Policy Research Working Paper 10061

Unlocking Sustainable Private Sector Growth in the Middle East and North Africa

Background Paper

Jobs, Access to Credit, and Informality
in MENA Countries
Emanuele Brancati
Michele Di Maio
Aminur Rahman

Middle East and North Africa Region
Office of the Chief Economist
June 2022
Policy Research Working Paper 10061

Abstract
This paper explores the link between jobs, access to Finance, from informal firms. As a possible mechanism underlying
and informality. Using longitudinal firm-level data for coun- this result, the paper provides evidence that firms that suffer
tries in the Middle East and North Africa, it documents that informal competition have worse expectations on future
jobs creation is positively associated with access to finance. sales growth, which in turn are associated with fewer loan
At the same time, the findings show that access to finance applications.
is lower for firms that are more exposed to competition

This paper is a product of the Office of the Chief Economist, Middle East and North Africa Region. It is part of a larger
effort by the World Bank to provide open access to its research and make a contribution to development policy discussions
around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The
authors may be contacted at emanuele.brancati@uniroma1.it, michele.dimaio@uniroma1.it, and arahman@worldbank.org.

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

Produced by the Research Support Team
Jobs, access to credit, and informality in MENA countries∗

†
Emanuele Brancati Michele Di Maio‡ Aminur Rahman§

Keywords: employment, access to credit, informality, MENA countries
JEL codes: J00, E26, O53

∗ We thank Frank Betz for useful comments and Roberta Gatti and Asif Mohammed Islam for preliminary discussion on a

previous version of this paper. We also thank for comments conference participants at EBRD / EIB/ World Bank Workshop
on joint MENA regional report (October 2021). The ﬁndings, interpretations, and conclusions expressed in this work do not
necessarily reﬂect the views of The World Bank, its Board of Executive Directors, or the governments they represent. All errors
are our own.
† emanuele.brancati@uniroma1.it, Sapienza University of Rome, and IZA Institute of Labor Economics
‡ michele.dimaio@uniroma1.it, Sapienza University of Rome, Italy
§ arahman@worldbank.org, World Bank
1 Introduction

The economic environment in the Middle East and North Africa (MENA) regions is characterized by a

long-lasting stagnation in job creation. Over the period 2016-2019, employment growth was about 1.4%,

which is well below the performance of lower-middle and upper middle-income countries (respectively 4,8%

and 3,3%). One crucial determinant of employment growth is ﬁnance. Recent evidence documents the

importance of such relationships for MENA countries, especially in regard to the positive eﬀect of access to

credit on employment and investment (Ayyagari et al., 2021). Yet, an obstacle to the virtuous role of the

ﬁnancial system is represented by the disconnectedness of private ﬁrms from the banking sector. This has

been shown to be a distinguishing feature of MENA countries and a possible element contributing to the

poor job creation of the private sector in the region (De Lima et al., 2016; Amin, 2021). Another important

feature of MENA countries is informality. Data from the World Bank Enterprise Survey (hereinafter, WBES)

shows that 29% of MENA ﬁrms report to be exposed to competition from informal ﬁrms, with this share

reaching more than 40% of ﬁrms in Lebanon and Tunisia. A large informal sector is a possible threat to the

proper functioning of the economy and to the operation of formal ﬁrms which are negatively aﬀected by the

competition from informal ones (Distinguin et al., 2016; Rozo and Winkler, 2021; Avenyo et al., 2021).

This paper explores the link between jobs, ﬁnance, and informality. We begin our analysis by looking

at the link between jobs and ﬁnance. Using longitudinal ﬁrm-level data for MENA countries, we document

that jobs creation is positively associated with access to ﬁnance. At the same time, we show that access to

ﬁnance is lower for formal ﬁrms more exposure to informal competition. As a possible mechanism underlying

this result, we provide evidence that formal ﬁrms that are more exposed to informal competition have worse

expectations on future sales growth, which in turn are associated with fewer loan applications.

Our analysis employs WBES data for a large sample of private companies from the MENA countries.

This data have two important features. First, they have a panel dimension that we exploit to account for

the evolution of ﬁrm’s economic performance and characteristics across time and deal with the simultaneity

bias. Second, the WBES is the only survey of ﬁrms in developing countries which provides - in addition to

a large set of comparable ﬁnancial variables and ﬁrms’ characteristics - information on ﬁrms’ expectations

on their future performance. This is an unique type of information that we use to provide evidence of the

existence of a demand channel explaining our main result.

We begin our analysis documenting the link between jobs and access to ﬁnance. Our results indicate that

employment, employment growth, productivity, and wage are positively associated with loan availability for

ﬁrms in MENA countries, conﬁrming previous studies showing the beneﬁcial eﬀect of access to ﬁnance on

job creation (Betz and Ravasan, 2016; Ayyagari et al., 2021). Next, we show that formal ﬁrms that report

1
suﬀering from competition of informal ﬁrms have a signiﬁcantly lower probability of accessing credit, as

proxied by loan availability. This result goes over and beyond standard measures for ﬁrms’ creditworthiness

and informational opacity such as age or size. We also document that exposure to competition from informal

ﬁrms signiﬁcantly reduces loan application. These results are robust to several checks of the estimation

strategy, including sample selection, and the use of an IV strategy and matching technique. Finally, we

explore a possible mechanism explaining this result. We provide suggestive evidence that the negative eﬀect

of competition from informal ﬁrms on formal ﬁrms’ loan applications operates through a reduction in the

ﬁrm’s expected future sales. To this end, we show that expectations on future sales growth are signiﬁcantly

lower for ﬁrms reporting to be more exposed to competition from informal ﬁrms. Importantly, this eﬀect

is unrelated to diﬀerential realized sales in the past, a proxy for growth opportunities. Then, we show that

expectations on future sales growth predict loan applications, which is positively correlated with employment

growth.

Our paper is related and contributes to three strands of literature. First, our paper relates to the vast

literature on the eﬀect of the informal sector on the economy (Perry et al., 2007; Maloney, 2004; La Porta and

Shleifer, 2014). Informality is a distinguishing characteristic of most developing economies and it is widely

shown to impact the behavior and performance of ﬁrms operating in the formal sector. Ulyssea (2018) show

that the coexistence and competition of informal ﬁrms with more productive (formal) companies lead to

a misallocation of resources and potentially large losses in total factor productivity. Moreover, a number

of studies have documented that informal competition has a negative eﬀect on formal ﬁrms in terms of

output (Rozo and Winkler, 2021), employment (Amin, 2021), productivity (Amin and Okou, 2020), quality

of products (Banerji and Jain, 2007), and innovation (Avenyo et al., 2021). Lastly, Distinguin et al. (2016)

provide evidence that the presence of informal competition makes formal SMEs’ more likely to be credit-

constrained. Our analysis contributes to this literature by showing that the impact of informality on the

formal sector depends on the perceived threat that formal ﬁrms associate to informal competition. This,

in turn, has relevant eﬀects on ﬁrms’ expectations, investment decisions, and borrowing choices. As such,

our paper provides an important piece of the puzzle in the understanding of the eﬀect of informality on the

functioning of the formal economy.

Second, our paper speaks to the literature on the determinants of ﬁrms’ access to ﬁnance in developing

countries. Several studies have analyzed how availability of ﬁnance is linked to ﬁrms’ characteristics (Beck

et al., 2005, 2008) and emphasized the existence of obstacles to the supply of credit (Banerjee and Duﬂo, 2014;

Kersten et al., 2017). Bigsten et al. (2003) use data from African countries to document how ineﬃciencies in

the credit market lead micro-sized and small ﬁrms to a lower probability of loan access compared to larger

companies. Kuntchev et al. (2014) make use of WBES data to show that credit availability is inversely

2
associated with ﬁrm size but positively related to productivity and the country’s ﬁnancial deepening. Betz

and Ravasan (2016) show that the characteristics of prevailing collateral practices aﬀect the allocation of

credit in MENA countries. Finally, Ayyagari et al. (2021) exploit the introduction of credit bureaus to

identify a positive (exogenous) credit supply shock and show its beneﬁcial eﬀect on ﬁrms’ access to ﬁnance.1

Our paper contributes to this strand of the literature by providing evidence that, in some contexts, the

demand side of the story may be equally important. More speciﬁcally, we show that the characteristics of

the economic environment –and in particular the perceived level and type of market competition– can have

signiﬁcant eﬀects on ﬁrms’ demand for credit.

Finally, this paper is also related to the small but growing literature on the role of expectations in

inﬂuencing ﬁrms’ decisions. The turmoil that followed the 2008-ﬁnancial crisis gave new impulses to this

ﬁeld of research, with a number of studies connecting ﬁrms’ economic outcomes with their forward-looking

expectations. Most of this literature called the attention to the role played by macroeconomic factors,2 while

only a few studies focused on ﬁrms’ expectations on their own future earnings. Within the latter, Gennaioli

et al. (2016) show that for US companies corporate investment plans and actual investments are well explained

by expected sales. Along the same line, Boneva et al. (2020) looks at UK ﬁrms to show substantial eﬀects

of expectations on pricing strategies and employment behavior. Finally, Enders et al. (2019a) study how

changes in the outlook of German ﬁrms impact their real decisions, even if expectations turn out to be

incorrect ex-post. We contribute to this literature by showing how ﬁrms’ expectations are aﬀected by the

competition of informal companies and that this eﬀect goes beyond diﬀerences in ﬁrms’ fundamentals or

realized performances. This may suggest that, even in absence of real obstacles, a ﬁrm’s lack of information

or biased perception can signiﬁcantly jeopardize its own growth through current investment decisions and

demand for credit. Importantly, this is the ﬁrst paper, to the best of our knowledge, providing evidence on

expectations of ﬁrms within developing countries.

The paper proceeds as follows. Section 2 describes the data, the variables, and the sample composition.

Section 3 describes the empirical analysis, presents the main results, and discusses the possible mechanisms

explaining them. Section 4 summaries the analysis and discusses some policy implications of our results

2 Data

Our main source of data is the WBES dataset, a large sample of privately-held companies constructed from

a standardized and globally comparable survey administrated by the World Bank. Because of our research
1 A companion literature uses randomized control trials to explore the eﬀect of interventions alleviating micro-entrepreneurs’

ﬁnancing constraints (de Mel et al., 2008; Banerjee et al., 2015; Crepon et al., 2015; Quinn and Woodruﬀ, 2019).
2 See Coibion et al. (2018), Enders et al. (2019b), Coibion et al. (2020), Coibion et al. (2020), and Tanaka et al. (2020).

3
question, we restrict the sample to Middle East and North Africa (MENA) regions, for which we have

establishment-level data in the Arab Republic of Egypt, Jordan, Lebanon, Morocco, Tunisia, West Bank

and Gaza. The survey is representative of the non-agricultural private sector of each country and provides

information on all size classes, including small ﬁrms with less than 20 employees. This feature is of high

importance for our analysis, as smaller companies are also more likely to suﬀer the competition from the

informal sector.

The analysis takes advantage of the longitudinal dimension of the WBES to deal with simultaneity bias

or unobserved factors in the empirical strategy. From the total sample of 13,000 company-year observations

we focus on about 2,000 ﬁrms, for which we are able to match at least two consecutive waves of the survey.3

We discuss possible selection issues in Section 2.

