, such as implementing innovations, where some degree of overconfidence can
even be helpful. Overall, optimism, confidence, and risk tolerance of entrepreneurs or top
managers facilitate business success but there can be too much of a good thing.
The paper is structured in five sections. Section 2describes the setting of the study, the
data, and the methodology. In Section 3, we present the main results and discuss potential
transmission channels and business survival. Section 4provides robustness checks while
Section 5concludes.
7
Electronic copy available at: https://ssrn.com/abstract=4692520
2 Setting, data, and methods
2.1 Setting of our study
The data of this research stem from a study in Western Uganda, where a financial edu-
cation treatment was tested. For our study, we mostly use the baseline data that were
collected in Spring 2019, i.e., before the treatment. When we conduct regressions on busi-
ness survival using the second survey wave, which was carried out between October 2020
and April 2021, we control (as robustness check) for the financial education intervention
that took place in Summer 2019 for the treatment group, but not for the control group.
The study conducted a survey of small businesses in all trading centers in the rural
Bunyangabu and Kabarole districts of Western Uganda. These districts are typical for
many rural areas in low-income countries, as they are characterized by a high share
of agricultural activities and a lack of functioning infrastructure. Small businesses in
this area cluster in villages where people live (or pass through via traffic) to provide a
minimal customer base. The accumulation of a few or many businesses forms a trading
center. These small businesses cover three sectors, i.e., retailing, manufacturing, and
services. Data collection considers all businesses from the small and medium sized trading
centers. However, it under-samples the seven large centers with more than 100 shops,
where every third shop is visited randomly due to resource constraints. Overall, the
survey team identified about 5,500 small businesses in 108 trading centers, of which 2,223
were interviewed (for a map see Appendix Figure A1). The remaining businesses were
either not open when the team visited or were left out in the big centers by design.
2.2 Data
The data we collect inform about the entrepreneurs, i.e., the business owners, their small
firms, and the last big adverse shock that these entrepreneurs faced. Our key individual
characteristic of interest is financial risk tolerance, as measured by responses to a standard
scale of willingness to take risk related to investing and borrowing, ranging between 0-
8
Electronic copy available at: https://ssrn.com/abstract=4692520
completely unwilling and 10-fully prepared (see Dohmen et al.,2011). This measure
has been widely applied in the literature, not just because it can be easily implemented
in surveys but also because it has been validated by experimentally derived measures of
risk tolerance. Moreover, its power in predicting field behavior is rather better than worse
compared to experimental measures, at least in developing countries (Menkhoff and Sakha,
2017). For robustness purposes, we also consider the analogous response to willingness
to take risk in general. Further characteristics that may influence entrepreneurship or
risk taking are gender (women operate businesses less often), age (younger individuals are
more risk tolerant), education (better educated are more risk tolerant), financial literacy
as a specific qualification to run a firm, the individual work experience (as an aspect of
human capital), and the ability to borrow a larger amount indicating access to credit
and/or the ability to bear risk (more details are provided in Appendix Table A1).
Regarding adverse idiosyncratic shocks, we use information about the estimated costs
of the last unexpected emergency. Taking the amount into the regressions ensures that
only severe shocks are considered. In Section 4 on robustness, we further consider the
shock amount relative to the consumption level in order to gain a sense about its relative
impact and other variables that are related to shocks.
The key firm characteristic for our research is profits of the small business, the number
of workers who are regularly working at the business, and the resulting profits per worker.
The respective survey item is an outright question about profits in the last four weeks,
which is shown to be reliable and easy to compile vis-à-vis alternatives (see De Mel et al.,
2009). Further variables of interest include business age, as older firms are less risky (here
approximated by the years the entrepreneur is working in the specific business), and the
sector of business (distinguishing between retail, service, and manufacturing). Finally, we
consider investments by the respective small business. Here, the information is about the
investment amount during the last 12 months prior to the survey.
We note that there are missing values in the following variables: age (0.4% of all
observations), profits (1.9%), investments (1.4%), and months the shop was open (0.1%
9
Electronic copy available at: https://ssrn.com/abstract=4692520
of all observations; this variable is used in the robustness analysis). Considering only
complete cases results in an estimation sample of 2,144 observations. The 79 incomplete
cases differ from the remaining estimation sample in some characteristics but, overall,
differences do not seem to be economically crucial (see Appendix Table A2). Moreover,
information between work experience and the age of the business is inconsistent in 53
out of the original 2,223 observations: respondents said that they have worked longer in
that recent shop than they reported to have worked in any shop, considering their whole
experience. However, the differences are mostly smaller than one year and might be due
to mistakes in memory. In these cases, we set the work experience equal to the age of the
business.
The descriptive statistics in Table 1show that the sample population is 64% female,
about 34 years old, the median has primary education, and, on average, has about seven
years of experience working in a shop.
Table 1: Descriptive statistics
Mean Median Std. Dev. Min. Max.
Profits last month (in 1000 UGX) 162.28 100.00 218.80 0.00 1500.00
No. of workers 1.43 1.00 0.70 1.00 12.00
Profits per worker (in 1000 UGX) 125.19 70.00 176.90 0.00 1500.00
Financial risk tolerance 4.94 5.00 2.55 0.00 10.00
Female 0.64 1.00 0.48 0.00 1.00
Age (in years) 33.69 30.00 11.42 16.00 82.00
Education level (0-5) 1.46 1.00 1.37 0.00 5.00
Financial literacy (std.) 0.01 0.22 1.00 -2.21 2.05
Work experience (in months) 82.51 55.50 90.91 0.00 660.00
Able to borrow 100,000 UGX (share) 0.86 1.00 0.35 0.00 1.00
Financial shock cost (in 1000 UGX) 134.98 5.00 386.21 0.00 3000.00
Age of business (in months) 54.38 36.00 69.27 0.00 612.00
Sector: retail (share) 0.69 1.00 0.46 0.00 1.00
Sector: services (share) 0.26 0.00 0.44 0.00 1.00
Sector: manufacturing (share) 0.04 0.00 0.21 0.00 1.00
Investments last 12 months (in 1000 UGX) 1422.75 500.00 2776.70 0.00 20000.00
Financial risk tolerance is measured on a scale from zero to ten, education level is measured from 0 (no
education or incomplete primary) to 5 (university). All variables measured in UGX are winsorized at the
value 0 and at 99%. Financial literacy is a standardized score based on seven knowledge questions.
10
Electronic copy available at: https://ssrn.com/abstract=4692520
About 86% are able to borrow 100,000 UGX (ca. 27$), which is about half of the
median monthly, equivalized household consumption. Most businesses operate in the retail
sector (69%), with 26% in the service sector and only a small fraction in manufacturing.
On average, the shops are less than five years old, with the median shop existing for only
three years, so far. The median amount invested in the last 12 months is five times the
median profit yielded in the last four weeks, indicating that these are primarily gross
investments, such as shop inventories.
As we are particularly interested in risk tolerance and profitability, we show the dis-
tribution for these variables in Figure 2. The left panel shows a conventional histogram
of the self-reported willingness to take financial risk on a scale between 0 and 10 with a
peak at and around 5, i.e., the middle categories, and some smaller peaks at the extremes
of the distribution. This measure of risk tolerance needs to be transformed into the three
common areas of risk preferences, i.e., averse, neutral, and seeking. Dohmen et al. (2011)
show the distributions of the self-reported scale and an incentivized experimental risk
measure for the identical population in Germany. Using the experimental measure, they
find that about 78% are risk averse, 13% risk neutral, and 9% risk seeking, but they do
not show the scale responses of those who are risk-neutral in the experiment. They refer
to an almost identical distribution for the U.S. Applying this to our case, the respective
answering categories reproducing the same distribution would be quite exactly 1 to 6 for
risk averse, 7/8 for risk neutral, and 9/10 for risk seeking. However, as poorer countries
have higher levels of risk tolerance (l’Haridon and Vieider,2019), we prefer to enlarge the
range of risk-seeking to category 8, implying that categories 8 to 10 indicate risk seeking
behavior, and 6/7 indicate risk neutrality. Then, it follows from this procedure that about
two-thirds of our population are regarded to be risk averse, 20% are risk neutral, and 15%
risk seeking.
The right panel of Figure 2provides the Kernel density estimate of profits per worker
and overall profits. We winsorize all variables measured in UGX at the value zero from
below and at the 99 percentiles from above to reduce the impact of outliers. The maximum
11
Electronic copy available at: https://ssrn.com/abstract=4692520
values for both absolute profits and profits per worker are then 1,500,000 UGX, while the
mean and median profit per worker is clearly smaller than the mean and median absolute
profit, respectively.
0 .05 .1 .15 .2 .25
Fraction
0 2 4 6 8 10
Financial risk tolerance
0 .002 .004 .006 .008
Density
0 500 1000 1500
Amount (in 1000 UGX)
Profits per worker
Profits
Figure 2: Histogram for the willingness to take financial risk from 0 to 10 (left) and
kernel density estimate for (winsorized) profits per worker and absolute profits (right)
2.3 Empirical approach
We relate small entrepreneurs’ risk taking and profitability to each other. Following
the main literature, risk taking can be assumed to be exogenous under three conditions
(Schildberg-Hörisch,2018): one should control for age, shocks, and emotions. We include
the first two control variables, while emotions at the time of the interviews remain un-
observed and, thus, contribute to noise in the relation. While we cannot exclude that
emotions may create a bias on our relations of interest, or that other unobserved charac-
teristics may impact these relations, we note that our data provide a rich set of controls.
12
Electronic copy available at: https://ssrn.com/abstract=4692520
Thus, we interpret the relation – in line with the literature – as a tentatively causal impact
of risk taking on profitability (keeping in mind the limitations of our identification). In
this respect, in the robustness section, we also show that risk taking has qualitatively the
same effect on the profitability of surviving businesses. Regarding a proper identification
of the profitability of the small businesses, we use controls for characteristics of the en-
trepreneur, the business, and the industry. In sum, for the multivariate analysis in the
next section, we use standard Ordinary Least Squares (OLS) regressions. The regressions
take the following form:
Y=α+β0R+β1RSQ +
P
X
j=2
βjXj+(1)
where Yis either absolute profits or profits per worker, Ris financial risk tolerance, RSQ
is financial risk tolerance squared, and Xjare the control variables. Standard errors are
clustered at the trading center level (108 clusters).
To address the heavily skewed values of profits and the other variables measured in
UGX, we winsorize these as previously noted. Winsorizing at 99% already reduces the
skewness tremendously. We use the winsorized variables throughout Section 3. In gen-
eral, we do not apply log-transformation because, on top of some entrepreneurs reporting
negative profits (5 observations), some reported zero profits (90 observations). These
observations cannot be kept after the log-transformation without assumptions. Further-
more, we do not opt for selection models as the share of zeros is still overall small (less
than 5%). Alternatively, in the robustness section, we apply the inverse hyperbolic sine
transformation to all variables measured in UGX, which strongly accounts for skewness.
