Ijfe 2370

Received: 7 July 2020 Revised: 17 November 2020 Accepted: 20 November 2020
DOI: 10.1002/ijfe.2370
RESEARCH ARTICLE
Investigating the determinants of commercial bank interest

rate spreads in Lesotho: Evidence from autoregressive
distributed lag (ARDL) and non-linear ARDL approaches
Moeti Damane
Research Department, Central Bank of

Lesotho, Maseru, Lesotho
Abstract
This article investigates the determinants of commercial bank interest rate
Correspondence spreads in Lesotho using monthly time series data from January 2009 to
Moeti Damane, Research Department,
Central Bank of Lesotho, P.O. Box 1184, December 2018. The Autoregressive Distributed Lag (ARDL) bounds testing
Maseru 100, Lesotho. approach is used to measure long-run co-integration while the non-linear
Email: [email protected]
ARDL (NARDL) model is used to test validity of long-run symmetric effects.
The bounds tests revealed existence of long-run co-integration between the
study variables. Inflation and the Treasury bill rate have a positive and statisti-
cally significant impact on interest margins while the deposit rate has a nega-
tive and significant effect. The pass-through of inflation and the deposit rate to
interest margins is less than one, respectively. This confirms that inflation
affects banks' lending rates with a second round effect while deposit liabilities
are not the only source of credit financing for banks. The null hypothesis of
long-run symmetry is rejected for Treasury bill rates. Authorities are advised to
ensure price and general macroeconomic stability while also pursuing policies
aimed at maximizing savings.
KEYWORDS
asymmetry, co-integration, commercial banks, interest rate spreads, Lesotho
1 | INTRODUCTION interest rate spread or interest margin is the difference

between the interest charged on loans and the interest
The importance of the financial intermediary role of paid on deposits. As profit-making institutional units,
commercial banks in the economy of any country cannot banks would want to pay lower interest on deposits while
be overstated. This is especially true in less developed charging higher interest on borrowers, thus creating a
countries where most individuals and firms rely heavily positive spread (Damane, 2019a, 2019b; Damane,
on bank credit for financing as opposed to a mix of Sekantsi, & Molapo, 2018; Khan & Jalil, 2020;
bank-based and market-based finance. The business of Männasoo, 2013; Molapo & Damane, 2016; Obeng &
commercial banking involves mobilizing funds from the Sakyi, 2017; Sheriff & Amoako, 2014; Tarus, Chekol, &
public in the form of demand, time and savings deposits Mutwol, 2012).
or in the form of borrowing from the public or other Given that the role of commercial banks in the econ-
banks then using the acquired funds, in whole or in part, omy is to transfer funds from lenders to borrowers, they
to extend loans, advances and credit facilities and/or for have to do so with efficiency and effectiveness to promote
investing by other means. Banks pay interest to the economic growth and the improvement of social welfare.
depositors and charge interest to the borrowers. The There is great value in the close monitoring of banks'
Int J Fin Econ. 2020;1–23. wileyonlinelibrary.com/journal/ijfe © 2020 John Wiley & Sons, Ltd. 1
2 DAMANE
interest rate spreads due to their inherent use as indica- evaluated on the basis of size, efficiency (as proxied by
tors of efficient price signals to market players. Narrow the interest rate spread), depth and reach, and compared
interest rate spreads are often associated with a relatively to its counterparts within the Common Monetary Area1
competitive and efficient banking sector. Such a sector (CMA) of Southern Africa, it consistently ranks lower,
promotes investment and savings on the back of afford- with the widest intermediation spread (Damane, 2019b;
able credit extensions to the borrowers and satisfying Khoabane, 2018). Figure 1 reflects the interest rate spread
returns to the depositors. Conversely, wider interest mar- in Lesotho from 1990 to 2018. On average, Lesotho's
gins impede the deepening of financial intermediation in interest rate spread was 8.32% over the period under con-
the country since lower deposit rates deter savings while sideration. The lowest point in the interest margin was in
high rates on loans reduce demand from borrowers. In 1995, at 3.03%, while the peak was reached in the year
order to promote economic growth, commercial banks 2000 at 12.19%. Although the bank spreads have shown
have to provide intermediation services at the lowest pos- volatility over time, they have generally exhibited an
sible cost. This is most important in developing econo- upward trend, as shown by the trend line in the figure.
mies where capital markets are underdeveloped and Numerous studies have investigated the determinants
commercial banks are the main supplier of financing for of commercial bank interest rate margins in a mix of
firms and individuals (Chirwa & Mlachila, 2004; Khan & developed and developing countries (see, Allen, 1988;
Jalil, 2020; Männasoo, 2013; Obeng & Sakyi, 2017; Sher- Brock & Suarez, 2000; Chirwa & Mlachila, 2004;
iff & Amoako, 2014; Tarus et al., 2012). Demirgüç-Kunt & Huizinga, 1999; Ho & Saunders, 1981;
Since the 1980s, Lesotho's financial sector has experi- Khan & Jalil, 2020; López-Espinosa, Moreno, & de
enced various reforms. These broadly consisted of finan- Gracia, 2011; Männasoo, 2013; McShane & Sharpe, 1985;
cial liberalization, financial sector law reforms as well as Obeng & Sakyi, 2017; Saunders & Schumacher, 2000;
structural reforms to strengthen credit disciple, lending Sheriff & Amoako, 2014; Tarus et al., 2012). The litera-
and the recovery environment. Noteworthy is the land ture advances that bank spreads are determined by fac-
title reform (legalization of married persons Act) that tors that include the macroeconomic environment, the
made it so that women can own land and use it as collat- banking sector's market structure, bank-specific factors
eral when seeking credit. It is complemented by the and financial regulation.
