Received: 10 November 2018 Revised: 8 March 2019 Accepted: 13 September 2019
DOI: 10.1002/ijfe.1800
R E S E A R CH AR T I C L E
Are competitive microfinance services worth
regulating? Evidence from microfinance institutions in
Sub-Saharan Africa
Amin Karimu1 | Samuel Salia2 | Javed G. Hussain3 | Ishmael Tingbani4
1Department of Economics, University of
Ghana Business School (UGBS), Legon, Abstract
Ghana In recent years, there is increasing appetite for regulation of microfinance ser-
2Department of Accounting and Finance, vices after the 2008 financial crisis. Policy questions such as whether competitive
De Montfort University, Leicester, UK microfinance institution (MFI) requires strong regulation to reduce, for example,
3Department of Accounting and Finance,
credit risk or competition and regulation operate in the opposite direction,
Birmingham City University,
Birmingham, UK which each tends to dampen the effect of the other, are an empirical issue that
4Department of Accounting and Finance, this paper provides answers based on data on Sub-Saharan Africa (SSA) for the
Bournemouth University, period 1995–2015, utilizing panel data approaches. Finding from the study indi-
Bournemouth, UK
cates that low competition increases credit risk among MFIs in SSA, which regu-
Correspondence lation helps reduce such behaviour. The effect of regulation on credit risk is
Samuel Salia, De Montfort University, conditional on the level of competition, at the first percentile of competition
Leicester, UK.
Email: s.salia@wlv.ac.uk (imply more competition); regulation does not reduce credit risk behaviour of
MFIs but does at competition level above the 25th percentile (imply less compe-
tition). Regulation, on the other hand, does not affect operational risk at any
level of competition. These findings have implications for policy formulation on
the regulation and operations of MFIs in SSA. Our findings suggest that the MFI
industry could be regulated efficiently if policymakers develop policies targeted
at reducing credit risk exposures of MFIs than their exposure to operational risk.
KEYWORD S
Competition, financial services, microfinance, portfolio risk, regulation
J E L C LA S S I F I C A T I ON
G23; G32; G38
1 | INTRODUCTION Saharan Africa (SSA). The severity of the implication of
such a vast size of the unbanked population on poverty
Microfinance institutions (MFIs) play an essential role in alleviation and lack of job creations especially for SSA is
most developing countries as they provide financial ser- that, majority, close to 90% of the unbanked population
vices, including poverty reduction intervention measures are in rural areas (Gentil and Servet, 2002), where pov-
to a significant share of the population that is unserved erty levels are endemic, with fewer job opportunities.
by the formal financial institutions. About 2.5 million Therefore, the lack of banking services to mobilize funds
adult population of the World is unbanked in 2014 at lower cost for the impoverished rural population to
(World Bank, 2016), which the majority live in Sub- create a small business, invest into agriculture to provide
Int J Fin Econ. 2019;1–17. wileyonlinelibrary.com/journal/ijfe © 2019 John Wiley & Sons, Ltd. 1
2 KARIMU ET AL.
the needed food requirement, and earn some income, fur- imposed, it will create a less favourable outcome than if
ther perpetuates the incidence of poverty in such areas. regulation is not imposed.
Over the recent decade, as a consequence of the prob- The recent financial crisis has increased appetite for
lems associated with the poor not having access to the for- more regulation towards the financial sector in general
mal banking services on their livelihood, poverty outcomes and may also be the case for MFIs for countries that have
and the associated social menace, coupled with the promis- experienced some Ponzi scheme-types of operations of
ing positive effects that MFIs are making, especially in serv- some MFIs such as in Ghana, DKM Diamond
ing the poor unbanked segment of the population, have Microfinance Company Limited that went bust in mid-
resulted in a plethora of different MFIs in developing coun- 2015 due to its Ponzi type of scheme it offered to clients.
tries, some with goals beyond the social intervention or However, the policymaker will have to assess the two
developmental goals such as pure profit motives. This phe- opposing effects to determine if regulation is necessary,
nomenon is partly a result of the success story of the especially in the case of MFIs given their core mandate to
microfinance model (Garrity and Martin, 2018), which leads pool resources to provide microloans to the segment of
to an increase in the commercial oriented type of MFIs to the society, who cannot access the main financial institu-
enter the microfinance segment of the financial market. tions such as banks for such microloans. Therefore,
The increase in MFIs from both types—development whether it is optimal for the government to regulate
oriented MFIs and commercial oriented MFIs—in recent MFIs is conditional on the level of competition and the
years in developing countries creates competition amongst consequences thereof relative to less competition.
these firms to provide financial services to the poor Whether the policymaker should regulate MFIs or not
(Bateman, 2019). The increase in competition amongst is an empirical question, which has not received much
MFIs due to the increase in the number of MFIs operating attention, especially about risk-taking behaviour. To the
in the World financial market from 10 million in 1997 to best of our knowledge, there is no study in the MFIs liter-
more than 100 million in 2007 (Assefa et al. 2013) creates ature that empirically examined the joint effect of regula-
some level of competition that may have negative conse- tion and competition on risk-taking behaviour of these
quences such as taking unnecessary risk in the quest to firms, especially in the SSA context. The closest studies
outcompete competitors for clients and markets. we have found in the literature are: Assefa et al. (2013),
Economic theory suggests that competition will result who looked at the effect of competition on performance;
in lower prices for products produced due to lower cost of Hartarska and Nadolnyk (2011), Purkayastha et al.
production, more output, and generally a welfare improve- (2014), Triki et al. (2017), and De Quidt et al. (2018), they
ment for the society relative to the less competitive market focussed on market structure, the effect of regulation on
environment. However, unhealthy competition may also performance, or growth of the MFIs.
result in competing firms taking unnecessary pricing, mar- This paper aims to provide some answers in that
keting, organisational, and overall business strategies that regard by providing an understanding of the relationships
expose them to more risk. On the other hand, having few between risks (portfolio risk and operational risk) and
firms with significant power may also create excessive both regulation and competition in the case of SSA. We
risk-taking behaviour in the absence of regulation as this achieved this by adopting a sample of 1,574 MFIs in SSA
was the case in the 2008 financial crisis. The question for the period 1995–2015. Evidence from the study sug-
whether competition is good or bad will depend on the gests that low competition increases credit risk amongst
strength in these two opposing effects of competition rela- MFIs in SSA, which regulation helps reduce such behav-
tive to few firms with significant market power, the level iour. In particular, we find that the effect of regulation on
of competition and whether the reference is to the firm, credit risk is conditional on the level of competition, at
consumer or clients, or society. If competition is creating the first percentile of competition (imply more competi-
more risk-taking behaviour relative to the lower prices tion); regulation does not reduce credit risk behaviours of
and increase in output (outreach in the case of MFIs) MFIs but does at competition level above the 25th per-
effect, it is prudent that authorities regulate the centile (imply less competition). On the other hand, we
microfinance market to curb competition and reduce the find that regulation does not affect operational risk at
unnecessary risk-taking behaviour of the MFIs. Therefore, any level of competition.
whether the government should regulate MFIs will This paper aims to contribute to the literature through
depend on whether competition was high and as a conse- assessing the sequencing impact of market concentration
quence, creating unhealthy competitive behaviours and regulation on the risk of MFIs in SSA. Contribution
amongst firms in the microfinance industry. If competition of this article is in three folds: first, to provide an under-
is not creating unhealthy outcomes and regulation is standing of the nature of relationships between credit
KARIMU ET AL. 3
risk, competition, and regulation of MFIs; second, to which has been reported to be a constraint to access
assess whether regulation and competition reinforces finance for entrepreneurial activities within developed
each other or substitute in terms of their effect on credit economies (Hussain et al. 2010).
