Small Bus Econ 
https://doi.org/10.1007/s11187-023-00792-0
RESEARCH ARTICLE
Africa’s businesswomen – underfunded 
or underperforming?
Charles Ackah  · Holger Görg  · 
Aoife Hanley  · Cecilia Hornok 
Accepted: 18 May 2023 
© The Author(s) 2023
Abstract While the recent success of Africa’s Suppliers to Africa’s businesswomen (e.g. sellers of 
‘Lionesses’ – successful female entrepreneurs – is cloth for garment manufacture) who offer the pos-
internationally celebrated, less is known about how sibility for delayed repayment, statistically boost the 
liquidity can fuel the success of the ‘Lionesses’ and survival of female-owned businesses. Africa’s ‘Lion-
other businesswomen. Using information from a esses’ – exceptional businesswomen – are a compara-
panel of over 800 male- and female-owned busi- tive rarity, a reason we explore the reasons behind 
nesses in Ghana (ISSER-IGC survey), we capture a their success. Using data from over 800 Ghanaian 
measure of underfunding, in addition to data on sup- businesses from 2011 to 2015, we examine the role of 
plier credit, equity and other finance sources. Our funding in explaining why the majority of business-
regressions reveal a female-to-male productivity gap women perform so poorly. Our findings highlight an 
of between − 11 and − 19 per cent, values similar to unusual fact – suppliers to female-owned businesses 
estimates for other African countries. However, when (e.g. sellers of cloth for manufacture into garments) 
financial constraints are taken into account, the gen- who offer their female customers the possibility to 
der performance gap disappears. Accordingly, female delay repayments are enormously beneficial in nar-
business owners who indicate that funding is not a rowing the gender gap. Targeting tax-cuts towards 
problem are associated with higher productivity than such suppliers would boost the emergence of future 
males, all things equal. In a finding new to the litera- ‘Lionesses’.
ture, our regressions reveal the importance of supplier 
credit for Africa’s businesswomen. Keywords Female-owned businesses · Liquidity · 
Productivity · Supplier credit · Africa · Ghana
Plain English Summary 300 African Busi-
nesswomen in Focus  - Suppliers Key to Success. JEL Classification D22 · J16 · L26
C. Ackah  1 Introduction
University of Ghana - ISSER, Accra, Ghana
H. Görg · A. Hanley ( ) · C. Hornok  The world has recently witnessed the emergence of *
Kiel Institute for the World Economy, Kiel, Germany successful female business leaders on the continent of 
e-mail: Aoife.Hanley@ifw-kiel.de Africa. These female business owners include Ethio-
H. Görg · A. Hanley  pian Bethlemen Alemu, director of SoleRebels who 
Kiel Centre for Globalization, Kiel, Germany grew up in a neighbourhood of textile artisans. While 
Vol.: (0123456789)
1 
 C. Ackah et al.
such African Lionesses are well known and cele- In sum, our analysis investigates the funding-per-
brated, Bethlemen Alemu may well be the exception formance gap for female business owners in Ghana, 
rather than the rule. In many countries, both devel- where we conjecture that – in the absence of credit 
oped and developing, fewer women than men succeed constraints – business performance (productivity or 
in starting a business (Kelley et al., 2010). And those exporting) should line up with ex ante funding for 
businesses that get started by women generally per- both males and females.
form worse than their male counterparts. Our analysis contains a further novelty – namely, 
Access to finance is blamed as an aggravating the use of firm-level panel data. Longitudinal micro-
factor (e.g. McKenzie & Paffhausen, 2019), where data is relatively uncommon in Africa, and our data-
women may find it comparatively difficult to obtain base provides a useful exception. Specifically, we 
funding from formal sources such as banks (e.g. Aris- use data extracted from the ISSER-IGC enterprise 
tei & Gallo, 2016; Chaudhuri et al., 2020,). However, survey collected by Ghana’s Institute of Statistical, 
bank lending is only one aspect of an overall story Social and Economic Research (ISSER). Addition-
where funding from family and friends or supplier ally, the focus on Ghana has further advantages, a 
credit may help to fill the funding gap (e.g. Beck country scoring highly across several metrics for 
et  al., 2008; Boudreaux et  al., 2022; Pham & Tala- female business owners (Kelley et al., 2010; MIWE, 
vera, 2018). 2022). Ghana ranks second in Africa for female entre-
Whether access to different funding sources helps preneurs and in 46th place in the global rankings.1 
to explain performance differences for female busi- But even in Ghana, female business owners perform 
ness owners in developing countries has, to the best worse than their male peers, with businesswomen 
of our knowledge, not been investigated to date. Our reporting reduced productivity (Owoo et  al., 2019) 
analysis aims to shed light on this issue. Addition- and sales (Agyire-Tettey et al., 2018).
ally, our analysis proposes to tie together two strands The granularity of our data allows us to probe 
of research – studies on the funding of developing aspects of financial constraints and finance sources in 
country females (e.g. Maden, 2015; Naguib and a way which has not – to our knowledge – been done 
Jamali, 2015; Taiwo, 2023) and studies document- to date. Our indicator for financial constraints meas-
ing the performance gap for these females (Agyire- ures the severity of such constraints (on a scale of 1 
Tettey et al., 2018; Campos & Gassier, 2017; Lange- to 9), as reported by the enterprises in the survey. In 
vang et al., 2015; De Mel et al., 2008; Owoo et al., the spirit of the studies emphasising the importance 
2019). By estimating an empirical model which of liquidity, we review differences in financial con-
takes both funding and performance into account, straints between female- and male-owned businesses, 
we can be more certain that any funding deficit for evaluating how they connect to differences in produc-
females compared to males is suggestive of credit tivity (for which we use alternative measurements). 
constraints. Moreover, our Ghanaian data for over We apply the data to consider a range of alterna-
800 male- and female-owned businesses from 2011 tive finance sources. We also use export activity as 
to 2015 represents one of the most comprehensive another indicator of firm performance. While there is 
investigations into this issue. To date, work has a small number of empirical studies on gender differ-
focussed on more limited data from focus groups or ences in exporting behaviour (e.g. McClelland et al., 
smaller surveys. 2005), such studies have not looked at the interaction 
Our analysis is prompted by an appeal (e.g. Halkias, between gender and access to finance as we do in this 
2011; Henry et  al., 2016; Verheul et  al., 2006) for a paper.
multivariate model which controls for confounding Our empirical work uncovers some interesting 
covariates (e.g. business sector) that might otherwise findings. Female business owners perform worse than 
skew evidence of a female funding-performance gap. their male peers, ceteris paribus, even if businesses 
Female business owners dominate in sectors where in the same industry, city, age and size category are 
businesses are traditionally small and not highly effi-
cient. For this reason, Henry et  al., (2016) highlight 
the need for researchers to control for industry sector 1 In a list of Middle East and North Africa (MENA) econo-
and apply proper sampling practices. mies.
Vol:. (1234567890)
1 3
Africa’s businesswomen – underfunded or underperforming? 
compared – a fact noted in previous studies (e.g. Bar- findings of these studies are pessimistic, whereby 
dasi et al., 2011; Owoo et al., 2019). In value terms, females perform significantly worse than their male 
the female-to-male productivity gap lies in the range peers. There are a few exceptions to this general 
of − 11 to − 19 per cent, depending on the productiv- rule, but these tend to be sector-specific (See Amin 
ity measure used – comparable with estimates for & Islam, 2014). But broadly speaking, at a concep-
other African countries (e.g. Aterido et  al., 2011). tual-theoretical level, differences in firm performance 
The gender gap in export propensity is estimated to depending on the gender of the business owner or 
be roughly − 2 percentage points – a substantial dif- manager are due to systematically different choices 
ference, given that the average enterprise in our data- made by males and females. As Croson and Gneezy 
base exports with a single-digit probability. (2009) argue, women are, on average, much more 
Considering constraints in access to finance reluctant than men to engage in competitive behav-
changes this story. We now find that females that are iour, which may lead to differences in the perfor-
severely financially constrained perform worse than mance of the firms that they run. However, these dif-
males, while this is not true for women that do not ferences in choices may be driven by differences in 
report financial constraints. In fact, our results indeed constraints between males and females, which put a 
show that female business owners reporting that fund- limit on female managers’ choices regarding invest-
ing is not a problem are associated with higher pro- ments, competitive behaviour or risk taking (Campos 
ductivity compared to men, all things equal. Interest- & Gassier, 2017).
ingly, access to finance is not an issue for male-owned A number of such constraints have been identified 
businesses. These results are robust to employing in the literature, e.g. underinvestment by female busi-
alternative productivity measures, the first lag of the ness owners needing to support their families (Faf-
financial constraint variable, or exporting status as the champs et al., 2014; McKenzie & Paffhausen, 2019), 
outcome variable. difficulties in raising external capital (Field et  al., 
Moreover, we also find that not all sources of busi- 2010; McKenzie, 2017), in leveraging family business 
ness liquidity are created equal. Female business networks (Aterido & Hallward-Driemeier, 2011), in 
owners, having a higher recourse to private savings to women being treated differently by investors (Kanze 
shore up their liquidity, are associated with reduced et al, 2018) or suffering from poorer education (Islam 
productivity. This hints at a worrying possibility & Amin, 2016; Islam et al, 2019). Generally, the con-
– Women relying on their own savings may be forced sensus view is that females face more severe con-
to do so from a lack of competitive alternatives. On straints on accessing finance from different sources.2
a positive note, we find robust evidence that females Empirical studies on the gender funding gap have 
sourcing credit from suppliers report higher produc- tended to focus on an evaluation of bank lending 
tivity, all things equal. (Aterido et  al., 2013). But a parallel literature has 
We structure our analysis in the following way. We highlighted the role of liquidity shocks (not exclu-
first describe the related literature to help motivate sively credit shocks) on the performance of enter-
and inform our hypotheses and the methodology we prises in developing countries (McKenzie, 2017; 
use to test the hypotheses. We then describe our data, Rotemberg, 2019). We now proceed to review these 
introducing the ISSER-IGC panel, and then follow studies, identifying two main ideas which have not 
the analysis section before we conclude with some been satisfactorily answered to date. First, does the 
comments on the implication of our findings. empirical evidence point towards a gender perfor-
mance gap, even when the idiosyncrasies of female- 
and male-owned businesses are considered? And 
2  Gender performance gap – studies 
and hypotheses
There is a small but well-organised literature report-
2
ing the underperformance of female-managed or  Of course, there are also other constraints which are not the 
female-owned businesses in developing countries focus of this paper. For example, Field et al. (2010) show that traditional cultural institutions in a country may put a con-
(e.g. Campos & Gassier, 2017). Generally, the straint on female entrepreneurship.