establishment of electronic land register and the simplifi- The continued economic and policy relevance of such
cation of the enforcement of a mortgage on land. These studies, especially for countries with financially liberal-
initiatives have resulted in reductions in the risks and ized economies, such as Lesotho, is without question.
costs associated with lending against leasehold titles. Policymakers care about banks' intermediation margins
They have also led to persistently rising mortgage loans since higher spreads are usually interpreted as a proxy of
to the private sector. The rise in private sector credit, inefficiency. Financial sector regulators stand to benefit
especially to households and small business has benefited from having an understanding of how banking specific
from civil law reform, which includes the establishment and/or macroeconomic factors affect interest rate spreads
of a commercial court. This has resulted in a streamline and whether such impacts, following either a positive or
of small claims procedures that improve lending environ- negative shock to one of the possible factors (say, infla-
ment for households and small businesses. Despite these tion), are symmetric. Such clarity will help to credibly
and other reforms, when the financial system is inform relevant policy responses. Previous empirical
F I G U R E 1 Interest rate
spread (%).
Source: Authors' own
calculations based on data from
World Bank Development
Indicators [Colour figure can be
viewed at
wileyonlinelibrary.com]
DAMANE 3
investigations into the factors that determine interest rate this, 56.68% were the assets of the country's four commer-
spreads in Lesotho have largely relied on a multi-country cial banks, three of which are subsidiaries of
panel data approach (see, Crowley, 2007; Folawewo & South African banks.5 In line with its dominance in the
Tennant, 2008; Ahokpossi, 2013 as well as Motelle & financial sector, the banking sector remains the primary
Biekpe, 2014). While worthwhile lessons can be garnered distributor of financial services and products in the coun-
from these studies, their analytical frameworks lend try. For example, the supply of investment finance to
themselves to the possibility that some country-specific firms in Lesotho is dominated by commercial bank
heterogeneities might not be adequately captured in the credit. The size of the banking system balance sheet grew
analysis. In addition, to the best of our knowledge, no by 8.3% from M16.1 billion in 2017 to M17.4 billion in
study has yet been undertaken to empirically investigate 2018. This was mainly on account of the credit portfolio.
the relationship between bank spreads, bank-specific fac- The credit portfolio of banks grew by 12. 4% from M5.8
tors and macroeconomic variables in Lesotho in both the billion in 2017 to M6.5 billion in 2018 (CBL, 2019;
short and long-run with a focus on the symmetric and Damane, 2019a; Damane et al., 2018; Molapo &
asymmetric nature of this relationship. The objective of Damane, 2016).
this study is, therefore, to investigate the short- and long- Lesotho's financial sector, much like that of other
run symmetric and asymmetric relationship between countries within the SADC region has undertaken a host
commercial bank interest rate spreads and a handful of of financial sector reforms since the 1980s with the inten-
macroeconomic and bank-specific variables in Lesotho. tion to improve its structure and efficiency. These finan-
Our study's main contribution is in the use of the latest cial system reforms were necessitated by such
linear and non-linear co-integration techniques and the developments as evolving macroeconomic dynamics
inclusion of high-frequency (monthly) narrow money (e.g., variable inflation rates, the establishment of a cen-
growth and inflation rate observations (these variables tral bank,6 growing overlap in services offered by finan-
are of great import in the Central Bank of Lesotho's mon- cial institutions, etc.), parliamentary legislative initiatives
etary policy regime2) among a list of variables to deter- to protect depositors and foster economic development,
mine commercial banks' interest rate spreads in Lesotho. conversion of the Post Office Savings Bank to the Lesotho
The remainder of the paper is organized as follows: Commercial Bank as well as changes in technology that
Section 2 gives an overview of Lesotho's financial sector; enables a broadening of financial services and geographi-
Section 3 provides a discussion of the relevant literature; cal areas over which financial services are delivered
Section 4 describes the data and analytical technique (Anchang, 2016; Ayaya, 1997; CBL, 2019;
used in the study; Section 5 presents and discusses the Damane, 2019a; Damane et al., 2018; Molapo &
results of the study. Section 6 offers robustness checks Damane, 2016; Mowatt, 2001). Between 2014 and 2018,
and, lastly, Section 7 concludes and offers policy domestic banks introduced and launched mobile and
recommendations. internet banking services that allow banks' clients to
access various financial services remotely. This took place
in the wake of announcements and deployments of
2 | OVERVIEW OF LESOTHO'S mobile money services (EcoCash in October 2012 and
FINANCIAL S ECTOR M-Pesa in July 2013) by telecoms companies in the
country. Between 2015 and 2018, registered mobile
Lesotho is a small and mostly mountainous country that money accounts in Lesotho have grown by around 26%, a
is largely rural with a population of approximately 2 mil- signal of growing access7 to financial services by the
lion people. It is completely enveloped by South Africa country's previously unbanked population. Similarly, in
and through its membership in the CMA,3 its currency; 2017, domestic deposit taking banks, including the Cen-
the Loti (plural Maloti), is pegged at par to the tral Bank of Lesotho (CBL), successfully updated their
South African Rand. Apart from CMA membership, the core banking8 systems in a move that signalled an impor-
country is also a member of Southern African Customs tant milestone in the country's digital transformation
Union (SACU)4 and Southern African Development journey (Damane & Sekantsi, 2020). Process innovations
Community (SADC). The lion's share of the financial sec- improve payment systems used in borrowing and lending
tor's total assets in Lesotho is accounted for by Other of funds and ultimately ameliorate risk, increase the
Depository Corporations (ODCs) followed by Other availability of credit to borrowers and provide financial
Financial Corporations (OFCs). In 2010, Lesotho's finan- institutions with a new and cost-effective way of raising
cial sector comprised of 450 corporations (except the Cen- capital (Bhatt, 1988; Tahir et al., 2018; Damane &
tral Bank of Lesotho) with total assets worth M14.1 Sekantsi, 2020). Despite the reforms, Damane (2019b)
billion, 60.16% of which were the assets of ODCs. Out of points out that when a handful of banking sector
4 DAMANE
F I G U R E 2 Interest rate
spread CMA (%).