risk; and third, the role of both competition and regula- Information asymmetry is the dominant factor for the
tion on operational risk of MFIs and whether their effects underdevelopment of the financial sector in SSA
are different in comparison with their effect on credit (Marcelin and Mathur, 2014; Smith, 2015; Domeher et al.,
risk. Literature review and the general narrative on SSA 2017). In the absence of information on clients' creditwor-
show that the factors examined by previous studies thy, lack of collateral and financial history, lending man-
(Kablan, 2014; Cull et al., 2015; Ayele, 2015) focussed agers are not equipped to evaluate risk, and this adversely
merely on measurement effect of portfolio risk on profit- affects the loan decision process. Therefore, information
ability, outreach, and repayment rates. The impact of reg- asymmetry gives rise to type 1 error, where a viable loan
ulation and market concentration jointly on credit application is rejected, and type two error, where a non-
(portfolio) risk of MFIs is omitted and therefore, policy viable loan application is approved (Deakins and Hussain,
questions such as whether having a competitive MFI 1994). This has the potential to give rise to defective loans
requires strong regulation to reduce, for example, portfo- that could lead to portfolio underperformance and a bar-
lio risk (credit risk), or they operate in the opposite direc- rier to the development of financial services.
tion, which each tends to dampen the effect of the other, Second, the extant literature (Kuku and Jakpal, 2015;
cannot be comprehensively answered based on the exis- Mujeri, 2015; Tumwine et al., 2015; Gohar and Batool,
ting literature. 2015) that indicates more impoverished populations do
The rest of the paper is structured as follows: not have the capacity to contract large loans but are
Section 2 presents the literature review for the study, slightly attracted to a large number of smaller size loans
followed by the study's conceptual framework in that have higher administrative cost, which serves as a
Section 3. Section 4 discusses the methodology; Section 5 barrier to the development of MFIs portfolio. This has a
presents the empirical evidence and sensitivity analysis. negative influence on the operation cost of MFIs and by
The conclusion and summary are presented in Section 6. extension, the overall portfolio profitability. Third, the
poor who operates at the margins of economic systems,
carries higher default risk, and lacks entrepreneurial
2 | RELATED LITERATURE capacity coupled with lack of financial track record and
collateral (Fafchamps, 2014; Fletschner and Kenney,
Both in its institutional range and in its penetration of 2014; Singh and Huang, 2016; Wellalage and Locke,
financial markets, MFIs have played a significant role in 2016; Matamanda and Chidoko, 2017; Otchere et al,
most developing economies (Mosley, 2009; Postelnicu 2017) adversely impacts on their chances of accessing
and Hermes, 2017) in the area of mobilizing funds for the finance. Limited earnings by the poor are broadly follow-
weak and small business and as a consequence, contrib- ing the economic interest of traditional banks to avoid
ute to the growth and development of developing coun- profound portfolio risk. According to Blazy and Will
tries. The financial services provided by MFIs enhance (2013) and Duarte et al. (2017), banks primarily rely on
the ability of the poor to become bankable (Ledgerwood collateral for reduction of loan loss in the event of
et al., 2013), leading to a positive impact on income and default. The existence of institutional and regulatory
asset accumulation level which potentially increase the rigidities suggests that the existence of banks alone does
standard of living of microfinance customers through the not benefit all section of the population equally.
establishment and expansion of business activities The above narrative highlights market failures, finan-
amongst these customers (Beisland, 2017). cial exclusion, and limited access to finance for the weak
Microfinance has the potential to empower a signifi- segment of the population in developing countries. To fill
cant portion of the workforce in developing countries this gap, MFIs reach out to financially excluded
and contribute towards financial sector deepening. In the populations, who have no collateral for loans to engage
case of SSA, about 90% of the population in rural areas in entrepreneurial activities (Islam, 2009). MFI shares a
are unbanked (Gentil and Servet, 2002), forcing a broad common objective of financial inclusion of the informal
swath of the population to operate without financial ser- sector within an economy to promote enterprise develop-
vices resulting in lack of capital and a considerable share ment. MFIs not only provide a loan, savings, and money
of the population, which is persistently impoverished. services, but also support them with social support, train-
Underdevelopment of the financial sector is attributable ing, and opportunity to network (Sengupta and
to three specific factors. First, information asymmetry, Aubuchon, 2008; Jain and Moore, 2003). The growth and
4 KARIMU ET AL.
presence of MFIs have increased over time. There are constitutes a significant threat to the overall survival of
currently around 3,700 MFIs, which provide collateral- the microfinance sector. Similarly, Assefa et al. (2013)
free loans to 230 million customers in more than found that for the periods 1995–2008, the competition
100 countries (Gul et al., 2017). Given the scope, impact, was negatively associated with the portfolio performance
and outreach of MFIs, many efforts have been spent on of 362 MFIs in 73 developing countries. Hartarska and
understanding how MFIs overcome challenges where Nadolnyk (2011) suggest that the jury is out on whether
traditional banks have failed to do so (Morduch, 2013; regulation improves the poor's access to finance but indi-
Weaver, 2016; Gan et al., 2017). rectly also enhances the sustainability of MFIs; however,
MFIs' role and function have evolved over the years regulation benefits organisations in obtaining and pro-
(Helms, 2006), and policymakers have accepted it as a moting savings amongst its members.
tool to alleviate poverty (Sengupta and Aubuchon, 2008; Empirical evidence is sparse on MFIs' regulations, but
Kanak and Liguni, 2007). The significant innovation that the study by Hartarska and Nadolnyk (2011) suggest that
gained traction over the last decade is the peer lending regulating MFIs does not directly affect the performance
that substitutes asset collateral with nontangible assets either regarding operational self-sustainability or out-
such as reputation, group standing, and community reach. Furthermore, Purkayastha et al. (2014) reported
cohesion. According to Wenner et al. (2007), group mem- that complying with prudential regulations and the asso-
bers undertake to enforce loan contracts; thus, anytime a ciated supervision stifled the growth of MFIs and inter-
group member defaults in repayments, the group is mediation efficiency in places such as India. This
obliged to pay the loan with their resources. If they do explanation has found support amongst views that
not, the group is at risk in losing access to future loans oppose excessive regulation of MFIs in SSA (Madestam,
(Al Mamun, 2012). These groups guarantee that practice 2014; Triki et al., 2017).
is similar to the concept of group insurance. In this case, Literature review on regulation of MFIs strand of
the MFI uses a compilation of people to reduce portfolio research in emerging markets has either shown discus-
risk; it is in every member's interest to ensure that the sions that assess the effect of portfolio risk on MFIs' per-
other members pay for their loans. The recipient interest formance (Magali, 2013; Castellani and Cincinelli, 2015;
deepens assurance and makes the MFIs more secured to Kusi, et al., 2017) or the impact of regulation on perfor-
issue noncollateralized loans. mance of MFIs (Barry and Tacneng, 2014; Yu et al.,
Through group collateral mechanism, MFIs have 2014; John, 2015; Spratt, 2016; Adams, 2017; Siwale and
emerged as an essential source of entrepreneurial finance Okoye, 2017). Studies by Pashkova et al. (2016) and
for the unbanked in SSA (Boateng et al., 2015). In the Chikalipah (2017) have used similar data from Africa to
past three decades, however, microfinance practice has analyse the environment of MFIs in SSA, but they too
evolved. On the one hand, crowding in the microfinance have focussed on business models and institutional envi-
industry highlights a more comprehensive and competi- ronments, respectively; regulation and market competi-
tive market mechanism. On the contrary, the financial tion and its impact on portfolio risk have not been
inclusion of the unbanked in low-income countries has explored to the best our knowledge. Furthermore, the
caught the attention of regulatory agencies that seek to extant literature has hardly explored operational risk,
protect the financial sector. This is consistent with the notably in conjunction with regulation and competition.
theory that competitive forces in finance increase scale The above review indicates that studies on the effect of
and homogeneity. Thus, there is the need to create a reg- regulation and market competition jointly on portfolio and
ulatory structure that can counteract this and maintain a operational risk of MFIs are rare in general and nonexis-
precarious ecosystem of financial institutions (Mwega, tence in the case of SSA, which makes it difficult to answer
2014; D'Espallier et al., 2017). policy questions such as whether having a competitive
The theory of competitive force has been tested and, MFI requires strong regulation to reduce, for example,
with few exceptions, found consistent with data in a wide credit risk (portfolio risk), or they operate in the opposite
variety of markets. For example, Anginer and Demirgüç- direction, which tends to dampen the effect of th1e other.
Kunt (2014) found systematic fragility to be less pro-
nounced in countries with institutions that allow for
better public and private monitoring of financial institu- 3 | CONCEPTUAL FRAMEWORK
tions in competitive conditions. The opposite is observed
in Appiah-Konadu et al.'s (2016) analysis; they found that MFIs operate at the margin of society, and borrowers oper-
MFIs' crowding coupled with limited regulatory supervi- ate at the fringes, often are not bankable, lack collateral,
sion generates high portfolio risk in Ghana and and suffer from financial asymmetry (Boateng et al., 2015).