Vol.: (0123456789)
1 3
 C. Ackah et al.
second, is there evidence that additional liquidity the unavailability of adequate funding to their firms. 
would help to mitigate this problem? If this is the case, then such financial constraints may 
explain part, if not all, of the female performance gap.
2.1 T he productivity premium – returns to liquidity While the literature generally agrees that 
for female- and male-owned businesses female-owned businesses perform worse than 
their male-owned peers, the empirical evidence 
There exists a relatively large literature on how access for credit constraints is more mixed. In a recent 
to finance affects the growth of small and medium- study, Chaudhuri et al. (2020) use data from busi-
sized enterprises, where growth is measured either ness owners in India, splitting the coefficient of 
as sales growth or employment growth (Fafchamps loan denial into an endowment component (female 
and Schündeln, 2013; Fowowe, 2017; Ayyagari et al., business owners constrained to exhibit the same 
2021). The gender performance gap is well docu- endowments as their male peers) and a character-
mented, with some researchers reporting statistical istics component (lenders apply the same criteria 
differences in male- and female-owned businesses to females as males).3 Their study reveals that 
for employment size (Bardasi et al., 2011; Chaudhuri the higher rejection rates on credit applications 
et al., 2020), productivity (Aterido & Hallward-Drie- from businesswomen are not a symptom of credit 
meier, 2011; Chaudhuri et al., 2020), growth (Belitski constraints. Rather, the higher rejection rates for 
& Desai, 2021; Chaudhuri et al., 2020; Coad & Tam- female applicants are underpinned by quality dif-
vada, 2012) or export participation rates (Presbitero ferences in the loan application. The characteris-
et  al., 2014). But, in the absence of highly granular tics of female loan applicants are so qualitatively 
data, it is difficult to grasp the severity of the prob- different – negatively so – from their male peers 
lem. If, for example, women are overrepresented in that the higher rejection rates cannot be blamed 
low-paying, informal or traditional sectors – charac- on gender discrimination. Other studies come to a 
terised by low productivity and earnings (see Klapper similar conclusion – Female loan applicants report 
& Parker, 2011) – then controlling for such sectoral higher rejection rates due to the inferior quality 
information might well cause the productivity gap to of their applications for finance rather than any 
disappear altogether. underlying, gender-based discrimination (Aterido 
In our paper, we focus on a performance measure & Hallward-Driemeier, 2011; Aterido et al., 2013; 
that has received less attention in the context of finan- Bardasi et al., 2011).
cial constraints: productivity. Productivity measures But other studies contradict these findings. At 
how efficiently inputs are transformed into output, least three of the most cited of these analyses reveal 
and, as such, it is a prominent performance measure substantial evidence for gender-biased credit con-
of a manufacturing enterprise. Sales or employment straints (Aristei & Gallo, 2016; Muravyev et  al., 
capture the size of a company, but not necessarily 2009; Presbitero et  al., 2014). Most recently, Aris-
production efficiency, as larger companies are not tei & Gallo (2016) uncover evidence of gender dis-
necessarily more productive. Our focus on productiv- crimination when business owners apply for credit. 
ity also draws on a large literature on heterogeneous Their data covers 28 transitional European coun-
firms (originating from the seminal paper of Melitz, tries. Here, the differences in denial rates are not due 
2003), which studies how productivity determines to covariates used in their regressions but to unex-
the success of manufacturers in domestic and interna- plained sources of variation (factors not picked up in 
tional markets. their estimations). Similarly, Muravyev et al. (2009) 
Having established (or otherwise rejected) the pos- also pick up variation in the error term consist-
sibility of a gender productivity gap, the next step is ent with a regime of gender-based credit rationing. 
to analyse whether funding (loans or other liquidity Finally, Presbitero et al. (2014) uncover evidence of 
sources) makes any difference in mitigating this gap. 
If, as argued above, gender-related constraints on 3
access to finance exist, then female owners/managers  To highlight the extent to which female business owners are credit constrained, Chaudhuri et al., (2020) employ a modified 
may make systematically different decisions due to version of the Oaxaca (1973) technique.
Vol:. (1234567890)
1 3
Africa’s businesswomen – underfunded or underperforming? 
credit constraints, employing data for around 360 broadening liquidity sources to include other sources of 
borrowers across 3 Caribbean countries.4 liquidity apart from bank loans. Specifically, Beck et al. 
From the studies reviewed so far, we find that evi- (2008) have argued that business owners in the mid-
dence is almost evenly split for and against gender dis- dle- and low-income countries have very different fund-
crimination in credit markets. But there is a caveat con- ing possibilities to their peers in Germany or the UK. 
nected to the existing work, its heavy emphasis on loans There are more commonly used alternatives to formal 
applied for and loans rejected. The literature is largely working capital loans, e.g. supplier credit. With respect 
silent on the overall liquidity position of female business to supplier credit, Beck et  al. examine the financing 
owners. Moreover, evidence by Bardasi et al. (2011) sug- patterns for firms across 48 countries, including many 
gests that the demand for loans by female borrowers is developing countries. Their findings build on evidence 
not accurately measured. This is due to the discouraged from the World Business Environment Survey, admin-
borrower effect, where female business owners in devel- istered by the World Bank. Supplier credit represents 
oping countries may be reluctant to apply for a bank loan the second-most important source of finance for small 
or line of credit, anticipating a rejection. Evidence for this firms in developing countries, after bank credit. Mean-
discouraged borrower effect is corroborated by Gonza- while, in a recent study using data for entrepreneurs in 
lez-Uribe & Leatherbee (2018).5 These studies highlight- Zambia, the role of supplier credit in reducing informa-
ing the discouraged borrower effect hint at the wisdom tion asymmetries is clear (Boudreaux et al., 2022). Sup-
of widening the definition of funding to include other pliers working close to the entrepreneur can gauge the 
sources of liquidity. This is because studies focussing on individual’s social capital, arguably better than a bank.
bank loans – due to the discouraged borrower effect – are This idea of formal vs informal finance is further 
likely to underestimate the real liquidity problem. developed by Pham & Talavera (2018) using data 
From this discussion, we can formulate the follow- across the size spectrum for Vietnam. On the basis of 
ing hypothesis. their estimations, they conclude that businesswomen 
are more successful in obtaining loans than their male 
H1: A gender-related productivity gap can be (at least peers. Additionally, businesswomen enjoy the privilege 
partly) explained by differences in access to funding of reduced interest rates. Pham and Talavera attribute 
between female- and male-owned businesses. the stronger loan performance of Vietnamese business-
women to a buoyant supply of informal finance (loans 
from friends and relatives). The Pham and Talavera 
2.2  Reliance on personal savings as a litmus test for study underpins the importance of viewing bank finance 
liquidity-constrained businesswomen as only one component in a wider and richer picture.
As Pham & Talavera (2018) have highlighted, 
There are two further hypotheses we can investigate supplier credit represents a widely used way to boost 
with our data. The first is connected with the idea of short-term liquidity – a cheap and less complex alter-
native to overdraft finance. The payback period is 
short (typically a month, in the case of Ghana), but 
4 Not all studies can be split into those reporting credit con- the loan is interest free. Importantly, the entrepreneur 
straints, and those which do not. There is even one study, find-
ing evidence of positive credit discrimination, where female- can bridge the time between procurement of materials, 
owned businesses are even preferred by banks (Hansen & working these materials to a final product which can 
Rand, 2014). Using data for 16 African countries and applying be sold for cash. Unlike banks (which are highly regu-
a method similar to Presbitero et al. (2014), Hansen and Rand lated and subject to public scrutiny), suppliers have 
demonstrate that small enterprises owned by females are less 
likely to be credit constrained compared to their male counter- much latitude in the terms they offer to their business 
parts. An effect reversed for medium-sized businesses, where customers (Fafchamps, 2000). Additionally, as noted 
no such favouritism exists. earlier, a recent study using data on 1971 entrepre-
5 We should note that although neither Bardasi, Sabarwal & Ter- neurs in Zambia has highlighted the role of suppliers 
rell (2011) nor Gonzalez-Uribe & Leatherbee (2018) highlight in providing credit to developing country entrepre-
differences between female and male applicants, they do suggest neurs (Boudreaux et al., 2022). Specifically, suppliers 
that type 2 selection error (rejection of promising candidates) is 
likely to be a problem – a point which underscores the need to look are in an excellent position to gauge the creditworthi-
beyond evidence on loans, considering a broader palette of finance. ness of their clients through repeated transactions.
Vol.: (0123456789)
1 3
 C. Ackah et al.
For businesses, it is a good thing to be able to exports of most African countries remain dominated 
source funding from a variety of providers (banks, by primary products, while manufacturing exports 
investors and suppliers) as it widens the set of fund- are historically low and likely to remain so for vari-
ing possibilities. The portion of funding from these ous reasons (Wood and Mayer, 2001). This hinders 
providers can vary from firm to firm. On one aspect, the continent’s economic development. Consequently, 
there is a universal consensus – A disproportionate policymakers in many African countries – including 
reliance on cash savings by any group suggests a defi- Ghana – have a strong interest in learning about the 
cit in the provision of formal finance (Guérin, 2006; drivers of manufacturing exports.
Loaba, 2022). Moreover, Guérin has argued that the Although, in our data, exporting is a small num-
reliance of females on informal finance is a conse- ber phenomenon, making it difficult to pick up 
quence of gender inequalities, an over-reliance that empirically, exporting is an activity often pushed by 
can perpetuate further inequality. Meanwhile, Loaba policymakers in developing countries. Exporting to 
has demonstrated that women are more reliant on other developing or even developed countries can 
informal sources of finance than males. One glimmer help indigenous businesses to buffer against demand 
of hope is offered by newer technological possibilities shocks in their home country, broaden their customer 
(e.g. mobile money), helping females to sidestep the base and motivate them to redouble their efforts 
perceived shortfall in bank credit. But many funding to reach the world technology frontier in order to 
sources, so long as they are competitively priced (e.g. remain competitive with a widened set of competi-
supplier credit), can be used to help the female entre- tors. The benefits of exporting are well documented 
preneur to expand her market share. In this way, she (e.g. Girma & Görg, 2022; Van Biesebroeck, 2005).
can improve her productivity. Indeed, the usefulness Furthermore, export activity is often used as an 
of bank credit may be overvalued. As has been shown alternative measure of firm performance, focusing 
using data for India, excessive reliance on bank credit on the international engagement of a firm. The over-
can point to cashflow problems within a firm (Sat- whelming evidence shows that exporting firms are 
pathy et  al., 2017). For this reason, it is not easy to more productive than non-exporters (e.g. Wagner, 
propose an ideal split between formal vs informal 2019). A small number of studies have looked at gen-
sources of liquidity. But broadly speaking, we expect der differences in export performance, showing that 
that a widened set of funding possibilities can help female-owned firms, on average, are less export-ori-
an entrepreneur to boost the productivity of her firm, ented than their male-owned counterparts (Manolova 
reducing the gender performance gap. et  al., 2002; McClelland, 2004; McClelland et  al., 
2005). One explanation given for these performance 
H2: Access to other informal sources of finance differences is restricted access to funding for female 
mitigates the gender performance gap. entrepreneurs (McClelland et al., 2005).