13
Electronic copy available at: https://ssrn.com/abstract=4692520
3 Results
3.1 Main results
While we show in the introduction the bivariate relation between financial risk tolerance
and profits per worker, we now estimate this relation using the control variables as in-
troduced in Section 2.2. Results in Table 2, columns (1) to (3) show effects on profits
per worker and columns (4) to (6) on absolute profits. Column (1) shows the non-linear
relation between risk tolerance and profits per worker as already depicted in Figure 1;
profits increase up to a moderate level of risk tolerance of about 6.7; i.e., a level being
close to risk neutrality. Subsequently there is indeed a turning point and risk tolerance is
thereafter related to decreasing profits; this non-linearity can also be seen when capturing
risk tolerance by dummy variables, i.e. not prescribing a functional form, as we show in
the robustness section below. While, on average, the jump from risk level 0 to 1 increases
profits per worker by around 12,000 UGX (i.e., about 0.05 standard deviation units), the
jump from level 5 to level 6 increases profits by only about 1,500 UGX. The jump from
level 6 to 7 then decreases profits per worker by 600 UGX.
In column (2), we add the full set of control variables, except investments¸ which
are expected to be a mediator variable, and those intended for robustness checks. The
coefficients of the two risk variables remain almost unchanged and significant, even though
many of the control variables are typically related to risk tolerance. The coefficients
of the other variables have the expected signs and are mostly statistically significant:
entrepreneurs have higher profits if they are male, younger (but not too young), and
better educated. However, the age effect is partially compensated by work experience
and – more importantly – by the age of business (which is highly correlated with work
experience; see Azoulay et al.,2020). Older businesses tend to be more profitable, while
– in this sample – firms in manufacturing and the service sector are less profitable than
those in the retail sector. Profits and ability to borrow are also positively related.
The main relations explaining profits per worker also hold in explaining absolute profits
14
Electronic copy available at: https://ssrn.com/abstract=4692520
as shown in columns (4) and (5). The results in columns (3) and (6) are discussed next.
Table 2: Profits per worker and profits – financial risk tolerance (scale from 0 to 10)
Profits per worker Profits
(1) (2) (3) (4) (5) (6)
Bivariate Controls Mediator Bivariate Controls Mediator
Financial risk tolerance 13.99*** 10.91*** 5.15 13.74** 9.85* 1.07
(3.72) (3.48) (3.62) (5.42) (5.31) (5.35)
Financial risk squared −1.04** −1.00** −0.48 −0.76 −0.76 0.03
(0.42) (0.39) (0.40) (0.59) (0.55) (0.57)
Female −51.91*** −44.01*** −79.37*** −67.97***
(8.07) (7.10) (9.96) (8.84)
Age (in years) 2.94* 1.01 3.79* 1.01
(1.63) (1.63) (1.97) (1.85)
Age squared −0.04** −0.02 −0.06*** −0.03
(0.02) (0.02) (0.02) (0.02)
Education level (0-5)=1 13.15** 5.14 14.88* 2.90
(6.29) (6.31) (7.71) (7.81)
Education level (0-5)=2 39.08*** 27.36** 42.91*** 26.58**
(11.22) (10.98) (11.70) (11.02)
Education level (0-5)=3 76.29*** 66.01*** 82.40*** 68.05***
(24.57) (22.94) (26.77) (24.69)
Education level (0-5)=4 61.80*** 39.23** 100.24*** 67.82***
(18.35) (17.77) (22.95) (21.17)
Education level (0-5)=5 148.50** 113.86** 193.88*** 143.66**
(58.44) (53.14) (64.41) (56.05)
Financial literacy (std.) 10.80*** 7.46* 14.08*** 8.93**
(4.10) (3.97) (4.54) (4.39)
Work experience (in months) 0.08 0.05 0.15** 0.12
(0.06) (0.06) (0.07) (0.08)
Able to borrow 100,000 UGX (share) 19.37** 13.36 25.34** 16.34
(8.31) (8.23) (11.20) (10.93)
Financial shock cost (in 1000 UGX) 0.00 0.00 0.01 0.01
(0.01) (0.01) (0.02) (0.02)
Age of business (in months) 0.23** 0.26*** 0.30*** 0.34***
(0.09) (0.10) (0.11) (0.11)
Sector: services −21.38*** −11.35* −26.66*** −10.87
(6.26) (5.89) (8.24) (7.17)
Sector: manufacturing −41.24** −31.59** −51.33*** −37.76**
(16.27) (14.39) (18.72) (15.89)
Investments next 12 months (in 1000 UGX) 0.02*** 0.03***
(0.00) (0.00)
Constant 88.10*** 33.67 66.23** 117.91*** 61.31 108.15***
(7.41) (33.30) (32.88) (12.01) (41.78) (39.85)
Adj. R-Squared 0.004 0.086 0.137 0.005 0.113 0.187
Observations 2144 2144 2133 2144 2144 2133
Dependent variables: profits is winsorized at the value zero and at 99%, profits per worker is profits divided by the number of persons
working regularly in the shop. Independent variables: education level is measured from 0 (no education or incomplete primary) to
5 (university) and financial literacy is a standardized score based on seven knowledge questions. All variables measured in UGX are
winsorized at the value 0 and at 99%. Omitted category for sector is retail. Standard errors clustered at trading center level (108
clusters).
* p <0.10, ** p <0.05, *** p <0.01
15
Electronic copy available at: https://ssrn.com/abstract=4692520
3.2 On the transmission channel
While a non-linear impact of risk tolerance on business profits seems plausible, here we
examine our expected main channel by which the willingness to take risk is transformed
into business success. As previously noted, risk aversion might lead to underinvestment.
Because investments are risky, more risk tolerance will motivate greater investment and
this should support business development. However, if risk tolerance increases further,
there could be too much investment, similar to the case of overconfidence and overin-
vestment. Thus, we add investment to the set of explanatory variables of profits and
hypothesize that investments serve as mediator between risk tolerance and profitability.
The consideration of this additional variable is expected to increase explanatory power
overall but to diminish the explanatory power of risk tolerance as risk tolerance is a main
driver of investing. We emphasize that the consideration of investments should not be
seen as a conventional control variable because investments are (partially) determined by
risk tolerance and, thus, an outcome such as profits. Instead, their consideration pro-
vides a simple form of mediation analysis (see Baron and Kenny,1986). Surprisingly,
investments show the same non-linear relation to risk tolerance as that shown in Figure
1 above (see Appendix Figure A2). Thus, risk seeking does not seem to foster too much
investment but again too few.
Results for the outcome variable profits per worker, where investments are added, are
provided in column (3) in Table 2and confirm expectations. The adjusted R-squared of
the regression increases by about 90% and the coefficients of risk tolerance become smaller
by more than a third, turning them marginally significant and insignificant, respectively.
When explaining absolute profits in column (6), both risk variables also become much
smaller and turn insignificant.
Further major changes in coefficients refer to the smaller coefficients of services and
manufacturing sectors, indicating that the reference sector, i.e., retail business, gets many
investments. This may seem somewhat surprising; it mainly refers to the stock of goods,
as larger and broader supply makes the business more attractive and reduces cases of
16
Electronic copy available at: https://ssrn.com/abstract=4692520
non-availability of goods. Another major change is the decreasing coefficient of financial
literacy. This indicates that the more financially literate also invest more and, thus,
operate more profitably. It seems reasonable that the capabilities coming with financial
literacy are related to more investments, while causality may go in both directions. Finally,
the coefficient on ability to borrow decreases by almost a third and significance gets lost,
indicating that borrowing ability supports investments.
Overall, investments seem to be determined by risk tolerance in a non-linear way and
including them in the regression of profits on risk tolerance shrinks the coefficients on risk.
Thus, investments serve as mediator for the final impact of risk tolerance on profitability.
3.3 Results looking forward
So far, we have analyzed the cross-section of small businesses. The non-linear impact
of risk tolerance raises the question whether this has an influence on the survival of
businesses. Is there evidence that those with lower profits will leave the market with
higher probability?
We extend information beyond the cross-section used in Section 3.1 and make use
of a follow-up survey wave. This wave contains, however, much fewer survey items, as
it was mainly conducted via telephone due to the COVID-19 pandemic. In particular,
the measures of risk tolerance and profits have not been surveyed again. Still, we know
whether the same small businesses remained in operation.1This pandemic caused the
most severe economic crisis in Uganda since 1985, as GDP fell by 1.5% while trend growth
is in the range between 4% to 6% p.a. (with population growth above 3% p.a.). As the
second wave was conducted after the pandemic had already hit economic activities, many
businesses were closed permanently; at the same time, the pandemic and the following
economic downturn limited the ability to open up new businesses.
From the considered sample of 2,144 entrepreneurs at baseline, 1,904 participated in
1Around 82% of the people who were interviewed in the follow-up were contacted by phone in fall 2020.
The other 18% were visited in spring 2021. Average profits at baseline do not differ significantly for
business survivors interviewed in 2020 and survivors interviewed in 2021. The analogue holds true for
those whose businesses did not survive.
17
Electronic copy available at: https://ssrn.com/abstract=4692520
the endline survey. Of these, 169 respondents report that they no longer operate their
old business.2For 171 more entrepreneurs, who did not participate in the endline survey,
we know that they have relocated. In addition to the 169, we also treat their businesses
as closed.3This gives us a total of 2,075 “observed” businesses at endline, 340 of which
were closed, i.e., about 16% of the total still observed (McCaig and Pavcnik,2021, report
in Vietnam 14-18% exit p.a.). Table 3describes the three samples (with characteristics
at baseline), i.e., (i) all 2,075 entrepreneurs and their businesses still observed at endline,
and we distinguish these using the follow-up between (ii) business still open at follow up
and (iii) business closed at follow up. There are many significant differences at baseline
between open and closed businesses as Table 3shows, indicating that the decision to close
is determined by systematic factors.
Most important among these are profitability and investments: still open businesses
had 37% higher profits per worker than closed ones and 47% higher investments. The
other, mostly significant, differences also point in the expected direction but not as
strongly: “surviving” entrepreneurs are more often male, older, have higher financial lit-
eracy, have longer work experience, and their businesses are older. Overall, qualified
entrepreneurs and solid businesses survive more often, according to our data. For our
purpose, the variable on risk tolerance is most interesting. This is somewhat higher for
the operating vs closing entrepreneurs but the mean difference is not significant.
Thus, we plot the relation of interest between risk tolerance and profitability. Is it
different for entrepreneurs of surviving vs closed businesses? Figure 3shows results at
baseline for the two groups (and for both groups together, for comparison). The main
difference is, as expected from Table 3, that operating businesses show higher profitability
than closed ones, suggesting that market forces drive out the less successful businesses.
Moreover, there remains the quadratic relationship between risk and profits for survivors.
2This sample divides into two groups, 86 persons no longer run a business and 83 opened a new business.