Source: Authors' own
calculations based on data from
IMF-FAS [Colour figure can be
viewed at
wileyonlinelibrary.com]
development indicators (indicators of banking sector the maturity of their assets and liabilities and thereby cre-
size,9 efficiency,10 depth11 and reach12)are used to com- ating a positive risk premium. Similarly, the expected
pare Lesotho's banking sector to that of its fellow mem- utility maximization models assume that the financial
bers in the CMA over the period 2007–2017, the country intermediary's objective is to maximize the expected util-
consistently lags behind the other countries in most of ity of terminal wealth, a function that is directly related
these indicators. Figure 2 presents the efficiency of the to the interest rate spread (Allen, 1988; Khan &
banking sector in CMA countries as measured by the Jalil, 2020; Maudos & De Guevara, 2004; McShane &
interest rate spread. Sharpe, 1985). The theoretical underpinning of the litera-
Between 2007 and 2017, the interest rate spread in ture reviewed in this study significantly draws from the
South Africa averaged 3.36% followed by Namibia with bank dealership model as proposed by Ho and
an average of 4.63% then Eswatini with a 6.51% average Saunders (1981). The model was the first to be used to
and, lastly, Lesotho with an average interest rate spread determine banks' interest rate spreads and it broadly
of 8.24%. When the growth rates13 in individual country remains a trusted workhorse. Its strength lies in the abil-
interest rate spreads are considered, the South African ity to combine the expected utility maximization models
and Namibian interest rate spreads experienced a 22.03 and the hedging hypothesis in the analysis of interest rate
and 22.39% decline between 2007 and 2017, respectively. spreads.
Conversely, Eswatini and Lesotho exhibited positive The bank dealership model's development borrows
growth rates in their interest rate spreads at a rate of from the literature on bid-ask prices for security market
23.52 and 35.06%, respectively. The findings reveal that dealers. In it, banks are considered as “dealers” that per-
Lesotho's banking sector is the least efficient out of the form the function of risk-averse intermediators in the
four CMA countries while South Africa's is the most credit market. They pay for funds (deposits) at a particu-
efficient. lar price (a “bid” price) and lend funds out at another
price (an “ask” price). Under this context, banks are faced
with two realities. First, uncertainty and costs arise from
3 | LITERATURE R EVIEW the stochastic behaviour of deposit suppliers and loan
demanders. For example, suppliers of deposits and
This section provides a brief review of theoretical and demanders of loans tend to arrive at dissimilar times,
empirical literature linking interest rate spreads and mac- causing banks to either hold a long or short position in
roeconomic variables from the perspective of developed the short-term money market. Second, banks deal with
and developing countries. loan demands and offers of deposits in money market
environments characterized by interest rate volatility. In
both cases, the assumption is that banks temporarily
3.1 | Theoretical review invest funds in the money market at given interest rates.
Should the interest rate on their money market invest-
The theoretical environment within which most investi- ments decline, they will be faced with reinvestment risk
gations of the determinants of banks' interest rate spreads since deposits might increase at a faster pace than the
are undertaken is usually framed in the context of two demand for loans. Similarly, in cases where the demand
approaches. Namely the hedging hypothesis and the for loans grows at a quicker rate than the supply of
expected utility maximization models. The hedging deposits, banks will be forced to borrow the shortfall
hypothesis explains the attempts of financial intermedi- from the money market. If the rate of interest in the
aries to minimize their shareholders' risk by matching money market rises, banks will be faced with refinancing
DAMANE 5
risk. Notwithstanding the aforementioned risks, further Huizinga (1999), their study includes a variety of bank-
credit risk can manifest itself in failure by loan specific factors as well as macroeconomic factors as pre-
demanders to pay back money lent to them in principal dictor variables of commercial banks' interest rate
and in interest. As a consequence, banks will set their spreads in their analysis. The findings of the study show
interest rate as a margin relative to the interest rate of the that monetary policy variables, namely reserve require-
money market. In addition, they will demand a positive ments and capital controls (as well as the control vari-
interest margin as a cost of providing intermediary ser- ables inflation and the corporate tax rate), play a more
vices in an uncertain environment caused by the asyn- significant role in the determination of interest margins,
chronous nature of deposit supplies and loan deposits accounting for over 76% of the variation, than do bank-
(Allen, 1988; Chirwa & Mlachila, 2004; Demirgüç-Kunt & specific factors such as bank size and provisions of non-
Huizinga, 1999; Ho & Saunders, 1981; Khan & performing loans.