But if female-owned businesses find it com-
paratively more difficult to access export markets 
2.3 E xporting as an alternative performance metric due to a lack of funding, then this bias needs to be 
corrected. We conjecture that female-owned busi-
Our final hypothesis concerns itself with exporting, nesses – lacking adequate liquidity – find it more 
another performance metric – apart from productivity. difficult to contest export markets. No study – to our 
We believe that export participation is also an impor- knowledge – has in detail examined exporting in the 
tant measure of firm performance. For one, following context of the interaction of gender and liquidity 
the paper of Bernard and Jensen (1999), a vast litera- constraints.
ture on heterogeneous firms and international trade This hypothesis is expressed as follows.
documents that productivity and exporting correlate 
strongly positively. Manufacturers need to achieve a H3: There is a gender export gap, which is driven 
certain level of productivity to enter export markets, by underfunded females.
while export participation can improve productiv-
ity further. Second, the export participation of Afri- Before moving on to the main empirical section, 
can manufacturers is of great policy importance. The we first describe our data.
Vol:. (1234567890)
1 3
Africa’s businesswomen – underfunded or underperforming? 
3  The Ghanaian ISSER‑IGC panel The sampled firms operate in 20 different two-digit 
manufacturing industries, applying the International 
We recall our initial research question to (1) investi- Standard Industrial Classification (ISIC) Revision 4 
gate Ghana’s gender performance gap and (2) ascer- classification. Table  1 illustrates the geographic and 
tain whether this performance gap explains any dif- sectoral breakdown of these firms, where industries 
ferences in the perceptions of business owners of both are grouped into four categories acknowledging the 
genders that they are underfunded. strong concentration of the firms in a few industries. 
To address these questions, we use data from the The Accra and neighbouring Tema area account for 
ISSER-IGC survey. This survey of micro, small and about half the firms. Sekondi-Takoradi, also on the 
medium-sized manufacturing enterprises in Ghana is coastline, represents about an eighth of the firms. The 
administered by the Institute of Statistical, Social and Eco- remaining firms are located further inland, in the city 
nomic Research (ISSER) based at the University of Ghana of Kumasi. In terms of the business sector, the over-
and funded by the International Growth Centre (IGC). In whelming majority of the businesses are active in the 
terms of timing, questionnaires were distributed in August/ Textiles and Clothing sector, followed by Wood Pro-
September 2016. The survey elicited information on the cessing and Food and Beverages.
characteristics of business owners and their businesses for Klapper & Parker (2011) noted the over-representation 
five consecutive years (2011 to 2015, inclusive).6 of females in the informal sector. Alternatively, in sectors 
The sample frame adopted for the questionnaire was with the least potential for growth and profits, while the 
extracted from the first phase of the Ghana Integrated Busi- food and beverages sector is dominated by female busi-
ness Establishment Survey (IBES). The latter represents an ness owners (67 per cent), Ghana’s main industry, textiles 
economic census of non-household enterprises conducted and clothing, exhibits almost equal proportions of male 
by the Ghana Statistical Service (GSS) from 2014 to 2015. and female business owners, with females comprising 55 
To undertake the survey, the sample frame was extracted per cent of this sector. Altogether, 43 per cent of the busi-
from the universe of manufacturing micro, small and ness owners in our sample are females.
medium-sized enterprises (MSMEs) located in the cities We continue with the discussion of the most impor-
of Accra, Tema, Kumasi and Sekondi-Takoradi. These cit- tant variables in our empirical analysis. A systematic 
ies represent the main industrial clusters of Ghana. To help description of all variables used is presented in Table 2. 
completeness, the data also includes firms from Ghana’s Basic descriptive statistics are shown in Table 3.
informal sector. From the IBES, all manufacturing MSMEs 
located in the four cities were selected. This amounted to 3.1 C alculating productivity
1244 firms in total. The interviewers conducting the survey 
encountered a reasonable response rate. However, there We measure productivity as total factor productivity 
was some sample attrition. This was due to firms declining (TFP) using a regression framework to estimate produc-
to participate (73 firms), business closure (55 firms) and tion functions.7 As we have signalled in the ‘Introduc-
failure to locate the business (231 firms). To sum up, alto- tion’, one innovation of our analysis is the provision of 
gether, 880 firms completed the questionnaire, correspond- alternative measures of TFP, allowing us to choose the 
ing to a 70 per cent response rate. TFP candidate which offers the most reliable estimates. 
Given the focus on productivity differences in our 
paper, we now outline our various TFP models using 
6 See Abeberese et al. (2019) for a detailed description and appli- alternative variants of the workhorse proxy variable 
cation of the panel. Because of the retrospective nature of the 
survey, a potential cause for concern is recall error. However, we 
should point out that data were collected through face-to-face 
interviews with respondents who were instructed to extract the 7 Of possible productivity measures, we opt for total factor pro-
information directly from the firm’s written records. In 60 per cent ductivity (TFP) and capitalise on the recent developments in 
and 30 per cent of the interviews, the respondent was the owner the econometrics of measuring TFP. As opposed to alternative 
or a senior manager, respectively. Furthermore, Abeberese et  al. productivity measures (e.g. labour productivity), TFP considers 
(2019) tested the robustness of their results by successively drop- the efficiency of the use of all inputs in the production process. 
ping earlier years from their estimation sample. Their estimation Moreover, thanks to available structural estimation methods, we 
results remained robust to these modifications, suggesting that are able to obtain an unbiased estimate for TFP, which accounts 
recall error does not seriously compromise the data. for the possible endogeneity of firm’s input decisions.
Vol.: (0123456789)
1 3
 C. Ackah et al.
Table 1  Number of firms Location of enterprise Gender of business 
by sector and location owner
Industry group Accra Tema Kumasi Sekondi- Total Male Female Total*
Takoradi
Food and beverages 40 20 41 16 117 38 78 116
Note: The sample of firms 
in the right part of the table Textiles and clothing 198 35 218 62 513 228 284 512
excludes 18 enterprises that Wood processing 59 22 84 14 179 164 3 167
are either owned by the Other manufacturing 28 3 31 9 71 61 6 67
state or do not report the Total 325 80 374 101 880 491 371 862
gender of the owner
estimation methods (also called as control function A well-known challenge in estimating production 
approach) first proposed by Olley & Pakes (1996). An functions is that input use is not independent of current 
advantage of estimating total factor productivity (TFP) productivity. The productivity term it is unobserved by 
in several different ways is to help raise confidence in us and hence becomes part of the error term in the estima-
the point estimates. On balance, our favoured estimation tion: uit = it + it . Firms, however, can obtain informa-
method is the most recent Ackerberg–Caves–Frazer, tion on their current productivity and adjust their contem-
short ACF method (Ackerberg et al., 2015). poraneous input use accordingly. This generates a positive 
Although these estimation methods are now consid- correlation between the input variables and the error term 
ered standard in the literature, we describe them here uit , leading to biased estimates when (1) is estimated by 
briefly. The total factor productivity (TFP) of a firm in OLS. Specifically, OLS coefficient estimates become 
a given year is measured as the residual from a produc- upward biased for the labour input and downward biased 
tion function estimation. We assume a Cobb–Douglas for capital (Olley & Pakes, 1996). To overcome this prob-
production function of the value-added output with lem, Olley and Pakes (OP) propose a two-step control 
two inputs – capital and labour – and standard Hicks- function estimation procedure, which was subsequently 
neutral technological change. Formally, improved by Levinsohn & Petrin (2003) and Acker-
berg et al. (2015). These methods are essentially a proxy 
yit = kkit + llit + it + it (1) for productivity by observable variables (investment or 
where y  is value-added output for firm i in year t, k  material use). In addition, Wooldridge (2009) proposes a it it
and l  are capital and labour, respectively,   is the potentially more efficient, one-step estimation procedure it it
unobserved total factor productivity and   is an idi- that yields the Levinsohn-Petrin estimator.
9
it
osyncratic error term. All variables are in logarithm. This paper applies three of the above estimation 
Value added is obtained as the value of gross output methods to generate total factor productivity: Levin-
less the cost of raw materials, capital is measured as sohn-Petrin (LP), Wooldridge and Ackerberg-Caves-
10
the replacement cost of capital items including land, Frazer (ACF).  Our preferred estimation method is, 
buildings and machinery, and labour is the number however, the most recent ACF procedure. Ackerberg 
of workers, as reported by firms in the survey. We et al. (2015) show that, due to functional dependence, 
deflate all nominal values to 2006 Ghanaian Cedis it is generally not possible to identify the labour coef-
using the manufacturing producer price index from ficient in the first step of the estimation procedure, 
the Ghana Statistical Service.8 which the OP and LP methods do. Labour use is 
namely functionally dependent on the other variables 
that are included in the first-stage regression to proxy 
8 Note that we opt for a value-added production function 
(instead of a gross output production function with capital, 
labour and materials as inputs) because the simultaneous use 
of materials as proxy variable and as regressor in the produc- 9 A detailed description of this literature is provided by Rovi-
tion function estimation causes identification problems (Gan- gatti and Mollisi (2017), among others.
dhi, Navarro and Rivers, 2020). Namely, production materials 10 We perform the production function estimation using the 
are assumed to be a flexible input, as is customary in the litera- prodest command in Stata, which was developed by Rovigatti 
ture, and are used as the proxy variable for productivity.. and Mollisi (2017).
Vol:. (1234567890)
1 3
Africa’s businesswomen – underfunded or underperforming? 
Table 2  Description of variables used
Variable Description Source/notes
ID Enterprise ID Original variable, ISSER-IGC panel
Year Calendar year (2011–2015) Original variable, ISSER-IGC panel
Industry Industry code ISIC rev. 4 (categorical, 1–24) Original variable, ISSER-IGC panel
Sector Broad industry (categorical, 1–4) industry grouped into four larger categories 
(food and beverages, textiles and clothing, 
wood processing, other manufacturing)
Location City in which the enterprise is located (cat- Original variable, ISSER-IGC panel
egorical, 1–4)
Female Primary owner of enterprise is female (binary) 0: male; 1: female
Age Age of enterprise (years) year – year of initial production (as reported in 
the survey)
Size Size of enterprise in terms of employment 0: micro (1–5 employees); 1 small (6–19 
(categorical, 0–2) employees); 2 medium (20 + employees)
Exporter Enterprise exports some of its production 0: no export; 1: export
output (binary)
Foreign At least 10% of the enterprise is owned by a 0: not foreign-owned 1: foreign-owned
foreign owner (binary)
tfp_ac Total Factor Productivity of enterprise (loga- Estimated by the Ackerberg-Caves-Frazer (ACF) 
rithm) estimator
tfp_lp Total Factor Productivity of enterprise (loga- Estimated by the Levinsohn-Petrin estimator
rithm)
tfp_wr Total Factor Productivity of enterprise (loga- Estimated by the Wooldridge estimator
rithm)
Finance constraint (FC) Access to finance as business constraint (rank The variable is based on the survey question, 
variable, 1–9, higher indicates more severe “Please rank the following nine obstacles in 
constraint) terms of their importance to the enterprise’s 
operations: access to finance, taxation, customs 
and regulation, security, bribery/informal 
payments, access to land, access to electricity, 
access to other infrastructure, market access.”