3In total, we could verify the whereabouts of 2,177 entrepreneurs. For the other 46, any kind of
information is missing. For 202 out of the 2,177, there is no further survey information as these
persons either relocated, were impaired, imprisoned, or deceased. We exclude all these 202 except
those who relocated for obvious reasons.
18
Electronic copy available at: https://ssrn.com/abstract=4692520
Table 3: Descriptive statistics for business survival
Observed endline Still open Closed Difference (3)-(2)
(1) (2) (3) (4)
Profits per worker (in 1000 UGX) 125.42 131.14 95.62 -35.52∗∗∗
(177.7) (184.2) (135.0)
Financial risk tolerance 4.95 4.98 4.78 -0.21
(2.550) (2.545) (2.576)
Female 0.64 0.63 0.69 0.06∗∗
(0.480) (0.483) (0.461)
Age (in years) 33.68 34.21 30.91 -3.29∗∗∗
(11.34) (11.43) (10.49)
Education level (0-5) 1.46 1.48 1.40 -0.08
(1.371) (1.373) (1.362)
Financial literacy (std.) 0.01 0.04 -0.11 -0.15∗∗
(0.994) (0.994) (0.984)
Work experience (in months) 82.57 84.47 72.65 -11.82∗∗
(90.87) (89.81) (95.69)
Able to borrow 100,000 UGX (share) 0.86 0.87 0.82 -0.06∗∗
(0.343) (0.334) (0.387)
Financial shock cost (in 1000 UGX) 135.24 143.60 91.70 -51.89∗∗∗
(389.2) (404.2) (295.4)
Age of business (in months) 54.77 57.64 39.77 -17.87∗∗∗
(69.48) (71.21) (57.42)
Sector: retail (share) 0.69 0.69 0.68 -0.01
(0.463) (0.462) (0.467)
Sector: services (share) 0.26 0.26 0.29 0.03
(0.441) (0.439) (0.455)
Sector: manufacturing (share) 0.05 0.05 0.03 -0.02∗
(0.208) (0.214) (0.171)
Investments last 12 months (in 1000 UGX) 1433.60 1511.36 1028.27 -483.09∗∗∗
(2795.5) (2932.7) (1880.6)
Share of borrowers 0.14 0.14 0.13 -0.00
(0.343) (0.344) (0.339)
Amount of Loan (in 1000 UGX) 57.87 62.36 34.46 -27.90∗
(502.9) (544.4) (160.9)
Observations 2075 1741 334 2075
* p <0.10, ** p <0.05, *** p <0.01; p-values are obtained using twosided t-tests.
Profits per worker is winsorized profits divided by the number of persons working regularly in the shop.
Observed endline are all entrepreneurs interviewed at baseline for which we have business information at
endline, still open are those who are still in business 15 months later and closed are those who do not own a
shop anymore, who opened another shop or who relocated. Financial risk tolerance is measured on a scale
from zero to ten. Education level is measured from 0 (no education or incomplete primary) to 5 (university).
All variables measured in UGX are winsorized at the value 0 and at 99%. Financial literacy is a standardized
score based on seven knowledge questions. Standard deviations in parentheses.
19
Electronic copy available at: https://ssrn.com/abstract=4692520
Interestingly, those who eventually closed their businesses have lower profitability at
low and moderate levels of risk tolerance, but not at high levels. The latter may indicate
that highly risk tolerant entrepreneurs left their businesses more vulnerable to the heavy
crisis of 2020 so that even relatively profitable businesses had to close.
60 80 100 120 140
Fitted values, profits per worker
0 2 4 6 8 10
Financial risk tolerance
Business still open Business closed
All observed at endline
Figure 3: Regressing profits per worker on risk tolerance, survivors vs non-survivors
Results so far indicate that risk tolerance is important for business survival, and this
relation seems to be non-linear. We test this by running a linear probability regression with
our standard set of control variables to explain survival (a Logit model approach provides
qualitatively the same results). As the quadratic non-linear relation may be not that
clearly given in the data, we additionally use three levels of risk tolerance as introduced
in Section 2.2, where the survey responses 6 and 7 are regarded as medium risk tolerant
(or: risk neutral). Results in Table 4, columns (1) and (2) show that the quadratic relation
between risk tolerance and survival is indeed difficult to see, but columns (3) and (4) are
helpful in this respect as they show that a medium risk tolerance goes along with highest
probability to survive, confirming Caliendo et al. (2010).
20
Electronic copy available at: https://ssrn.com/abstract=4692520
Table 4: Business survival, regressions
Fin. risk tolerance Fin. risk tolerance groups
(1) (2) (3) (4)
No mediator Mediator No mediator Mediator
Financial risk tolerance 0.00 0.00
(0.01) (0.01)
Financial risk squared 0.00 0.00
(0.00) (0.00)
Fin. tolerance, level 6-7 0.06*** 0.06***
(0.02) (0.02)
Fin. tolerance, level 8-10 0.01 0.01
(0.03) (0.03)
Female −0.03** −0.03* −0.03** −0.03**
(0.02) (0.02) (0.02) (0.02)
Age (in years) 0.02*** 0.02*** 0.02*** 0.02***
(0.00) (0.00) (0.00) (0.00)
Age squared −0.00*** −0.00*** −0.00*** −0.00***
(0.00) (0.00) (0.00) (0.00)
Education level (0-5)=1 0.02 0.02 0.02 0.02
(0.02) (0.02) (0.02) (0.02)
Education level (0-5)=2 0.04** 0.04* 0.04** 0.04*
(0.02) (0.02) (0.02) (0.02)
Education level (0-5)=3 0.03 0.03 0.03 0.02
(0.04) (0.04) (0.04) (0.04)
Education level (0-5)=4 0.06** 0.05* 0.06** 0.06*
(0.03) (0.03) (0.03) (0.03)
Education level (0-5)=5 −0.02 −0.02 −0.01 −0.02
(0.05) (0.05) (0.05) (0.05)
Financial literacy (std.) 0.01 0.01 0.01 0.01
(0.01) (0.01) (0.01) (0.01)
Work experience (in months) −0.00*** −0.00*** −0.00*** −0.00***
(0.00) (0.00) (0.00) (0.00)
Able to borrow 100,000 UGX (share) 0.04* 0.04* 0.04* 0.04*
(0.02) (0.02) (0.02) (0.02)
Financial shock cost (in 1000 UGX) 0.00** 0.00** 0.00** 0.00**
(0.00) (0.00) (0.00) (0.00)
Age of business (in months) 0.00*** 0.00*** 0.00*** 0.00***
(0.00) (0.00) (0.00) (0.00)
Sector: services 0.00 0.00 0.00 0.01
(0.02) (0.02) (0.02) (0.02)
Sector: manufacturing 0.04 0.05 0.04 0.05*
(0.03) (0.03) (0.03) (0.03)
Investments last 12 months (in 1000 UGX) 0.00** 0.00**
(0.00) (0.00)
Constant 0.43*** 0.43*** 0.43*** 0.43***
(0.08) (0.08) (0.08) (0.08)
Adj. R-Squared 0.029 0.030 0.032 0.033
Observations 2075 2075 2075 2075
Linear probability model. Dependent variable: Still open is a dummy variable that equals 1, if the
business was still open at the follow-up interview and zero, if otherwise. Independent variables: profits
per worker is winsorized profits divided by the number of persons working regularly in the shop, education
level is measured from 0 (no education or incomplete primary) to 5 (university) and financial literacy is
a standardized score based on seven knowledge questions. All variables measured in UGX are winsorized
at the value 0 and at 99%. Omitted category for sector is retail. Standard errors clustered at trading
center level (108 clusters).
* p <0.10, ** p <0.05, *** p <0.01 21
Electronic copy available at: https://ssrn.com/abstract=4692520
3.4 Results looking backward
We complement our analysis by also looking backwards at businesses that have existed
for several years, analyzing the role of risk tolerance for them. Due to the kind of data we
use, this analysis is limited because of “survivorship bias.” Different from the preceding
Section 3.3, where we observe surviving and failing businesses, here we only see survivors.
This implies a positive selection of entrepreneurs and their businesses. That means these
businesses perform probably better than those which were closed, and which we cannot
observe anymore. So what we can see is whether there is a difference between those which
survived “long-term” vs those surviving rather “short-term.” However, all the information
we have about entrepreneurs and their businesses are from the last year only, so that we
hesitate to draw conclusions about the development of businesses over time.
As a proxy for business age, we rely on information regarding how many months the
entrepreneur has been running their current business. This proxy underestimates true
age, for example when current entrepreneurs take over a business from within the family.
Still, we take this information and see that about half of the sample already existed three
years ago (probably the true share is larger). Basically, we expect the same three relations
that we know from above: (i) older businesses, i.e., those operating for at least three years,
are more profitable; (ii) these entrepreneurs may be somewhat more risk tolerant; and
(iii) the inferior combination of high risk tolerance and low profitability may tentatively
disappear over time.
Table 5replicates Table 3, but divides the sample by business age. Results are similar
to those in Section 3.3: older businesses are characterized by higher profitability, as the
profits per worker are, on average, 148,000 UGX for older businesses and 103,000 UGX
for younger ones; further, they are characterized by higher investments. Entrepreneurs of
older businesses are also older and more experienced, but not better educated. Regard-
ing risk tolerance, there is not a strong difference: the mean value of risk tolerance for
entrepreneurs operating older businesses is slightly higher at 4.98 relative to 4.90 for the
other entrepreneurs.
22
Electronic copy available at: https://ssrn.com/abstract=4692520
Table 5: Descriptive statistics for business age
At least 3 years Younger than 3 years Diff.
(1) (2) (3)
Profits per worker (in 1000 UGX) 147.59 102.62 −44.97∗∗∗
Profits last month (in 1000 UGX) 194.42 129.89 −64.53∗∗∗
Financial risk tolerance 4.98 4.90 −0.09
Female 0.60 0.68 0.08∗∗∗
Age (in years) 37.15 30.20 −6.94∗∗∗
Education level (0-5) 1.37 1.56 0.20∗∗∗
Financial literacy (std.) 0.04 −0.03 −0.07
Work experience (in months) 124.70 40.01 −84.70∗∗∗
Able to borrow 100,000 UGX (share) 0.87 0.85 −0.03∗
Financial shock cost (in 1000 UGX) 160.10 109.68 −50.42∗∗∗
Age of business (in months) 94.92 13.55 −81.37∗∗∗
Sector: retail (share) 0.71 0.68 −0.03
Sector: services (share) 0.24 0.29 0.05∗∗∗
Sector: manufacturing (share) 0.06 0.03 −0.02∗∗
Investments last 12 months (in 1000 UGX) 1639.14 1204.75 −434.39∗∗∗
Observations 1076 1068 2144
* p <0.10, ** p <0.05, *** p <0.01; p-values are obtained using twosided t-tests.
Profits per worker is winsorized profits divided by the number of persons working regularly in the shop.