Jalil, 2020; Männasoo, 2013; McShane & Sharpe, 1985; Saunders and Schumacher (2000) extend on the deal-
Obeng & Sakyi, 2017; Saunders & Schumacher, 2000; ership model by Ho and Saunders (1981) to investigate
Sheriff & Amoako, 2014; Tarus et al., 2012). the determinants of bank margins in a sample of banks
Since it was introduced, the bank dealership model in seven of the major countries of the Organisation of
has benefited from various extensions at the hands of Economic Corporation and Development (OECD) over
researchers. In general, it could be concluded that the the period 1988–1995. Much like Demirgüç-Kunt and
various model extensions sort to explore how the intro- Huizinga (1999) as well as Brock and Suarez (2000),
duction of additional bank-specific and/or macroeco- Saunders and Schumacher (2000) underscore that inter-
nomic factors would affect banks' interest rate spreads. est margins can be affected by bank-specific factors as
For instance, McShane and Sharpe (1985) augment the well as macroeconomic factors. Their analysis decom-
model by replacing the interest rates on deposits and poses determinants of bank interest margins into three
credits, as the source of interest rate risk, with uncer- components. Namely (a) indirect “tax”/regulatory effect
tainty in the money market. Furthermore, Allen (1988) (i.e., reserve requirements); (b) market structure effect
opted to include different forms of credits and deposits, (reflecting relative degree of monopoly power) and
whereas Maudos and De Guevara (2004) extended the (c) the risk premium effect (reflecting the extent of inter-
variable list by including operating costs into the model. est rate risk borne by bank in intermediation). The find-
ings of the paper were similar to those by Brock and
Suarez (2000); in that, they indicate that an increase in
3.2 | Empirical review reserve requirements and interest rate volatility leads to a
widening of interest margins.
Demirgüç-Kunt and Huizinga (1999) make use of regres- Chirwa and Mlachila (2004) focus their attention on
sion analysis and bank-level data (income statements and financial sector reform and its impact on interest rate
balance sheets) for 80 industrial and developing coun- spreads in Malawi's commercial banking system. Their
tries, spanning 1988–1995, to investigate the determi- study uses monthly panel data from five Malawian com-
nants of bank interest margins and profitability. The mercial banks and covers the period 1989–1999. The ana-
findings of the study reflect that interest margins are lytical framework is similar to that taken by Demirgüç-
affected by determinants that include bank characteris- Kunt and Huizinga (1999), Brock and Suarez (2000), as
tics, macroeconomic conditions, deposit insurance regu- well as Saunders and Schumacher (2000) in that it
lation, overall financial structure, explicit and implicit explains interest margins as a function of bank- and
bank taxation and the underlying legal and institutional market-specific factors, the regulatory environment and
indicators. Specifically, increases in the ratio of bank macroeconomic characteristics. One of the main contri-
assets to GDP and lower market concentration ratio were butions is in the use of both narrow and wide definitions
discovered to lead to lower interest margins. Further evi- of interest margins as dependent variables. Notably, the
dence suggests that an increase in the level of inflation as results of the study show that increases in bank-specific
well as credit risk leads to higher interest rate margins. factors such as the provision for doubtful debts widen
Similarly, higher corporate tax burden on banks is fully interest margins. The same goes for regulatory factors
passed onto bank customers (i.e., lead to higher interest such as the liquidity reserve requirement and macroeco-
margins) while a higher reserve requirements are not. nomic factors such as price instability, represented by
Brock and Suarez (2000) investigate the determinants inflation.
of bank spreads in the Caribbean14 using unbalanced Crowley (2007) recognizes that most studies on inter-
panel data regression techniques and data that spanned est margins in the literature pay very little attention to
1989–2004. In a similar way to Demirgüç-Kunt and evidence from the African continent. Their study fills this
6 DAMANE
gap by investigating the determinants of interest rate consumer price inflation leads to a widening of interest
spreads in English-speaking African countries using rate spreads in the short and long-run. The relationship
panel data regression techniques and data for the period between total banking sector deposits and the interest
1975–2004. Just as Demirgüç-Kunt and Huizinga (1999) margin was found to be positive and significant in the
and Chirwa and Mlachila (2004), Crowley (2007) notes long run while insignificant in the short-run. Conversely,
the presence of various definitions of interest rate spreads T-bill rates exert a negative influence on the interest rate
and elects to use a narrow definition. The study also spread in the short-run but are statistically insignificant
includes macroeconomic factors such as inflation as well determinants of the interest margin in the long-run.