FC categories Access to finance as business constraint (cat- It is generated from the finance constraint vari-
egorical, 0–2) ables. It takes value 0 if the finance constraint 
is 1, 2 or 3 (low), value 1 if the finance 
constraint is 4, 5 or 6 (medium), and 2 if the 
finance constraint is 7, 8 or 9 (high)
FS bank loan Bank loan from formal institutions (% of work- FS variables are based on the survey question, 
ing capital) “What percentage of the enterprise’s work-
FS own resources Personal savings and retained earnings (% of ing capital was obtained from the following 
working capital) sources?”
FS friends and relatives Loan from friends and relatives (% of working 
capital)
FS suppliers credit Suppliers credit (% of working capital)
FS equity and bond Issuance of equity and bonds (% of working 
capital)
FS other Other finance sources (% of working capital)
Variables used in TFP estimation
Y Value added of production (2006 cedis) Generated as the value of production output 
minus the value of raw materials used in 
production (both from the ISSER-IGC panel, 
deflated by PPI)
Vol.: (0123456789)
1 3
 C. Ackah et al.
Table 2  (continued)
Variable Description Source/notes
L Number of workers (both production and non- Original variable, ISSER-IGC panel
production)
K Estimated resale value of capital (land, build- Original variable, ISSER-IGC panel, deflated 
ings, machinery and equipment) (2006 cedis) by PPI
PPI Producer Price Index (2006 = 1) for the manu- Ghana Statistical Services
facturing sector in Ghana
Table 3  Descriptive statistics of key variables
Variable Total Female-owned Male-owned
Obs Mean Std. dev Obs Mean Std. dev Obs Mean Std. dev
Female 4310 0.430 0.495 1855 1.000 0.000 2455 0.000 0.000
Age 4105 13.621 9.644 1760 13.335 8.859 2345 13.836 10.190
Size 4207 0.766 0.603 1824 0.737 0.575 2383 0.788 0.622
Exporter 3548 0.035 0.184 1549 0.019 0.138 1999 0.048 0.213
Foreign 4310 0.017 0.131 1855 0.008 0.090 2455 0.024 0.154
tfp_ac 3891 7.570 1.254 1674 7.414 1.101 2217 7.688 1.346
tfp_lp 3891 8.139 1.294 1674 7.904 1.165 2217 8.317 1.357
tfp_wr 3891 8.071 1.468 1674 7.825 1.352 2217 8.257 1.523
Finance Constraint (FC) 4233 7.250 1.877 1820 7.219 1.887 2413 7.273 1.869
FC categories 4233 1.701 0.581 1820 1.700 0.581 2413 1.701 0.581
FS bank loan 4206 3.011 12.726 1822 2.142 10.226 2384 3.675 14.313
FS own resources 4206 80.713 33.212 1822 81.049 33.205 2384 80.456 33.222
FS friends and relatives 4206 3.369 15.133 1822 2.953 13.792 2384 3.688 16.079
FS suppliers credit 4206 5.345 16.202 1822 5.113 16.090 2384 5.523 16.288
FS equity and bond 4206 2.357 12.686 1822 1.965 11.631 2384 2.657 13.431
FS other 4206 5.204 19.745 1822 6.778 22.460 2384 4.002 17.294
Note: The sample excludes observations of enterprises that are either owned by the state or do not report the gender of the owner
for productivity. Consequently, the ACF procedure formal modelling of the relationship between gen-
estimates both the capital and labour coefficients in der, funding sources and performance below.
the second step.
The TFP measures we obtain as a result of 3.2  Exports
these estimations are sufficiently similar, as indi-
cated by the pairwise correlations of around 0.9 As reported in our literature section, firm-level 
(Table  4). After netting out industry means from exporting is an under-researched performance 
the TFP variables, we also plot the TFP distribu- metric in many developing countries. Fortunately, 
tions in Fig. 1 for male and female business own- the ISSER-IGC panel includes information on the 
ers, respectively. The kernel density estimates look firm’s export engagement. Specifically, the ques-
visually similar for all TFP methods used, with tionnaire asked respondents to report the share of 
the exception of Wooldridge. Of course, a simple output that is exported annually (export intensity). 
visual comparison of distributions cannot con- Based on this continuous measure, we generate an 
sider other potential differences between these two export dummy for the firm’s export status in any 
groups (for example, firm age). Nor do they allow given year. In our sample, only 3.5 per cent of the 
us to infer causality. We, therefore, turn to a more firm-year observations are exporters, reflecting 
Vol:. (1234567890)
1 3
Africa’s businesswomen – underfunded or underperforming? 
Fig. 1  Kernel density estimates for the TFP variables
low export incidence in this representative sample criteria – is what underpins the extension of credit 
of firms (Table  3). As unconditional correlations from banks, equity investors and other creditors (e.g. 
in Table 4 reveal, it is mainly the larger and more suppliers). Precisely, this is the problem making it 
productive enterprises which sell abroad. Low hard to identify credit constraints – the absence of vis-
export incidence is, therefore, not uncommon in a ible performance metrics for a firm. Because of this, 
sample of MSMEs. investors can plausibly argue that any applicant group 
(e.g. females or marginalised applicants from certain 
3.3  Underfunding religious, ethnic or socio-economic backgrounds) are 
excluded on the basis of market criteria. However, 
One of our most important variables is our proxy for all things equal, if the performance of marginalised 
underfunding – the extent to which respondents of applicants is similar to those of applicants from the 
both sexes report that lack of access to finance repre- non-excluded group, this argument no longer holds 
sents a serious obstacle to their business operations. and these female applicants may well be credit con-
Why do we need to consider funding and perfor- strained. Accordingly, the case for credit-constrained 
mance in the same empirical model of underfund- females hinges on the business performance of these 
ing? The reason has to do with the concept of credit females since it must be viewed through the lens of 
constraints. Credit constraints only hold sway when the lender.
funding is denied to an entrepreneur on the basis of Since banks (and other investors) should be 
reasons other than standard investment criteria. And strictly guided by market criteria awarding or declin-
standard investment criteria – rather than any other ing an entrepreneur’s application for credit, the 
Vol.: (0123456789)
1 3
 C. Ackah et al.
Vol:. (1234567890)
1 3
Table 4  Pairwise correlation coefficients between key variables
Female Age Size Exporter Foreign tfp_ac tfp_lp tfp_wr FC
Female rho 1
N 4330
Age rho  − 0.024 1
N 4125 4125
Size rho  − 0.0449*** 0.0914*** 1
N 4226 4021 4226
Exporter rho  − 0.0749*** 0.0091 0.1186*** 1
N 3563 3412 3523 3563
Foreign rho  − 0.0613***  − 0.0599*** 0.1018*** 0.0388** 1
N 4330 4125 4226 3563 4330
tfp_ac rho  − 0.1071***  − 0.0042 0.0136 0.0436** 0.0147 1
N 3905 3743 3870 3317 3905 3905
tfp_lp rho  − 0.1579*** 0.0283* 0.1659*** 0.0815*** 0.0657*** 0.9619*** 1
N 3905 3743 3870 3317 3905 3905 3905
tfp_wr rho  − 0.1459*** 0.0278* 0.1600*** 0.0823*** 0.0673*** 0.8464*** 0.8878*** 1
N 3905 3743 3870 3317 3905 3905 3905 3905
FC rho  − 0.0145  − 0.0242  − 0.0552***  − 0.0124  − 0.1075***  − 0.0289*  − 0.0318**  − 0.0419*** 1
N 4253 4048 4210 3552 4253 3894 3894 3894 4253
Note: Pairwise correlation coefficients and the corresponding sample sizes. The sample size varies between variable pairs due to varying data availability
* Significant at 10%
** Significant at 5%
* Significant at 1%
Africa’s businesswomen – underfunded or underperforming? 
entrepreneur’s performance metrics (e.g. productiv- Table 5  Wilcoxon tests for key business constraints
ity or exporting) are key signals of the strength of the Access to Taxation Market access
entrepreneur’s enterprise and accordingly whether finance
the investment can be repaid, or not. Where such per- H0: equal distributions
formance metrics are not easily observed (e.g. total 
factor productivity cannot be obtained as a back-of- Z  − 2,0130  − 0,3310 0,4370
envelope calculation), research can still empirically Prob >|z| 0,0441 0,7409 0,6625
calculate such measures to shed light on whether an Pr (male > female) 0,4820 0,4970 0,5040
entrepreneur is underfunded or not. Note: Before running the tests, we removed the industry means 
In the ISSER-ICG panel, respondents were asked from the business constraint variables to ensure the results 
to rank obstacles to the firm’s operations across are not driven by industry differences in the ratio of male-to-female firms. Values in bold refer to significance at the 95 per-
the years 2011–2015. The obstacle rank variables cent level of confidence
that we generate from the responses assume inte-
ger values between 1 and 9, where 9 indicates the 
highest and 1 is the lowest importance. ‘Access to and relatives, supplier credit, and equity and bond 
finance’ is one of the obstacles on the list. It is also finance (Table  2). Table  6, which reports correla-
one of the obstacles where male and business own- tions between the FC variable and the use of differ-
ers report such differences in the severity of this ent finance sources, illustrates that female business 
problem that we must reject the null of equal dis- owners who receive a loan are disproportionately 
tribution in the breakdowns of responses. To illus- less likely (versus their male peers) to report severe 
trate the severity of access to finance as a problem financial constraints. The same intuition applies to 
for female business owners, Table  5 supplies the businesswomen who receive suppliers’ credit.
results of Wilcoxon rank-sum tests that compare the Although male business owners also seem to 
distributions of male and female rankings for three value suppliers’ credit in helping to mitigate the 
obstacles impacting the day-to-day running of the perception of underfunding, the correlation is 
business, (1) access to finance, (2) taxation and 3) higher for females. As we indicated before, suppli-
market access. The test reports a statistically signifi- ers’ credit represents a key source of liquidity for 
cant difference between male and female rankings many developing-world businesses (Beck et  al., 
of access to finance. Specifically, female-owned 2008). Similarly, both male and female business 
firms are significantly more likely than male-owned owners who dig into their own resources to fund 
ones to designate access to finance as important. their businesses appear to experience difficulties in 
This suggests that this obstacle deserves further tapping appropriate finance. However, the correla-
investigation within a regression framework. tion for supplier credit is higher for businesswomen. 