At least 3 years are all businesses that exist for at least three years, younger than 3 years are all businesses
that are younger than that. Financial risk tolerance is measured on a scale from zero to ten. Education
level is measured from 0 (no education or incomplete primary) to 5 (university). All variables measured
in UGX are winsorized at the value 0 and at 99%. Financial literacy is a standardized score based on
seven knowledge questions.
Finally, we analyze the (ex post) relation between risk tolerance and profitability in
more detail, by splitting the age groups into finer intervals. The results presented in Figure
4tend to confirm expectations about the effect of market forces: First, the lines for older
businesses lie quite consistently above those for younger businesses, reflecting their higher
profitability; an exception are the two bottom lines, i.e., profit levels of businesses less than
a year old and those one to two years old, which are almost identical. Second, regarding
the curvature of these lines, the lines for the oldest businesses tend to be somewhat less
curved at the risk tolerant end. This applies to businesses that are four to six years old
and is very obvious for those older than six years as there is no longer any curvature. This
seems to indicate another impact of market forces: in the longer run the profitability of
23
Electronic copy available at: https://ssrn.com/abstract=4692520
surviving small businesses operated by highly risk tolerant entrepreneurs becomes similar
to the level of business operated by moderately risk tolerant entrepreneurs.
Thus, this analysis indicates two consequences of market dynamics, plausibly at work
in the low-income rural area: first, market forces drive out businesses with lower profitabil-
ity and, second, market forces drive out businesses which are characterized by very high
risk tolerance of their entrepreneurs and profitability being below that of entrepreneurs
with moderate risk tolerance.
50 100 150 200
Fitted values, profits per worker
0 2 4 6 8 10
Financial risk tolerance
Less than a year 1−2 years
2−4 years 4−6 years
More than 6 years
Figure 4: Regressing profits per worker on risk tolerance separately for different ages
24
Electronic copy available at: https://ssrn.com/abstract=4692520
4 Robustness
We provide a set of robustness checks that tentatively confirm the above findings and
which go into eight directions: (i) including risk tolerance in three or four categories in-
stead of using the quadratic term; (ii) using a measure for general risk tolerance; (iii) using
other information to consider a possible influence from financial shocks; (iv) taking the
skewness of variables more into account by using hyperbolic sine transformation instead
of winsorizing; (v) testing main results for surviving entrepreneurs; (vi) considering the
financial education treatment as an additional variable to predict business survival; (vii)
also considering information about borrowing to see whether an exit is possibly related to
overborrowing; (viii) testing how risk tolerance and profits are related to planned instead
of past investments.
(i) Applying a quadratic functional relation between risk tolerance and profits is to
some extent arbitrary, so that we also group the survey responses on risk tolerance into
categories, thus avoiding any functional relation. In line with the other experimental work
(as argued in Section 2.2) and as proposed by referees, we apply three categories. These
cover entrepreneurs with little risk tolerance (survey responses 0 to 5), entrepreneurs with
medium (6 and 7) and high risk tolerance (8 to 10). Results in Table 6show what we have
seen in earlier figures and regressions: profits are highest for the group of entrepreneurs
with medium risk tolerance and they are lower for those with little or high risk tolerance.
This also holds when we split those with little risk tolerance in two groups or if we shift
the borders of groups by one category (see Appendix Tables A3 and A4).
(ii) In the main text, we use the risk measure that asks for the stance toward financial
risk. We think that this is an appropriate risk measure in the realm of entrepreneurship.
However, one may argue that financial aspects are a bit narrow – even in the context of
running a small business – and that a general risk measure provides a broader measure of
risk tolerance (see Dohmen et al.,2011). Thus, we replace the financial risk measure in
the main specifications of Table 2with the general risk measure. Results in Table 7show
that this does not make a qualitative difference.
25
Electronic copy available at: https://ssrn.com/abstract=4692520
Table 6: Profits per worker and profits – financial risk tolerance in three groups
Profits per worker Profits
(1) (2) (3) (4) (5) (6)
Fin. tolerance, level 6-7 12.17 5.31 4.81 20.08* 10.95 10.27
(9.18) (8.27) (7.50) (11.07) (10.31) (8.98)
Fin. tolerance, level 8-10 4.60 −9.37 −8.07 24.76 3.81 5.57
(10.73) (9.84) (10.04) (15.07) (13.03) (13.58)
Female −52.24*** −41.76*** −79.82*** −65.62***
(8.10) (6.76) (9.95) (8.16)
Age (in years) 3.23** 1.97 4.09** 2.39
(1.61) (1.31) (1.97) (1.58)
Age squared −0.05** −0.03** −0.06*** −0.05***
(0.02) (0.02) (0.02) (0.02)
Education level (0-5)=1 13.39** 5.38 15.00* 4.15
(6.28) (6.31) (7.71) (8.04)
Education level (0-5)=2 39.25*** 27.85** 42.94*** 27.50**
(11.20) (11.65) (11.69) (12.08)
Education level (0-5)=3 76.85*** 68.79*** 82.68*** 71.78***
(24.37) (20.96) (26.59) (22.04)
Education level (0-5)=4 61.97*** 43.23** 100.17*** 74.78***
(18.39) (19.02) (22.90) (22.60)
Education level (0-5)=5 148.22** 123.40** 193.48*** 159.86**
(58.33) (58.42) (64.51) (64.11)
Financial literacy (std.) 11.27*** 5.44 14.56*** 6.65
(4.03) (3.50) (4.48) (4.15)
Work experience (in months) 0.07 0.05 0.14* 0.11
(0.06) (0.06) (0.07) (0.08)
Able to borrow 100,000 UGX (share) 21.38** 15.01* 27.23** 18.60*
(8.31) (8.28) (11.05) (10.90)
Financial shock cost (in 1000 UGX) 0.00 −0.00 0.01 −0.00
(0.01) (0.01) (0.02) (0.01)
Age of business (in months) 0.23** 0.23*** 0.30*** 0.29***
(0.10) (0.08) (0.11) (0.10)
Sector: services −21.30*** −6.68 −26.61*** −6.81
(6.29) (6.48) (8.32) (8.40)
Sector: manufacturing −41.49** −24.62* −51.33*** −28.48*
(16.48) (14.75) (18.90) (16.90)
Investments last 12 months (in 1000 UGX) 0.02*** 0.03***
(0.00) (0.00)
Constant 121.87*** 50.35 54.93* 154.15*** 77.09* 83.30**
(6.08) (33.80) (28.03) (7.14) (41.17) (34.65)
Adj. R-Squared -0.000 0.085 0.168 0.001 0.112 0.212
Observations 2144 2144 2144 2144 2144 2144
Dependent variables: profits is winsorized at the value zero and at 99%, profits per worker is profits divided by the number of persons
working regularly in the shop. Independent variables: education level is measured from 0 (no education or incomplete primary) to
5 (university) and financial literacy is a standardized score based on seven knowledge questions. All variables measured in UGX are
winsorized at the value 0 and at 99%. Omitted category for sector is retail. Standard errors clustered at trading center level (108
clusters).
* p <0.10, ** p <0.05, *** p <0.01
26
Electronic copy available at: https://ssrn.com/abstract=4692520
Table 7: Profits per worker and profits – general risk tolerance (scale from 0 to 10)
Profits per worker Profits
(1) (2) (3) (4) (5) (6)
General risk tolerance 12.81*** 10.06*** 6.73* 13.84*** 10.64** 6.12
(3.49) (3.27) (3.45) (4.87) (4.72) (4.93)
General risk squared −0.84** −0.83** −0.47 −0.65 −0.70 −0.22
(0.38) (0.36) (0.38) (0.53) (0.51) (0.53)
Female −51.74*** −41.20*** −79.03*** −64.71***
(8.01) (6.68) (9.88) (8.09)
Age (in years) 3.10* 1.91 3.96** 2.34
(1.60) (1.31) (1.95) (1.57)
Age squared −0.04** −0.03** −0.06*** −0.05***
(0.02) (0.02) (0.02) (0.02)
Education level (0-5)=1 12.40* 4.60 13.67* 3.08
(6.27) (6.24) (7.64) (7.93)
Education level (0-5)=2 38.14*** 26.96** 41.41*** 26.22**
(11.25) (11.64) (11.77) (12.09)
Education level (0-5)=3 75.96*** 68.19*** 81.27*** 70.71***
(24.54) (21.12) (26.78) (22.19)
Education level (0-5)=4 59.90*** 41.23** 97.48*** 72.11***
(18.27) (18.89) (22.89) (22.46)
Education level (0-5)=5 148.05** 122.57** 193.76*** 159.13**
(58.41) (58.82) (64.31) (64.24)
Financial literacy (std.) 10.91*** 5.18 14.14*** 6.36
(4.07) (3.54) (4.53) (4.20)
Work experience (in months) 0.08 0.05 0.15** 0.11
(0.06) (0.06) (0.07) (0.08)
Able to borrow 100,000 UGX (share) 17.59** 11.80 22.92** 15.06
(8.15) (8.18) (11.10) (11.07)
Financial shock cost (in 1000 UGX) 0.00 −0.00 0.01 −0.00
(0.01) (0.01) (0.02) (0.01)
Age of business (in months) 0.22** 0.22*** 0.29*** 0.29***
(0.09) (0.08) (0.11) (0.10)
Sector: services −20.62*** −6.35 −25.69*** −6.30
(6.33) (6.52) (8.32) (8.43)
Sector: manufacturing −39.27** −22.40 −48.80** −25.88
(16.40) (14.69) (18.94) (17.07)
Investments last 12 months (in 1000 UGX) 0.02*** 0.03***
(0.00) (0.00)
Constant 87.54*** 31.64 39.62 112.64*** 54.63 65.48*
(7.76) (32.87) (27.55) (11.53) (41.48) (35.59)
Adj. R-Squared 0.006 0.087 0.170 0.008 0.114 0.215
Observations 2144 2144 2144 2144 2144 2144
Dependent variables: profits is winsorized at the value zero and at 99%, profits per worker is profits divided by the number of persons
working regularly in the shop. Independent variables: education level is measured from 0 (no education or incomplete primary) to
5 (university) and financial literacy is a standardized score based on seven knowledge questions. All variables measured in UGX are
winsorized at the value 0 and at 99%. Omitted category for sector is retail. Standard errors clustered at trading center level (108
clusters).
* p <0.10, ** p <0.05, *** p <0.01
27
Electronic copy available at: https://ssrn.com/abstract=4692520
(iii) As risk tolerance depends on the experience of recent shocks, we consider such
shocks by the estimated financial costs of the shock, as shown in Table 2. Now we replace
this variable with three alternatives. These are, first, the costs of the financial shock
relative to monthly consumption expenses; second, the number of months the business
was open in the last 12 months (for businesses older than 12 months); and, third, the
self-assessment whether the last four weeks before the survey brought low or high profits
relative to the average of the last year, measured on a scale from 0 to 10. Results are
shown in Table 8. We do not find that shock amount relative to monthly consumption
expenses is significantly related to profitability. For the other two shock alternatives,
effects are as expected: “months the shop was open” is positively related to profits and
the assessment that “last month was a good month” as well. The U-shaped relationship
between risk tolerance and profitability as well as the mediating effect of investments
remain qualitatively unchanged.