KARIMU ET AL. 5
Whereas MFIs have financial constraints, and their goals optimum regulation assists to manage competition
are to mitigate adverse effects of poverty, impact posi- amongst MFIs that support them to develop optimum
tively on the welfare of society as a whole. In achieving portfolios, operational procedures, and reduce risk. The
these objectives, MFIs aim to develop lending strategies study builds lessons and the inferences of mainstream
to ensure that risk is managed, and capital pilferage is banks (Dewatripont, 2014; Berger et al., 2016) and lends
minimized. This section of the paper examines (a) effect support to the argument that portfolio adequacy, efficient
of regulation, (b) the role of competition amongst MFIs, and appropriate regulation, and the market discipline
and (c) portfolio approach employed by MFIs to manage affect the performance of MFIs. Therefore, there is a case
risk through adopting lending portfolio methods and bet- to measure the joint effect of regulations and market
ter operational procedures. competition on portfolio risk of MFIs in SSA.
For the efficient functioning of MFIs, competition
plays a central role, and regulation is a prerequisite for 4 | METHODOLOGY
this purpose as it aids the structural development of the
market at large and the institution in particular. Efficient 4.1 | Sample and data
and fair regulation brings transparency in its conduct
and enables performance to be measured. Within this The source of data for the study is from the Microfinance
context, competition fosters the efficient allocation of Information Exchange (MIX) Market dataset that covers
resources and negates imperfections. Pragmatisms dictate for the period 1995–2015 for 3856 microfinance firms for
lenders to make lending decisions to alleviate poverty SSA countries. The period was chosen based on data
through the entrepreneurial financing of individual and availability for many microfinance firms in SSA coun-
groups which are not bankable, yet at the same time pre- tries. The dataset is a panel but due to differences in the
serve MFIs' assets, earn a return to ensure continuity of year of operations across different MFIs within and
funding provision. To achieve this objective, MFIs adopt between countries in the dataset, we have an unbalanced
a portfolio approach that manages its lending portfolio. panel. Also, due to missing observations for some of the
Figure 1 illustrates the relationship between regulation, variables, our final sample reduced to 1,574 firms. The
competition, and risk. This connectivity between the summary statistics for each of the variables for the analy-
three pillars serves as a continuous loop to enhance one sis are presented in Table A1 of the Appendix, which
another's performance. reflects the average values and variability within a coun-
Regulation is integral to issues related to corporate try and across countries for each of the variables that are
governance of MFIs; it ensures that processes and proce- defined in Section 4.2.
dures applied to sanction loans are consistent across
institutions. The consistency in lending methodology pro- 4.2 | Variables definitions
motes operational and allocational efficiency. Competi-
tion serves to lower the price of a loan, to improve MFI credit risk is measured as impaired loans to gross
services, and to provide choice to the borrower. At the loans and advances and used as the dependent variable
same time, effectiveness and transparency enable MFIs in this study. Chaibi and Ftiti (2015) argue that credit risk
to adopt a sectoral portfolio approach to reduce measured as impaired loans divided by gross loans is a
unsystematic risk and also use advanced due diligence better representation of credit risk as it reflects actual
and screening methods to lend to viable borrowers and credit risk or loss that pertains to a specific time. The
implement better operational procedures. However, this portfolio at risk is used as a proxy and is estimated as the
method risks alienating or further disadvantaging the proportion of the loan portfolio of an MFI that is overdue
impoverished section of the population. for 30 days and is at risk of not being settled. Differently
The central proposition of this empirical study is that phrased, the portfolio at risk >30 variable reflects the
MFIs portfolio risk is not independent of market compe- actual risk of a delinquency problem because it takes into
tition and the regulatory environment. Efficient and account the full amount of the loan at risk predominantly
when the loan payments are small (Ledgerwood, 2000).
Portfolio in itself specifies the aggregate incomes accessi-
Risk Competition ble for the MFIs to disburse it as credits to its customers.
Portfolio quality is a way of determining how best the
organization can safeguard its portfolio. It is a crucial
Regulation aspect of performance evaluation, as it is the most signifi-
cant source of risk for most business organisations that
FIGURE 1 Link between risk, competition, and regulation exist in their assets portfolio. Hence, to their best effort,
6 KARIMU ET AL.
MFIs need to sustain the value of their investments. For faced by MFIs might be influenced dramatically due to the
our study, we consider portfolio at risk over 30 days (PAR competitive nature of the market in which they operate.
>30 days) as used in Assefa et al. (2013). We include this The Lerner index is our measure of MFI-level of com-
variable to determine how well an MFI is managing its petition (see Aghion et al., 2005). The index ranges
risks as it provides services to its clients. between 0 and 1, where a value close to zero is an indica-
Operational risk is defined as the loss resulting from tion of strong competition, whilst close to 1 suggests less
inadequate or failed internal processes, people, and sys- competition. The index is of form LI = P−MCP , where p is
tems or external events (Chavez-Demoulin et al., 2006).
the output price proxy by yield on gross loan portfolio,
This is proxied by computing the coefficient of variation
and MC is the marginal cost of the firms. High (low)
(CV) of write-off ratio of loans by MFIs. The CV is then
used as a proxy for operational risk. index implies low (high) competition. In estimating the
The dataset also contains information on whether the Lerner index, we follow an approach by Assefa et al.
MFI is regulated or not. Regulation is measured as a (2013) by constructing a translog cost function as follows:
1 X2 1X2 X2 X2
lnTCit = β0 + β1ln yit + β2ln y
2
it + γklnwk,it + ϕ jln y2 2 it
 lnw j,it + θk, jlnwk,it
X k=1 j=1 k=1 j=12
 1 2 1
XX
lnw j,it + φ jlnw j,it + ρtrend+ τtrend
2 + ζjklnw j,it + δtrend+ εit ð1Þ
j=12 2 j< k
dummy variable that takes a value of 1 if the MFI is regu- where TC is the total cost of firm i for time t; yit
lated and 0 if it is not regulated. This is to determine denotes the put of the firms; wit is a vector of firms' input,
whether regulated MFIs are exposed to more risk than which in this study constitutes labour and capital; trend
their counterparts. Gietzen (2017) found no association is the time trend to capture technological progress, whilst
between regulation (regulatory quality from the World εit is random error term.
Bank governance indicators) and risk exposure and thus In order to get the marginal cost (MC), we take the
conclude that regulators might see no need intervening first derivatives with respect to the output to obtain the
in the sector due to seemingly lower liquidity risk. Their MC function as shown below
regulatory index is generic at the country level, not MFIs " #
specific regulation and therefore, considering MFIs TC X2
MC itit = β + β ln y + ϕ lnw j,it + δtrendit ð2Þ
related regulation will likely provide a better understand- y 1 2 it jit j=1
ing of the role of regulation on risk-taking by MFI. To the
best of our knowledge, the only available MFI regulatory We then estimate the MC function because it cannot
variable is the dummy variable in the MIX Market be inferred directly from the data.
dataset that indicates whether the MFI is regulated or The advantages of the Lerner index relative to other
not. We, therefore, rely on this variable as our regulation measures of competition are: (a) Lerner index enables us
variable. to investigate competition at the firm-level, and (b) it var-
Competition is one of the critical variables for our ies over time which again gives us the opportunity to
study. However, this variable is not readily available; we measure competition over some years (Assefa
preferably have to compute it by considering existing liter- et al., 2013).
ature on the best measure for competition. There is no In estimating the MC in the equation above, the fol-
unanimity in the literature of the optimal way to measure lowing variables were used. First, the total cost for each
competition. The Lerner index is our primary choice to firm (TC), which is the aggregate of all expenses incurred
proxy competition due to its relatively good properties as by an MFI in a given financial year. It consists of both
presented in the next paragraph. The inclusion of competi- operating and financial expenses that the firm incurred
tion for the analysis is to assess whether the risk-taking in running the affairs of the business. The sum of operat-
behaviour of MFIs increases with competition or other- ing and financial expenses incurred by the firm is used to
wise. It is crucial to include competition because the risk proxy for this.