For further analysis, we define a financial con- Finally, male business owners able to source equity 
straint (FC) variable on the basis of the obstacle finance are far less troubled by underfunding issues. 
rank for ‘access to finance’. Variable FC, there- However, because the numbers involved are very 
fore, takes integer values between 1 and 9, where a small (In 2015, for example, there were only 31 
higher value means more serious finance constraints male and 12 female businesses that financed some 
perceived by the enterprise. of their working capital through equity or bond), it 
Since financial constraints pose a particular prob- is difficult to do more than remark on the potential 
lem for female business owners, the next step is to uplift for these more sophisticated forms of finance, 
delve into funding patterns for both genders – how in helping Africa’s business community to scale up 
these breakdowns can be linked to a gender fund- their business capacity.
ing gap. One point worth exploring in this context Summing up, both lender and supplier credit 
is the sources of funding used by male and female appear to attenuate the problem of underfunding for 
business owners – providing a possible intuition businesswomen. The bivariate correlations above 
for a gender funding gap. The data provides infor- can help to shape our expectations about the role 
mation on a number of different funding sources of credit supply and gender. We now embark on a 
(FS), namely own resources, loans from friends fuller examination of funding using a regression 
Vol.: (0123456789)
1 3
 C. Ackah et al.
Table 6  Correlations between financial constraint and different finance sources
Bank loan Own resources Friends and Suppliers credit Equity and bond Other
relatives
Female owned (N = 1812)
  Correlation  − 0,126 0,116 0,055  − 0,137  − 0,008  − 0,046
  Significance 0,000 0,000 0,019 0,000 0,729 0,052
Male owned (N = 2383)
  Correlation 0,013 0,067  − 0,008  − 0,038  − 0,045  − 0,061
  Significance 0,542 0,001 0,701 0,067 0,027 0,003
Total (N = 4195)
  Correlation  − 0,035 0,088 0,017  − 0,080  − 0,030  − 0,054
  Significance 0,022 0,000 0,260 0,000 0,052 0,001
Note: Correlation coefficients and their significance levels between the use of different finance sources (FS) and the finance con-
straint (FC) variable. Finance source variables are percentages of working capital financed from a given source. A negative correla-
tion means that firms that report to have more serious problems with access to finance use less of the given finance source. Own 
resources = retained earnings + personal savings. Other is a residual category, which also includes loans from moneylenders
framework which relates underfunding to gender, productivity gap is captured by the parameter 1 , meas-
firm performance and business metrics for the busi- uring the difference in productivity for female-owned 
ness owner respondents. relative to male-owned enterprises, having controlled 
for the characteristics of the enterprise and other 
observables. Controlling for this vector of covariates 
4 G ender performance gap and liquidity is key to our estimation strategy since these observable 
constraints characteristics of the enterprise may influence produc-
tivity and, at the same time, correlate with the gender of 
We first proceed to our main regressions, exploring the owner. Covariates we consider include the age and 
the role of gender differences in productivity. First, the size of the enterprise, a binary variable denoting 
we document the female-to-male productivity gap for foreign ownership, the enterprise’s industry s (2-digit 
Ghanaian MSMEs. Observing these firms for a 5-year NACE) and location l, as well as year effects common 
annual panel (2011–2015) allows us to estimate the to all enterprises to remove macro trends. We estimate 
following regression equation using pooled OLS: (2) with pooled OLS and robust standard errors.
We now turn to the results for our estimation of the 
tfpit = 1femalei + Xit + s + l + t + it (2) gap in female-entrepreneur TFP and perceived finan-
The dependent variable is the logarithm of total cial constraints (Table 7).
factor productivity (tfp) of enterprise i in year t. We We report a statistically significant and negative 
use either of the three alternative productivity meas- female-to-male productivity gap, within the range 
ures as the dependent variable. Our preferred meas- of − 11 to − 19 per cent (columns 1 in Table  7), 
11
ure is tfp_ac, while the remaining two are reported for regardless of the measures used.  Hence, our evi-
robustness. The variation in productivity is explained dence suggests that a gender productivity gap exists 
by a dummy denoting whether the primary business and remains, even after controlling for many idiosyn-
owner is female. Apart from the female ownership cratic sources of variation in productivity.
dummy, we also include several other enterprise char- In order to look at Hypothesis 1, we next explore 
acteristics in vector X. Industry, location and time the role of finance constraints in explaining the 
effects are also included in the δ terms. observed gender gap. To do so, we include a variable 
The time-constant binary variable female takes 
the values of 1 and 0 for enterprises with female and 11 This percentage change is calculated as 100*exp(coefficient) − 100. 
male primary owners, respectively. (A few state-owned The coefficient − 0.118 therefore converts to 11%, and the coeffi-
enterprises are excluded from the sample.) The gender cient − 0.212 to − 19%.
Vol:. (1234567890)
1 3
Africa’s businesswomen – underfunded or underperforming? 
Vol.: (0123456789)
1 3
Table 7  Gap in female-entrepreneur TFP and perceived financial constraints
Dependent variable: tfp_ac tfp_ac tfp_ac tfp_lp tfp_lp tfp_lp tfp_wr tfp_wr tfp_wr
(1) (2) (3) (1) (2) (3) (1) (2) (3)
Female  − 0.118**  − 0.117** 0.407**  − 0.207***  − 0.207*** 0.347*  − 0.212***  − 0.212*** 0.293
(0.0501) (0.0501) (0.181) (0.0508) (0.0509) (0.185) (0.0509) (0.0509) (0.187)
Age of firm  − 0.00478**  − 0.00475**  − 0.00438*  − 0.00298  − 0.00295  − 0.00257  − 0.00251  − 0.00247  − 0.00212
(0.00225) (0.00225) (0.00225) (0.00226) (0.00226) (0.00226) (0.00227) (0.00227) (0.00226)
Firm size categories (benchmark is micro)
  Small  − 0.0707  − 0.0739  − 0.0777 0.153*** 0.151*** 0.147*** 0.173*** 0.170*** 0.166***
(0.0506) (0.0506) (0.0505) (0.0506) (0.0505) (0.0505) (0.0506) (0.0505) (0.0505)
  Medium sized 0.0715 0.0637 0.0659 0.707*** 0.701*** 0.703*** 0.763*** 0.755*** 0.757***
(0.0918) (0.0912) (0.0907) (0.0959) (0.0953) (0.0947) (0.0974) (0.0968) (0.0962)
Foreign  − 0.00609  − 0.0341 0.0419 0.137 0.114 0.195 0.179 0.150 0.223
(0.167) (0.166) (0.169) (0.162) (0.161) (0.165) (0.167) (0.165) (0.169)
Lagged FC  − 0.0148 0.0179  − 0.0120 0.0226  − 0.0154 0.0161
(0.0125) (0.0174) (0.0127) (0.0174) (0.0128) (0.0176)
Female × lagged FC  − 0.0714***  − 0.0756***  − 0.0688***
(0.0240) (0.0246) (0.0248)
Observations 2917 2917 2917 2917 2917 2917 2917 2917 2,917
R-squared 0.090 0.090 0.093 0.150 0.150 0.153 0.350 0.350 0.352
Note: OLS estimation with industry, location and year effects. The dependent variables are different TFP estimates in logarithm: tfp_ac stands for the Ackerberg-Caves-Frazer 
estimate, tfp_lp for the Levinsohn-Petrin estimate and tfp_wr for the Wooldridge estimate. Firm size categories are micro (0–5 employees), small (6–19 employees) and medium 
(20 + employees). Robust standard errors in parentheses
*** p < 0.01
** p < 0.05
* p < 0.1
 C. Ackah et al.
for finance constraints and its interaction with the increases with the severity of the finance constraints 
female variable in the regression equation reported by females. Moreover, the inclusion of the 
interaction term seems to explain away the gender gap 
tfpit = 1femalei + 2FCi,t−1 + 3femalei
(3) entirely, as the estimate for 1 turns statistically insig-
× FCi,t−1 + Xit + s + l + t + it nificant and even positive in some regressions once the 
Variable  is a self-reported measure show- interaction term is included. This is in line with H1.FC
ing how important an enterprise ranks its access to Graphically, we can depict the main information from 
finance from a list of 9 business constraints listed in Table 7 how financial constraints underpin the predicted 
the survey.  can adopt integer values from 1 to 9, TFP of female- and male-owned businesses (Fig. 2).FC
where a larger value indicates more serious finance In Fig. 2, the problem of gender-biased underfund-
constraints. To mitigate concerns over the simultane- ing is thrown into sharp focus. The figure plots predic-
ity between productivity and finance constraints, we tive margins from a regression in which the financial 
exploit the fact that  is a time-varying variable and constraint variable is split into three binary variables: FC
includes its first lag in the regression. low, medium and high. A low financial constraint cor-
When  is included but without any interaction responds to values up to and including 3. For medium FC
term (columns 2 in Table  7), we find no significant financial constraints, we consider values between 4 
relationship between productivity and the finance and 6 inclusive. And high financial constraints cor-
constraints reported a year earlier. When FC is inter- respond to values above 6. Female business owners 
acted with  (columns 3 in Table 7), the estimate who categorise access to finance as relatively unprob-female
for  remains statistically zero, while the estimate for lematic (financial constraint = low) are associated 2
the interaction term ( 3 ) is found to be significantly 
with predicted TFP (black dotted line), outstripping 
negative at around − 0.07. that predicted for the male control group (grey-dotted 
How do we interpret this finding in connection with line). The reverse is true for businesswomen whose 
our first hypothesis, H1, stating that within each gender business operations are most severely hamstrung by 
group (female vs male), differences in funding explain a lack of finance (financial constraint = high). Here, 
productivity differences? Reading off the results for their similarly underfunded male peers are predicted 
the interaction term and the female dummy, we can to have higher TFP rates, all things equal. We can 
conclude the following: Males reporting financial con- only hypothesise the reasons for this pattern – for the 
straints perform (statistically) no worse nor no better comparatively high TFP predicted for males versus 
than those that do not. For females, this situation is dif- females – when funding is restricted. Males may be 
ferent. Here, the better-funded female-owned businesses able to tap alternative sources of funding – a possi-
are associated with the highest productivity levels. Anal- bility not open to females. Or males may be able to 
ogously, their weaker-funded peers perform significantly manage on a tighter budget, aligning the scale of 
worse. Interpreting these findings, by looking across their operations to match their funding. Whatever the 
the gender category, financial constraints only seem to reason, the TFP of female business owners responds 
bind for the female-owned businesses (significance of most adversely to funding problems. The same is less 
the female × lagged finance covariate). More concretely, true for male business owners.
among female-owned enterprises, ranking finance con-
straints by one place higher (on a scale of 1 to 9) asso- 4.1  Different sources of finance
ciates with an almost 7 per cent dip in productivity, all 
things equal. In other words, the negative female-to- Next, we explore the relationship between productiv-
male productivity gap increases with the severity of the ity and the use of different finance sources in order 
finance constraints reported by females. to deal with Hypothesis 2. In the survey, enterprises 
We can also express these findings in a different are asked what percentage of their working capital 
way. While self-reported finance constraints do not was obtained from the following sources: bank loans, 
explain productivity differences between male-owned own resources, friends and relatives, supplier credit, 
enterprises, they do so for female-owned ones. As a equity and bond and others. Based on this informa-
result, the negative female-to-male productivity gap tion, we generate finance source variables (abbrevi-ated as FS) that measure the percentage of working 
Vol:. (1234567890)
1 3
Africa’s businesswomen – underfunded or underperforming? 