(iv) We address the fact of a few extreme values and skewness in some nominal vari-
ables, e.g., profits, by winsorizing them at the 99% level. This is a standard procedure
with data where values may be distorted because of incorrect entrances, misleading mem-
ories of participants, misunderstandings, or just extreme cases (outliers). However, it does
not reduce skewness tremendously. An alternative procedure that takes better account of
this latter issue and can transform zero values is the hyperbolic sine transformation. Ap-
plying the transformation to all variables measured in UGX, we again replicate Table 2.
In Table 9, we report marginal effects calculated as proposed by Norton (2022). The table
shows that the main results are robust to the transformations. However, the mediating
effect of investments seems a bit muted.
(v) We test if our main results hold if only those businesses that were still open at the
follow-up are considered. This considers the fact that the data including the risk measure
stem from the baseline survey while the survivorship information is more recent, further
increasing the credibility of an exogenous measure of risk tolerance. Table 10 shows that
results also hold for this group as all coefficients are very similar to those in Table 2.
28
Electronic copy available at: https://ssrn.com/abstract=4692520
Table 8: Additional controls for shocks
Shock/consump. Shop open Bad-Good month
(1) (2) (3) (4) (5) (6)
Financial risk tolerance 10.89*** 7.04* 10.92** 6.55 7.15** 3.86
(3.48) (3.60) (4.84) (4.89) (3.54) (3.61)
Financial risk squared −0.99** −0.64 −0.88 −0.52 −0.65 −0.35
(0.39) (0.40) (0.54) (0.55) (0.40) (0.41)
Female −52.02*** −41.57*** −53.99*** −43.85*** −46.78*** −37.39***
(8.07) (6.75) (9.71) (8.46) (7.57) (6.41)
Age (in years) 2.98* 1.79 2.20 0.98 3.60** 2.43*
(1.64) (1.33) (2.04) (1.68) (1.67) (1.36)
Age squared −0.04** −0.03** −0.03 −0.02 −0.05** −0.04**
(0.02) (0.02) (0.02) (0.02) (0.02) (0.02)
Education level (0-5)=1 13.06** 5.22 14.02* 5.11 12.14* 4.69
(6.27) (6.29) (8.31) (7.98) (6.24) (6.13)
Education level (0-5)=2 39.17*** 27.74** 38.24*** 28.03* 38.00*** 27.30**
(11.25) (11.68) (14.08) (14.60) (11.17) (11.60)
Education level (0-5)=3 76.34*** 68.56*** 59.24** 52.46** 73.30*** 66.28***
(24.48) (21.05) (28.00) (24.04) (23.46) (20.26)
Education level (0-5)=4 61.88*** 43.10** 67.63*** 50.02** 61.49*** 43.56**
(18.38) (19.03) (22.90) (23.34) (18.23) (18.72)
Education level (0-5)=5 148.74** 123.13** 183.59** 161.78** 140.55** 117.28**
(58.70) (58.86) (75.93) (75.75) (56.61) (57.27)
Financial literacy (std.) 10.73** 5.13 10.67** 4.23 7.20* 2.19
(4.10) (3.59) (5.30) (4.68) (4.03) (3.48)
Work experience (in months) 0.08 0.05 −0.01 −0.03 0.03 0.01
(0.06) (0.06) (0.07) (0.07) (0.06) (0.06)
Able to borrow 100,000 UGX (share) 19.44** 13.66 25.15** 19.45* 12.28 7.59
(8.32) (8.27) (10.67) (10.56) (8.02) (8.11)
Financial shock cost (in 1000 UGX) −0.00 −0.01 −0.00 −0.01
(0.01) (0.01) (0.01) (0.01)
Age of business (in months) 0.23** 0.23*** 0.23** 0.25*** 0.26*** 0.25***
(0.09) (0.08) (0.10) (0.09) (0.10) (0.08)
Sector: services −21.65*** −6.84 −31.17*** −15.06* −19.27*** −5.56
(6.27) (6.45) (7.85) (8.19) (6.34) (6.60)
Sector: manufacturing −41.61** −24.37* −54.47*** −35.74** −39.23** −23.28
(16.32) (14.63) (17.83) (16.04) (16.11) (14.89)
Investments last 12 months (in 1000 UGX) 0.02*** 0.02*** 0.02***
(0.00) (0.00) (0.00)
Shock cost rel. to mth. consumption −0.19 −0.63
(2.06) (2.18)
Shop open prev. year (in months) 4.17*** 3.66***
(1.44) (1.33)
Bad-good month from 1-10 14.42*** 12.85***
(1.83) (1.71)
Constant 33.41 44.27 10.68 27.94 −37.35 −19.74
(33.32) (27.86) (41.73) (37.13) (35.97) (29.82)
Adj. R-Squared 0.086 0.169 0.083 0.170 0.118 0.194
Observations 2144 2144 1560 1560 2144 2144
Dependent variables: Profits per worker is winsorized profits divided by the number of persons working regularly in the shop. Inde-
pendent variables: education level is measured from 0 (no education or incomplete primary) to 5 (university) and financial literacy is
a standardized score based on seven knowledge questions. All variables measured in UGX are winsorized at the value 0 and at 99%.
Omitted category for sector is retail. Standard errors clustered at trading center level (108 clusters).
* p <0.10, ** p <0.05, *** p <0.01
29
Electronic copy available at: https://ssrn.com/abstract=4692520
Table 9: Inverse hyperbolic transformation instead of winsorizing
Profits per worker Profits
(1) (2) (3) (4) (5) (6)
Financial risk tolerance 38627.16*** 41407.60*** 37364.58*** 79534.50*** 80313.39*** 75677.44***
(12659.41) (14198.34) (14116.18) (21626.59) (28828.33) (29089.32)
Financial risk squared −4141.34*** −4585.43*** −4154.39*** −6477.18*** −7291.78*** −6733.37***
(1133.50) (1258.41) (1245.10) (1830.42) (2458.93) (2458.99)
Female 54600.13** 54753.76** −165589.04***−173887.62***
(24299.06) (24326.73) (31412.38) (32691.43)
Age (in years) −4684.15 −5089.23 666.51 −93.72
(5171.72) (5093.32) (6641.47) (6862.47)
Age squared 67.84 72.97 −59.74 −52.71
(63.94) (62.27) (85.54) (87.45)
Education level (0-5)=1 28060.72 25604.13 64252.42** 62766.00*
(23546.20) (23924.00) (31974.00) (33802.20)
Education level (0-5)=2 81896.54** 78184.93** 132043.09*** 131166.51***
(39276.19) (38814.76) (38693.16) (38606.72)
Education level (0-5)=3 421.14 −5105.99 46194.93 36868.63
(40445.15) (39722.14) (69560.14) (71209.74)
Education level (0-5)=4 722.48 −6678.36 86330.75* 72720.17
(32245.29) (31819.41) (50959.41) (52548.63)
Education level (0-5)=5 207723.11 181370.44 237931.29* 208520.65*
(195003.98) (182005.89) (126736.17) (120491.10)
Financial literacy (std.) 25379.20** 22713.01* 79409.91*** 77675.22***
(12444.19) (12070.74) (18986.67) (19505.35)
Work experience (in months) −161.02 −173.05 331.99 325.01
(221.15) (223.34) (286.66) (302.19)
Able to borrow 100,000 UGX (share) 51438.86 43120.52 190392.88*** 182362.12***
(38920.95) (38682.39) (51304.11) (53843.77)
Financial shock cost (in 1000 UGX) −61.50* −67.41** −82.07 −97.76
(34.17) (33.89) (80.82) (84.09)
Age of business (in months) 164.80 215.72 548.66 678.81*
(269.09) (267.74) (368.92) (386.69)
Sector: services 6763.15 7214.18 −7361.28 −6817.04
(33195.84) (33652.05) (40819.26) (43306.08)
Sector: manufacturing −61490.62* −56128.99 −107760.26** −100690.73**
(35556.28) (37239.14) (44797.62) (50048.65)
Investments last 12 months, ihs 8608.82*** 17635.58***
(2892.28) (4872.60)
Observations 2144 2144 2144 2144 2144 2144
Marginal effects reported. Dependent variables: profits is transformed using the inverse hyperbolic sine transformation, profits
per worker is ihs-transformed profits divided by the number of persons working regularly in the shop. Independent variables:
education level is measured from 0 (no education or incomplete primary) to 5 (university) and financial literacy is a standardized
score based on seven knowledge questions. All variables measured in UGX are transformed using the inverse hyperbolic sine
transformation. Omitted category for sector is retail. Standard errors clustered at trading center level (108 clusters).
* p <0.10, ** p <0.05, *** p <0.01
30
Electronic copy available at: https://ssrn.com/abstract=4692520
Table 10: Main results only for business survivors
Profits per worker Profits
(1) (2) (3) (4) (5) (6)
Financial risk tolerance 14.82*** 11.78*** 8.33** 15.56** 11.83** 7.30
(4.44) (4.11) (4.18) (6.05) (5.77) (5.63)
Financial risk squared −1.21** −1.19** −0.85* −1.02 −1.06* −0.61
(0.49) (0.46) (0.47) (0.64) (0.61) (0.60)
Female −56.10*** −45.05*** −87.41*** −72.86***
(8.51) (7.35) (10.74) (9.13)
Age (in years) 2.56 1.61 3.49 2.24
(2.00) (1.59) (2.32) (1.78)
Age squared −0.04 −0.03 −0.06** −0.05**
(0.02) (0.02) (0.03) (0.02)
Education level (0-5)=1 14.22** 5.26 15.75* 3.95
(7.11) (7.48) (8.54) (9.31)
Education level (0-5)=2 37.81*** 25.71* 42.96*** 27.04*
(12.77) (13.36) (13.92) (14.18)
Education level (0-5)=3 83.38*** 76.48*** 92.08** 83.00***
(31.13) (28.11) (35.31) (31.14)
Education level (0-5)=4 59.20*** 40.12* 96.12*** 71.03***
(20.84) (21.61) (24.12) (23.80)
Education level (0-5)=5 133.72* 105.16 179.14** 141.57*
(69.57) (69.19) (76.01) (74.67)
Financial literacy (std.) 11.19** 4.62 12.74** 4.10
(5.15) (4.62) (5.81) (5.46)
Work experience (in months) 0.06 0.00 0.09 0.02
(0.07) (0.07) (0.09) (0.09)
Able to borrow 100,000 UGX (share) 21.47** 14.18 30.42** 20.83
(10.11) (9.77) (13.12) (12.82)
Financial shock cost (in 1000 UGX) 0.00 −0.00 0.00 −0.01
(0.01) (0.01) (0.02) (0.01)
Age of business (in months) 0.26** 0.27*** 0.37*** 0.39***
(0.11) (0.09) (0.13) (0.12)
Sector: services −23.15*** −7.71 −31.86*** −11.56
(7.46) (7.84) (8.87) (9.31)
Sector: manufacturing −42.33** −23.37 −51.12** −26.18
(18.53) (16.62) (20.74) (18.52)
Investments last 12 months (in 1000 UGX) 0.02*** 0.03***
(0.00) (0.00)
Constant 95.06*** 49.11 55.09 123.64*** 74.16 82.04**
(9.61) (41.52) (34.39) (14.05) (48.18) (40.11)
Adj. R-Squared 0.003 0.080 0.169 0.004 0.112 0.219
Observations 1741 1741 1741 1741 1741 1741
Dependent variables: profits is winsorized at the value zero and at 99%, profits per worker is profits divided by the number of persons
working regularly in the shop. Independent variables: education level is measured from 0 (no education or incomplete primary) to
5 (university) and financial literacy is a standardized score based on seven knowledge questions. All variables measured in UGX are
winsorized at the value 0 and at 99%. Omitted category for sector is retail. Standard errors clustered at trading center level (108
clusters).