Our main measure of interest, Constrained by informal, is a binary variable identifying ﬁrms that perceive

the competition from the informal sector as a major constraint. The original WBES survey explicitly asks

to what degree practices of competitors in the informal sector are an obstacle to the current operations of

the ﬁrm. We classify a company to be constrained by the informal sector if it declares such practices to be a

major or very severe obstacle (the top two categories).4 Our measure has the great advantage of capturing

an idiosyncratic component that goes beyond the mere diﬀusion of informal practices within the operating

sector of a country. Instead, it represents a speciﬁc proxy for how much the company perceives informal

competition as a jepardizing factor for its own business. As such, it is likely to be accounted for when ﬁrms

make their decisions and form their own expectations about the future.5

In this regard, the WBES dataset represents the only available survey providing forward-looking expec-

tations on ﬁrms’ future sales for MENA regions. This is a critical piece of information that made our study

possible and allows to shed light on the mechanism at stake. We employ a continuous measure for ﬁrms

expected sales growth in the next year (E(Sales growth )), as well as an ordinal measure capturing increasing,

decreasing, or stable expected earnings (respectively, E(Sales growth ): Positive, Negative, or Stable ).

We match all this information with a wide set of ﬁnancial characteristics of the company. First of all,

we exploit data on the availability of outstanding loans or credit lines (Loan availability ) at the time of the

survey, which is a synthetic measure for ﬁrms’ access to bank ﬁnance. Because we are interested in the
3 The original number of waves for countries in the MENA region ranges between two and ﬁve, from 2007 to 2020. However,

only in more recent years (i.e., the global dataset) ﬁrms are attached to a consistent panel identiﬁer across waves that allows
for longitudinal analyses. For most countries, two waves with panel identiﬁers are available: 2013–2019 for Jordan, Lebanon,
Morocco, West Bank and Gaza, and 2013–2020 for Tunisia. The only exception is Egypt, for which three waves are available:
2013, 2016, and 2020. We account for diﬀerences across countries and timing with the inclusion of country-speciﬁc time ﬁxed
eﬀects that purge the model from economy-wide factors that vary over time.
4 We collapsed the two answers because there is no clear ranking between the available options and they are both identifying

a signiﬁcant perceived obstacle for the company. As a robustness check, we employed a categorical measure with unchanged
results. The exact formulation of the questionnaires is provided in Table B1 of the Data Appendix.
5 We also control for a dummy identifying those that started their activity as unregistered ﬁrms (Originally informal ) or a

continuous measure for the number of years since the ﬁrm was formally registered (Years of formality ).

4
overall degree of connectedness of the company, our baseline speciﬁcation does not impose any constraint on

the original issuance of the loan. However, our results are broadly robust if we restrict the analysis to loans

that are issued within shorter horizons (ten years, seven, ﬁve, two, or even one year before the interview).

Second, we employ information on loan applications in the last year (Loan application ) to provide preliminary

evidence on whether the heterogeneous availability of funds is due to ﬁrms’ credit demand or is, instead,

driven by a diﬀerential probability of banks’ acceptance. Our extensive set of controls include information on

ﬁrms’ belonging sector (Manufacturing, Retail, Other services ) and form of proprietorship (Listed company,

LLC, Sole proprietorship, Partnership, Ltd Partnership ), structural characteristics (Size, Age ), exporting

status (Export ), realized past performance (Sales growth ), and number of competitors (N competitors ). All

variables are deﬁned in Tables B1 and B2 of the Data Appendix.

Selection and attrition Since we rely on the longitudinal dimension of the WBES dataset, it is worth

discussing possible selection issues aﬀecting our estimating sample. This has clearly to do with the non-

random probability of response and the self-selection of companies that kept answering the survey in following

a-vis ﬁrms that dropped out of the sample. Indeed, if such selection is somewhat simultaneously
waves, vis-`

correlated with our main regressors of interest and dependent variables, it may create a bias driving our

conclusions. In Table A1 of the Online Appendix, we tackle this issue by focusing on the full set of respondents

in the original 2013-waves and testing the correlation between ﬁrms’ likelihood of being interviewed a second

time and the variables employed in the analysis. Our estimates assuage concerns about systematic biases

by showing no signiﬁcance between ﬁrms’ probability of belonging to the panel and our main variables of

interest (Constrained by informal, Loan availability, Loan application, and Investment).6 Nevertheless, we

also provide additional robustness to our results by presenting Heckman selection models that deal with

endogenous sampling selection and propensity score matching techniques (see Section 3.3.2).

3 Empirical analysis

3.1 Descriptive evidence

Table 1 presents some descriptive statistics for the main variables in the sample. Loans and credit lines are

available only for 20% of the ﬁrms in the sample, suggesting that access to ﬁnance is underdeveloped. Yet,

the share of ﬁrms applying for a loan is only mildly larger. We interpret this as evidence of the degree of

disconnectedness from the banking sector of ﬁrms in the MENA region. Most ﬁrms ﬁnance working capital
6 Results are presented in Table A1 of the Online Appendix. The only exception is represented by Age, whereby older

ﬁrms are more likely to belong to the panel than younger companies, possibly because of their diﬀerent probability of survival.
However, since we always control for such a characteristic in our estimating regression, our estimates should still be unbiased.

5
through internally generated cash ﬂow, and tend not to rely on external sources of funding. At the same

time, alternative sources like private loans of owners are not often used to ﬁnance a ﬁrm’s business. All this

translates into a relatively low impact of ﬁnancial constraints and rationing: these are regarded to be relevant

issues by a small share of ﬁrms (14% to 27% of constrained ﬁrms, depending on the deﬁnition). While all ﬁrms

in our sample are formal, most of them started as unregistered businesses (90% were originally informal),

which conﬁrms the relevance of the informal sector in MENA countries. Among ﬁrms in our sample over the

period 2013-2019, 30% report competition from the informal sector as a major constraint to their activity.

At the country level, the share of ﬁrms constrained by informal competition is 20% in Jordan, 26% in West

Bank and Gaza, 27% in Egypt, 33% in Morocco, 41% in Tunisia and 44% in Lebanon.

Table 1: Descriptive statistics

Variable Average Stdev Min Max
Employment (log) 3.375 1.317 0.693 8.294
Employment growth 1.578 13.72 -84.2 100
Loan availability 0.202 0.401 0.000 1.000
Loan application 0.332 0.471 0.000 1.000
Turned down 0.045 0.206 0.000 1.000
Account 0.782 0.413 0.000 1.000
No need 0.598 0.490 0.000 1.000
Rationing: not rationed 0.630 0.483 0.000 1.000
Rationing: partially rationed 0.151 0.358 0.000 1.000
Rationing: fully rationed 0.110 0.313 0.000 1.000
Investment 0.221 0.415 0.000 1.000
Age 2.864 0.756 0.693 5.094
Size 3.365 1.375 0.000 10.59
Export 0.181 0.385 0.000 1.000
Number of competitors 4.172 1.692 0.000 5.204
Manufacturing 0.586 0.493 0.000 1.000
Retail 0.092 0.289 0.000 1.000
Other services 0.322 0.467 0.000 1.000
Listed company 0.061 0.240 0.000 1.000
LLC 0.221 0.415 0.000 1.000
Sole proprietorship 0.380 0.486 0.000 1.000
Partnership 0.192 0.394 0.000 1.000
Ltd Partnership 0.137 0.344 0.000 1.000
E(Sales growth) 0.005 0.255 -1.000 1.000
E(Sales growth): Positive 0.489 0.499 0.000 1.000
E(Sales growth): Stable 0.257 0.437 0.000 1.000
E(Sales growth): Negative 0.254 0.435 0.000 1.000
Sales growth -3.986 21.24 -98.99 99.83
Originally informal 0.898 0.303 0.000 1.000
Years of formality 2.842 0.758 0.000 5.357
Constrained by informal 0.293 0.455 0.000 1.000

Notes: Descriptive statistics for the main variables in the sample.

The focus of our analysis is on understanding how competition from informal ﬁrms shapes formal ﬁrms’

ﬁnancial choices and future prospects. Table 2 we report distributions conditional on whether ﬁrms perceive

the competition from the informal sector as a major constraint, i.e. whether Constrained by informal takes

value 1. Firms suﬀering informal competition have a lower probability to have a loan or credit line. Yet, this is

not linked with a larger rejection rate from the banking sector (Turned down ). Instead, preliminary evidence

shows that most of the diﬀerence is driven by the application process, whereby ﬁrms that are constrained by

informal companies have a signiﬁcantly lower credit demand compared to their unconstrained counterparts.

6
This heterogeneity is not reﬂected in other components of the loan covenant, suggesting that this evidence

is not merely related to a diﬀerential risk or creditworthiness. Notice that ﬁrms that are suﬀering from

informal competition are not even associated with a diﬀerent probability of being credit rationed. As for

other structural characteristics, constrained ﬁrms are somewhat smaller, tend to export relatively less, and

are more concentrated in the manufacturing sector. As such, our estimating regression always accounts for

a rich set of additional controls to account for these diﬀerences across the two groups.

Another dimension along which there is a diﬀerence between the two groups of ﬁrms is expectations

on future sales growth: ﬁrms suﬀering from informal competition have signiﬁcantly worse prospects on

their future earnings (-3.15% vs. 2.13%, on average). Interestingly, such heterogeneity in expectations does

not ﬁnd a match in realized sales, which are somewhat similar across groups (and even less negative for

constrained ﬁrms). This evidence suggests that it is unlikely that our proxy for informality simply reﬂects

diﬀerent fundamentals (i.e., good vs. bad ﬁrms).

Table 2: Conditional averages: Unconstrained vs constrained by informal competition

Unconstrained Constrained Diﬀ mean
Variable by informal by informal p-value
Employment (log) 3.439 3.227 0.000
Employment growth 2.491 0.254 0.000
Loan availability 0.198 0.155 0.022
Loan application 0.327 0.259 0.002
Turned down 0.051 0.044 0.491
Account 0.828 0.806 0.232
No need 0.603 0.603 0.990
Rationing: not rationed 0.647 0.632 0.519
Rationing: partially rationed 0.145 0.159 0.432
Rationing: fully rationed 0.101 0.125 0.121
Age 2.772 2.794 0.903
Size 3.385 3.179 0.000
Export 0.186 0.168 0.022
Manufacturing 0.569 0.619 0.000
E(Sales growth) 2.129 -3.158 0.000
E(Sales growth): Positive 0.523 0.390 0.000
E(Sales growth): Stable 0.237 0.313 0.000
E(Sales growth): Negative 0.240 0.296 0.000
Sales growth -3.739 -3.665 0.875
Investment 0.206 0.156
Originally informal 0.894 0.893 0.762
Years formality 2.839 2.841 0.925

Notes: conditional distributions of the main variables employed. In Column 1, we report averages for the sample of ﬁrms
declaring no major constraints from the informal sector, while in Column 2 we focus on the subsample of constrained ﬁrms
only. Column 3 reports the p-value of the t-test on equality of means.

3.2 Jobs and ﬁnance

We begin our analysis by looking at the role of ﬁnance as a determinant of job creation. To this end, we

estimate the following model:

yi,t = α + β Loan availabilityi,t + γ Xi,t + γi + λt + εi,t (1)

7
where yi,t is the outcome variable, alternatively, the number of employees (Employment ), the labor produc-

tivity (Productivity ), and the ﬁrm-level wage rate (Wage ). Xi,t−1 is a vector of structural controls for ﬁrms’

size, age, exporting behavior, and sector. γi and λt are, respectively, ﬁrm and time ﬁxed eﬀects.

Table 3 shows the results for regression model ?? reporting in each column one of our labor market

outcomes of interest. Estimates in column 1 indicate that employment is positively correlated with access

to ﬁnance. That having access to bank credit has a positive and signiﬁcant eﬀect on the (log) number of

employees conﬁrms how external sources of funding may have a critical role in aﬀecting ﬁrms’ employment

growth. This eﬀect goes beyond structural controls for the age of the ﬁrm, the exporting activity of the ﬁrm,

operating sector of the company, common time components that capture cyclical factors (time ﬁxed eﬀects),

as well as ﬁrm-speciﬁc ﬁxed eﬀects absorbing any time-invariant characteristic (observable or not). Results

also indicate evidence of a stronger job creation for younger and export-oriented companies. Noticeably,

the positive eﬀect of loan availability is not limited to the raw number of employees but extends to labor

productivity (as proxied by sales per worker) and the average salary paid.