KARIMU ET AL. 7
The output variable (y) for each MFI is the gross loan is measured as a dummy variable that takes a value of
portfolio, which consists of all outstanding principal for all 1 if the MFI is regulated and 0 for unregulated MFI; com-
outstanding client loans, including current, delinquent, and petition is measured using two different competition indi-
restructured loans, but not loans that have been written off. ces (Lerner index and Herfindahl-Hirschman Index
It does not include interest receivable and employee loans. (HHI)), X is a vector of controls that include business
In constructing the cost function, we considered only two size, financial cost of the microfinance firm, operational
inputs, which are very crucial to the operations of MFIs. efficiency of each microfinance firm, financial strength of
These include the cost of labour and the cost of capital. The microfinance firms, financial revenue, and both firm (ηi)
cost of capital refers to the cost of equity and debt used in and time fixed effects (μt) to account for unobserved het-
financing the microfinance business. It is the opportunity erogeneity; εit is a random error term.
cost of taking a specific investment. It is measured as the The firm size is measured as the natural log of gross
ratio of financial expenditure to total liabilities of the firm loan portfolio (Barry and Tacneng 2014). Following the
within the financial accounting year. economies of scale and diseconomies of scale theories,
The cost of labour, on the other hand, consists of both the study expects a positive or negative effect of MFIs size
direct and indirect cost incurred by employees for render- on credit risk. That is, following the economies of scale,
ing services to the firm. In estimating the labour cost, the larger MFIs have the needed resources, both financial
study took the ratio of personnel expenses to total assets and human, and the capacity to monitor and supervise
as a proxy, with the assumption that the primary compo- their customers or borrowers; thus, reduction in credit
nent of operating costs is the personnel salaries. To con- risk. However, following the diseconomies of scale, larger
trol for important unobservable such as technology, we MFIs are overwhelmed by their size causing replication
included a time trend to take care of technological of functions and idle resources to monitor clients, which
change or capture movement of the cost function over could result in increased credit risk. For instance,
time and MFI-specific fixed effects. This is to cater for Williamson (1967) and Himmelberg et al. (1999) prove
related variances in the cost structures amongst MFIs that as the size of a financial institution becomes too
and unobserved MFI heterogeneity.
large, it results in inefficiencies in monitoring and evalu-
ation of clients due to the massive cost of operation;
4.3 | Econometric specifications and therefore, leading to increased credit risk.
method In addition to the size of the firm, we also control for
4.3.1 | Empirical model the financial cost of MFI. This is the cost of the firm
incurs in disbursing loans to their clients (Ceb and Traca,
Based on the previous literature as discussed in Section 2 2009). Once, loans are the primary product for
and coupled with Section 3, the following reduced-form microfinance activity; we proxy financial cost with the
model is formulated for the empirical analysis to answer cost per loan. It is measured as the ratio of financial
the research questions raised in Section 1 of the paper. The expenses to gross loan portfolio to determine per unit cost
extant literature on determinants of a portfolio (credit) risk of distributing loans to customers. The essence of this is
suggests that it is influenced by firm-specific factors such as to indicate the efficiency of MFIs in its loans
the size of the business, the financial cost of the MFI, opera- disbursement.
tional efficiency of the MFI, financial strength, and financial Operational efficiency of MFIs is also controlled in
revenue of the MFI. Also, both theory and policy discus- our estimations. This is a performance measure that
sions suggest that both competition and regulation are key shows how well the MFIs are rationalizing its operations
market and policy variables that influence the risk-taking and takes into consideration the cost of the input and the
behaviour of firms in general, including MFIs. Based on price of output (Barry and Tacneng 2014; Kinde, 2012).
this, the reduced-form model is specified as Efficiency in expense management should ensure more
efficient use of MFIs loanable resources. It is proxy with
the write off ratio. It is the ratio of the total amount of
lnRiskit = β0 + β1Regulationit + β2lnCompetionit
 0 loans written off to gross loan portfolio (Kinde, 2012).+ β3Regulationit lnCompetitionit + λ Xit High (low) ratio indicates a low (high) efficiency of
management.
+ μt + ηi + εit ð3Þ The last controlled variables in the model are financial
strength and financial revenue. Financial strength mea-
Where risk in this study will focus on two different sures the soundness or profitability of a company
aspects of risk, credit risk and operational risk, regulation (Chavez-Demoulin et al. 2006; Ceb and Traca 2009). It
8 KARIMU ET AL.
measures the firm's ability to generate positive net Regulationit = α0 + α1lnRiskit + α2lnCompetitionit
incomes for a given level of investment. This variable also + θ0Xit + μt + ηi + eit ð4Þ
determines how well management is running the affairs
of the business in the interest of shareholders. We proxy where all the variables are the same as defined in Equa-
financial strength with the yield on gross loan portfolio tion (1); eit is a random error term.
obtained from the MIX market database (Serrano-Cinca
and Gutiérrez-Nieto, 2014). The yield is the net incomes
from gross loans of an MFI. Financial revenue, on the 4.3.2 | Empirical strategy
other hand, is some incomes that a firm generates Our estimation strategy follows three steps. In the first
through its activities within a specific period. It includes step, we estimate Equation (4) using fully parametric
revenue generated from both the gross loan portfolio and econometric methods (panel probit model because regu-
investments (Gutiérrez-Nieto et al. 2009). It measures the lation is a dummy variable) appropriate for panel analysis
total amount of money that accrues to an MFI in a given to generate the residuals for the main Equation (3) of
financial year. It determines how well management will interest to control potential endogeneity problem. We
be able to meet their financial obligation. The variable is then estimate Equation (3) to assess our fundamental
proxy with interest and fee income on transactions. The questions.
descriptive statistics for the entire critical variables The models presented in both Equations (3) and (4)
described above is presented in the Appendix (Table A1). are estimated using the fixed effect estimation approach.
From Equation (3), the total impact of regulation on The estimation strategy is in two steps. In the first step,
risk is given by taking the partial derivative of risk con- we estimate the regulation model and save the residuals
cerning the regulation, which is express as to be included in the risk model. The purpose of this is
to reduce potential endogeneity problems due to the
dlnRiskit = β + β  lnCompetition interdependence between risk and regulation. In the
dRegulation 1 3 itit second step, we estimate the risk model, both for credit
risk and operational risk. In the final stage, we perform
On the other hand, the total effect of competition on sensitivity analysis on our main results by relaxing the
risk is given by taking the partial derivative of risk con- static structure imposed in estimating the Equations (3)
cerning the competition based on Equation (3), which is and (4) to a dynamic structure. In the case of the
specified as dynamic model, the usually fixed effect model will not
be appropriate because the included lag dependent vari-
dlnRiskit = β2 + β3  lnRegulation able as a regressor will be correlated with the fixed:dCompetition itit effect, creating a dynamic panel bias (Nickel bias),
which is severe in small panels. Because our panel time
We adopt a fixed effect approach that controlled for dimension is less than 30 years, the threshold level
potential endogeneity problem associated with risk and where Nickel bias is not critical (Judson and Owen,
regulation. The potential endogeneity problem due to 1999), we need to apply the appropriate methods to
possible reverse causation between risk and regulation reduce the effects of Nickel bias.
is resolved by estimating a second-reduced form equa- In the literature, both the least squares dummy vari-
tion for regulation as specified in Equation (4), where able corrected (LSDVC) and the generalized method of
the residuals for this equation is generated and added moments (GMM) estimators were designed purposely to
into Equation (3) as an additional covariate (two-stage handle dynamic panels to correct for Nickel bias, espe-
residual inclusion approach). The purpose is to control cially in panels with short periods, where the bias is
the endogeneity problem caused by the reverse causa- severe. In panels with period above 30 years, the bias cre-
tion between risk and regulation. This approach has ated by the correlation between the lagged dependent
been suggested and used by prior studies such as variable and fixed effects is small (Judson and Owen,
Hausman (1978), Das et al. (2003), Blundell and Powel 1999). In such instances, the fixed effects (FE) estimator
(2004), and Terza et al. (2008) to deal with issues of performs well relative to both the GMM and the LSDVC.
endogeneity when there are no suitable available instru- In this study, we opted the LSDVC to correct the bias cre-
ments. We assumed that risk, competition, and firm ated by the lagged dependent variable in the dynamic
characteristics are vital factors that influence the level model estimation due to its superior performance over
of regulation of MFIs and therefore specify the reduced- the GMM showed by Judson and Owen (1999), especially
form model as: in panels with 20 years' time period.