Fig. 2  Gap in female-entre-
preneur TFP and perceived 
financial constraints
capital obtained from each of the above sources. The noting, however. First, the interaction term in the case 
FS variables are time-varying, taking values between of ‘own resources’ is significantly negative, indicat-
0 and 100. ing that the size of the gender-related productivity 
Simple pairwise correlations between the finance gap is more pronounced and the higher the share of 
constraint variable (FC) and the finance source (FS) own resources is for finance. This deterioration in the 
variables (Table  6) reveal that enterprises reporting gender productivity gap, with an increased reliance 
more severe finance constraints typically use a greater of female business owners on their own savings as 
portion of their own resources, and, in the case of a source of funding, is consistent with the idea that 
female-owned businesses, resources from friends female business owners face more severe difficulties 
and relatives to finance their liquidity. In contrast, the than males in sourcing external finance. This find-
use of bank loans, supplier credit and equities/bonds ing ties in with our second hypothesis (H2), which 
is associated with enterprises reporting less severe conjectures that an overreliance on cash savings by 
finance constraints. female business owners is strongly connected to the 
To examine how the different finance sources cor- gender productivity gap.
relate with the gender productivity gap, we repeat our The second issue worth noting relates to sup-
baseline regression, with tfp_ac as our chosen pro- plier credit. Here, the positive and significant inter-
ductivity measure (Table 8). action term indicates that the gender gap seems to 
Unlike the earlier regression, we now include the decrease with increased usage of supplier credit. 
financial source (FS variables) in lieu of the financial This is potentially an interesting finding. While the 
constraints (FC) variables. The columns now report use of supplier credit is associated with lower pro-
our regression results for different sources of finance ductivity among male-owned enterprises (as sug-
– bank borrowing, own resources, financial support gested by the significantly negative coefficient for 
from friends and relatives, supplier credit, and equity lagged FS), this is not the case for female-owned 
and bonds. enterprises. Female-owned businesses with higher 
Our regressions do not reveal strong associations usage of a supplier credit report a narrower produc-
between sources of finance and the gender produc- tivity gap vs similar male-owned businesses. This 
tivity gap. The coefficients for the interaction terms result suggests that better access to supplier credits 
in Table 8 are only statistically significant (at the 10 can play a role in levelling the playing field between 
per cent level) in two cases. Two things are worth credit-constrained male and female businesses.
Vol.: (0123456789)
1 3
 C. Ackah et al.
Table 8  TFP, gender and finance sources
Dependent variable: tfp_ac tfp_ac tfp_ac tfp_ac tfp_ac
Finance source: Bank loan Own resources Friends and relatives Suppliers credit Equity and bond
Female  − 0.125** 0.0799  − 0.118**  − 0.143***  − 0.119**
(0.0509) (0.126) (0.0513) (0.0508) (0.0506)
Lagged FS 0.000710 0.00165  − 0.00640***  − 0.00462** 0.000565
(0.00206) (0.00100) (0.00211) (0.00235) (0.00285)
Female x Lagged FS 0.00492  − 0.00241*  − 0.000524 0.00644* 0.000864
(0.00408) (0.00139) (0.00324) (0.00336) (0.00353)
Age of firm  − 0.00496**  − 0.00516**  − 0.00541**  − 0.00510**  − 0.00478**
(0.00225) (0.00228) (0.00226) (0.00225) (0.00226)
Firm size categories (benchmark is micro)
  Small  − 0.0754  − 0.0721  − 0.0877*  − 0.0622  − 0.0706
(0.0507) (0.0508) (0.0511) (0.0504) (0.0507)
  Medium sized 0.0618 0.0793 0.0625 0.0904 0.0722
(0.0919) (0.0914) (0.0922) (0.0906) (0.0919)
Foreign 0.00214 0.0102  − 0.0219 0.0729  − 0.0102
(0.168) (0.167) (0.167) (0.168) (0.164)
Observations 2917 2917 2917 2917 2917
R-squared 0.090 0.091 0.095 0.091 0.090
Note: OLS estimation with industry, location and year effects. The dependent variable is tfp_ac, which is the logarithm of TFP esti-
mated using the Ackerberg-Caves-Frazer method. Finance source variables are percentages of working capital financed from a given 
source. Firm size categories are micro (0–5 employees), small (6–19 employees) and medium (20 + employees). Robust standard 
errors in parentheses
*** p < 0.01
** p < 0.05
* p < 0.1
4.2 E xports constraint (FC) variable, which, for ease of interpre-
tation, is again split into three categories (FC is low, 
Our analysis now moves to the topic of exports, FC is medium, and FC is high) as before.
a metric of considerable interest to policymak- Due to the small number of exporting enterprises 
ers in LDCs. In concrete terms, we consider in our data, we complement conventional logit esti-
whether the enterprise is able to sell its prod- mations with penalized maximum likelihood logis-
ucts abroad. This section of our analysis maps to tic regressions. The conventional logistic regression 
our third hypothesis (H3) that the gender export- tends to yield biased estimates when the occurrence 
ing gap is driven by underfunded female-owned of events (e.g. exporting among Ghanaian MSMEs) 
businesses. are rare events. The penalized logit, also known as a 
Using maximum likelihood logit estimations, Firth Logit after its first application by Firth (1993), 
we explain the propensity to export, controlling for applies a correction to reduce the above bias.12
enterprise characteristics. Our ultimate aim is to The marginal effect estimates from the export-
shed light on the conjectured gender gap in export- ing regressions are reported in Table  9. The estimates 
ing and the role finance constraints might play in this from the conventional logit and the Firth Logit are very 
relationship.
Our dependent variable is a time-varying binary 
12
variable exporter, taking the value of 1 if an enter-  To perform the penalized logit estimation, we use the firth-
prise sells some of its output abroad and 0 otherwise. logit command in STATA, which was written by Joseph Cov-eney. Note that robust standard error estimation is not allowed 
Finance constraints are measured by the financial by this command.
Vol:. (1234567890)
1 3
Africa’s businesswomen – underfunded or underperforming? 
similar, suggesting that the bias under the conventional even this small difference is economically mean-
method is not substantial. ingful. Moreover, the size of this gender gap 
We document a statistically significant gender remains virtually unchanged when controlling for 
gap in exporting. The marginal effect is − 0.02 lagged productivity (columns 2 of Table 9). Thus, 
(significant at a 5 per cent level), indicating that the gender gap in exporting cannot be explained 
female-owned enterprises are 2 percentage points simply by gender differences in productivity.
less likely to export than male-owned ones (col- Figure 3 illustrates the situation graphically. As 
umns 1 of Table 9). But because Ghanaian micro- the variable ‘financial constraint’ gets ranked from 
enterprises export with a single-digit probability, low to high, so too do female and male business 
Table 9  Gap in female-entrepreneur export participation and perceived financial constraints
Method: ML logit Penalized ML logit (Firth)
(1) (2) (3) (1) (2) (3)
Dependent variable: Exporter Exporter Exporter Exporter Exporter Exporter
Female  − 0.0200**  − 0.0190** 0.0688  − 0.0214**  − 0.0203** 0.0648
(0.0085) (0.0085) (0.0683) (0.0089) (0.0090) (0.0568)
Lagged tfp_ac 0.0084*** 0.0069** 0.0089*** 0.0075**
(0.0030) (0.0030) (0.0034) (0.0034)
Lagged FC categories (benchmark is low)
  Lagged FC is medium 0.0221 0.0207
(0.0190) (0.0203)
  Lagged FC is high 0.0189 0.0164
(0.0152) (0.0163)
Female × lagged FC categories
  Female × lagged FC is medium  − 0.0360  − 0.0350
(0.0351) (0.0335)
Female × lagged FC is high  − 0.1000***  − 0.1024***
(0.0361) (0.0338)
Age of firm  − 0.0004  − 0.0004  − 0.0002  − 0.0004  − 0.0003  − 0.0002
(0.0004) (0.0004) (0.0004) (0.0004) (0.0004) (0.0004)
Firm size categories (benchmark is micro)
  Small 0.0210** 0.0202* 0.0186* 0.0215* 0.0206* 0.0192*
(0.0106) (0.0104) (0.0105) (0.0114) (0.0114) (0.0114)
  Medium sized 0.0730*** 0.0727*** 0.0693*** 0.0768*** 0.0765*** 0.0734***
(0.0130) (0.0128) (0.0122) (0.0141) (0.0140) (0.0139)
Foreign 0.0400** 0.0462** 0.0501*** 0.0440** 0.0507** 0.0557**
(0.0177) (0.0180) (0.0172) (0.0222) (0.0222) (0.0229)
Observations 2,298 2,298 2,298 2,298 2,298 2,298
Pseudo R2 0.136 0.146 0.176
McFadden R2 0.149 0.161 0.195
Note: Maximum likelihood (ML) logit and penalized ML logit (Firth’s method) estimates. All regressions include industry, loca-
tion and year dummies. Financial constraint categories are small (1 to 3), medium (4 to 6) and high (7 to 9). Firm size categories are 
micro (0–5 employees), small (6–19 employees) and medium (20 + employees). The table reports marginal effects. Standard errors 
are in parentheses and obtained with the Delta method. Robust standard errors for the logit regressions
*** p < 0.01
** p < 0.05
* p < 0.1
Vol.: (0123456789)
1 3
 C. Ackah et al.
Fig. 3  Gap in female-entre-
preneur export participation 
and perceived financial 
constraints
owners differ in their propensity to export. Simi- the sector and other confounding variables which 
lar to the pattern for TFP, female business owners would otherwise skew our comparisons. Our esti-
are predicted to have far higher export propensi- mation model builds on the intuition that funding 
ties when access to finance is less of a constraint. supports ex post business performance, all things 
When finance is ranked as an exceptionally high equal. Now, we view our funding-performance 
problem, this export propensity falls to near zero. model through the lens of gender. Where compari-
Returning to our hypothesis concerning exports sons of credit-unconstrained females outperform 
and gender-biased liquidity constraints, we can their male peers from the ‘unconstrained’ group, 
conclude that our consideration of the finance there is a case to be made that the more promis-
constraint indicators and their interactions with ing females are inadequately provided for by banks 
female suggests that the export gender gap is (market failure) since the marginal productivity of 
driven by those female-owned enterprises report- finance is higher for this subgroup.