Table 3: Access to ﬁnance and jobs

Dependent variable: Employment Productivity Wage
(1) (2) (3)
Loan availability 0.132*** 0.256*** 0.155**
[0.0392] [0.0838] [0.0672]
Age -10.49*** -6.827 -0.776
[2.821] [5.066] [3.855]
Export 0.208*** 0.114 0.155*
[0.0565] [0.110] [0.0898]
Manufacturing 0.139 0.352** 0.246**
[0.0969] [0.152] [0.122]
Retail -0.131* 0.221* 0.203**
[0.0682] [0.134] [0.0995]
Model OLS OLS OLS
Time FE yes yes yes
Firm FE yes yes yes
R2 0.0445 0.117 0.107
Observations 12605 11159 11581

Notes: within estimator with ﬁrm and time ﬁxed eﬀects. Measures are deﬁned in Table B4. Robust standard errors in brackets.
*, **, *** indicate statistical signiﬁcance at the 10%, 5%, and 1%, respectively.

3.3 Finance and informality

Next, we explore the determinants of ﬁrms’ access to ﬁnance and how this is related to competition from

informal ﬁrms. We begin by looking at the characteristics of ﬁrms having access to a loan or credit-line from

a bank. Then, we focus on the determinants of the ﬁrm’s decision to apply for a loan.

Our baseline speciﬁcation reads as follows:

yi,t = α + β Constrained by informali,t−1 + γ Xi,t−1 + λt + µa(i) + εi,t (2)

8
where yi,t is the outcome variable. In our analysis this is a dummy taking value 1 if ﬁrm i at time t has

a loan or credit-line and zero otherwise (Loan availability ) or a dummy indicating if the ﬁrm has made an

application for bank funding (Loan application ).Constrained by informal is our dummy of interest, as deﬁned

in Section 2, and Xi,t−1 is a vector of structural controls for ﬁrms’ size, age, exporting behavior, number of

competitors, belonging industry, and form of proprietorship. λt and µa(i) are, respectively, time ﬁxed eﬀects

and a set of granular indicators for the geographical area of the company (41 in total).7 These are meant to

capture time varying common factors and persistent heterogeneities linked with the operating environment

of the ﬁrm. In our analysis, all regressors are lagged to rule out simultaneity bias. When yi,t is a binary

measure (Loan availability, Loan application) equation 2 is estimated via a logistic model while when it is a

continuous measure (Expected sale growth) it is estimated using a linear probability model. In all tables, we

report White’s heteroscedasticity-consistent standard errors but our results are largely robust to alternative

clusterings.

Discussion of the empirical strategy In assessing the eﬀect of informal competition on ﬁrms’ connect-

edness to the banking sector we face two main empirical challenges.

The ﬁrst one has to do with omitted variable bias, whereby relevant characteristics excluded from the

model may drive the relationship between ﬁrms’ availability of bank ﬁnance and informal competition. For

instance, if we missed to properly account for ﬁrms’ fundamentals that characterize the perception about

informal pressure, and if this aﬀects demand and/or supply of credit, our estimates may be biased. In

order to assuage such concern, we augment our baseline speciﬁcation with an extensive set of additional

regressors that virtually cover any dimension available in the WBES survey. In our robustness checks,

we account for past sales and productivity growth, capacity utilization, as well as granular information

on main destination markets (local, national, or international), managerial characteristics (experience of the

managers, female managers) and ownership (government ownership and female owners). All of which capture

diﬀerent dimensions of ﬁrms’ fundamentals. Moreover, we control for relevant characteristics of the local

environment that may be spuriously associated with informality: the size of the city, measures of corruption

(bribery depth or gifts to oﬃcials) and crime (loss from thefts and vandalism), as well as other shocks linked

with the localization of the company (number and length of electric outages), together with geographic area

ﬁxed eﬀects interacted with more granular sector controls (2-digit ISIC Rev. 3.1 level). Furthermore, in

some speciﬁcations we enrich the vector X with controls for the issuance of the loan, whether the company

was originally informal, and the number of years since formal registration occurred (see below). Finally,
7 Geographical areas are deﬁned as the localization group used as a stratum in the WBES sampling scheme. From the

sampling methodology note of the WBES: “Geographical distribution is deﬁned to reﬂect the distribution of the non-agricultural
economic activity of the country; for most countries this implies including the main urban centers or regions of the country.”

9
we account for the introduction of ﬁrm-speciﬁc ﬁxed eﬀects (in our benchmark results) through conditional

logistic and linear probability (within estimators) models. This allows us for purging all ﬁrms’ characteristics

(observable and unobservable) that are stable over time.8

The second interrelated issue has to do with endogeneity and, in particular, with possible reverse causality

(i.e., whereby it is access to credit that drives the perception of informal competition and not the other

way round; Friesen and Wacker, 2019). For instance, ﬁrms that are constrained by banks may be unable

to fund potentially proﬁtable projects, increase their production scale, or upgrade toward higher levels of

productivity. While this is unlikely to impact the actual operating environment of the company (i.e., access

to credit of a speciﬁc ﬁrm does not aﬀect its real competition), we cannot a priori exclude an eﬀect on ﬁrms’

perception about the magnitude and relevance of the informal competition faced.

Notice that, if this were the case, we should observe signiﬁcantly-diﬀerent patterns in a ﬁrm’s probability

of rationing. However, as shown in Table 2, ﬁrms that suﬀer informal competition are linked neither with

a diﬀerential likelihood of credit constraints (full or partial rationing), nor with actual or expected rejection

rates on loan applications. Thus, our descriptive evidence seems to suggest that, at least in our sample of

MENA countries, reverse causality should not be a relevant concern.

Nevertheless, we adopt a number of alternative approaches to deal with this potential endogeneity issue.

First of all, we always employ lagged regressors so as to rule out simultaneity bias. This, however, does

not address endogeneity if both the dependent variable and our measure of interest displays high degrees of

persistence. Hence, we further shed light on this issue by restricting our estimating sample to companies

with no credit access in t − 1. Even focusing on “switchers” only (i.e., ﬁrms with new loans that were

disconnected in the previous waves), our results prove to be extremely robust (see Table A2). Moreover,

unreported regressions linking past loan availability with current perception of informal competition show

no correlation between the two, further suggesting that reverse causality is unlikely to drive our ﬁndings.9

Most importantly, we employ an IV approach to further take care of endogeneity. We rely on a cell-

average method wherein we instrument informal competition with the proportion of all other ﬁrms that are

constrained by informality and operate within the same 2-digit sector and geographical area of each company

(at a given time). This approach, widely used in the literature (see, among many others, Distinguin et al.,

2016; Dollar et al., 2006; Fisman and Svensson, 2007; Amin and Soh, 2021; Amin, 2021), allows us to capture

an environmental component of the informal competition faced by a company that is, however, unrelated to
8 Because of the structure of the dataset, we can only introduce ﬁrms’ ﬁxed eﬀects when dealing with loan availability

and application. Notice that, in such case, identiﬁcation is achieved exclusively exploiting data from Egypt, which is the only
country for which we have three waves (as discussed in footnote 3). For expectations, the absence of a panel structure does not
allow for such analysis.
9 Notice that, while being largely insigniﬁcant (p value of 0.210), the coeﬃcient of past loan availability on informal compe-

tition is even of opposite sign: positive rather than negative.

10
its speciﬁc characteristics, including fundamentals and past availability of banking funds. Notice that, by

computing averages at the stratum level (intersection of industry and geographical area), we still document

eﬀects that go beyond sector and location ﬁxed eﬀects. First stage regressions conﬁrm the sizable power of

our instrument.

Finally, we further take care of self-selection and endogeneity by means of matching techniques. We rely

on radius matching or a bias-corrected nearest neighbor matching estimator (as in Abadie and Imbens, 2011)

to recover a subsample of companies with the same ex ante probability of being constrained by the informal

sector. We then explore the average treatment eﬀects on our main dependent variables. Results show good

balancing properties of our procedures and are, again, largely consistent with the main analyses.

3.3.1 Loan availability

Table 4 shows the estimates for our model 2 when the dependent variable is Loan Availability, a dummy

taking value 1 if the ﬁrm has a loan or a credit line and zero otherwise. Column 1 presents the marginal

eﬀects estimates using the pooled logistic estimator for the baseline speciﬁcation.10 Access to ﬁnance is

also signiﬁcantly and positively associated with ﬁrms’ size, while all other coeﬃcients are insigniﬁcant. In

column 2, we enrich the speciﬁcation with two additional important controls: Account (a dummy for ﬁrms

with checking or savings account) and No need (a dummy for ﬁrms that do not apply for a loan because of

a lack of ﬁnancial needs).11 Results indicate that having an account does not inﬂuence loan availability and

that ﬁrms not needing a loan are - reassuringly - signiﬁcantly less likely to have a bank credit line. In column

3, we make sure that the eﬀect of informality is not simply arising from a higher degree of competition. Even

after augmenting the model with the approximate number of competitors, our results are virtually unchanged,

suggesting that the eﬀect of competition form informal ﬁrms goes over and beyond the eﬀect of competition

per se. In terms of magnitude of the eﬀect, ﬁrms more constrained by competition from the informal sector

have a 8.4% lower probability of accessing bank ﬁnance. Finally, we fully exploit the longitudinal dimension of

the data set and allow for ﬁrm-speciﬁc ﬁxed eﬀects employing linear probability (within estimator in column

4) and conditional logistic models (column 5).12 Even if we purge the speciﬁcation from any observable and
10 In all speciﬁcations, we control for a rich set of ﬁrm’s structural characteristics to capture heterogeneities in ﬁrms’ cred-
itworthiness that can be potentially correlated with our measure of interest. To this aim, we always account for ﬁrms’ size
and age, which are essential determinants of ﬁrms’ choices and are traditionally used as direct proxies for ﬁnancial constraints.
Moreover, we control for the form of proprietorship to absorb any heterogeneity in banks’ willingness to grant credit associated
to diﬀerent forms of governance (which may entail a diﬀerent degrees of ﬁnancial solidity and opacity). Finally, we always
account for ﬁrms’ exporting status, sectoral classiﬁcation, as well as time and geographical-area ﬁxed eﬀects so as to capture
environmental factors that can impact a ﬁrm’s fundamentals and result into a diﬀerent creditworthiness (or credit demand),
such as idiosyncratic shocks to the demand for goods in certain periods, industries, or countries.
11 Note that Account captures demand for credit by the ﬁrm as long as that comes with an overdraft facility. At the same

time, No need entails a spurious relationship with informal competition if this decision is itself inﬂuenced by the exposure to
informal competition.
12 Notice that the sample size with conditional logistic that rely on the time-variation in the dependent variable models drops

substantially. Nevertheless, our main results are qualitatively similar to the ones presented.

11
unobservable time-invariant characteristic (including the ﬁrm’s location and any other persistent factor that

may aﬀect the ﬁrm-bank relationship), ﬁrms that are constrained by informal competition are found to have

a signiﬁcantly lower probability of access to bank loans.