as bank-specific factors such as non-performing loans as In a similar way to Sheriff and Amoako (2014), Obeng
predictor variables in the regression analysis. A notable and Sakyi (2017) study the factors that affect interest rate
finding is that unlike the conclusions of Demirgüç-Kunt spreads in Ghana in the short and long-run using the
and Huizinga (1999) and Chirwa and Mlachila (2004), ARDL bounds test approach to co-integration. They
the results reflected that an increase in inflation led to extend on the work by Sheriff and Amoako (2014) by
lower interest rate margins in English-speaking African introducing, among the predictor variables, money
countries. growth and a measure of institutional quality. The study
Tarus et al. (2012) use panel data regression tech- uses annual data covering the period 1980–2013. Its find-
niques and data spanning from 2000 to 2009 on a sample ings reaffirm the existence of a long-run relationship
of 44 banks to investigate the determinates of interest between the study's variables in Ghana. In addition,
margins of commercial banks in Kenya. Just as other inflation, fiscal deficit, crowding out effect, exchange rate
studies reviewed previously, the research explains inter- volatility, deposit interest rate volatility, money growth as
est rate spreads in terms of macroeconomic together with well as the monetary policy inters rate exert a positive
bank- and market-specific factors. Operating expectations and statistically significant influence on the interest mar-
and credit risk are found to have a positive impact on the gin in the short and long-run.
interest rate margin. On the same token, the study dis- Much like Ahokpossi (2013), Tarus and Man-
covers that high levels of inflation translate in wider com- yala (2018) investigate the determinants of bank interest
mercial bank interest rate spreads in Kenya. This rate spreads in sub-Saharan African countries using fixed
evidence is contrary to the findings of Crowley (2007) effects estimations and data drawn from a sample of
and similar to those by Demirgüç-Kunt and Hui- 20 sub-Saharan African countries for a period spanning
zinga (1999) coupled with Chirwa and Mlachila (2004). 10 years from 2003 to 2012. The determinants of interest
Ahokpossi (2013) examines the determinants of bank margins are categorized into macro-specific, bank-
interest margins in sub-Saharan Africa using a sample of specific and institutional variables. Unlike the study by
456 banks in 41 sub-Saharan African countries coupled Ahokpossi (2013), findings of the research show that
with panel data regression techniques and data spanning inflation has a negative and significant effect on the
1995–2008. Interest margins are explained as a function interest rate spread whereas operating costs and bank
of bank-specific factors, market structure and macroeco- concentration have a positive and significant effect on
nomic factors. The study finds that when market concen- the interest margin.
tration is interacted with bank efficiency, its impact on Khan and Jalil (2020) explore the determinants of the
bank interest spreads is positive and significant while net interest margin (NIM) in Pakistan using a two-step
positive and insignificant when it is not interacted with system generalized method of moments (GMM) and
bank efficiency. This shows that the impact of bank con- unbalanced quarterly panel data from 46 commercial
centration on interest margins depends on bank effi- banks spanning from 2003Q3 to 2017Q. The data com-
ciency. Inflation was also found to have a positive and prise a combination of commercial banks' data
significant impact on interest rate margins while the (e.g., operation cost, credit risk, managerial efficiency)
growth in GDP was insignificant. and data on macroeconomic indicators (growth in broad
Sheriff and Amoako (2014) use of monthly time series money, national saving, inflation, taxation). Notable find-
data from 1999 to 2010 and the Autoregressive Distrib- ings of the study are that the money supply, operating
uted Lag (ARDL) model to investigate the macroeco- costs, T-bill rate and national savings are positively and
nomic determinants of commercial bank spreads in statistically significantly related to the NIM whereas
Ghana in the short and long run. The study's variables managerial efficiency and risk aversion have no signifi-
included consumer price inflation, Treasury bill rates, cant effect on the NIM in Pakistan's banking sector.
total banking sector deposits and a proxy for public sector The studies reviewed in the literature reflect that the
crowding out. Findings reveal the existence of co- pricing behaviour of commercial banks can be affected
integration between the variables. An increase in the by a wide array of factors that include the structure of the
DAMANE 7
TABLE 1 Variable names and sources
Nature of variable Name of variable Abbreviation Variable state Source of data

Dependent variable Interest rate spread SP Percentage CBL
Independent variables Macroeconomic factors Consumer price inflation INF Percentage BOS
91-day Treasury bill rate TB Percentage CBL
Growth in narrow money M1g Percentage CBL
1 year deposit rate DR Percentage CBL
Credit risk CR Percentage CBL
Abbreviations: BOS, Bureau of Statistics; CBL, Central Bank of Lesotho.
capital market, the nature of the regulatory environment factors. Under the macroeconomic factors, the rate of
as well as bank-specific and macroeconomic factors. An inflation is measured by changes in the consumer price
overarching consistency in the literature is the impor- index (CPI). Narrow money comprises the currency out-
tance of the role of macroeconomic and bank-specific fac- side banks and demand deposits. Considering the bank-
tors in the determination of commercial banks' interest specific factors, the credit risk is proxied by the ratio of
rate spreads. To capture the impact of such factors on total loans by banks to total bank assets. All variables are
banks' pricing, interest spread equations often include in percentage terms. Table 2 presents the study's a priori
money market rates (Treasury bill and discount rates), expectations.