* p <0.10, ** p <0.05, *** p <0.01
31
Electronic copy available at: https://ssrn.com/abstract=4692520
(vi) Here, we show the results on business survival while controlling for the financial
education treatment. In the first three columns of Table 11, risk tolerance is measured
continuously, in the following three columns it is measured in three groups. For both
measures we show the survival determinants if we do not control for the treatment at all
(columns 1 and 4), if we control for being invited to the treatment (intention to treat effect,
ITT, in columns 2 and 4), and if we control for actually participating in the treatment
(treated effect, TOT, in columns 3 and 6). The number of observations goes down in the
later columns because not all individuals were invited in some treated trading centers and,
additionally, some individuals did not respond to the invitation. We find that moderate
risk tolerance (levels 6 to 7) seems to support business survival. Moreover, the financial
education treatment does not impact any of the relations of interest, but interestingly has
a significantly positive impact on survival.
(vii) We add information about the share of borrowers in each group (survivors vs.
non-survivors) alongside the average amount of loans to former Table 3, and show the
result in Table 12. This result may be unexpected from the viewpoint of overborrowing
concerns. Closed shops are characterized by the same share of borrowers but by much
smaller loan amounts, indicating rather a shortage of loans than overborrowing.
(viii) We investigate how stable the relationship between risk tolerance and investments
is by also looking at planned investments. In the survey, entrepreneurs were asked how
much they plan to invest in their business in the next 12 months. The correlation between
past and future investments is very high and significant. Reassuringly, Figure A3 in
the appendix shows that planned investments have the inverted u-shape relation to risk
tolerance, as past investments used above.
32
Electronic copy available at: https://ssrn.com/abstract=4692520
Table 11: Business survival, controlling for financial education training
Fin. risk tolerance Fin. risk tolerance groups
(1) (2) (3) (4) (5) (6)
No ITT TOT No ITT TOT
Financial risk tolerance 0.00 0.01 0.01
(0.01) (0.01) (0.01)
Financial risk squared 0.00 −0.00 −0.00
(0.00) (0.00) (0.00)
Fin. tolerance, level 6-7 0.06*** 0.06*** 0.06**
(0.02) (0.02) (0.02)
Fin. tolerance, level 8-10 0.01 0.02 0.02
(0.03) (0.03) (0.03)
Female −0.03* −0.02 −0.01 −0.03** −0.02 −0.01
(0.02) (0.02) (0.02) (0.02) (0.02) (0.02)
Age (in years) 0.02*** 0.01*** 0.02*** 0.02*** 0.01*** 0.02***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Age squared −0.00*** −0.00*** −0.00*** −0.00*** −0.00*** −0.00***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Education level (0-5)=1 0.02 0.02 0.04 0.02 0.02 0.04
(0.02) (0.02) (0.02) (0.02) (0.02) (0.02)
Education level (0-5)=2 0.04* 0.04* 0.06** 0.04* 0.04* 0.06**
(0.02) (0.02) (0.03) (0.02) (0.02) (0.03)
Education level (0-5)=3 0.03 0.03 0.05 0.02 0.03 0.05
(0.04) (0.04) (0.05) (0.04) (0.04) (0.05)
Education level (0-5)=4 0.05* 0.07** 0.08** 0.06* 0.07** 0.08**
(0.03) (0.03) (0.04) (0.03) (0.03) (0.04)
Education level (0-5)=5 −0.02 −0.02 0.03 −0.02 −0.01 0.03
(0.05) (0.06) (0.06) (0.05) (0.06) (0.06)
Financial literacy (std.) 0.01 0.01 0.01 0.01 0.01 0.01
(0.01) (0.01) (0.01) (0.01) (0.01) (0.01)
Work experience (in months) −0.00*** −0.00*** −0.00*** −0.00*** −0.00*** −0.00***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Able to borrow 100,000 UGX (share) 0.04* 0.03 0.01 0.04* 0.03 0.01
(0.02) (0.03) (0.03) (0.02) (0.03) (0.03)
Financial shock cost (in 1000 UGX) 0.00** 0.00*** 0.00** 0.00** 0.00** 0.00**
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Age of business (in months) 0.00*** 0.00*** 0.00*** 0.00*** 0.00*** 0.00***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Sector: services 0.00 0.00 −0.01 0.01 0.00 −0.00
(0.02) (0.02) (0.03) (0.02) (0.02) (0.03)
Sector: manufacturing 0.05 0.06* 0.03 0.05* 0.06* 0.03
(0.03) (0.04) (0.04) (0.03) (0.03) (0.04)
Investments last 12 months (in 1000 UGX) 0.00** 0.00*** 0.00** 0.00** 0.00*** 0.00**
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Invited to training 0.03* 0.03*
(0.02) (0.02)
Took training 0.06*** 0.06***
(0.02) (0.02)
Constant 0.43*** 0.40*** 0.40*** 0.43*** 0.41*** 0.41***
(0.08) (0.09) (0.11) (0.08) (0.09) (0.11)
Adj. R-Squared 0.030 0.033 0.030 0.033 0.037 0.033
Observations 2075 1743 1472 2075 1743 1472
Linear probability model. Dependent variable: Still open is a dummy variable that equals 1, if the business was still open at
the follow-up interview and zero, if otherwise. Independent variables: profits per worker is winsorized profits divided by the
number of persons working regularly in the shop, education level is measured from 0 (no education or incomplete primary)
to 5 (university) and financial literacy is a standardized score based on seven knowledge questions. All variables measured in
UGX are winsorized at the value 0 and at 99%. Omitted category for sector is retail. Standard errors clustered at trading
center level (108 clusters).
* p <0.10, ** p <0.05, *** p <0.01
33
Electronic copy available at: https://ssrn.com/abstract=4692520
Table 12: Descriptive statistics for business survival, including borrowing
Observed endline Still open Closed Difference
(1) (2) (3) (4)
Profits per worker (in 1000 UGX) 125.42 131.14 95.62 −35.52∗∗∗
Financial risk tolerance 4.95 4.98 4.78 −0.21
Female 0.64 0.63 0.69 0.06∗∗
Age (in years) 33.68 34.21 30.91 −3.29∗∗∗
Education level (0-5) 1.46 1.48 1.40 −0.08
Financial literacy (std.) 0.01 0.04 −0.11 −0.15∗∗
Work experience (in months) 82.57 84.47 72.65 −11.82∗∗
Able to borrow 100,000 UGX (share) 0.86 0.87 0.82 −0.06∗∗
Financial shock cost (in 1000 UGX) 135.24 143.60 91.70 −51.89∗∗∗
Age of business (in months) 54.77 57.64 39.77 −17.87∗∗∗
Sector: retail (share) 0.69 0.69 0.68 −0.01
Sector: services (share) 0.26 0.26 0.29 0.03
Sector: manufacturing (share) 0.05 0.05 0.03 −0.02∗
Investments last 12 months (in 1000 UGX) 1433.60 1511.36 1028.27 −483.09∗∗∗
Share of borrowers 0.14 0.14 0.13 −0.00
Amount of Loan (in 1000 UGX) 57.87 62.36 34.46 −27.90∗
Observations 2075 1741 334 2075
* p <0.10, ** p <0.05, *** p <0.01; p-values are obtained using twosided t-tests.
Profits per worker is winsorized profits divided by the number of persons working regularly in the shop. Observed
endline are all entrepreneurs interviewed at baseline for which we have business information at endline, still
open are those who are still in business 15 months later and closed are those who do not own a shop anymore,
who opened another shop or who relocated. Financial risk tolerance is measured on a scale from zero to
ten. Education level is measured from 0 (no education or incomplete primary) to 5 (university). All variables
measured in UGX are winsorized at the value 0 and at 99%. Financial literacy is a standardized score based
on seven knowledge questions.
5 Conclusion
This research addresses the role of risk tolerance for the success of small business en-
trepreneurs. It is known that the willingness to take risks is crucial for most entrepreneurs,
especially if they are not forced to become an entrepreneur, the so-called necessity en-
trepreneurs. Thus, entrepreneurs tend to be more risk tolerant than the average popula-
tion, but does this imply that being more risk tolerant is always superior? Alternatively,
is it conceivable that risk tolerance is beneficial but that there can be too much of a good
34
Electronic copy available at: https://ssrn.com/abstract=4692520
thing? In theory, there should be no reason why risk seeking behavior leads to higher
profitability than risk neutral decision making. Rather, high risk tolerance may be sim-
ilar to the cases of confidence or optimism, both being characteristics that are useful in
moderate form but can become detrimental if there is too much confidence and optimism.
Analyzing this issue, we make use of survey data for about 2,100 small entrepreneurs
in Western Uganda. We find that both low and high risk tolerance come with lower
profitability than moderate risk tolerance, which yields graphically an inverted U-shape
relationship between risk tolerance and profitability. Reassuringly, this relation is con-
firmed when we control for a set of individual and business characteristics. An important
transmission channel from risk tolerance to profitability appears to go via investments.
Thus, we contribute to the literature by analyzing the impact of risk tolerance on prof-
itability, allowing for non-linearity, controlling for relevant variables, and identifying a
transmission channel.
Adding insights about market dynamics, we also observe these entrepreneurs during a
second survey wave more than 18 months later and find that those closing their business
are mainly less profitable (and invested less) than the survivors. Additionally, other
characteristics hint in the expected direction; for example, that closing entrepreneurs
have less education. Regarding risk tolerance, we find that those surviving more often
tend to have a moderate degree of risk tolerance. Moreover, the businesses already being
in operation for a long time, i.e., four years or longer, do not show the strong decrease in
profitability for high risk tolerant entrepreneurs. This indicates that market forces may
work against the high price, i.e., relatively low profitability, to be paid by risk seeking
entrepreneurs.
These findings are relevant for active and future entrepreneurs. While earlier trainings
often motivated to take risks in order to run a successful small business, our results suggest
to also limit the willingness to take risks and that this limitation is not just a theoretical
concept but applies to real world data. Findings also, cautiously, suggest to emphasize
the necessity of sufficient investments for a successful and sustainable enterprise.