Finally, one may be worried that our measure of access to bank funds collapses credit lines and loans

that are issued in diﬀerent times and that, potentially, can date back to many years in advance. To tackle

this issue we restricted loan availability to a shorter time horizon for the issuance (10, 7, 5, 2, or 1 year)

obtaining results that are virtually unchanged. Results are provided in Table A3 of the Online Appendix.13

Table 4: Competition from informal ﬁrms and loan availability

Dependent variable: Loan availability
(1) (2) (3) (4) (5)
Constrained by informalt−1 -0.0632*** -0.0714*** -0.0840*** -0.175* -2.187*
[0.0192] [0.0193] [0.0225] [0.0942] [1.163]
Accountt−1 0.0374 0.00199 -0.165* -0.719
[0.0252] [0.0277] [0.0953] [0.937]
No needt−1 -0.0448*** -0.0467** -0.0530 0.404
[0.0165] [0.0192] [0.0738] [0.806]
Aget−1 0.00293 0.00594 -0.00164 -4.018** -21.45
[0.0135] [0.0137] [0.0158] [2.037] [19.82]
Sizet−1 0.0383*** 0.0344*** 0.0374*** 0.527 24.85
[0.00645] [0.00672] [0.00802] [1.074] [33.83]
Exportt−1 0.00420 0.00916 0.0186 0.143 0.176
[0.0197] [0.0196] [0.0263] [0.174] [1.357]
Number of competitorst−1 -0.00550
[0.00561]
Model Logit Logit Logit Within Cond. logit
Time FE yes yes yes yes yes
Geographic area FE yes yes yes no no
Firm FE no no no yes yes
Observations 1982 1931 1379 1398 62
Pseudo R2 (R2) 0.181 0.186 0.195 0.139 0.355

Notes: logit marginal eﬀects, within estimator, and conditional logistic models. Variables are deﬁned in Table B4. Robust
standard errors in brackets. *, **, *** indicate statistical signiﬁcance at the 10%, 5%, and 1%, respectively.

3.3.2 Loan application

Table 5 shows the estimates for our model when we use as dependent variable a dummy taking value 1 if

the ﬁrm makes a loan application and zero otherwise. Results of the logit estimates indicate that the eﬀect

of being exposed to informality competition on the probability of a loan application is consistently negative

across various econometric speciﬁcations and samples. In particular, the eﬀect of Constrained by informal

is negative when controlling for whether or not the ﬁrm does need a loan and the ﬁrm has an account

(which may imply that the ﬁrm can use the overdraft) (column 2), or if we control for the number of ﬁrm’s

competitors (column 3). In terms of magnitude of the eﬀect, ﬁrms more constrained by competition from
13 Notice that the dataset also contains information on the size and duration of the most recent loan or credit line, as well as

the required collateral and its approximate value. Unfortunately, such measures are not populated and the estimating sample
size drops substantially (fewer than 200 observations). With the available data no eﬀect of informality emerges.

12
the informal sector have a 7.2% lower probability of making a loan application. Results do not change when

we use the within ﬁrm estimator (column 4) and when we estimate the model using the conditional logit

(column 5): being exposed to competition from informal ﬁrms reduces the loan application by formal ﬁrms.

Table 5: Competition from informal ﬁrms and loan application

Dependent variable: Loan application
(1) (2) (3) (4) (5)
Constrained by informalt−1 -0.0483** -0.0530** -0.0715*** -0.354*** -3.371**
[0.0206] [0.0209] [0.0239] [0.0895] [1.491]
Accountt−1 0.0223 0.00938 -0.0777 -1.310
[0.0253] [0.0287] [0.0990] [1.348]
No needt−1 -0.0399** -0.0399* 0.170** -0.0892
[0.0185] [0.0216] [0.0755] [1.441]
Aget−1 0.00682 0.00778 0.00566 -7.575*** 1.724
[0.0145] [0.0147] [0.0170] [2.094] [1.191]
Sizet−1 0.0348*** 0.0307*** 0.0372*** 0.810 0.889
[0.00735] [0.00762] [0.00909] [1.166] [0.778]
Exportt−1 0.00865 0.0111 0.0171 0.178 0.787
[0.0229] [0.0232] [0.0316] [0.171] [1.354]
Number of competitorst−1 0.00204
[0.00639]
Model Logit Logit Logit Within Cond. logit
Time FE yes yes yes yes yes
Geographic area FE yes yes yes no no
Firm FE no no no yes yes
Additional controls yes yes yes yes yes
Observations 2065 2008 1443 1446 64
Pseudo R2 (R2) 0.197 0.199 0.222 0.262 0.481

Notes: logit marginal eﬀects, within estimator, and conditional logistic models. Unreported controls follow the speciﬁcation in
Table 5. Variables are deﬁned in Table B4. Robust standard errors in brackets. *, **, *** indicate statistical signiﬁcance at
the 10%, 5%, and 1%, respectively.

Robustness checks We perform a number of robustness tests to check the validity of our results. To begin,

we take care of selection issues by employing Heckman-type selection models (Heckman, 1976; Lewis, 1974;

Gronau, 1974). In essence we model a ﬁrms’ probability of belonging to the panel (i.e., being interviewed in

two consecutive waves of the WBES survey) adding as an excluded regressor ﬁrms’ belonging stratum (the

intersection of sector and country). We then augment the original speciﬁcation with the inverse Mill’s ratio

of the selection regression. As shown in Table A4, our main conclusions are largely unchanged.

One possible concern with our results is that ﬁrms that report to be constrained most by informal

competition are the ones that operated as informal ﬁrms in the past and only recently switched to a formal

form of business. If this were the case, the lower connectedness with the banking sector may only capture

a matter of timing whereby constrained ﬁrms are those having had a shorter periods to establish a banking

connection. To assuage this concern we augment our baseline speciﬁcation with a direct indicator about

whether the company was originally operating informally, together with a measure for number of years

since the formal registration. Results for both loan availability (reported in Table A5 column 1) and loan

13
application (reported in Table A6 column 1) show that the coeﬃcient for our main variable of interest

does not change and that these measures turn out to be largely insigniﬁcant. Moreover, our ﬁndings hold

when we add a large set of additional regressors that assuage concerns about possible confounding factors.

This includes measures for past investment in physical assets and growth opportunities (past productivity

and sales growth), more granular controls for the destination market of ﬁrms’ products (local, national,

or international), controls for management characteristics and ownership (existence of a board of directors,

years of past experience of the management, presence of females in the board, share of government ownership,

female owners), characteristics of the location of the ﬁrm and exposure to local shocks (dummies for the

size of the city, as well as the number and length of electric outages), the degree of capacity utilization of

the company, and measures for problems related to the local environment (including the share of bribery

depth, the share of gifts, the share of losses from theft and vandalism, or whether the company feels to be

constrained by corruption or by crime). Results for loan availability (reported in Table A5 column 2-5) and

for loan application (in Table A6 column 2-5) show that our main ﬁnding continues to hold: ﬁrms more

exposed to informal competition are less likely to have a loan or a credit line and to apply for a loan.

Despite our results are robust to a large number of controls for ﬁrms’ fundamentals, one residual concern

may still be reverse causality. In particular, it is possible that our ﬁndings merely reﬂect the negative eﬀect

of access to ﬁnance onto ﬁrms’ perception about the pressure exerted by informal competitors. This is linked

with the very deﬁnition of our variable of interest, which is subjective in nature and may capture some

spurious relationships with past performances and growth, rather than actual heterogeneities in the informal

competition faced by the company. While descriptive evidence we discussed above already showed this is

unlikely to be the case, we performed two additional exercises to further assuage such concern.

First, we develop an instrumental variable approach to directly tackle endogeneity issues. In particular,

we rely on a cell-average method widely used in the literature (Distinguin et al., 2016; Dollar et al., 2006;

Fisman and Svensson, 2007) and instrument ﬁrm-speciﬁc perception about informal competition with the

share of other ﬁrms declaring relevant constraints from the informal sector within the same environment.

At each point in time, we deﬁne the averaging cell at the intersection of the 2-digit sector and geographical

region of the company. This approach is meant to isolate a component of informal competition that is related

to the operating environment of the ﬁrm and, as such, is not subject to swings due to ﬁrm-speciﬁc factors.

Indeed, a ﬁrm’s access to banking ﬁnance is unlikely to drive other ﬁrms’ perception about the diﬀusion and

relevance of informal competitors, therefore addressing our primary concern. In Table A7, we present the

estimates of IV-linear probability models. Results largely conﬁrm our previous ﬁndings and point at a very

negative and signiﬁcant eﬀect of informal competition onto ﬁrms’ availability of loans.14 Again, this eﬀect
14 Notice that, because we deﬁne averaging cells at the intersection of industry and geographical area, we are still documenting

14
seems to operate through a 10%-lower probability of application, possibly implying that the eﬀect mainly

operates through a reduction in credit demand (we further discuss this issue in the following sections). Notice

that, underidentiﬁcation and weak-identiﬁcation tests conﬁrm the power of our instrument, which, in line

with prior expectations, is positively correlated with Constrained by informal.

As an additional exercise, we further take care of self-selection and endogeneity by means of matching

estimators. In essence, we employ propensity-score techniques to select a sample of ﬁrms that are constrained

by the informal sector or not (i.e., treated and control group) and that are similar along a broad set of

characteristics but diﬀer for their actual condition of facing informal competition (i.e., with the same ex-ante

probability of being treated). We then implement two diﬀerent estimators for the average treatment eﬀect

(ATT), one based on nearest neighbor matching with bias correction as in Abadie and Imbens (2011), and

the other based on the radius matching. In computing the propensity score, we exploited the full set of ﬁrms’

characteristics employed so far. Table A8 reports the balancing properties of the procedure and shows no

diﬀerence in ﬁrms’ characteristics between the treated and control group after the matching, thus reassuring

about the success of the balancing. In Table A9, we present the estimated ATT and conﬁrm the negative

eﬀect of informal competition on loan availability and loan applications.

3.3.3 Channel: Expectations on future sales

Firms that report to be more constrained by competition from informal ﬁrms are less likely to demand a

loan (see Table 5). As we documented in the previous section, this choice is not related to diﬀerences in

ﬁrm’s age, size, export activity, or to the fact that the ﬁrm is diﬀerent in its need for a loan or has access to

a bank account (which may be a substitute for loans if the ﬁrm has an overdraft facility) or by self-selection,

endogeneity, or reverse causality.

One possible channel explaining our main result is that diﬀerence among ﬁrms in terms of how much

they are threatened by competition from informal ﬁrms aﬀect the level of their expectations for future

performance. To test for this, we exploit the response to a question introduced in the most recent wave

of the WBES where the ﬁrm is asked whether the growth in sales are expected to be negative, stable, or

positive. Table 6 reports the estimates of a multinomial regression in which these three alternatives are

regressed on our explanatory variable (Constrained by informal ) controlling for area and time ﬁxed eﬀects

and our full set of controls.15 Results in column 1-3 indicate that ﬁrms which report to be more exposed
eﬀects that go above the average dynamic at the sector and location levels.
15 Since the question on expected expectations is only available for the last wave of the survey, this estimation is performed

cross-sectionally (i.e., matches current expectations with competition from the informal sector in the same wave). Notice that
because expectations are formed in time t and regard ﬁrms’ earnings at the one-year horizon, and since Constrained by informal
refers to the recent past, there is no overlap in the timing of the two variables so that we can safely avoid the use of lagged
regressors.

15
to competition from informal ﬁrms are signiﬁcantly more likely to report negative or stable expected sales

growth and signiﬁcantly less likely to report positive expected sales growth.16 This ﬁnding is conﬁrmed

when we consider as outcome a continuous version of the variable measuring expected sales growth (column

4) and when we consider only the sample of ﬁrms in the panel (column 5): ﬁrms more exposed to informal

competition have worse expectations on future sales growth. One important observation concerns the actual

level of ﬁrm-sales growth (Sales growth ): these results holds controlling for the realized growth of sales,

which turn out to be positively correlated with growth expectations.