credit risk, consumer inflation and monetary aggregates
(i.e., growth of money supply) as control variables. In
general, findings show that increases in these variables 4.2 | Model specification
lead to wider commercial bank spreads across both devel-
oped and developing economies. The generalized model for examining the relationship
between interest rate spread and a handful of macroeco-
nomic variables in Lesotho is presented as follows;
4 | EMPIRICAL FRAMEWORK
SPt = α0 + β1 INFt + β2 TBt + β3 M1gt + β4 DRt + β5 CRt + εt ,
4.1 | The data ð1Þ
The study makes use of monthly data from January 2008

to December 2018, making it a total of 120 months of where variables are as explained in Table 1, α0 is the con-
data. The choice of the timeline is dictated by the avail- stant term, β1, β2, β3, β4 and β5 are the long-run coeffi-
ability of data pertaining to the variables of interest. cients while εt is the error term, which is assumed to be
Table 1 shows the names of the variables included in the white noise.
study as well as their sources. According to Demirgüç- The Autoregressive Disturbed Lag (ARDL) bounds
Kunt and Huizinga (1999), Brock and Suarez (2000) as testing approach to co-integration and Error Correction
well as Chirwa and Mlachila (2004), the interest rate Model (ECM) based on the ARDL procedure developed
spread can be defined both in a narrow and wide way, by Pesaran and Shin (1999) and advanced by Pesaran,
with the main distinction lying in the inclusion or exclu- Shin, and Smith (2001) is used to investigate the symmet-
sion of fees and commissions relating to the loan and ric relationship between the interest rate spread and a
deposit transactions in each case. Our study defines the handful of macroeconomic variables. The ARDL bounds
commercial bank interest rate spread in Lesotho on a testing approach offers the following advantages over
narrow basis in the same way as Demirgüç-Kunt and other alternative co-integration techniques such as the
Huizinga (1999) and Crowley (2007). Specifically, it is Johansen (1988) and the Johansen and Juselius (1990).
defined as the difference between the industry average First, unlike the Johansen conventional co-integration
prime interest rate earned on loans and the industry aver- method that estimates the long-run relationship under
age interest rate paid on 1 year deposits by banks. This the restrictive assumption that all of the model's variables
definition is chosen due to its wide applicability. are integrated of order 1, that is it [I(1)], the ARDL
From the table, the independent variables are approach can be applied irrespective of whether the
grouped into macroeconomic factors and bank-specific underlying regressors are integrated of order one [I(1)],
8 DAMANE
TABLE 2 A priori expectations
Relationship between Expected sign Rationale Source

SP and INF Positive The inflation rate is considered to be an Demirgüç-Kunt and Huizinga (1999),
indicator of the cost of doing business. Thus, Brock and Suarez (2000), Chirwa and
higher levels of inflation are expected to lead Mlachila (2004), Crowley (2007),
to higher lending rates especially in Folawewo and Tennant (2008),
developing countries where inflation is high Ahokpossi (2013), Sheriff and
and variable. Amoako (2014), Obeng and
Sakyi (2017), Khan and Jalil (2020)
SP and TB Positive/Negative Positive: Owing to its strong relationship with Demirgüç-Kunt and Huizinga (1999),
the commercial bank rates as a benchmark Brock and Suarez (2000), Chirwa and
for rates charged by banks, the 91-day T-Bill Mlachila (2004), Folawewo and
rate is considered an indicator of Tennant (2008), Sheriff and
government interest rate policy. The Amoako (2014), Tarus and
expectation is that a higher T-Bill rate will Manyala (2018), Khan and Jalil (2020)
lead to higher interest rate spreads through
its positive influence on the lending rate for
a given level of the deposit rate.
Negative: Treasury bills act as alternative
investment vehicles for savers. As such,
increases in the T-bill rate negatively affect
banks' interest margins. Savers, with the
option of choosing between low savings and
high T-bill returns will force banks to offer
competitive savings rates to attract them
(given that banks cannot attract any more
lenders due to tight competition). This will
lead to a narrowing of the interest rate
margin.
SP and M1g Positive The growth in narrow money creates Sheriff and Amoako (2014), Obeng and
inflationary pressures and thus increases the Sakyi (2017), Khan and Jalil (2020)
risk premium of banks which will lead to an
increase in the lending rates and therefore
the interest rate spread at a given level of the
deposit rate.
SP and DR Negative An increase in the deposit rate is expected to Sheriff and Amoako (2014), Obeng and
encourage greater demand for deposits that Sakyi (2017)
will translate into a greater supply of
loanable funds and a reduction in the
lending rate and thus the interest margin.