KARIMU ET AL. 9
5 | EMPIRICAL EVIDENCE percentile level increases, suggesting amongst other
things that a very low competitive microfinance industry
5.1 | Results of the empirical estimation should be regulated if the policy target is to reduce credit
risk exposure. However, if the level of competition is high
We first present the results based on a fixed effect model
as demonstrated by 1% percentile level of competition,
for both credit risk and operational risk in that sequence
regulation is bad for credit risk. This is because regula-
and provide some discussion on the results and later, pre-
tion of competitive MFIs may induce some market power
sent sensitivity analysis by relaxing the static model
for the existing firms, which could result in more risk-
imposed to obtain our main findings and also by using a
taking behaviour for pure profit motives. Also, in a high
different index to measure competition and the implica-
competitive market, with very many firms, effective regu-
tion of the sensitivity analysis on our primary results.
lation may be difficult to achieve and in such an environ-
ment, an ineffective regulation may instead induce
5.2 | Credit risk results reckless credit risk-taking behaviour by competing MFIs,
where the MFIs in the industry will not adhere to rules
In Table 1, we present the credit risk results based on a and regulation provided by the policymaker or regulator
fixed effect model. Table 1 contains three columns; each
to ensure a smooth and less risk-taking activities
represents a unique version of the fixed effect model; first
amongst MFIs.
column (1) is a model without both the interactions
The estimated direct effect of competition is positive
between regulation and competition, and time dummies;
and significant at any of the conventional significance
the second column (2) is based on a model without time
levels. The estimated direct elasticity between credit risk
dummies; and the third column (3) is based on our speci-
and competition is about 1.8, which is also the total
fied model presented in Equation (3).
effect of competition on credit risk for nonregulated
In all cases, we find a significant positive direct effect
MFI because the interaction effect evaluated for non-
of regulation on credit risk across the three different spec-
regulated MFI is zero. On the other hand, the total
ifications, with an increasing magnitude as one moves
from column (1) to column (3). This is just the direct impact of competition for regulated MFI is 0.213, which
effect of regulation because the estimated coefficient of is calculated by adding the coefficient of the interaction
the interaction term in column (3) is negative but signifi- term (−1.589) to the coefficient of competition (1.802)
cant at any of the conventional levels of significance; it, via the partial derivative of credit risk with respect to
therefore, implies that the estimated regulation coeffi- competition as expressed in the model section.
cient presented in columns (1) and (2) is tentatively the This implies that nonregulated MFIs tend to take more
direct effects of regulation, but the indirect impact via its risk if they operate in a less competitive environment rela-
interaction with competition is not captured by the tive to regulated MFIs. The transmission mechanism is as
models estimated and presented in columns (1) and (2), follows, without regulation, MFIs enter the industry for all
respectively. The negative coefficient of the interaction manner of reasons including serving the poor and for com-
term between regulation and competition implies that mercial purposes; as a consequence, these MFIs tend to
the total effect of regulation on credit risk could be posi- take more risk for profit motives due to the relaxed rules
tive or negative depending on the level of competition via governing their operations. Few big MFIs can utilize
taking the partial derivative of credit risk with respect to unfair competitive strategies to dominate the market to
regulation, which is presented in the model section after gain some market power. Given the power, they will be
Equation (3). taking excessive risk in the absence of regulation. This
The estimated total effect of regulation evaluated at means that, given a less competitive environment, regula-
different percentiles, 1st, 25th, 50th, 75th, and 95th tion will tend to reduce credit risk exposure.
respectively is all significant at the 5% significance level Amongst the control variables, only the estimated
except at the 25th percentile, where it is not significant. coefficients on financial revenue, financial strength,
This result is reported in Table 2 and revealed that the and the residuals from estimating a regulation model
total effect of regulation on credit risk is conditional on (included to control potential endogeneity of regula-
the level of competition. The impact is positive at the 1st tion in our credit risk model) are statistically signifi-
percentile level of competition (high competition) and cant. The estimated elasticity between credit risk and
turns negative (significant) on the 50th, 75th, and 95th financial revenue is −0.05, whilst the estimated elastic-
percentiles (low level of competition) of competition ity between credit risk and financial strength is −0.09.
proxy by Lerner index. The table further revealed that the These results imply that MFIs tend to take less credit
magnitude of the negative interaction effect increases as risk when their financial revenue position is high.
10 KARIMU ET AL.
TABLE 1 Regression results from estimating a fixed effect static model for portfolio risk
(1) (2) (3)
Variables Credit risk (log) Credit risk (log) Credit risk (log)
Regulation 0.742*** 1.354*** 1.395***
(0.245) (0.347) (0.337)
Competition (log Lerner index) 0.125 0.251 1.802***
(0.214) (0.206) (0.433)
Regulation*Competition (log Lerner index) −1.589***
(0.393)
Size (log) 0.157*** −0.079 −0.088
(0.056) (0.076) (0.075)
Financial cost (log) −0.031 −0.033 −0.033
(0.027) (0.027) (0.026)
Operational efficiency 0.041 0.027 0.029
(0.144) (0.141) (0.145)
Financial revenue (log) −0.065*** −0.054** −0.049**
(0.025) (0.024) (0.024)
Financial strength (log) −0.100** −0.097* −0.086*
(0.051) (0.051) (0.050)
Residual (regulation residual) −0.296*** −0.533*** −0.742***
(0.107) (0.147) (0.156)
Constant −5.154*** −2.766*** −2.659***
(0.881) (0.997) (0.989)
Time dummies No Yes Yes
Observations 1,574 1,574 1,574
Number of firms 444 444 444
CVS 1.155 1.122 1.105
Note. Robust standard errors that correct for heteroskedasticity are in parentheses. Competition has inverse interpretation as higher values implies lower
competition, whilst lower values denote high competition. “Yes” on Time dummies row indicates that time dummies are included in the regression and are
statistically significant, whilst “No” indicates that it is not included in the estimation.
Abbreviation: CVS, cross validation score.
***p < 0.01; **p < 0.05; *p < 0.1.
MFIs with excellent financial strength also tend to excellent financial power will not take unnecessary
make less credit risk, which is very intuitive. This is credit risk exposures. Besides, MFIs with such good
because MFI with good financial revenue position and and excellent financial revenues and financial strength
are more likely to make strict screening measures to
TABLE 2 Total effect of regulation evaluated at different reduce risk exposures relative to those without such
percentiles of competition
financial standing, as they are not under severe reve-
Percentile of nue and liquidity pressure to venture into taking
competition unnecessary credit risk.
(Lerner
index) 1% 25% 50% 75% 95%
Total 1.330*** −0.344 −1.811** −2.760** −2.792** 5.3 | Operational risk results
regulation (0.30) (0.54) (0.89) (1.12) (1.13)
effect Next, we assess whether regulation and competition mat-
(FE model) ter regarding the operational risk of MFIs in SSA. Thus,
given the finding that both competition and regulations
Note. Robust standard errors that correct for heteroskedasticity are in
parentheses. are essential factors to consider when implementing poli-
***p < 0.01; **p < 0.05; *p < 0.1. cies to reduce credit risk amongst MFIs, does this also
KARIMU ET AL. 11
TABLE 3 Regression results from estimating a fixed effect static model for operational risk
(1) (2) (3)
Variables Operational risk (log) Operational risk (log) Operational risk (log)
Regulation −0.012 0.010 0.021
(0.066) (0.095) (0.094)
Competition (log Lerner index) 0.215*** 0.205*** 0.283**
(0.060) (0.063) (0.111)
Regulation*Competition −0.081
(0.095)
Size (log) −0.102*** −0.092*** −0.092***
(0.013) (0.017) (0.017)
Financial cost (log) 0.005 0.003 0.003
(0.008) (0.008) (0.008)
Operational efficiency −0.057* −0.064* −0.062*
(0.033) (0.033) (0.033)
Financial revenue (log) 0.008 0.007 0.007
(0.008) (0.008) (0.008)
Financial strength (log) 0.016 0.013 0.013
(0.012) (0.012) (0.012)
Residual (regulation residual) 0.017 0.011 −0.005
(0.027) (0.039) (0.041)
Constant −0.076 0.076 0.081
(0.215) (0.258) (0.258)
Time dummies No Yes Yes
Observations 2,144 2,144 2,144
R-squared 560 560 560
CVS 0.147 0.147 0.147
Note. Robust standard errors that correct for heteroskedasticity are in parentheses. “Yes” on Time dummies row indicates that time dummies are included in
the regression and are statistically significant, whilst “No” indicates that it is not included in the estimation.