35
Electronic copy available at: https://ssrn.com/abstract=4692520
Overall, high risk tolerance leads to lower entrepreneurial profitability in the cross-
section, as does low risk tolerance. Consequently, reflecting market forces, relatively
profitable businesses survive more often than others. In the long run, market forces
may even drive out businesses characterized by high risk tolerance but only medium
profitability, so that the observed inverted U-shape relation between risk tolerance and
profitability describes the cross-section but not necessarily a longer-term equilibrium of
business development.
References
Azoulay, Pierre, Benjamin F. Jones, J. Daniel Kim, and Javier Miranda, 2020, “Age and
high-growth entrepreneurship.” American Economic Review: Insights,2(1), 65–82.
Baron, Reuben M. and David A. Kenny, 1986, “The moderator–mediator variable distinction in
social psychological research: Conceptual, strategic, and statistical considerations.” Journal of Per-
sonality and Social Psychology,51 (6), 1173.
Begley, Thomas M. and David P. Boyd, 1987, “Psychological characteristics associated with perfor-
mance in entrepreneurial firms and smaller businesses.” Journal of Business Venturing,2(1), 79–93.
Bernstein, Shai, Emanuele Colonnelli, Davide Malacrino, and Tim McQuade, 2022, “Who
creates new firms when local opportunities arise?” Journal of Financial Economics,143 (1), 107–130.
Caggese, Andrea, 2012, “Entrepreneurial risk, investment, and innovation.” Journal of Financial Eco-
nomics,106 (2), 287–307.
Caliendo, Marco, Frank M. Fossen, and Alexander S. Kritikos, 2009, “Risk attitudes of nascent
entrepreneurs–new evidence from an experimentally validated survey.” Small Business Economics,32
(2), 153–167.
Caliendo, Marco, Frank M. Fossen, and Alexander S. Kritikos, 2010, “The impact of risk atti-
tudes on entrepreneurial survival.” Journal of Economic Behavior & Organization,76 (1), 45–63.
Caliendo, Marco, Frank M. Fossen, and Alexander S. Kritikos, 2014, “Personality characteristics
and the decisions to become and stay self-employed.” Small Business Economics,42, 787–814.
36
Electronic copy available at: https://ssrn.com/abstract=4692520
Chanda, Areendam and Bulent Unel, 2021, “Do attitudes toward risk taking affect entrepreneurship?
Evidence from second-generation Americans.” Journal of Economic Growth,26 (4), 385–413.
De Blasio, Guido, Maria De Paola, Samuele Poy, and Vincenzo Scoppa, 2021, “Massive earth-
quakes, risk aversion, and entrepreneurship.” Small Business Economics,57, 295–322.
De Mel, Suresh, David J. McKenzie, and Christopher Woodruff, 2009, “Measuring microen-
terprise profits: Must we ask how the sausage is made?” Journal of Development Economics,88 (1),
19–31.
De Meza, David, Christopher Dawson, Andrew Henley, and G. Reza Arabsheibani, 2019,
“Curb your enthusiasm: Optimistic entrepreneurs earn less.” European Economic Review,111, 53–69.
Dohmen, Thomas, Armin Falk, Bart H. H. Golsteyn, David Huffman, and Uwe Sunde, 2017,
“Risk attitudes across the life course.” The Economic Journal,127 (605), F95–F116.
Dohmen, Thomas, Armin Falk, David Huffman, Uwe Sunde, Jürgen Schupp, and Gert G
Wagner, 2011, “Individual risk attitudes: Measurement, determinants, and behavioral consequences.”
Journal of the European Economic Association,9(3), 522–550.
Hamilton, Barton H., Nicholas W. Papageorge, and Nidhi Pande, 2019, “The right stuff?
Personality and entrepreneurship.” Quantitative Economics,10 (2), 643–691.
Hvide, Hans K. and Georgios A. Panos, 2014, “Risk tolerance and entrepreneurship.” Journal of
Financial Economics,111 (1), 200–223.
Kerr, Sari Pekkala, William R. Kerr, and Margaret Dalton, 2019, “Risk attitudes and personality
traits of entrepreneurs and venture team members.” Proceedings of the National Academy of Sciences,
116 (36), 17712–17716.
Knight, Frank H.,Risk, Uncertainty, and Profit, Boston: Houghton Mifflin, 1921.
Koellinger, Philipp, Maria Minniti, and Christian Schade, 2007, “’I think I can, I think I can’:
Overconfidence and entrepreneurial behavior.” Journal of Economic Psychology,28 (4), 502–527.
Kraft, Priscilla S., Christina Günther, Nadine H. Kammerlander, and Jan Lampe, 2022,
“Overconfidence and entrepreneurship: A meta-analysis of different types of overconfidence in the
entrepreneurial process.” Journal of Business Venturing,37 (4), 106207.
37
Electronic copy available at: https://ssrn.com/abstract=4692520
Kreiser, Patrick M., Louis D. Marino, Donald F. Kuratko, and K. Mark Weaver, 2013,
“Disaggregating entrepreneurial orientation: the non-linear impact of innovativeness, proactiveness
and risk-taking on SME performance.” Small Business Economics,40, 273–291.
Kritikos, Alexander S., “Personality and entrepreneurship.” in Klaus F. Zimmermann, ed., Handbook
of labor, human resources and population economics, Springer Cham, 2022.
l’Haridon, Olivier and Ferdinand M. Vieider, 2019, “All over the map: A worldwide comparison
of risk preferences.” Quantitative Economics,10 (1), 185–215.
Malmendier, Ulrike and Geoffrey Tate, 2005, “CEO overconfidence and corporate investment.” The
Journal of Finance,60 (6), 2661–2700.
Malmendier, Ulrike and Geoffrey Tate, 2015, “Behavioral CEOs: The role of managerial overconfi-
dence.” Journal of Economic Perspectives,29 (4), 37–60.
Malmendier, Ulrike and Stefan Nagel, 2011, “Depression babies: Do macroeconomic experiences
affect risk taking?” The Quarterly Journal of Economics,126 (1), 373–416.
Manso, Gustavo, 2016, “Experimentation and the returns to entrepreneurship.” The Review of Financial
Studies,29 (9), 2319–2340.
McCaig, Brian and Nina Pavcnik, 2021, “Entry and exit of informal firms and development.” IMF
Economic Review,69, 540–575.
McKenzie, David and Christopher Woodruff, 2017, “Business practices in small firms in developing
countries.” Management Science,63 (9), 2967–2981.
Meier, Armando N., 2022, “Emotions and risk attitudes.” American Economic Journal: Applied Eco-
nomics,14 (3), 527–58.
Menkhoff, Lukas and Sahra Sakha, 2017, “Estimating risky behavior with multiple-item risk mea-
sures.” Journal of Economic Psychology,59, 59–86.
Moskowitz, Tobias J and Annette Vissing-Jørgensen, 2002, “The returns to entrepreneurial in-
vestment: A private equity premium puzzle?” American Economic Review,92 (4), 745–778.
Nieß, Christiane and Torsten Biemann, 2014, “The role of risk propensity in predicting self-
employment.” Journal of Applied Psychology,99 (5), 1000–1009.
38
Electronic copy available at: https://ssrn.com/abstract=4692520
Norton, Edward C., 2022, “The inverse hyperbolic sine transformation and retransformed marginal
effects.” The Stata Journal,22 (3), 702–712.
Panousi, Vasia and Dimitris Papanikolaou, 2012, “Investment, idiosyncratic risk, and ownership.”
The Journal of Finance,67 (3), 1113–1148.
Pikulina, Elena, Luc Renneboog, and Philippe N. Tobler, 2017, “Overconfidence and investment:
An experimental approach.” Journal of Corporate Finance,43, 175–192.
Puri, Manju and David T. Robinson, 2007, “Optimism and economic choice.” Journal of Financial
Economics,86 (1), 71–99.
Schildberg-Hörisch, Hannah, 2018, “Are risk preferences stable?” Journal of Economic Perspectives,
32 (2), 135–54.
Sharma, Smriti and Finn Tarp, 2018, “Does managerial personality matter? Evidence from firms in
Vietnam.” Journal of Economic Behavior & Organization,150, 432–445.
Vereshchagina, Galina and Hugo A. Hopenhayn, 2009, “Risk taking by entrepreneurs.” American
Economic Review,99 (5), 1808–1830.
Willebrands, Daan, Judith Lammers, and Joop Hartog, 2012, “A successful businessman is not
a gambler. Risk attitude and business performance among small enterprises in Nigeria.” Journal of
Economic Psychology,33 (2), 342–354.
39
Electronic copy available at: https://ssrn.com/abstract=4692520
Appendix (Electronic Supplementary Material)
Figure A1: Maps of sampled trading centers in Western Uganda
40
Electronic copy available at: https://ssrn.com/abstract=4692520
Table A1: Description of independent variables
Financial risk
tolerance
Answer to the question “When thinking about investing and borrowing,
are you a person who is fully prepared to take risks or do you try to
avoid taking risk?” on a scale from 0-completely unwilling to take risks
to 10-fully prepared to take risk.
General risk
tolerance
Answer to the question “Are you generally a person who is fully prepared
to take risks or do you try to avoid taking risk?” on a scale from 0-
completely unwilling to take risks to 10-fully prepared to take risk.
Female Dummy that takes the value 1 if the respondent is female and 0 otherwise.
Age (in years) Age of the respondent in years.
Education level
(0-5)
Educational attainment of the respondent: 0-none or incomplete primary
education, 1-primary education, 2-lower secondary education, 3-upper
secondary education, 4-some tertiary education, 5-completed tertiary ed-
ucation.
Financial
literacy (std.)
Standardized financial literacy score based on the following questions:
1 What are 10 percent of 200,000 UGX?
DO NOT READ ANSWERS
-Gives right answer (20,000 UGX) quickly
-Gives right answer after a while
-Gives wrong answer
-Does not try to answer
2 Suppose you borrow 100,000 UGX at an interest rate of 2% per month,
with no repayment for 3 months. After 3 months, do you owe
-less than 102,000 UGX,
-exactly 102,000 UGX,
-or more than 102,000 UGX?
41
Electronic copy available at: https://ssrn.com/abstract=4692520
3 Imagine you have 100,000 UGX in a savings account earning 1% interest
per year, and prices for goods and services rise 2% over a 1-year period.
After one year, how much could you buy with this money?
-More than I could buy today
-Less than I could buy today
-The same amount that I could buy today
4 Is it riskier to plant
-multiple crops or
-one crop?
5 Suppose you need to borrow 500,000 UGX. Two people offer you a loan.
Which loan represents a better deal for you?
-One loan requires you to pay back 600,000 UGX in 1 month.
-The second loan requires you to pay back in 1 month 500,000 UGX plus
15% interest.
6 Suppose you owe 3,000,000 UGX to a bank. You pay a minimum pay-
ment of 30,000 UGX each month. At a monthly interest rate of 1%, how
many years would it take to eliminate debt if you took no additional
loan??