SP and CR Positive The higher the credit risk ratio (as, defined in Demirgüç-Kunt and Huizinga (1999),
our study as the ratio of loans to total Maudos and De Guevara (2004), Tarus
assets—to capture the risks in lending) the et al. (2012), Ahokpossi (2013),
more banks are exposed to loan default risk. Jamaludin, Klyuev, and
This can also be used to measure asset Serechetapongse (2015)
quality. An increase in the credit risk ratio
translates into weaker asset quality and as
such banks will want to make up for past
losses by charging higher lending rates for a
given level of the deposit rate and cover this
risk.
order zero [I(0)] or mutually co-integrated. This means it provides for the possibility that different variables have
that it avoids the pre-testing of variables to identify their different optimal lags, which is impossible under the con-
order of integration (although pre-testing is rec- ventional Johansen approach. Third, it produces robust
ommended to ensure the variables are not [I(2)]). Second, results even in cases of small samples. Fourth, this
DAMANE 9
technique has finite-sample critical values as opposed to Once co-integration is proven to exist between the
other co-integration approaches for which the distribu- variables of interest, the long-run and error correction
tion of the test statistic may not be known in finite sam- models are estimated using the ARDL framework as
ples. Narayan (2005) develops a set of sample-specific shown in Equations (3) and (4), respectively.
critical value bounds for the sample sizes ranging from
30 to 80 using the same approach and GAUSS code used X
m X
n
SPt = α0 + Ω1 SPt − i + Ω2 INFt − i
by Pesaran et al. (2001) in generating the asymptotic
i=1 i=1
values. Furthermore, it provides unbiased estimates of X
r X
v X
w
the long-run model and valid t-statistics even in the pres- + Ω3 TBt − i + Ω4 M1gt − i + Ω5 DRt − i
ence of endogenous regressors. This will prove important i=1 i=1 i=1
in the study because of the potential endogeneity of Xx

+ Ω6 CRt − i + εt , ð3Þ
regressors such as inflation. Last but not least, ARDL i=1
framework allows the derivation of an ECM through a
simple transformation, which generates short-run adjust- Xm Xn
ΔSPt = α0 + փ ΔSPt − i +
i=1 1
փ ΔINFt − i
i=0 2
ment with long-run equilibrium without losing the long- Xr Xv
run information (Nkoro & Uko, 2016; Obeng & + փ3 ΔTBt − i + փ4 ΔM1gt − i
Sakyi, 2017; Sheriff & Amoako, 2014; Zhang, Tsai, & Xiw= 0 Xix= 0
+ փ ΔDRt − i +
i=0 5 i=0 6
փ ΔCRt − i
Chang, 2017).
In order to carry-out the bounds test, the following + փ7 ECTt − 1 + μt , ð4Þ
unrestricted error correction model (UECM) is estimated:
where all the variables are as previously defined. εt and μt
ΔSPt = α0 + Ω1 SPt − 1 + Ω2 INFt − 1 + Ω3 TBt − 1 are the random error terms while Ω1…. Ω6 and փ1…. փ7
+ Ω4 M1gt − 1 + Ω5 DRt − 1 + Ω6 CRt − 1 are the parameters of each model while m, n, r, v, w and
Xp Xq x are the maximum lag lengths15 and փ7 is the coefficient
+ β ΔSP − + β2 ΔINFt − i
Xiq= 1
1 t i
Xiq= 1 of the lagged error correction term (ECTt − 1). The փ7
+ β3 ΔTBt − i + β4 ΔM1gt − i coefficient measures the speed of adjustment to long-run
Xiq= 1 Xiq= 1 equilibrium following a shock to the system in the previ-
+ β ΔDRt − i +
i=1 5
β ΔCRt − i + εt ,
i=1 6
ð2Þ
ous period.
where all variables are as previously defined, Δ is the first

difference operator, p and q are the lag lengths, α0 is the 4.3 | Non-linear ARDL approach
drift component and εt is the random error term. The
ARDL framework, evaluates the long-run relationship Equation (1) is a representation of the long-run relation-
between variables in the model through the Wald test of ship between the interest rate spread and selected macro-
coefficient restriction by testing the null hypothesis; economic variables with the assumption that the
H0 : Ω1 = Ω2 = Ω3 = Ω4 = Ω5 = = 0 against the alternative explanatory variables linearly affect the dependent vari-
hypothesis; H1 : Ω1 ≠ Ω2 ≠ Ω3 ≠ Ω4 ≠ Ω5 ≠ Ω6 ≠ 0. able.16 However, according to Shin, Yu, and Greenwood-
The computed F-statistic derived from the Wald test is Nimmo (2014), Michis (2016), Lacheheb and Sirag (2016),
compared with two sets of critical values (lower and Zhang et al. (2017), Kwasi Obeng (2018), as well as
upper bound values) for a given level of significance Qamruzzaman and Jianguo (2018), the assumption of a
reported in Pesaran et al. (2001) and Narayan (2005) for linear relationship between dependent and independent
large samples and small sample sizes, respectively. The variables may not always be sufficient to properly explain
lower bound values assume that all variables in the macroeconomic relationships. In light of this, the study
ARDL model are I(0) while the upper bound values further examines the asymmetric relationship between
assume that the variables are I(1). Therefore, if the com- the interest rate spread and the handful of macroeco-