Abbreviation: CVS, denotes cross validation score.
***p < 0.01; **p < 0.05; *p < 0.1.
apply to operational risk? In addressing this objective, we a consequence, in such a case, regulation is likely to be
similarly estimated an operational risk model as done in associated with loan and credit activities of these institu-
the case of credit risk. We assessed three different ver- tions but less to operational activities.
sions, which are reported under three different columns Competition, on the other hand, increases the opera-
in Table 3. Column (3) is estimated based on the model tional risk of MFIs because the estimated coefficient is
presented in Equation (1), whilst column (2) excluded positive and significant at least at the 5% significance
the interaction between regulation and competition, and level, meaning that a less competitive MFIs industry is
column (1) excluded both the interaction term and time associated with high operational risk. The interaction
dummies. The reported results indicate that regulation is term between regulation and competition is however
not essential for operational risk of the MFIs because the insignificant, further supporting the finding from the
estimated coefficient is statistically insignificant across direct effect of regulation on operational risk. In a nut-
the three different versions at any of the conventional sig- shell, the regulation does not affect the operational risk
nificance levels. A possible explanation for this is may be of MFIs in our sample, both direct and indirect.
that most of the regulation of MFIs is directed towards The estimated coefficient of size is negative and sig-
loan activities but less towards how the MFIs operate. As nificant, implying that the size of the MFI has an
12 KARIMU ET AL.
impact on operational failures and hence, operational 6 | SUMMARY AND CONCLUSION
risk. The mechanism for this is as follows, large firms
can afford better systems and implement relatively bet- The study highlights the sequencing impact of portfolio
ter policies and procedures on the average relative to risk, market concentration, and regulation of MFIs in
small firms. The implication of this is that large firms SSA. To establish this, we use both fixed effect and
on the average can reduce employee errors due to the dynamic panel regression models on a sample of 1,574
better screening process and monitoring procedures, microfinance firms from SSA countries for the period
reduce system failures, and in general, reduce events 1995–2015. Evidence from our extensive panel fixed effect
that are likely to create problems for the firm's and dynamic models suggests a significantly positive
operations. direct impact of regulation on credit risk. The result
The results also revealed that operational efficiency implies that regulation substantially affects credit risk
has a significant negative effect on operational risk of positively. In similar evidence, the findings also suggest a
MFIs, which amongst other things means that if the firm significantly negative relationship between the interac-
is operating efficiently, the firm tends to be less prone to tive term of regulation and competition on risk. This indi-
failures in procedure, systems, and policies, and as a con- cates that a low competitive MFI industry could be
sequence, reduce employee errors, system failures, reduc- efficiently regulated if the policy target is to reduce credit
tion in criminal activities, and any action that will risk exposure. This is because regulation will control
disrupt the firm's business process. This ultimately reckless credit risk-taking behaviour by powerful MFIs to
reduces the cost associated with operational failures and ensure that MFI in the industry adheres to rules and reg-
hence, operational risk. ulation provided by the policymaker or regulator to aid
The other controls such as financial revenue, financial smooth and less risk-taking activities amongst MFIs.
strength, and financial cost are not statistically significant Contrary to the above evidence, we did not find any
at any of the conventional significance level, which significant relations between regulation and operational
implies that these controls have no impact on MFIs' oper- risk. A possible explanation for this is that most of the
ational risk exposures, contrary to the findings from regulation of MFIs is directed towards loan activities but
credit risk. less towards how the MFIs operate. Consequently, regu-
lation is likely to associate with loan and credit activities
5.4 | Sensitivity analysis of these institutions but less to operational activities.
However, we find the estimated coefficient of competi-
Our primary results reported in Tables 1 and 3 may be tion on operational risk to be positive and significant,
sensitive to the type of structure imposed on the model which suggests that low competitive MFIs are very much
(a static model for the primary results). To assess the exposed to high operational risk. Our general conclusion
implication of the imposed structure of the model on the based on this sensitivity analysis is that the results are
results, we relax the static nature of the model by estimat- robust to the model structure (static or dynamic) for both
ing a dynamic model. The results based on a dynamic the credit risk and operational risk models. The results
model are reported in Table A2 of the Appendix. The remain consistent after controlling for model structure
results revealed that, in general, they are qualitatively (static or dynamic).
similar to our primary results for both the credit risk However, the findings of our study should be inter-
results and the operational risk results. They were only preted in light of some limitations. For instance, due to
slightly different regarding the size of the coefficients. data availability, our study is limited to a sample of 1,574
Our general conclusion based on this sensitivity analysis microfinance firms from SSA countries. As a result, we
is that the results are robust to the model structure (static caution scholars against generalization using the findings
or dynamic) for both the credit risk and the operational of this paper. Also, despite the findings that regulation
risk models. and competition have different effects on credit risk and
The general conclusion from the sensitivity analysis is operational risk, it is possible that there may be other
that, in general, the type of model structure imposed explanatory variables not included in our model. This is
(static versus dynamic) does not significantly influence especially particular in SSA countries, where so many
the model results. In the case of the choice of proxy for other variables can contribute to credit risk and opera-
competition, we found that the estimates on the critical tional risk-taking behaviour of firms.
variables of interest (regulation and competition) are sen- Further and more extensive analyses in multiple con-
sitive to the proxy used for competition (Lerner index ver- texts and countries will help to establish causal effects
sus HHI). between the variables. Finally, it is possible that the
KARIMU ET AL. 13
impact of regulation and competition on credit risk and Anginer, D. and Demirgüç-Kunt, A. (2014). Bank capital and sys-
operational risk might be best captured at a sectorial temic stability.
level, as this will enable the characteristics of individual Appiah-Konadu, P., Churchill, R. Q., Agbodohu, W., &
sectors to be modelled; due to data availability, our study Frimpong, H. K. (2016). Evaluating the credit risk management
practices of microfinance institutions in Ghana: Evidence from
could not achieve this. This could be an avenue for future Capital Line Investment Ltd. and Dream Finance Ltd. Archives of
research. Business Research (ABR) Vol, 4(5).
Despite these limitations, the study makes a signifi- Assefa, E., Hermes, N., & Meesters, A. (2013). Competition and the
cant contribution to academic literature and policy impli- performance of microfinance institutions. Applied Financial
cation. First, we offer new evidence of the relationship Economics, 23(9), 767–782.
between regulation and competition on risk-taking Ayele, G. T. (2015). Microfinance institutions in Ethiopia, Kenya
behaviour of MFIs. Whilst prior studies provide extensive and Uganda: Loan outreach to the poor and the quest for finan-
empirical evidence on the impact of regulation on risk- cial viability. African Development Review, 27(2), 117–129.
Barry, T. A., & Tacneng, R. (2014). The impact of governance and
taking behaviour of firms, there is no evidence that dem-
institutional quality on MFI outreach and financial performance
onstrates how regulation and competition are likely to in Sub-Saharan Africa. World Development, 58, 1–20.
have different effects on credit risk and operational risk Bateman, M. (2019). Impacts of the microcredit model: Does theory
amongst firms. Our model reveals that regulation is likely reflect actual practice? Blankenburg and Kozul-Wright:
to associate with loan and credit activities of these insti- Bateman.
tutions but less to operational activities. Against this Beisland, A., D’Espallier, B.,  Roy Mersland, R. (2017). The com-
backdrop, we suggest further studies to control for these mercialization of the microfinance industry: Is there a ‘personal
conditions to derive reliable conclusions. mission drift’ among credit officers? Journal of Business Ethics,
pp1-16.
In terms of the policy implications, our findings sug-
Berger, A. N., Bouwman, C. H., Kick, T., & Schaeck, K. (2016).
gest that policymakers should be concerned about the Bank liquidity creation following regulatory interventions and
economic consequences of regulation on credit risk- capital support. Journal of Financial Intermediation, 26,
taking behaviour of MFIs. Our findings suggest that the 115–141.