Table 6: Competition from informal ﬁrms and expectations on future sales

Dependent variable: E(Sales growth) E(Sales growth)
Negative Stable Positive Continuous
(1) (2) (3) (4) (5)
Constrained by informalt−1 0.0412*** 0.0509*** -0.0921*** -4.103*** -2.812**
[0.0122] [0.0145] [0.0153] [0.735] [1.424]
Aget−1 -1.362** -2.094*** 3.456*** 81.53** 80.39
[0.616] [0.791] [0.864] [40.01] [71.88]
Sizet−1 -0.0251*** -0.0279*** 0.0531*** 1.507*** 1.608***
[0.00552] [0.00630] [0.00626] [0.299] [0.553]
Sales growtht−1 -0.00471*** -0.00120*** 0.00591*** 0.232*** 0.130***
[0.000370] [0.000420] [0.000435] [0.0194] [0.0352]
Exportt−1 -0.0451*** 0.000915 0.0442** 2.054** 2.169
[0.0169] [0.0201] [0.0208] [0.989] [1.756]
Sample Full Full Panel
Model Multinomial logit OLS OLS
Time FE yes yes yes
Geographic area FE yes yes yes
Additional controls yes yes yes
Observations 4313 4191 1265
Pseudo R2 (R2) 0.231 0.310 0.367

Notes: OLS estimates and multinomial logistic marginal eﬀects. The dependent variable in columns 1-3, we employ a categorical
variable taking value -1, 0, and +1 in case of negative, stable, and positive expected sales growth, respectively. In Column
4, we employ a continuous measure for ﬁrms’ expected sales growth in the following year. Because expectations are forward
looking, all regressors are simultaneous. Unreported controls follow the speciﬁcation in Table 4. Variables are deﬁned in Table
B4. Robust standard errors in brackets. *, **, *** indicate statistical signiﬁcance at the 10%, 5%, and 1%, respectively.

As a ﬁnal step in our argument, we explore the relation between future sales expectations and loan

application. In Table 7, columns 1 and 2 consider this link using the categorical version of the the expected

sales growth variable. Column 1 indicates that - one controlled for ﬁrm size - better expected sales growth

does not seems to increase the probability of loan application. Yet, when expectations are interacted with

the ﬁrm size, a diﬀerent pattern emerges. As shown in column 2, having positive sales growth expectation

increases the probability of loan application even though the eﬀect decreases with the size of the ﬁrm. Since

ﬁrm size is positively correlated with loan application, this suggests that the role of expectations is important

but becomes less important for ﬁrms which have characteristics which make them more likely to need a loan.

These same ﬁndings are conﬁrmed when we use the continuous variable for measuring expected sales growth

(see column 3 and 4).17

This ﬁnding adds to previous evidence in the literature on the eﬀect of informal competition on formal
16 Expectations are speciﬁc to the ﬁrm and unknown to the bank. Controlling for ﬁrms’ fundamentals and realized growth

16
Table 7: Expectations on future sales and loan application

Dependent variable: Loan application
(1) (2) (3) (4)
E(Sales growth): Stable -0.0343 -0.00274
[0.0352] [0.0929]
E(Sales growth): Stable × Size -0.00864
[0.0258]
E(Sales growth): Positive -0.0165 0.145*
[0.0333] [0.0784]
E(Sales growth): Positive × Size -0.0462**
[0.0210]
E(Sales growth) -0.0393 0.225*
[0.0531] [0.121]
E(Sales growth)× Size -0.0745**
[0.0308]
Size 0.0319*** 0.0623*** 0.0328*** 0.0346***
[0.00943] [0.0181] [0.00939] [0.00941]
Model Logit Logit Logit Logit
Time FE yes yes yes yes
Geographic area FE yes yes yes yes
Additional controls yes yes yes yes
Observations 1263 1263 1263 1263
Pseudo R2 0.189 0.189 0.189 0.193

Notes: logit marginal eﬀects. The dependent variable is a dummy measure for loan application in the last year. Because of
data availability, expectations refer to sales growth for the following year. Other regressors are timed consistently with previous
analyses. Unreported controls follow the speciﬁcation in Table 4. Variables are deﬁned in Table B4. Robust standard errors in
brackets. *, **, *** indicate statistical signiﬁcance at the 10%, 5%, and 1%, respectively.

ﬁrms. While in general increased competition between ﬁrms is expected to be welfare enhancing, this is

not obvious when competition is between formal and informal ﬁrms. In this case, competition may indeed

be detrimental to formal ﬁrms and to the overall economy (Friesen and Wacker, 2019). As noted by (Rozo

and Winkler, 2021), a larger informal sector may end up hurting formal ﬁrms’ performance in several ways,

including (unfair) cost competition and by reducing the quality of local public goods. Our results indicate

another negative eﬀect due to informal competition: by worsening (sales growth) expectations it reduces the

demand for credit by the ﬁrm.

Taken together, these results suggest that a mechanism explaining the poor employment performance of

formal ﬁrms in MENA countries is due to the negative eﬀect of informal competition on access to ﬁnance,

the latter being an important determinant of employment creation.

we make sure that we are not capturing a spurious relationship.
17 Interestingly, our results are in line with previous analyses of the relationship between competition from informal ﬁrms
and ﬁrm size. La Porta and Shleifer (2014) highlights that ﬁrms perceive the threat from informal ﬁrms as a diﬀerent obstacle
to their business depending on ﬁrm characteristics. Gonzalez and Lamanna (2007) show that the formal ﬁrms most aﬀected by
informal competition are those that resemble informal ﬁrms the most. Our results conﬁrm these predictions. By showing that
the induced reduction in loan application due to lower expected sales growth is diﬀerentially larger for small ﬁrms, we provide
evidence that the negative eﬀect of informal competition varies depending on ﬁrm’s characteristics and it i is more negative for
smaller ﬁrms, i.e. those more similar to informal ﬁrms.

17
4 Concluding remarks

This paper explores the link between jobs, ﬁnance, and exposure to informal competition for ﬁrms in MENA

countries. Using longitudinal data from the WBES survey, we document that job creation is facilitated by

access to ﬁnance, which - in turn - is reduced when the ﬁrm is exposed to competition from informal ﬁrms.

We provide suggestive evidence that a possible mechanism explaining this result is that ﬁrms that are more

exposed to competition from informal ﬁrms have worsen expectations for future sales growth that in turn

make them less likely to apply for a loan, depressing job creation.

Our results have some clear policy implications. The ﬁrst concerns which policy to implement to favour

employment creation in MENA countries. Because job creation is inﬂuenced by the functioning of the ﬁ-

nance sector, our ﬁndings indicate that reducing the disconnectedness which characterizes ﬁrms in MENA

countries is a possible strategy to foster employment growth. Our results provide a novel demand-side view

on the determinants of disconnectedness, documenting that the presence of informal competition negatively

inﬂuences the demand for ﬁnance. This implies that increasing the supply of credit and easing the access to

credit should not be considered the only possible strategies to increase the use of ﬁnance by ﬁrms in MENA

countries. Second, our results provide a novel motivation for reducing informality. Reducing informality

would beneﬁt the overall economy by inducing formal ﬁrms to increase loan applications and thus employ-

ment in response to a lower competition threat from informal ﬁrms. Third, our results show that policy

interventions to support ﬁrms should be designed taking into account that the perception of a constraint

is as important as the existence of an actual constraint in driving ﬁrms’ behavior. As we document in our

analysis, a ﬁrm’s decision not to apply for a loan is inﬂuenced by the perceived competition threat from

informal ﬁrms, which is not necessarily correlated with an existing actual threat.

References

Abadie, A. and G. W. Imbens (2011). Bias-corrected matching estimators for average treatment eﬀects.

Journal of Business & Economic Statistics 29 (1), 1–11.

Amin, M. (2021). Does competition from informal ﬁrms hurt job creation by formal ﬁrms? evidence using

ﬁrm-level survey data. World Bank Policy Research Working Paper 9515.

Amin, M. and C. Okou (2020). Casting a shadow: Productivity of formal ﬁrms and informality. Review of

Development Economics, 24 (4), 1610–1630.

18
Amin, M. and Y. C. Soh (2021). Does greater regulatory burden lead to more corruption? evidence using

ﬁrm-level survey data for developing countries. The World Bank Economic Review 35 (3), 812–828.

Avenyo, E. K., M. Konte, and P. Mohnen (2021). Product innovation and informal market competition in

sub-saharan africa. Journal of Evolutionary Economics, 84 (C), 605–637.

Ayyagari, M., P. Juarros, M. S. Martinez Peria, and S. Singh (2021). Access to ﬁnance and job growth:

ﬁrm-level evidence across developing countries. Review of Finance, forthcoming.

Banerjee, A. and E. Duﬂo (2014). Do ﬁrms want to borrow more? testing credit constraints using a directed

lending program. Review of Economic Studies, 81, 572–607.

Banerjee, A., E. Duﬂo, R. Glennerster, and C. Kinnan (2015). The miracle of microﬁnance? evidence from

a randomized evaluation. American Economic Journal: Applied Economics 7 (1), 22–53.

Banerji, A. and S. Jain (2007). Quality dualism. Journal of Development Economics 84 (1), 234–250.

u¸
Beck, T., A. Demirg¨ c-Kunt, and V. Maksimovic (2005). Financial and legal constraints to growth: Does

ﬁrm size matter? The Journal of Finance 60 (1), 137–177.

u¸
Beck, T., A. Demirg¨ c-Kunt, and V. Maksimovic (2008). Financing patterns around the world: Are small

ﬁrms diﬀerent? Journal of Financial Economics 89 (3), 467–487.

Betz, F. and F. Ravasan (2016). Collateral regimes and missing jobcreation in the mena region.

Bigsten, A., P. Collier, S. Dercon, M. Fafchamps, B. Gauthier, J. W. Gunning, A. Oduro, R. Oostendorp,

oderbom, F. Teal, and A. Zeufack (2003). Credit Constraints in Manufacturing Enterprises
C. Patillo, M. S¨

in Africa. Journal of African Economies 12 (1), 104–125.

Boneva, L., J. Cloyne, M. Weale, and T. Wieladek (2020). Firms’ price, cost and activity expectations:

Evidence from micro data. Economic Journal 130 (627), 555–586.

Coibion, O., Y. Gorodnichenko, and S. Kumar (2018). How do ﬁrms form their expectations? New survey

evidence. American Economic Review 108 (9), 2671–2713.

Coibion, O., Y. Gorodnichenko, S. Kumar, and M. Pedemonte (2020). Inﬂation expectations as a policy

tool? Journal of International Economics 124, 103297.

Coibion, O., Y. Gorodnichenko, and T. Ropele (2020). Inﬂation expectations and ﬁrm decisions: New causal

evidence. Quarterly Journal of Economics 135 (1), 165–219.

19
e (2015). Estimating the impact of microcredit on those who
Crepon, B., F. Devoto, E. Duﬂo, and W. Parient´

take it up: Evidence from a randomized experiment in morocco. American Economic Journal: Applied

Economics 7 (1), 123–50.

De Lima, P., D. Revoltella, J. L. Rodriguez Meza, and H. Schweiger (2016). What’s holding back the private

sector in MENA? Lessons from the enterprise survey. World Bank.

de Mel, S., D. McKenzie, and C. Woodruﬀ (2008). Returns to Capital in Microenterprises: Evidence from

a Field Experiment. The Quarterly Journal of Economics 123 (4), 1329–1372.

Distinguin, I., C. Rugemintwari, and R. Tacneng (2016). Can informal ﬁrms hurt registered smes’ access to

credit? World Development 84, 18–40.

Dollar, D., M. Hallward-Driemeier, and T. Mengistae (2006). Investment climate and international integra-

tion. World development 34 (9), 1498–1516.

unnekes, and G. J. M¨
Enders, Z., F. H¨ uller (2019a). Firm expectations and economic activity. CESifo

Working Paper 7623.

unnekes, and G. J. M¨
Enders, Z., F. H¨ uller (2019b). Monetary policy announcements and expectations:

Evidence from German ﬁrms. Journal of Monetary Economics 108, 45–63.

Fisman, R. and J. Svensson (2007). Are corruption and taxation really harmful to growth? ﬁrm level

evidence. Journal of development economics 83 (1), 63–75.