-Less than 5 years
-Between 5 and 10 years
-Between 10 and 15 years
-Never, you will continue to be in debt
7 If you were offered a loan with 5% monthly interest rate and a loan with
20% annual interest rate, which loan would offer better value?
-5% monthly interest rate
-20% annual interest rate
Work
experience (in
months)
Answer to the question “How long have you been working in any shop in
general?” in months.
Able to borrow
100,000 UGX
Answer to the question “Do you think you could be able to borrow
100,000 UGX in case you want to?” (dummy).
42
Electronic copy available at: https://ssrn.com/abstract=4692520
Financial shock
cost (in 1,000
UGX)
Answer to the question “How many months ago was the last time you had
an unexpected emergency that required for you to pay your own money
(such as a burial, a fire or a family member or friend falling sick)? How
much did you have to pay?” in UGX.
Age of business
(in months)
Answer to the question “How long have you been working in this partic-
ular shop?” in months.
Sector Sector the business operates in: 1-Retail and Wholesale, 2-Services and
3-Manufacturing
Investments
last 12 months
(in 1,000 UGX)
Answer to the question “How much money have you invested in your busi-
ness in total during the past 12 months? Investments could be for exam-
ple new equipment, new furniture or signs for advertising. Investments
should not include regular expenditures for buying new supplies/stock.
Investments should however include expenditures for restocking that you
do on top of your regular restocking.” in UGX.
Number of
workers
Answer to the question “How many people including yourself are regu-
larly working in this shop?”
Shop open
prev. year (in
months)
Answer to the question “How many months was the business in operation
during the past 12 months?” in months.
Bad-good
month from
1-10
Answer to the question “Compared to the average profits of your business
over the past year, would you say these past four weeks were a very bad
month, a normal month or a very good month? Please choose one number
from a 0 to 10 scale, where 0 means “very bad month, profits very low
compared to usual profits” and 10 means “very good month, profits very
high compared to usual profits.”
43
Electronic copy available at: https://ssrn.com/abstract=4692520
Table A2: Descriptive statistics for observations with and without missing values
No missings Missings Difference
(1) (2) (3)
Financial risk tolerance 4.94 3.63 −1.31∗∗∗
Female 0.64 0.56 −0.08
Age (in years) 33.69 38.94 5.26∗∗
Education level (0-5) 1.46 1.11 −0.35∗∗
Financial literacy (std.) 0.01 −0.23 −0.24∗∗
Work experience (in months) 82.51 117.23 34.72∗∗
Able to borrow 100,000 UGX (share) 0.86 0.76 −0.10∗∗
Financial shock cost (in 1000 UGX) 134.98 238.23 103.24
No. of workers 1.43 1.44 0.02
Age of business (in months) 54.38 69.16 14.78
Sector: retail (share) 0.69 0.73 0.04
Sector: services (share) 0.26 0.20 −0.06
Sector: manufacturing (share) 0.04 0.06 0.02
Observations 2144 79 2223
* p <0.10, ** p <0.05, *** p <0.01; p-values are obtained using twosided t-tests.
No missings are observations for which all variables used in the analysis are not missing, missings are
those observations for which some variables have missing values. The table only compares those variables
that are never missing for any of the observations except for age, which is missing for eight observations.
Financial risk tolerance is measured on a scale from zero to ten. Education level is measured from 0 (no
education or incomplete primary) to 5 (university). All variables measured in UGX are winsorized at the
value 0 and at 99%. Financial literacy is a standardized score based on seven knowledge questions.
44
Electronic copy available at: https://ssrn.com/abstract=4692520
800 1000 1200 1400 1600
Fitted values, investments last 12 months
0 2 4 6 8 10
Financial risk tolerance
Linear fit Quadratic fit
Figure A2: Relation between risk tolerance and past investments
500 1000 1500 2000
Fitted values, investments next 12 months
0 2 4 6 8 10
Financial risk tolerance
Linear fit Quadratic fit
Figure A3: Relation between risk tolerance and future investments
45
Electronic copy available at: https://ssrn.com/abstract=4692520
Table A3: Profits per worker and profits – financial risk tolerance in four groups
Profits per worker Profits
(1) (2) (3) (4) (5) (6)
Fin. tolerance, level 4-5 21.37** 12.20 5.04 20.49** 8.27 −1.45
(8.78) (8.33) (8.17) (9.94) (9.42) (9.03)
Fin. tolerance, level 6-7 24.81** 12.65 7.84 32.20*** 15.93 9.40
(9.83) (9.32) (8.78) (12.13) (11.98) (10.90)
Fin. tolerance, level 8-10 17.25 −2.06 −5.06 36.89** 8.77 4.70
(12.40) (11.70) (11.20) (16.40) (14.65) (14.52)
Female −52.03*** −41.69*** −79.68*** −65.64***
(8.04) (6.73) (9.91) (8.16)
Age (in years) 3.03* 1.89 3.95* 2.41
(1.64) (1.34) (2.00) (1.61)
Age squared −0.04** −0.03** −0.06*** −0.05***
(0.02) (0.02) (0.02) (0.02)
Education level (0-5)=1 13.46** 5.42 15.05* 4.13
(6.28) (6.33) (7.72) (8.04)
Education level (0-5)=2 39.28*** 27.89** 42.96*** 27.49**
(11.25) (11.68) (11.71) (12.09)
Education level (0-5)=3 77.08*** 68.91*** 82.84*** 71.74***
(24.41) (20.98) (26.62) (22.03)
Education level (0-5)=4 62.37*** 43.44** 100.44*** 74.72***
(18.41) (19.05) (22.92) (22.61)
Education level (0-5)=5 148.01** 123.36** 193.34*** 159.87**
(58.11) (58.34) (64.40) (64.15)
Financial literacy (std.) 10.89*** 5.29 14.30*** 6.70
(4.06) (3.54) (4.51) (4.17)
Work experience (in months) 0.07 0.05 0.15* 0.11
(0.06) (0.06) (0.07) (0.08)
Able to borrow 100,000 UGX (share) 20.19** 14.53* 26.43** 18.73*
(8.31) (8.30) (11.18) (11.10)
Financial shock cost (in 1000 UGX) 0.00 −0.00 0.01 −0.00
(0.01) (0.01) (0.02) (0.01)
Age of business (in months) 0.23** 0.23*** 0.30*** 0.29***
(0.09) (0.08) (0.11) (0.10)
Sector: services −21.38*** −6.74 −26.67*** −6.80
(6.28) (6.48) (8.31) (8.38)
Sector: manufacturing −41.06** −24.48* −51.05*** −28.52*
(16.31) (14.67) (18.86) (16.91)
Investments last 12 months (in 1000 UGX) 0.02*** 0.03***
(0.00) (0.00)
Constant 109.22*** 47.56 53.77* 142.02*** 75.20* 83.63**
(6.68) (33.33) (27.69) (8.59) (40.75) (34.38)
Adj. R-Squared 0.002 0.085 0.168 0.002 0.112 0.212
Observations 2144 2144 2144 2144 2144 2144
Dependent variables: profits is winsorized at the value zero and at 99%, profits per worker is profits divided by the number of persons
working regularly in the shop. Independent variables: education level is measured from 0 (no education or incomplete primary) to
5 (university) and financial literacy is a standardized score based on seven knowledge questions. All variables measured in UGX are
winsorized at the value 0 and at 99%. Omitted category for sector is retail. Standard errors clustered at trading center level (108
clusters).
* p <0.10, ** p <0.05, *** p <0.01
46
Electronic copy available at: https://ssrn.com/abstract=4692520
Table A4: Profits per worker and profits – financial risk tolerance in three groups
(shifted by one)
Profits per worker Profits
(1) (2) (3) (4) (5) (6)
Fin. tolerance, level 7-8 10.89 1.77 −2.44 32.90*** 20.16* 14.48
(9.61) (8.74) (8.32) (12.51) (10.93) (9.33)
Fin. tolerance, level 9-10 2.88 −10.23 −5.27 19.23 −0.94 5.76
(14.71) (13.89) (14.08) (20.92) (18.94) (19.51)
Female −52.29*** −41.69*** −80.02*** −65.68***
(8.09) (6.76) (9.96) (8.16)
Age (in years) 3.27** 2.00 4.18** 2.46
(1.61) (1.31) (1.98) (1.60)
Age squared −0.05** −0.03** −0.07*** −0.05***
(0.02) (0.02) (0.02) (0.02)
Education level (0-5)=1 13.24** 5.42 14.41* 3.83
(6.28) (6.27) (7.68) (7.97)
Education level (0-5)=2 39.25*** 27.83** 43.34*** 27.90**
(11.22) (11.67) (11.63) (12.11)
Education level (0-5)=3 76.88*** 69.25*** 82.20*** 71.88***
(24.69) (21.32) (26.88) (22.35)
Education level (0-5)=4 61.52*** 42.87** 99.84*** 74.61***
(18.35) (18.98) (23.01) (22.75)
Education level (0-5)=5 147.63** 122.68** 192.78*** 159.04**
(58.76) (58.92) (64.69) (64.40)
Financial literacy (std.) 11.32*** 5.49 14.66*** 6.77
(4.07) (3.54) (4.50) (4.17)
Work experience (in months) 0.07 0.04 0.14* 0.11
(0.06) (0.06) (0.07) (0.08)
Able to borrow 100,000 UGX (share) 21.23** 15.07* 26.55** 18.22*
(8.32) (8.26) (11.08) (10.92)
Financial shock cost (in 1000 UGX) 0.00 −0.00 0.01 −0.00
(0.01) (0.01) (0.02) (0.01)
Age of business (in months) 0.23** 0.23*** 0.30*** 0.29***
(0.09) (0.08) (0.11) (0.10)
Sector: services −21.40*** −6.89 −26.31*** −6.68
(6.25) (6.49) (8.19) (8.34)
Sector: manufacturing −41.24** −24.44 −50.77*** −28.06*
(16.46) (14.74) (18.74) (16.80)
Investments last 12 months (in 1000 UGX) 0.02*** 0.03***
(0.00) (0.00)
Constant 123.36*** 50.46 55.14* 155.71*** 76.54* 82.87**
(5.66) (34.10) (28.24) (6.64) (41.29) (34.79)
Adj. R-Squared -0.000 0.084 0.168 0.002 0.112 0.213
Observations 2144 2144 2144 2144 2144 2144
Dependent variables: profits is winsorized at the value zero and at 99%, profits per worker is profits divided by the number of persons
working regularly in the shop. Independent variables: education level is measured from 0 (no education or incomplete primary) to
5 (university) and financial literacy is a standardized score based on seven knowledge questions. All variables measured in UGX are
winsorized at the value 0 and at 99%. Omitted category for sector is retail. Standard errors clustered at trading center level (108
clusters).
* p <0.10, ** p <0.05, *** p <0.01
47
Electronic copy available at: https://ssrn.com/abstract=4692520