MFI industry could be regulated efficiently if Blazy, R., & Weill, L. (2013). Why do banks ask for collateral in
policymakers develop policies targeted at reducing credit SME lending? Applied Financial Economics, 23(13), 1109–1122.
risk exposures of MFIs than their exposure to Blundell, R. W., & Powell, J. L. (2004). Endogeneity in semi-
operational risk. parametric binary response models. The Review of Economic
Studies, 71(3), 655–679.
Finally, our findings also offer a guide to business
Boateng, G. O., Boateng, A. A. and Bampoe, H. S. (2015).
owners on the type of risk exposure they may be exposed Microfinance and poverty reduction in Ghana: Evidence from pol-
to under different market conditions. Our model reveals icy beneficiaries.
that low competitive MFIs are very much likely to be Boateng, S. D., Amoah, I. A., & Anaglo, J. N. (2015). The influence
exposed to high operational risk. of demographic factors, products and service characteristics of
microfinance institutions on repayment performance among
farmers in the eastern region. Journal of Ghana Science Associa-
ORCID tion, 16(2), 19.
Bustamante M. C. and Donangelo A. (2014). Product market com-
Samuel Salia https://orcid.org/0000-0002-5651-5009
petition and industry returns. Financial Markets Group Discus-
Ishmael Tingbani https://orcid.org/0000-0002-4012- sion Paper No 728.
1224 Castellani, D., & Cincinelli, P. (2015). Dealing with drought-related
credit and liquidity risks in MFIs: Evidence from Africa. Strate-
gic Change, 24(1), 67–84.
REFERENCES Ceb, H. M and Traca, D. (2009). On the efficiency effects of subsi-
Adams, A. (2017). Performance of microfinance institutions in Sub- dies in microfinance: An empirical inquiry, World Development,
Saharan Africa: A cross-country analysis of outreach, sustainabil- 39, 6.
ity, efficiency and regulation. University of Zululand: Doctoral Cull, R., Harten, S., Nishida, I., Rusu, A. B., & Bull, G. (2015).
dissertation. Benchmarking the financial performance, growth, and outreach
Aghion, P., Bloom, N., Blundell, R., Griffith, R., & Howitt, P. of greenfield MFIs in Africa. Emerging Markets Review, 25,
(2005). Competition and innovation: An inverted-U relation- 92–124.
ship. The Quarterly Journal of Economics, 120(2), 701–728. Chavez-Demoulin, V., Embrechts, P., & Nesˇlehova,´ J. (2006).
Appiah-Konadu, P., Churchill, R. Q., Agbodohu, W., & Quantitative models for operational risk: Extremes, dependence
Frimpong, H. K. (2016). Evaluating the credit risk management and aggregation. Journal of Banking & Finance, 30, 2635–2658.
practices of microfinance institutions in Ghana: Evidence from Chikalipah, S. (2017). Institutional environment and microfinance
Capital Line Investment Ltd. And Dream Finance Ltd. Archives performance in sub-Saharan Africa. African Development
of Business Research (ABR) Vol, 4(5). Review, 29(1), 16–27.
14 KARIMU ET AL.
Chaibi, H., & Ftiti, Z. (2015). Credit risk determinants: Evidence study of Morogoro municipality. The Open University Of Tanza-
from a cross-country study. Research in Internation Business and nia: Doctoral dissertation.
Finance, 33(2015), 1–16. Judson, R. A., & Owen, A. L. (1999). Estimating dynamic panel data
Das, M., Newey, W. K., & Vella, F. (2003). Nonparametric estima- models: A guide for macroeconomists. Economics Letters,
tion of sample selection models. The Review of Economic Studies, 65, 9–15.
70(1), 33–58. Kablan, S. (2014). Microfinance efficiency in West African Economic
Deakins, D., & Hussain, G. (1994). Risk assessment with asymmet- and Monetary Union (WAEMU): Have reforms promoted sustain-
ric information. International Journal of Bank Marketing, 12(1), ability or outreach?
24–31. Kinde, B. A. (2012). Financial sustainability of microfinance institu-
Dewatripont, M. (2014) "Effects of regulation, the safety net, and tions (MFIs) in Ethiopia. European Journal of Business and
other government gGuarantees." The social value of the financial Management, 4(15), 1–10.
sector: Too big to fail or just too big?. 267-272. Koku, P. S., & Jagpal, S. (2015). Do payday loans help the working
D’Espallier, B., Goedecke, J., Hudon, M., & Mersland, R. (2017). poor? International Journal of Bank Marketing, 33(5), 592–604.
From NGOs to banks: Does institutional transformation alter Kusi, B. A., Agbloyor, E. K., & Ansah-Adu, K. (2017). Bank credit
the business model of microfinance institutions? World Develop- risk and credit information sharing in Africa: Does credit infor-
ment, 89, 19–33. mation sharing institutions and context matter? Research in
De Quidt, J., Fetzer, T., & Ghatak, M. (2018). Market structure and International Business and Finance.
borrower welfare in microfinance. The Economic Journal, 128 Ledgerwood, J., Earne, J., & Nelson, C. (2013). The new
(610), 1019–1046. microfinance handbook: A financial market system perspective.
Domeher, D., Musah, G., & Hassan, N. (2017). Inter-sectoral differ- World Bank Publications.
ences in the SME financing gap: Evidence from selected sectors Magali, J. J. (2013). The impacts of credits risk management on prof-
in Ghana. Journal of African Business, 18(2), 194–220. itability of Rural Savings and Credits Cooperative Societies
Duarte, F. D., Gama, A. P. M., & Esperança, J. P. (2017). Collateral- (SACCOs): The case study of Tanzania. Browser Download This
based in SME lending: The role of business collateral and per- Paper.
sonal collateral in less-developed countries. Research in Interna- Matamanda, S. H., & Chidoko, C. (2017). Accessing formal financ-
tional Business and Finance, 39, 406–422. ing by small and medium enterprises in Zimbabwe: The case of
Fafchamps, M. (2014). Credit constraints, collateral and lending to SMEs and banks in Chiredzi Urban. Cell, 263(773), 555908.
the poor. Revue d'économie du développement, 22(HS01), 71-89. Marcelin, I., & Mathur, I. (2014). Financial development, institu-
Fletschner, D. and Kenney, L. (2014). Rural women's access to tions and banks. International Review of Financial Analysis, 31,
financial services: Credit, savings, and insurance. In Gender in 25–33.
agriculture (pp. 187-208). Springer Netherlands. Madestam, A. (2014). Informal finance: A theory of moneylenders.
Gan, C., Nartea, G. V. and Xia, J. L. (2017). An overview of Journal of Development Economics, 107, 157–174.
microfinance. In Microfinance in Asia (pp. 1-22). Morduch, J. (2013). How microfinance really works. Milken Insti-
Garrity, P., & Martin, C. (2018). Developing a microfinance model tute Review Second Quarter, 51–59.
to break the cycle of poverty. Business Horizons, 61(6), 937–947. Mujeri, M. K. (2015). Improving access of the poor to financial ser-
Gietzen, T. (2017). The exposure of microfinance institutions to vices. Bangladesh: General Economics Division of the Planning
financial risk. Review of Development Finance xxx, 2017, xx. Commission.
Gul, F. A., Podder, J., & Shahriar, A. Z. M. (2017). Performance of Mwega, F. M. (2014). Financial regulation in Kenya: Balancing
microfinance institutions: Does government ideology matter? inclusive growth with financial strength. Overseas Development
World Development, 100, 1–15. Institute.
Gutiérrez-Nieto, B., Serrano-Cinca, C., & Molinero, C.M. (2009). Ordás Criado, C., Valente, S., & Stengos, T. (2011). Growth and pol-
Social efficiency in microfinance institutions, Journal of the Oper- lution convergence: Theory and evidence. Journal of Environ-
ational Research Society,60(1) 104-119. mental Economics and Management, 62(2), 199–214.