Friesen, J. and K. Wacker (2019). Who is afraid of informal competition? the role of ﬁnance for ﬁrms in

developing and emerging economies. The European Journal of Development Research 108 (31), 1126–1146.

Gennaioli, N., Y. Ma, and A. Shleifer (2016). Expectations and investment. NBER Macroeconomics An-

nual 30 (1), 379–431.

Gonzalez, A., J. E. Lopez-Cordova, and E. E. Valladares (2007). The incidence of graft on developing-country

ﬁrms. World Bank Policy Research Working Paper 4394.

Gonzalez, A. S. and F. Lamanna (2007). Who fears competition from informal ﬁrms ? evidence from latin

america. World Bank Policy Research Working Paper 4316.

Gronau, R. (1974). Wage comparisons–a selectivity bias. Journal of political Economy 82 (6), 1119–1143.

Heckman, J. J. (1976). The common structure of statistical models of truncation, sample selection and

limited dependent variables and a simple estimator for such models. Annals of economic and social

measurement 5 (4), 475–492.

20
Kersten, R., J. Harms, K. Liket, and K. Maas (2017). Small ﬁrms, large impact? a systematic review of the

sme ﬁnance literature. World Development 97, 330–348.

ıguez-Meza, and J. S. Yang (2014). What have we learned from the
Kuntchev, V., R. Ramalho, J. Rodr´

enterprise surveys regarding access to credit by smes? World Bank Policy Research Working Paper 6670.

La Porta, R. and A. Shleifer (2014). Informality and development. Journal of Economic Perspectives, 28 (3),

109–126.

Lewis, H. G. (1974). Comments on selectivity biases in wage comparisons. Journal of Political Econ-

omy 82 (6), 1145–1155.

Maloney, W. (2004). Informality revisited. World Development, 32 (7), 1159–1178.

Perry, G., W. Maloney, O. Arias, P. Fajnzylber, A. Mason, and J. Saavedra-Chanduvi (2007). Informality:

Exit and Exclusion. World Bank, Washington, DC.

Quinn, S. and C. Woodruﬀ (2019). Experiments and entrepreneurship in developing countries. Annual

Review of Economics 11 (1), 225–248.

Rozo, S. and H. Winkler (2021). Is informality good for business? the impacts of idp inﬂows on formal ﬁrms.

Journal of Human Resources, forthcoming.

Tanaka, M., N. Bloom, J. M. David, and M. Koga (2020). Firm performance and macro forecast accuracy.

Journal of Monetary Economics 114, 26–41.

Ulyssea, G. (2018). Firms, informality, and development: Theory and evidence from brazil. American

Economic Review 108, 2015–47.

21
Online Appendix

Table A1: Attrition

Dependent variable: Panel
(1) (2) (3) (4) (5)
Size 0.000673 -0.00199 0.000206 -0.000120 0.00131
[0.00568] [0.00633] [0.00588] [0.00574] [0.00582]
Age 0.0242*** 0.0317*** 0.0265*** 0.0240*** 0.0232***
[0.00792] [0.00862] [0.00807] [0.00792] [0.00799]
Export 0.0137 0.0198 0.0112 0.0135 0.0143
[0.0166] [0.0181] [0.0170] [0.0166] [0.0167]
Manufacturing 0.148 0.286 0.149 0.151 0.150
[0.153] [0.273] [0.152] [0.153] [0.153]
Retail 0.213 0.346 0.213 0.215 0.216
[0.154] [0.274] [0.154] [0.154] [0.154]
Other services 0.213 0.341 0.213 0.216 0.215
[0.152] [0.272] [0.152] [0.152] [0.153]
Listed company -0.00170 0.000831 0.00520 -0.000125 -0.0104
[0.0637] [0.0773] [0.0651] [0.0637] [0.0643]
LLC -0.0169 0.00524 -0.00870 -0.0153 -0.0217
[0.0611] [0.0740] [0.0624] [0.0611] [0.0616]
Sole proprietorship -0.0379 -0.00927 -0.0290 -0.0355 -0.0433
[0.0611] [0.0739] [0.0625] [0.0611] [0.0616]
Partnership -0.0158 0.0128 -0.00760 -0.0137 -0.0203
[0.0617] [0.0745] [0.0631] [0.0617] [0.0623]
Ltd Partnership -0.0378 -0.00804 -0.0278 -0.0360 -0.0428
[0.0620] [0.0747] [0.0634] [0.0620] [0.0626]
Constrained by informal -0.00239
[0.0154]
Loan availability -0.00155
[0.0161]
Loan application 0.0141
[0.0140]
Investment 0.000102
[0.0144]
Model Logit Logit Logit Logit Logit
Geographic area FE yes yes yes yes yes
Observations 5219 4347 5063 5219 5181
Pseudo R2 0.0328 0.0367 0.0331 0.0328 0.0327

Notes: logit marginal eﬀects. The estimating sample is composed of the entire set of ﬁrms interviewed in the 2013-wave of the
WBES. The dependent variable is a dummy taking value of 1 if the ﬁrm is included in our panel estimation (i.e., it is interviewed
in the following wave), and 0 otherwise. All regressors are timed at the beginning of period. Measures are deﬁned in Table B4.
Robust standard errors in brackets. *, **, *** indicate statistical signiﬁcance at the 10%, 5%, and 1%, respectively.

22
Table A2: Loan availability and loan application: Restricting the sample to ﬁrms with no loan in t − 1

Dependent variable: Loan availability Loan application
(1) (2)
Constrained by informal -0.0714*** -0.0831***
[0.0269] [0.0278]
Account -0.0285 -0.00862
[0.0250] [0.0283]
No need -0.00700 -0.0163
[0.0223] [0.0246]
Age -0.00498 0.00962
[0.0158] [0.0180]
Size 0.0277*** 0.0231**
[0.00847] [0.00995]
Export 0.0356 0.0452
[0.0293] [0.0375]
N competitors -0.00816 0.00353
[0.00604] [0.00722]
Model Logit Logit
Time FE yes yes
Geographic area FE yes yes
Additional controls yes yes
Observations 1002 1067
Pseudo R2 0.162 0.220

Notes: logit marginal eﬀects. This table replicates the analysis in Column 3 of Tables 4 and 5, while restricting the sample to
ﬁrms with no access to credit in the previous wave (i.e., switchers only). All regressors are lagged once. Unreported controls
follow the speciﬁcation in Table 4. Measures are deﬁned in Table B4. Robust standard errors in brackets. *, **, *** indicate
statistical signiﬁcance at the 10%, 5%, and 1%, respectively.

Table A3: Loan availability: Restricting the timing of the issuance

Dependent variable: Loan availability
Issuance: 10 years 7 years 5 years 2 years 1 year
(1) (2) (3) (4) (5)
Constrained by informal -0.0764*** -0.0730*** -0.0690*** -0.0582*** -0.0663***
[0.0222] [0.0219] [0.0216] [0.0203] [0.0254]
Account 0.00839 0.00602 0.00274 0.0215 0.0457
[0.0279] [0.0280] [0.0281] [0.0283] [0.0388]
No need -0.0464** -0.0462** -0.0487*** -0.0439** -0.0599**
[0.0192] [0.0191] [0.0188] [0.0183] [0.0233]
Age -0.00655 -0.00700 -0.00600 -0.00281 0.0135
[0.0157] [0.0156] [0.0156] [0.0148] [0.0198]
Size 0.0382*** 0.0374*** 0.0364*** 0.0289*** 0.0235**
[0.00792] [0.00779] [0.00768] [0.00734] [0.00962]
Export 0.0195 0.0222 0.0154 0.0103 0.0265
[0.0259] [0.0254] [0.0252] [0.0237] [0.0304]
N competitors -0.00520 -0.00621 -0.00683 -0.00548 -0.00959
[0.00558] [0.00549] [0.00537] [0.00509] [0.00650]
Model Logit Logit Logit Logit Logit
Time FE yes yes yes yes yes
Geographic area FE yes yes yes yes yes
Additional controls yes yes yes yes yes
Observations 1371 1362 1351 1277 906
Pseudo R2 0.197 0.202 0.199 0.227 0.209

Notes: logit marginal eﬀects. This table replicates the analysis in Column 3 of Table 4, while restricting the availability of
loans to an issuance occurring within the last 10, 7, 5, or 2 years (respectively in columns 1, 2, 3, and 4). All regressors are
lagged once. Unreported controls follow the speciﬁcation in Table 4. Measures are deﬁned in Table B4. Robust standard errors
in brackets. *, **, *** indicate statistical signiﬁcance at the 10%, 5%, and 1%, respectively.

23
Table A4: Heckman selection model

Dependent variable: Loan availability Loan application
(1) (2)
Constrained by informal -0.0566*** -0.0534**
[0.0187] [0.0210]
Account 0.0155 0.00799
[0.0221] [0.0247]
No need -0.0495*** -0.0413**
[0.0177] [0.0198]
Age -0.0121 -0.00371
[0.0129] [0.0145]
Size 0.0342*** 0.0373***
[0.00773] [0.00864]
Export 0.0394* 0.0405
[0.0230] [0.0256]
Manufacturing 0.0346* 0.0469**
[0.0191] [0.0214]
Retail 0.00516 -0.0176
[0.0303] [0.0336]
Listed company 0.0491 0.0731
[0.183] [0.208]
LLC 0.0559 0.103
[0.182] [0.206]
Sole proprietorship 0.0229 0.0820
[0.181] [0.205]
Partnership 0.0328 0.0878
[0.181] [0.206]
Ltd Partnership 0.0284 0.0884
[0.182] [0.206]
Model Heckman Heckman
Time FE yes yes
Geographic area FE yes yes
Additional controls yes yes
Observations 10386 10444
Selected 1911 1969
Not selected 8475 8475
Wald χ2 296.31 498.86
Inverse Mill’s ratio -0.0309* -0.0308

Notes: Heckman selection model. In this table, we explicitly model the probability of being included in our analysis in a
two-step Heckman-type selection model (Heckman, 1976; Lewis, 1974; Gronau, 1974). The selection equation models a ﬁrms’
probability of belonging to the panel (i.e., being interviewed in two consecutive waves of the WBES survey) depending on ﬁrms’
age, size, and belonging stratum (the intersection of sector and country, excluded in the main speciﬁcation). The inverse Mill’s
ratio is included as an additional regressor in the original speciﬁcation (reported in the bottom panel). All regressors are timed
consistently with previous analyses. Unreported controls follow the speciﬁcation in Table 4. Measures are deﬁned in Table B4.
Robust standard errors in brackets. *, **, *** indicate statistical signiﬁcance at the 10%, 5%, and 1%, respectively.