ing high finance constraints. These enterprises are In line with other studies, our estimations do 
almost 10 percentage points less likely to export indeed reveal a robustly significant gender perfor-
than female businesses reporting low finance con- mance gap. In terms of magnitude, we estimate the 
straints, whose propensity, in turn, does not differ performance differential lies somewhere between 
from the propensity to export of male-owned busi- 11 and 19 per cent, depending on the TFP measure 
nesses (columns 3 in Table 9). used. Interestingly, the severity of the finance con-
straints reported by females is critical. This perfor-
mance gap disappears when gender is interacted 
5  Conclusions with the magnitude of funding constraints. On one 
end of the funding spectrum, there is no significant 
In this first comprehensive study of the funding-per- performance difference between males and females, 
formance gap of Africa’s female business owners, where females receive adequate funding. But female 
we analyse the productivity and exports of females business owners towards the higher end of the fund-
compared to their male peers. Since women are dis- ing constraints scale (one placing higher on a scale 
proportionately active in economic sectors where of 1 to 9) report an almost 7 per cent productiv-
businesses are traditionally smaller and more labour ity dip, all things equal. For male business owners 
intensive (e.g. clothes manufacture), we control for (control group), there is no equivalent productivity 
Vol:. (1234567890)
1 3
Africa’s businesswomen – underfunded or underperforming? 
dip, as financial constraints increase in severity. tackled? Apart from an open confrontation with lend-
Perhaps males can divert funding from alternative ers – urging them to reconsider their lending strate-
sources. But for females, where household and busi- gies – there are other possibilities. These include 
ness are likely to share a common budget, investing greater recognition of the role of supplier credit in 
in her business can mean a zero-sum game. underpinning the liquidity of businesses. Tax conces-
An additional result connects to the source of fund- sions for suppliers which provide these credits might 
ing – not all funding sources are created equal. For represent one avenue for supporting the liquidity of 
females reliant on their own savings, the absolute size female-owned businesses. This policy would most 
of the gender gap is more pronounced. This negative likely help businesswomen at the middle-lower end 
connection between the use of savings by business- of the performance distribution. A further possibility 
women and performance hints at a worrying possibil- for helping the female business owners at the top end 
ity – Females reliant on their own resources to finance of the performance distribution (Africa’s Lionesses) 
their business may be forced to do so from a lack of is examining alternative funding structures – how 
competitive alternatives. best to target equity and bond packages towards these 
On a more positive note, females using supplier exceptional businesswomen. In this way, female busi-
credit report a narrower productivity gap compared ness owners such as Bethlemen Tilahun Alemu, crea-
to their male peers. This result suggests that better tor of the international footwear phenomenon soleRe-
access to supplier credits can play a role in levelling bels, could more easily ramp up their sales capacity, 
the playing field between credit-constrained female- reduce their average costs and target foreign markets.
and male-owned businesses. Conversations with prac-
titioners in Ghana have underscored the importance Funding Open Access funding enabled and organized by 
of this trust-based relationship between business- Projekt DEAL.
women and their suppliers. Supply credit can offer Data Availability For any questions regarding the data used 
something of a lifeline – allowing women to align (ISSER-IGC Panel), please e-mail Charles Ackah (ackah@
their cash receipts to their cash outgoings. ug.edu.gh).
In terms of policy, we can derive a few conclu-
sions. Policymakers are understandably wary about Open Access  This article is licensed under a Creative Com-
dictating to banks and other lenders, which lending mons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any 
criteria they should apply when judging between good medium or format, as long as you give appropriate credit to the 
and bad business risks. Only in the case of systemic original author(s) and the source, provide a link to the Crea-
market failure is there a prima facie case for taking a tive Commons licence, and indicate if changes were made. The 
policy step. Our evidence hints that such market fail- images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated 
ure is indeed a possibility. Despite controlling for var- otherwise in a credit line to the material. If material is not 
ious characteristics of Ghanaian businesses, industry, included in the article’s Creative Commons licence and your 
location and year, the fact still remains that Ghanaian intended use is not permitted by statutory regulation or exceeds 
female business owners seem more seriously impeded the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit 
by a lack of credit access than their male counterparts. http:// creati vecom mons.o rg/ licens es/b y/4. 0/.
For female business owners, this obstacle translates 
into compromised productivity and export potential.
As with any study of this kind, there are caveats. First 
and foremost, some of the phenomena examined are References
still very much marginal activities – what could be con-
sidered atypical for the average firm. Exporting is only Abeberese, A. B., Ackah, C. G., & Asuming, P. O. (2019). 
pursued by 3–4 per cent of firms. Yet, exporting is con- Productivity losses and firm responses to electricity short-ages: Evidence from Ghana. The World Bank Economic 
sidered a key avenue for building the most exceptional Review, 1–18. https://d oi. org/1 0. 1093/ wber/ lhz027
firms – those capable of contesting the global stage. Ackerberg, D. A., Caves, K., & Frazer, G. (2015). Identifica-
We now return to the original question of female tion properties of recent production function estimators. 
business owners and our evidence of a funding gap. Econometrica, 83(6), 2411–2451. https:// doi. org/1 0. 3982/ ecta1 3408
The question remains, how can this problem be 
Vol.: (0123456789)
1 3
 C. Ackah et al.
Agyire-Tettey, F., Ackah, C. G., & Asuman, D. (2018). Gender Economics, 39(2), 383–400. https:// doi. org/ 10. 1007/ 
and returns to entrepreneurship in Africa. International s11187- 011-9 318-7
Journal of Social Economics, 45(12), 1609–1630. https:// Croson, R., & Gneezy, U. (2009). Gender differences in pref-
doi. org/ 10.1 108/ IJSE- 11- 2017-0 549 erences. Journal of Economic Literature, 47(2), 448–474. 
Amin, M., & Islam, A. (2014). Are there more female managers https://d oi.o rg/1 0. 1257/ jel.4 7.2.4 48
in the retail sector? Evidence from survey data in develop- De Mel, S., McKenzie, D., & Woodruff, C. (2008). Returns to 
ing countries. Journal of Applied Economics, 17(2), 213– capital in microenterprises: Evidence from a field experi-
228. https:// doi. org/ 10.1 016/ S1514-0 326(14) 60010-6 ment. The Quarterly Journal of Economics, 123(4), 1329–
Aristei, D., & Gallo, M. (2016). Does gender matter for firms’ 1372. https://d oi.o rg/1 0.1 596/ 1813-9 450-4 230
access to credit? Evidence from international data. Fafchamps, M. (2000). Ethnicity and credit in African manufac-
Finance Research Letters, 18, 67–75. https:// doi. org/ 10. turing. Journal of Development Economics, 61(1), 205–235. 
1108/ ijse- 11-2 017- 0549 https:// doi.o rg/1 0. 1016/ s0304-3 878(99)0 0068-1
Aterido, R., & Hallward-Driemeier, M. (2011). Whose busi- Fafchamps, M., & Schündeln, M. (2013). Local financial devel-
ness is it anyway? Small Business Economics, 37(4), 443. opment and firm performance: Evidence from Morocco. 
https:// doi. org/ 10.1 007/ s11187-0 11- 9375-y Journal of Development Economics, 103, 15–28. https:// 
Aterido, R., Hallward-Driemeier, M., & Pagés, C. (2011). Big doi.o rg/ 10. 1016/j. jdeve co.2 013.0 1. 010
constraints to small firms’ growth? Business environment Fafchamps, M., McKenzie, D., Quinn, S., & Woodruff, C. (2014). 
and employment growth across firms. Economic Develop- Microenterprise growth and the flypaper effect: Evidence 
ment and Cultural Change, 59(3), 609–647. https:// doi. from a randomized experiment in Ghana. Journal of Devel-
org/ 10. 1086/ 658349 opment Economics, 106, 211–226. https:// doi. org/1 0. 1016/j. 
Aterido, R., Beck, T., & Iacovone, L. (2013). Access to finance jdevec o.2 013. 09. 010
in Sub-Saharan Africa: Is there a gender gap? World Field, E., Jayachandran, S., & Pande, R. (2010). Do traditional 
Development, 47, 102–120. https:// doi. org/ 10. 1016/j. institutions constrain female entrepreneurship? A field 
worldd ev. 2013.0 2. 013 experiment on business training in India. American Eco-
Ayyagari, M., Demirguc-Kunt, A., & Maksimovic, V. (2021). nomic Review, 100(2), 125–129. https:// doi. org/ 10. 1257/ 
Are large firms born or made? Evidence from develop- aer. 100.2.1 25
ing countries. Small Business Economics, 57, 191–219. Firth, D. (1993). Bias reduction of maximum likelihood estimates. 
https://d oi. org/ 10. 1007/ s11187-0 19-0 0303-0 Biometrika, 80, 27–38. https://d oi.o rg/1 0.1 093/b iomet/ 80.1. 
Bardasi, E., Sabarwal, S., & Terrell, K. (2011). How do female 27
entrepreneurs perform? Evidence from three develop- Fowowe, B. (2017). Access to finance and firm performance: 
ing regions. Small Business Economics, 37(4), 417–441. Evidence from African countries. Review of Development 
https:// doi.o rg/ 10. 1007/ s11187-0 11- 9374-z Finance, 7(1), 6–17. https://d oi. org/1 0. 1016/j.r df.2 017. 01. 
Beck, T., Demirgüç-Kunt, A., & Maksimovic, V. (2008). 006
Financing patterns around the world: Are small firms dif- Gandhi, A., Navarro, S., & Rivers, D. A. (2020). On the iden-
ferent? Journal of Financial Economics, 89(3), 467–487. tification of gross output production functions. Journal of 
https://d oi. org/ 10. 1016/j.j fine co. 2007.1 0. 005 Political Economy, 128(8), 2973–3016. https:// doi. org/ 10. 
Belitski, M., & Desai, S. (2021). Female ownership, firm age 1086/7 07736
and firm growth: A study of South Asian firms. Asia Girma, S., & Görg, H. (2022). Productivity effects of processing 
Pacific Journal of Management, 38(3), 825–855. https:// and ordinary export market entry: A time-varying treatments 
doi.o rg/ 10. 1007/ s10490-0 19-0 9689-7 approach. Review of International Economics, 30, 836–853. 