Hartarska, V., & Nadolnyk, D. (2011). Do regulated microfinance insti- Otchere, I., Senbet, L. and Simbanegavi, W. (2017). Financial sector
tutions achieve better sustainability and outreach? Cross-country development in Africa—An overview.
evidence. Journal of Applied Economics, 39(7), 1207–1222. Pashkova, N., Trujillo-Barrera, A., Apostolakis, G., Van Dijk, G.,
Hausman, J. A. (1978). Specification tests in econometrics Drakos, P. D., & Baourakis, G. (2016). Business management
(pp. 1251–1271). Econometrica: Journal of the econometric models of microfinance institutions (MFIs) in Africa: A study of
society. their enabling environments. International Journal of Food and
Himmelberg, C. P., Hubbard, R. G., & Palia, D. (1999). Understand- Beverage Manufacturing and Business Models (IJFBMBM), 1(2),
ing the determinants of managerial ownership and the link 63–82.
between ownership and performance. Journal of financial eco- Postelnicu, L., & Hermes, N. (2017). Microfinance performance and
nomics, 53(3), 353–384. social capital: A cross-country analysis 153(2). Journal of Busi-
Hussain, J. G., Scott, J. M., Harrison, R. T., & Millman, C. (2010). ness Ethics, pp, 427–445.
“Enter the dragoness”: Firm growth, finance, guanxi, and gen- Purkayastha, D., Tripathy, T., & Das, B. (2014). The impact of regu-
der in China. Gender in Management: An International Journal, lations on microfinance industry: A strategic perspective. IUP
25(2), 137–156. Journal of Business Strategy, 11(3), 24.
John, M. (2015). The impact of financial regulations to the financial Mosley, P. (2009). Assessing the success of microinsurance
sustainability of micro-financial institutions in Tanzania: A case programmes in meeting the insurance needs of the poor.
KARIMU ET AL. 15
Racine, J., & Li, Q. (2004). Nonparametric estimation of regression Van Leuvensteijn, M., Bikker, J. A., Van Rixtel, A. A., &
functions with both categorical and continuous data. Journal of Sørensen, C. K. (2011). A new approach to measuring competi-
Econometrics, 119(1), 99–130. tion in the loan markets of the euro area. Applied Economics, 43
Serrano-Cinca, C. and Gutiérrez-Nieto, B (2014), Microfinance, the long (23), 3155–3167.
tail and mission drift, International Business Review,23(1), 181-194. Weaver, F. (2016). The effects of Grameen-model microfinance insti-
Singh, R. J., & Huang, Y. (2016). Financial channels, property rights tutions and self-help group microfinance on women's economic
and poverty: A Sub-Saharan African perspective. Economic empowerment in rural India.
Notes, 45(3), 327–351. Wellalage, N. H., & Locke, S. (2016). Informality and credit con-
Siwale, J., & Okoye, N. (2017). Microfinance regulation and social straints: Evidence from Sub-Saharan African MSEs. Applied
sustainability of microfinance institutions: The case of Nigeria Economics, 48(29), 2756–2770.
and Zambia. Annals of Public and Cooperative Economics. Williamson, O. E. (1967). Hierarchical control and optimum firm
Smith, S. C. (2015). Ending global poverty: A guide to what works. St. size. The Journal of Political Economy, 123–138.
Martin's Press. Yu, V., Damji, R., Vora, V., & Anand, L. (2014). Regulation on
Spratt, S. (2016). Financial regulation, stability and growth in low- microfinance: Effect upon profitability and loan diversity. Univer-
income countries. Achieving Financial strength and Growth in sity of Chicago Undergraduate Business Journal.
Africa, 8, 21.
Terza, J. V., Basu, A., & Rathouz, P. J. (2008). Two-stage residual
inclusion estimation: Addressing endogeneity in health econo-
metric modeling. Journal of health economics, 27(3), 531–543. How to cite this article: Karimu A, Salia S,
https://doi.org/10.1016/j.jhealeco.2007.09.009 Hussain JG, Tingbani I. Are competitive
Tumwine, S., Akisimire, R., Kamukama, N., & Mutaremwa, G. microfinance services worth regulating? Evidence
(2015). A borrowing cost model for effective performance of from microfinance institutions in Sub-Saharan
SMEs in Uganda. World Journal of Entrepreneurship, Manage- Africa. Int J Fin Econ. 2019;1–17. https://doi.org/
ment and Sustainable Development, 11(2), 74–89.
10.1002/ijfe.1800
Triki, T., Kouki, I., Dhaou, M. B., & Calice, P. (2017). Bank regula-
tion and efficiency: What works for Africa? Research in Interna-
tional Business and Finance, 39, 183–205.
16 KARIMU ET AL.
APPENDIX
TABLE A1 Descriptive statistics for the dataset for the analysis for overall, where data are pooled cross firms and time period, within
firms and between firms
Variable Mean Std. Dev. Minimum Maximum Observation
Portfolio risk Overall 0.100 0.200 0.000 6.843 N = 240
Between 0.197 0.000 2.440 n = 62
Within 0.140 −2.264 4.504
Regulated Overall 0.773 0.419 0.000 1.000 N = 385
Between 0.324 0.000 1.000 n = 81
Within 0.327 −0.164 1.607
Gross loan portfolio Overall 16300000.000 107000000.000 0.000 3400000000.000 N = 369
Between 53100000.000 0.000 1010000000.000 n = 80
Within 78000000.000 −952000000.00 2410000000.000
Cost per loan Overall 261.810 482.417 4.000 6822.000 N = 117
Between 453.854 5.000 4164.000 n = 52
Within 305.979 −2542.690 4367.477
Write of ratio Overall 0.048 0.648 −0.023 25.711 N = 189
Between 0.213 0.000 3.723 n = 52
Within 0.599 −3.675 22.037
Financial strength Overall 2074.507 91062.650 −1.959 4100000.000 N = 202
Between 40741.080 −0.799 1025000.000 n = 63
Within 78886.220 −1022926.000 3077074.000
Financial revenue Overall 52137.630 407119.200 −2057.860 15000000.000 N = 385
Between 274067.700 0.000 5000507.000 n = 81
Within 339063.100 −4948369.000 10100000.000
Lerner index Overall 0.702 0.156 0.041 0.974 N = 235
Between 0.148 0.091 0.974 n = 59
Within 0.084 0.139 1.026
HHI Overall 0.002 0.009 0.000 0.115 N = 385
Between 0.007 0.000 0.115 n = 81
Within 0.007 −0.044 0.107
Total cost Overall 5581175.000 30800000.000 4.885 820000000.000 N = 238
Between 18400000.000 97.214 385000000.000 n = 60
Within 18500000.000 −281000000.00 441000000.000
Labour cost Overall 12937.260 11423.950 0.851 244348.800 N = 215
Between 10401.210 1.136 90979.570 n = 54
Within 6782.434 −22465.180 221594.000
Capital cost Overall 0.063 0.207 0.000 6.644 N = 219
Between 0.319 0.001 6.644 n = 56
Within 0.088 −1.654 2.139
Market share Overall 0.010 0.032 0.000 0.628 N = 369
Between 0.022 0.000 0.240 n = 80
Within 0.028 −0.151 0.546
Abbreviation: HHI, Herfindahl-Hirschman index.
KARIMU ET AL. 17
TABLE A2 Regression results from estimating a least squares
dummy variable corrected dynamic model for operational risk and
portfolio risk
(1) (3)
Operational risk Credit
Variables (log) risk (log)
Lag credit risk (log) 0.460***
(0.033)
Lag operational risk (log) 0.080***
(0.016)
Regulation −0.009 0.853*
(0.121) (0.493)
Competition 0.254** 1.879***
(Lerner index (log)) (0.112) (0.491)
Size (log) −0.096*** −0.087
(0.020) (0.107)
Regulation*Competition −0.051 −1.491***
(0.174) (0.539)
Financial cost (log) 0.004 −0.022
(0.007) (0.052)
Operational efficiency −0.059* −0.130
(0.034) (0.254)
Financial revenue (log) 0.006 −0.044*
(0.004) (0.024)
Financial strength (log) 0.011 −0.053**
(0.007) (0.022)
Residual 0.009 −0.509**
(regulation residual) (0.037) (0.224)
Time dummies Yes Yes
Observations 2,017 1,174
Number of firms 521 324
Note. Robust standard errors that correct for heteroskedasticity are in
parentheses. “Yes” on Time dummies row indicates that time dummies are
included in the regression and are statistically significant.
***p < 0.01; **p < 0.05; *p < 0.1.