24
Table A5: Loan availability: Additional controls

Dependent variable: Loan availability
(1) (2) (3) (4) (5)
Constrained by informal -0.0720*** -0.0723*** -0.0714*** -0.0706*** -0.124***
[0.0193] [0.0222] [0.0224] [0.0226] [0.0368]
Years formality -0.0453 -0.0660 -0.0726 -0.0184 -0.158
[0.0699] [0.0775] [0.0782] [0.0945] [0.235]
Originally informal 0.0110 -0.0125 -0.00828 -0.00496 0.0147
[0.0290] [0.0340] [0.0341] [0.0354] [0.0548]
Sales growth 0.00131 0.00113 0.000964 -0.000519
[0.000861] [0.000883] [0.000890] [0.00141]
Productivity growth -0.00142* -0.00125 -0.00106 0.000286
[0.000851] [0.000875] [0.000879] [0.00137]
Investment -0.0213 -0.0175 -0.0172 -0.0534
[0.0212] [0.0213] [0.0215] [0.0353]
Local market 0.00185 -0.000510 -0.0401
[0.0322] [0.0326] [0.0552]
National market -0.00431 -0.000114 -0.0165
[0.0274] [0.0277] [0.0488]
Years manager experience 0.000594 0.000270 0.000942
[0.000896] [0.000917] [0.00146]
Government ownership -0.00112 -0.000787 0
[0.00128] [0.00126] [.]
Female top manager -0.00299 -0.0121 -0.0392
[0.0430] [0.0439] [0.0576]
Female owner -0.00838 -0.0113 -0.00817
[0.0223] [0.0232] [0.0356]
City 1 0.0763 0.00338
[0.0684] [0.0934]
City 2 0.129* 0.0707
[0.0660] [0.0907]
City 3 0.0881 -0.0251
[0.0651] [0.0847]
City 4 0.0872 -0.0354
[0.0645] [0.0879]
Electric outages (N) 0.000427 -0.000343
[0.000297] [0.000714]
Electric outages (length) 0.00394 0.0135
[0.00492] [0.00863]
Bribery depth -0.00111**
[0.000498]
Gifts (share) 0.000534
[0.000471]
Constrained by corruption 0.0518
[0.0387]
Loss from theft -0.00124
[0.00540]
Constrained by crime -0.000215
[0.000480]
Capacity utilization -0.0000916
[0.000737]
Model Logit Logit Logit Logit Logit
Time FE yes yes yes yes yes
Geographic area FE yes yes yes yes yes
Additional controls yes yes yes yes yes
Observations 1914 1411 1388 1359 527
Pseudo R2 0.187 0.216 0.215 0.219 0.273

Notes: logit marginal eﬀects. All regressors are lagged once. Unreported additional regressors follow the speciﬁcation in Column
3 of Table 4. Measures are deﬁned in Table B4. Robust standard errors in brackets. *, **, *** indicate statistical signiﬁcance
at the 10%, 5%, and 1%, respectively.

25
Table A6: Loan application: Additional controls

Dependent variable: Loan application
(1) (2) (3) (4) (5)
Constrained by informal -0.0506** -0.0400* -0.0415* -0.0413* -0.117***
[0.0209] [0.0242] [0.0248] [0.0250] [0.0412]
Years formality -0.0945 -0.122 -0.124 -0.0890 0.0505
[0.0745] [0.0900] [0.0919] [0.100] [0.244]
Originally informal 0.0347 0.0141 0.00520 0.00511 0.00398
[0.0325] [0.0387] [0.0391] [0.0402] [0.0609]
Sales growth 0.000866 0.000669 0.000774 -0.000582
[0.000922] [0.000962] [0.000968] [0.00165]
Productivity growth -0.00152* -0.00129 -0.00136 -0.000467
[0.000888] [0.000928] [0.000927] [0.00152]
Investment -0.0375 -0.0354 -0.0349 -0.0654
[0.0242] [0.0247] [0.0250] [0.0439]
Local market -0.0122 -0.0195 -0.0146
[0.0369] [0.0371] [0.0709]
LMain national -0.00123 -0.00419 0.0182
[0.0322] [0.0324] [0.0649]
Years manager experience -0.000142 -0.000413 0.0000948
[0.00102] [0.00102] [0.00169]
Government ownership 0.000321 0.000592 -0.000388
[0.00177] [0.00179] [0.00244]
Female top manager 0.0545 0.0313 0.0311
[0.0506] [0.0494] [0.0702]
Female owner -0.00986 -0.00993 -0.0114
[0.0254] [0.0261] [0.0410]
City 1 0.0373 -0.0764
[0.0689] [0.105]
City 2 0.119* 0.0429
[0.0647] [0.0983]
City 3 0.0709 -0.0519
[0.0612] [0.0848]
City 4 0.107* -0.0106
[0.0582] [0.0881]
Electric outages (N) 0.000673* 0.00175
[0.000393] [0.00177]
Electric outages (length) 0.00301 -0.00419
[0.00667] [0.0132]
Bribery depth -0.00145**
[0.000650]
Gifts (share) -0.0000411
[0.000553]
Constrained by corruption -0.0217
[0.0422]
Loss from theft -0.000938
[0.00632]
Constrained by crime 0.000290
[0.000525]
Capacity utilization -0.000147
[0.000782]
Model Logit Logit Logit Logit Logit
Time FE yes yes yes yes yes
Geographic area FE yes yes yes yes yes
Additional controls yes yes yes yes yes
Observations 1987 1464 1439 1409 561
Pseudo R2 0.196 0.213 0.206 0.211 0.199

Notes: logit marginal eﬀects. All regressors are lagged once. Unreported additional regressors follow the speciﬁcation in Column
3 of Table 5. Measures are deﬁned in Table B4. Robust standard errors in brackets. *, **, *** indicate statistical signiﬁcance
at the 10%, 5%, and 1%, respectively.

26
Table A7: IV estimates

Dependent variable: Loan availability Loan application
(1) (2)
Constrained by informal -0.590** -0.614**
[0.273] [0.307]
Account 0.0431 0.0390
[0.0281] [0.0308]
No need -0.108*** -0.109**
[0.0392] [0.0440]
Age 0.0244 0.0282
[0.0181] [0.0203]
Size 0.0174 0.0153
[0.0119] [0.0133]
Export 0.00551 0.00503
[0.0274] [0.0302]
Model 2SLS 2SLS
Time FE yes yes
Geographic area FE yes yes
Additional controls yes yes
Observations 1951 2011
Underidentiﬁcation (p-value) 0.000 0.000

Notes: 2SLS estimates. In this table, we instrument Constrained by informal with the average of ﬁrms’ belonging stratum
deﬁned at the intersection of macro-sector (manufacturing vs services), geographical area, in the previous wave of the ﬁrm. All
regressors are timed consistently with previous analyses. Unreported controls follow the speciﬁcation in Table 4. Measures are
deﬁned in Table B4. Robust standard errors in brackets. *, **, *** indicate statistical signiﬁcance at the 10%, 5%, and 1%,
respectively.

27
Table A8: Balancing properties of the matching

Unmatched Mean % Reduct. t-test
Variable Matched Treated Control % Bias Bias t p >| t |
U 7.5972 7.5971 1.1 – 0.16 0.876
Age
M 7.5972 7.5961 12.4 -990.8 1.27 0.205
U 3.1893 3.5172 -24.3 – -3.27 0.001
Size
M 3.1889 3.1367 3.9 84.1 0.49 0.621
U 0.22887 0.2338 -1.2 – -0.17 0.868
Export
M 0.23485 0.20833 6.3 -437.9 0.73 0.464
U 0.88732 0.80986 21.7 – 2.96 0.003
Account
M 0.87879 0.84091 10.6 51.1 1.25 0.211
U 0.54577 0.64366 -20.0 – -2.88 0.004
No need
M 0.56818 0.60227 -7.0 65.2 -0.79 0.428
U 0.38732 0.66338 -57.5 – -8.24 0.000
Egypt
M 0.41667 0.41288 0.8 98.6 0.09 0.930
U 0.04577 0.03099 7.7 – 1.14 0.254
Jordan
M 0.04924 0.04924 0.0 100.0 0.00 1.000
U 0.23239 0.10986 32.9 – 5.02 0.000
Lebanon
M 0.24242 0.24242 0.0 100.0 0.00 1.000
U 0.07394 0.0493 10.2 – 1.52 0.128
Morocco
M 0.06439 0.05682 3.2 69.3 0.36 0.716
U 0.03873 0.06056 -10.1 – -1.37 0.170
State of Palestine
M 0.04167 0.04545 -1.7 82.6 -0.21 0.832
U 0.22183 0.08592 38.3 – 5.96 0.000
Tunisia
M 0.18561 0.19318 -2.1 94.4 -0.22 0.825
U 0.59155 0.59155 0.0 – 0.000 1.000
Manufacturing
M 0.61364 0.54167 14.6 – -0.82 0.094
U 0.09859 0.07042 10.1 – 1.49 0.136
Retail
M 0.0947 0.08712 2.7 73.1 0.30 0.763
U 0.30986 0.33803 -6.0 – -0.85 0.116
Other services
M 0.29167 0.37121 -17.0 -182.4 -1.9 0.112
U 0.25 0.23803 2.8 – 0.40 0.691
LLC
M 0.25 0.25758 -1.8 36.7 -0.20 0.842
U 0.34155 0.38028 -8.1 – -1.14 0.253
Sole proprietorship
M 0.33333 0.375 -8.7 -7.6 -1.00 0.318
U 0.16901 0.14507 6.6 – 0.95 0.343
Partnership
M 0.16667 0.1553 3.1 52.5 0.35 0.723
U 0.16197 0.14085 5.9 – 0.85 0.396
Ltd Partnership
M 0.17045 0.17045 0.0 100.0 0.00 1.000
U 0.84155 0.89296 -15.2 – -2.24 0.025
Originally informal
M 0.83712 0.81818 5.6 63.2 0.58 0.565
U -3.4194 -4.5228 5.9 – 0.82 0.414
Sales growth
M -3.7487 -3.9356 1.0 83.1 0.13 0.899
U 0.39789 0.45493 -11.5 – -1.64 0.102
Local market
M 0.39394 0.37121 4.6 60.2 0.54 0.592
U 0.52817 0.44366 16.9 – 2.42 0.016
National market
M 0.52652 0.56439 -7.6 55.2 -0.87 0.383
U 0.5493 0.61408 -13.1 – -1.88 0.060
Board of directors
M 0.53788 0.56439 -5.4 59.1 -0.61 0.541
U 25.884 23.604 19.5 – 2.76 0.006
Years manager experience
M 25.867 25.367 4.3 78.1 0.49 0.622
U 0.94014 0.37183 8.7 – 1.43 0.154
Government ownership
M 0.35606 0.125 3.6 59.3 0.89 0.373
U 0.05282 0.04507 3.6 – 0.52 0.604
Female top manager
M 0.05303 0.06061 -3.5 2.2 -0.38 0.708
U 0.23944 0.1169 32.4 – 4.92 0.000
Female owner
M 0.22348 0.24621 -6.0 81.5 -0.62 0.539
U 0.14437 0.09296 15.9 – 2.37 0.018
City 1
M 0.14773 0.1553 -2.3 85.3 -0.24 0.809
U 0.29225 0.19155 23.6 – 3.48 0.001
City 2
M 0.2803 0.32955 -11.6 51.1 -1.23 0.220
U 0.17958 0.14507 9.4 – 1.36 0.175
City 3
M 0.17424 0.19318 -5.1 45.1 -0.56 0.575
U 0.30282 0.53239 -47.8 – -6.69 0.000
City 4
M 0.31818 0.26894 10.3 78.6 1.24 0.215
U 0.61972 0.35915 53.9 – 7.70 0.000
Constr. corruption
M 0.59848 0.60985 -2.4 95.6 -0.27 0.790
U 22.535 7.4648 43.1 – 6.81 0.000
Constr. crime
M 17.424 17.424 0.0 100.0 -0.00 1.000

Notes: Balancing properties from radius matching (0.2 stdev) in Table A9.

28
Table A9: Matching estimator: Average Treatment Eﬀect

Abadie and Imbens (2002) estimator Radius Matching (0.2 stdev)

Outcome variable: Loan availability Loan application Loan availability Loan application
(1) (2) (3) (4)
Constrained by informal -0.0845*** -0.0819*** -0.0434** -0.0502**
[0.0253] [0.0290] [0.0185] [0.0219]

Notes: Average Treatment Eﬀects for Constrained by informal (i.e., our treatment variable). In the left panel, we perform the
Abadie and Imbens (2011) estimator, while in the right panel, we employ radius matching with a 0.2-stdev caliper. Balancing
properties are provided in Table A8 of the Online Appendix. All regressors are timed consistently with previous analyses.
Measures are deﬁned in Table B4. Robust standard errors in brackets. *, **, *** indicate statistical signiﬁcance at the 10%,
5%, and 1%, respectively.