Bernard, A. B., & Jensen, J. B. (1999). Exceptional exporter https://d oi. org/1 0. 1111/ roie.1 2580
performance: cause, effect, or both? Journal of Interna- Gonzalez-Uribe, J., & Leatherbee, M. (2018). The effects of busi-
tional Economics, 47(1), 1–25. https:// doi.o rg/ 10.1 016/ ness accelerators on venture performance: Evidence from 
S0022- 1996(98)0 0027-0 start-up Chile. The Review of Financial Studies, 31(4), 1566–
Boudreaux, C., Clarke, G., & Jha, A. (2022). Social capital and 1603. https://d oi.o rg/1 0. 5465/a mbpp. 2017.1 1659a bstr act
small informal business productivity: The mediating roles Guérin, I. (2006). Women and money: Lessons from Senegal. 
of financing and customer relationships. Small Business Development and Change, 37(3), 549–570. https://d oi. org/ 10. 
Economics, 59(3), 955–976. https:// link. spring er. com/ 1111/j. 0012-1 55X. 2006. 00490.x
articl e/1 0.1 007/ s11187-0 21- 00560-y Halkias, D., Nwajiuba, C., Harkiolakis, N., & Caracatsanis, 
Campos, F., & Gassier  M. (2017). Gender and enterprise S. M. (2011). Challenges facing women entrepreneurs in 
development in sub-Saharan Africa: A review of con- Nigeria. Management Research Review, 34(2), 221–235. 
straints and effective interventions. World Bank Policy https:// doi.o rg/ 10.1 108/0 1409 17111 11028 21
Research Working Paper no.8239. https://d oi. org/ 10. 1596/ Hansen, H., & Rand, J. (2014). The myth of female credit dis-
1813- 9450-8 239 crimination in African manufacturing. Journal of Devel-
Chaudhuri, K., Sasidharan, S., & Raj, R. S. N. (2020). Gender, opment Studies, 50(1), 81–96. https:// doi.o rg/1 0. 1080/ 
small firm ownership, and credit access: Some insights 00220 388. 2013.8 49337
from India. Small Business Economics, 54(4), 1165–1181. Henry, C., Foss, L., & Ahl, H. (2016). Gender and entre-
https:// doi. org/ 10. 1007/ s11187-0 18-0 124-3 preneurship research: A review of methodological 
Coad, A., & Tamvada, J. P. (2012). Firm growth and barriers approaches. International Small Business Journal, 34(3), 
to growth among small firms in India. Small Business 217–241. https:// doi. org/ 10.1 177/0 2662 42614 549779
Vol:. (1234567890)
1 3
Africa’s businesswomen – underfunded or underperforming? 
Islam, A., & Amin, M. (2016). Women managers and the gender- Melitz, M. J. (2003). The impact of trade on intra-industry 
based gap in access to education: Evidence from firm-level reallocations and aggregate industry productivity. Econo-
data in developing countries. Feminist Economics, 22(3), metrica, 71(6), 1695–1725. https:// doi. org/ 10. 1111/ 1468- 
127–153. https:// doi. org/ 10. 1080/ 13545 701. 2015.1 0817 05 0262.0 0467
Islam, A., Muzi, S., & Amin, M. (2019). Unequal laws and the MIWE Report (2022).  The mastercard index of women entre-
disempowerment of women in the labour market: Evi- preneurs. How targeted support for women-led business 
dence from firm-level data. The Journal of Development can unlock sustainable economic growth. https://w ww. 
Studies, 55(5), 822–844. https:// doi.o rg/ 10.1 080/0 0220 maste rcard. com/ news/ media/ phwev xcc/ the- maste rcard- 
388. 2018. 14870 55 index- of- women- entrep rene urs. pdf. Accessed 05/06/2023
Kanze, D., Huang, L., Conley, M. A., & Higgins, E. T. (2018). Muravyev, A., Talavera, O., & Schäfer, D. (2009). Entrepre-
We ask men to win and women not to lose: Closing the neurs’ gender and financial constraints: Evidence from 
gender gap in startup funding. Academy of Management international data. Journal of Comparative Economics, 
Journal, 61(2), 586–614. https://d oi. org/1 0. 5465/a mbpp. 37(2), 270–286. https://d oi. org/ 10. 1016/j. jce.2 008. 12. 001
2017. 10010 abstra ct Naguib, R., & Jamali, D. (2015). Female entrepreneurship in 
Kelley D, Bosma N, Amorós JE (2010) Global entrepreneur- the UAE: a multi-level integrative lens. Gender in Man-
ship monitor: 2010 Global Report. https:// entre preno agement: An International Journal, 30(2), 135–161. 
rskap sforum. se/ wp- conte nt/ uploa ds/ 2011/ 02/ GEM- 2010- https:// doi. org/1 0. 1108/ GM- 12-2 013- 0142
Global-R eport. pdf Oaxaca, R. L. (1973). Male-female wage differentials in urban 
Klapper, L. F., & Parker, S. C. (2011). Gender and the busi- labor markets. International Economic Review, 14, 693–
ness environment for new firm creation. The World Bank 709. https:// doi. org/ 10. 2307/2 5259 81
Research Observer, 26(2), 237–257. https:// doi. org/ 10. Olley, G. S., & Pakes, A. (1996). The dynamics of productivity 
1093/ wbro/ lkp032 in the telecommunications equipment industry. Economet-
Langevang, T., Gough, K., Yankson, P. W. K., Owusu, G., & rica, 64, 1263–1297. https:// doi. org/ 10. 3386/ w3977
Osei, R. (2015). Bounded entrepreneurial vitality: The Owoo, N. S., Oduro, A. D., & Ackah, C. (2019). Exploring 
mixed embeddedness of female entrepreneurship. Journal firm performance and growth among own-account and 
of Economic Geography, 91(4), 449–473. https://d oi.o rg/ micro enterprises in Ghana. International Journal of 
10. 1111/ ecge.1 2092 Entrepreneurship and Small Business, 38(3), 294–311. 
Levinsohn, J., & Petrin, A. (2003). Estimating production func- https:// doi.o rg/ 10. 1504/ ijesb.2 019.1 03431
tions using inputs to control for unobservables. Review of Pham, T., & Talavera, O. (2018). Discrimination, social capi-
Economic Studies, 70, 317–341. https:// doi. org/ 10.3 386/ tal, and financial constraints: The case of Viet Nam. World 
w7819 Development, 102, 228–242. https:// doi. org/ 10. 1016/j. 
Loaba, S. (2022). The impact of mobile banking services on sav- world dev.2 017.1 0.0 05
ing behavior in West Africa. Global Finance Journal, 53, Presbitero, A. F., Rabellotti, R., & Piras, C. (2014). Barking up 
100620. https:// doi. org/1 0. 1016/j.g fj.2 021.1 00620 the wrong tree? Measuring gender gaps in firm’s access 
Maden, C. (2015). A gendered lens on entrepreneurship: to finance. Journal of Development Studies, 50(10), 1430–
women entrepreneurship in Turkey. Gender in Man- 1444. https:// doi. org/ 10. 1080/ 00220 388.2 014. 940914
agement, 30(4), 312–331. https:// doi. org/ 10. 1108/ Rotemberg, M. (2019). Equilibrium effects of firm subsidies. 
GM-1 1- 2013- 0131 American Economic Review, 109(10), 3475–3513. https:// 
Manolova, T. S., Brush, C. G., Edelman, L. F., & Green, P. doi. org/ 10. 1257/a er. 20171 840
G. (2002). Internationalization of small firms: Personal Rovigatti, G., & Mollisi, V. (2018). Theory and practice of 
factors revisited. International Small Business Jour- total-factor productivity estimation: The control function 
nal, 20(1), 9–31. https://d oi. org/ 10. 1177/ 02662 42602 approach using Stata. The Stata Journal, 18(3), 618–662. 
201003 https:// doi. org/ 10.1 177/ 15368 67x180 1800 307
McClelland, E. (2004). Irish female entrepreneurs: Mapping Satpathy, L., Chatterjee, B., & Mahakud, J. (2017). Financial 
the route to internationalisation. Irish Journal of Manage- constraints and total factor productivity: Evidence from 
ment, 25, 92–109. Indian manufacturing companies. Journal of Management 
McClelland, E., Swail, J., Bell, J., & Ibbotson, P. (2005). Fol- Research, 17(3), 146–162.
lowing the pathway of female entrepreneurs: A six-coun- Taiwo, A. (2023). Entrepreneurial financing options of female 
try investigation. International Journal of Entrepreneurial businesses and enterprise performance in Nigeria. In: New 
Behaviour & Research, 11(2), 84–107. https:// doi. org/1 0. Horizons and global perspectives in female entrepreneur-
1108/1 3552 550510 59052 7 ship research (pp. 75–92). Emerald Publishing Limited. 
McKenzie, D. (2017). Identifying and spurring high-growth https://d oi. org/1 0.1 108/ 978-1- 83982- 780- 820231 003
entrepreneurship: Experimental evidence from a business Van Biesebroeck, J. (2005). Exporting raises productivity in 
plan competition. American Economic Review, 107(8), sub-Saharan African manufacturing firms. Journal of 
2278–2307. https://d oi. org/ 10. 1596/ 29067 International Economics, 67(2), 373–391. https:// doi.o rg/ 
McKenzie, D., & Paffhausen, A. L. (2019). Small firm 10.1 016/j. jinte co. 2004.1 2.0 02
death in developing countries. Review of Economics Verheul, I., Stel, A. V., & Thurik, R. (2006). Explaining 
and Statistics, 101(4), 645–657. https://d oi. org/ 10. female and male entrepreneurship at the country level. 
1596/1 813-9 450-8 236
Vol.: (0123456789)
1 3
 C. Ackah et al.
Entrepreneurship and Regional Development, 18(2), 151– Wooldridge, J. (2009). On estimating firm-level production 
183. https://d oi. org/ 10. 1080/ 089856 2050 05320 53 functions using proxy variables to control for unobserva-
Wagner J (2019) International trade in goods: Evidence from bles. Economics Letters, 104, 112–114. https:// doi. org/ 10. 
transaction data. Published by WSPC. https:// doi. org/ 10. 1016/j.e conle t.2 009. 04. 026
1142/1 1175
Wood, A., & Mayer, J. (2001). Africa’s export structure in a Publisher’s note Springer Nature remains neutral with regard 
comparative perspective. Journal of Economics, 25(3), to jurisdictional claims in published maps and institutional 
369–394. https:// doi.o rg/1 0.1 093/c je/ 25.3.3 69 affiliations.
Vol:. (1234567890)
1 3