29
Data Appendix

Table B1: Deﬁnition: Dependent variables
Variable name Deﬁnition
question k8: “At this time, does this establishment have a line of credit or a loan from a ﬁnancial
Loan availability
institution?”. Loan availability=1 if k8=yes and 0 otherwise.
if k8=no, question bmk7: “What is the reason for not having a loan or line of credit at the
Loan application moment?”. Answer bmk7a: “Because this establishment did not apply for a loan or line of credit”.
Loan application=0 if bmk7a=yes and 1 otherwise (even if k8=yes).
question bmd1a: “Considering the next year, are this establishment’s total sales expected to in-
E(Sales growth): Positive
crease, decrease, or stay the same?”. Positive expectations=1 if bmd1a=“increase” and 0 otherwise.
E(Sales growth): Stable Stable expectations=1 if bmd1a=“stay the same” and 0 otherwise.
E(Sales growth): Negative Negative expectations=1 if bmd1a=“decrease” and 0 otherwise.
question bmd1b: “In percentage terms, what is the expected change in total sales?”.
E(Sales growth) E(Sales growth)=bmd1b if E(Sales growth): Positive=1, E(Sales growth)=–bmd1b if
E(Sales growth): Negative=1, and 0 otherwise.
if k8=no, question bmk7: “What is the reason for not having a loan or line of credit at the
Turned down moment?”. Answer bmk7b: “Because the last application for a loan or line of credit was turned
down”. Turned down=1 if bmk7b=yes and 0 otherwise.
variable constructed as in Kuntchev et al. (2014). Rationing=2 (fully constrained) if the ﬁrm does
not have external sources of ﬁnance and applied for a loan and was rejected (question bmk7b) or did
Rationing not apply because of the terms and conditions (question k17). Rationing=1 (partially constrained)
if the ﬁrm has external sources of ﬁnance and the loan was approved in part, it was rejected, or
because of the terms and conditions. Rationing=0 (not constrained) otherwise.

30
Table B2: Deﬁnition: Main regressors
Variable name Deﬁnition
question e30: “To what degree are practices of competitors in the informal sector an obstacle to
the current operations of this establishment?”. Available options: i. no obstacle, ii. minor obstacle,
Constrained by informal
iii. moderate obstacle, iv. major obstacle, or v. very severe obstacle. Constrained by informal=1
if e30= iv. or v., and 0 otherwise.
question b6a: “Was this establishment formally registered when it began operations?”. Originally
Originally informal
informal=1 if b6a=yes, and 0 otherwise.
question b6b: “In what year was this establishment formally registered?”. Years of
Years of formality
formality=ln(1+T–b6b), where T is the year of the survey.
question k6: “Now let’s talk about the establishment’s current situation. At this time, does this
Account
establishment have a checking or savings account?”. Account =1 if k6=yes and 0 otherwise.
question k16: “Referring again to the last ﬁscal year, did this establishment apply for any loans
or lines of credit?”. If k16=no, question k17: “What was the main reason why this establishment
No need
did not apply for any line of credit or loan?”, answer k17a: “No need for a loan – establishment
had suﬃcient capital”. No need=1 if k17a=yes, and 0 otherwise.
question b5: “In what year did this establishment begin operations?”. Age = ln(1+T–b5), where
Age
T is the year of the survey.
question l2 “Looking back, at the end of two ﬁscal years ago, how many permanent, full–time
Size individuals worked in this establishment? Please include all employees and managers”. Size=
ln(1+l2).
question d3: “Coming back to the last ﬁscal year, what percentage of this establishment’s sales
Export were: national sales [d3a], indirect exports (sold domestically to third party that exports products)
[d3b], direct exports [d3c]?”. Export=1 if d3c > 10%.
question d2: “In the last ﬁscal year, what were this establishment’s total annual sales for ALL
products and services?”. Question n3: “Three ﬁscal years ago, what were total annual sales for
Sales growth this establishment?”. Sales growth is measured as a percentage change in sales between the last
completed ﬁscal year and the previous period. All sales values are deﬂated to 2009 using each
country’s GDP deﬂators.
question e2: “In the last ﬁscal year, for the main market in which this establishment sold its main
product, how many competitors did this establishment’s main product face?”. The original answer
was a cardinal measure distinguishing the following classes: i. 0, ii. 1, iii. 2–3, iv. 4–5, v. 6–10, vi.
Number of competitors 11–180, or vii. too many to count. For conciseness, we generated a continuous measure by imposing
the median number of each class and assuming the lowerbound of 181 for the last category vii. We
then took the augmented log (1+). Our analysis is not sensitive to alternative choices or to the
direct use of the original categorical measure.

31
Table B3: Deﬁnition: Additional regressors
Variable name Deﬁnition
annual labor productivity growth is measured by a percentage change in labor productivity between
the last completed ﬁscal year and a previous period. Labor productivity is deﬁned as the ratio
Productivity growth
between sales and the number of full-time permanent workers. All sales values are deﬂated to 2009
using each country’s GDP deﬂators.
question k4: “In the last ﬁscal year, did this establishment purchase any new or used ﬁxed as-
Investment sets, such as machinery, vehicles, equipment, land or buildings?”. Investment=1 if k4=yes, and 0
otherwise.
question e1: “In the last ﬁscal year, which of the following was the main market in which this
establishment sold its main product?”. Available answers: i. Local (main product sold mostly in
Local market
same municipality where establishment is located), ii. National (main product sold mostly across
the country where establishment is located), and iii. International. Local market=1 if e1=i.
National market National market=1 if e1=ii.
question bmb4: “Does the ﬁrm have a board of directors or a supervisory board?”. Board of
Board of directors
directors=1 if bmb4=yes, and 0 otherwise.
question b7: “How many years of experience working in this sector does the top manager have?”.
Years manager experience
Years manager experience=log(1+b7).
question b2: “What percentage of this ﬁrm is owned by each of the following”. Government
Government ownership
ownership=b2c, “% Government or State”.
Female top manager question b7a: “Is the Top Manager female?”. Female top manager=1 if b7a=yes, and 0 otherwise.
question b4: “Amongst the owners of the ﬁrm, are there any females?”. Female owner=1 if b4=yes,
Female owner
and 0 otherwise.
question a3: “Size of locality”. Available answers: i. “City with population above 1 Million”, ii.
City 1 “Over 250.000 to 1 million”, iii. “50,000 to 250,000”, iv. “Less than 50,000”. City 1=1 if a3=iv,
and 0 otherwise.
City 2 City 2=1 if a3=iii, and 0 otherwise.
City 3 City 3=1 if a3=ii, and 0 otherwise.
City 4 City 4=1 if a3=i, and 0 otherwise.
question c7: “In a typical month, over the last ﬁscal year, how many power outages did this
Electric outages (N)
establishment experience?”. Electric outages (N)=log(1+c7).
question c8: “How long did these power outages last on average?”. Electric outages
Electric outages (lenght)
(lenght)=log(1+c8).
Bribery depth is computed similarly as the Graft Index from Gonzalez et al. (2007). it is con-
structed from the following questions. Question c5: “In reference to that application for an elec-
trical connection, was an informal gift or payment expected or requested?”. Question c14: “In
reference to that application for a water connection, was an informal gift or payment expected or
requested?”. Question g4: “In reference to that application for a construction-related permit, was
Bribery depth
an informal gift or payment expected or requested?”. Question j5: “In any of these inspections or
meetings (with tax oﬃcials) was a gift or informal payment expected or requested?”. Question j12:
“In reference to that application for an import license, was an informal gift or payment expected
or requested?”. Question j15: “In reference to that application for an operating license, was an
informal gift or payment expected or requested?”.
question j7: “It is said that establishments are sometimes required to make gifts or informal
payments to public oﬃcials to “get things done” with regard to customs, taxes, licenses, regulations,
Gifts (share)
services etc. On average, what percentage of total annual sales do establishments like this one pay
in informal payments or gifts to public oﬃcials for this purpose?”.
question j30: “As I list some factors that can aﬀect the current operations of a business, please
look at this card and tell me the degree to which you think each factor is an obstacle to the current
Constrained by corruption operations of this establishment”. Available options: i. no obstacle, ii. minor obstacle, iii. moderate
obstacle, iv. major obstacle, or v. very severe obstacle. Constr. corruption=1 if j30= iv. or v., and
0 otherwise.
question i4: “In the last ﬁscal year, what were the estimated losses as a result of theft, robbery,
Loss from theft vandalism or arson that occurred on this establishment’s premises either as a percentage of total
annual sales?”.
question i30: “To what degree is Crime, Theft and Disorder an obstacle to the current operations
Constrained by crime of this establishment?”. Available options: i. no obstacle, ii. minor obstacle, iii. moderate obstacle,
iv. major obstacle, or v. very severe obstacle. Constr. crime=1 if i30= iv. or v., and 0 otherwise.
question f1: “In the last ﬁscal year, what was this establishment’s output produced as a percentage
Capacity utilization
of the maximum output possible if using all the resources available (capacity utilization)?”.

32
Table B4: Variable description: Main measures
Variable name Description
Dependent variables
Loan availability dummy for ﬁrms with an outstanding loan or credit line.
Loan application dummy for ﬁrms that applied for a loan or credit line (independently of the outcome).

E(Sales growth) continuous measure for ﬁrms’ expected sales growth over the following year.
E(Sales growth): Positive dummy for ﬁrms expecting increasing sales in the following year.
E(Sales growth): Stable dummy for ﬁrms expecting stable sales in the following year.
Regressors
Account dummy for ﬁrms with a checking or savings account.
Constrained by informal dummy for ﬁrms identifying practices of competitors in the informal sector as a major constraint.
Originally informal dummy for ﬁrms originally starting their activity without being formally registered.
Years of formality log–years since the ﬁrm was formally registered.
Age log–age (1+).
Size log–employees (1+).
Export dummy for exporting ﬁrms.
Sales growth realized sales growth over the last three years.
Number of competitors log–number of competitors (1+).
Manufacturing dummy for ﬁrms operating in the manufacturing sector.
Retail dummy for ﬁrms operating in the retail sector.
Listed company dummy for listed companies.
LLC dummy for LLC ﬁrms.
Sole proprietorship dummy for sole proprietorship ﬁrms.
Partnership dummy for partnership ﬁrms.
Ltd Partnership dummy for Ltd partnership ﬁrms.

Table B5: Variable description: additional regressors
Variable name Description
Productivity growth annual labor productivity growth.
Local market dummy for ﬁrms mainly selling products to local markets.
National market dummy for ﬁrms mainly selling products to national markets.
Board of directors dummy for ﬁrms having a board of directors or a supervisory board.
Years manager experience number of years of experience of the manager (in log).
Government ownership share of the ﬁrm owned by the government.
Female top manager dummy for ﬁrms with a female as a top manager.
Female owner dummy for ﬁrms with a female owner.
City 1 dummy for ﬁrms operating in cities with population below 50,000.
City 2 dummy for ﬁrms operating in cities with population between 50,000 and 250,000.
City 3 dummy for ﬁrms operating in cities with population between 250,000 and 1,000,000.
City 4 dummy for ﬁrms operating in cities with population above 1,000,000.
Electric outages (N) number of electric outages experienced in the last year (in log).
Electric outages (lenght) average duration of electric outages experienced in the last year (in log).
percentage of instances in which a ﬁrm was either expected or requested to provide a gift or
Bribery depth
informal payment during solicitations for public services, licenses or permits.
Gifts (share) ﬁrms’ informal payment (gifts to public oﬃcials to get things done) as a percentage of total sales.
Constrained by corruption dummy for ﬁrms identifying corruption as a major constraint.
Loss from theft losses due to theft and vandalism against the ﬁrm as a percentage of total sales.
Constrained by crime dummy for ﬁrms identifying crime, theft and disorder as a major constraint.
Capacity utilization percentage of capacity utilization.