puted F-statistic is less than the lower bound value, the nomic variables identified in Table 1. To do this, the
null hypothesis of no co-integration cannot be rejected. study adopts the non-linear ARDL (NARDL) co-
On the other hand, if the computed F-statistic is greater integration approach advanced by Shin et al. (2014). The
than the upper bound value, the null hypothesis of no NARDL is an asymmetric extension of the ARDL model
co-integration is rejected and it is concluded that the that is capable of simultaneously and coherently model-
variables are co-integrated. Nonetheless, the test ling asymmetries in both the short and long run. More-
becomes inconclusive in cases where the computed over, it can be used to derive asymmetric cumulative
F-statistic falls between the two critical bound values. dynamic multipliers that allow for the tracing out of the
10 DAMANE
asymmetric adjustment patterns of the dependent vari- non-linear long-run and non-linear error-correction
able given positive and negative shocks to the indepen- models are estimated
dent variables (Karantininis, Katrakylidis, &
Persson, 2011; Shin et al., 2014; Zhang et al., 2017). Simi- X
m X
n
SPt = α0 + Ω1 SPt − i + Ω2+ INFt+− i
lar to Shin et al. (2014), Qamruzzaman and
i=1 i=1
Jianguo (2018), Zhang et al. (2017) as well as Kwasi X
q X
r X
v
Obeng (2018), the asymmetric pass-through of specific + Ω3− INFt−− i + Ω4 TBt − i + Ω5 M1gt − i
macroeconomic variables on the interest rate pass- i=1 i=1 i=1
through is represented by decomposing the macroeco- Xw X

x
+ Ω6 DRt − i + Ω7 CRt − i + φt , ð8Þ
nomic variables into positive and negative shocks, where, i=1 i=1
for example, considering only the proxy for inflation,
INF; INF+ and INF− represent the partial sums of posi- Xm Xn
ΔSPt = α0 + փ 1 ΔSP t − i + փ2+ ΔINFt+− i
tive and negative changes in the INF variable. These par- Xq i=1
Xr i = 0
tial sums are calculated as follows: + փ − ΔINFt−− i +
i=0 3
փ4 ΔTBt − i
Xv Xw i = 0
+ փ5 M1gt − i + փ DRt − i
i=0 6
P P ) Xix= 0
INFt+ = ti − 1 ΔINFt+ = ti − 1 maxðΔINFi, 0Þ + փ CRt − i + փ8 ECTt − 1 + μt : ð9Þ
P P : ð5Þ i=0 7
INFt− = ti − 1 INFt− = ti− 1 minðINFi, 0Þ
The existence of long-run and short-run symmetry is
tested using the Wald test with the null hypothesis H 0 :
P P
Based on the partial sums presented in Equation (5), Ω2+ = Ω3− and H 0 : qi= 0 փ2+ = qi= 0 փ3− , respectively.
the long-run asymmetric relationship between the inter- Should the null hypothesis of symmetry be rejected, the
est rate spread and inflation is estimated as follows: asymmetric dynamic multipliers17 of a change in INF−
and INF+ can be developed, respectively. The cumulative
dynamic multiplier effects of INF− and INF+on SP are
SPt = α0 + β1+ INFt+ + β2− INFt− + β3 TBt evaluated using the following formulas:
+ β4 M1gt + β5 DRt + β6 CRt + εt , ð6Þ
Ph ∂SPt + i Ph ∂SPt + i
mh+ = i = 0 ∂INFt+ ; mh− = i = 0 ∂INFt− : (10)
where variables are as explained in Table 1, α0 is the
constant term, β1+ , β2− , β3 , β4 … β6 are the long-run coef- According to Shin et al. (2014), as h!∞ then mh+ !
ficients while εt is the error term, which is assumed Ω2+ , mh− ! Ω3− , where Ω2+ and Ω3− are the asymmetric
to be white noise. To test the existence of asymmetric long-run coefficients.
long-run relationship between the interest rate
spread and the inflation rate, the study estimates the
following non-linear conditional error correction 5 | D A T A A N A LY S I S A N D
model (NECM): I NT E R P R E T A T I O N
ΔSPt = α0 + π 1 SPt − 1 + π 2+ INFt+− 1 + π 3− INFt−− 1

5.1 | Descriptive statistics
+ π 4 TBt − 1 + π 5 M1gt − 1 + π 6 DRt − 1 + π 7 CRt − 1
Xp Xq A graphical depiction of the variables is provided in
+ i=1 1
ψ ΔSPt − i + i=0 2
ψ + ΔINFt+− i Figure A1. Table 3 highlights the descriptive statistics
Xq X q (means and standard deviations as well as tests for nor-
+ i=0 3
ψ − ΔINFt−− i + ψ 4 ΔTBt − i
Xq Xqi = 0 mality) for all the variables under consideration.
+ ψ 5 ΔM1gt − i + i=1 6
ψ ΔDRt − i From the table, the average interest margin over the
Xiq= 1
+ ψ ΔCRt − i + ϕt , ð7Þ review period is 7.73% while the average level of infla-
i=1 7
tion is 5.22%.18 The mean 91-day T-Bill rate is 0.97%
higher than the average rate of inflation. Holders of T-
where ψ 1…. ψ 7 are the short-run coefficients and the Bills during the period under consideration were com-
long-run coefficients are represented by π 1…. π 7. pensated for the loss of value resulting from increases in
Just as was done under the ARDL model, after esta- prices. The banking sector's total loans as a ratio of its
blishing the existence of co-integration using the NARDL total assets measures the credit risk and it averaged
bounds testing procedure, which is similar to the ARDL approximately 35.17% with a maximum of 48% and a
bounds testing procedure outlined earlier, the following minimum of 20.17%. This shows that credit extension is
DAMANE 11
TABLE 3 Descriptive statistics

SP INF TB M1G DR CR
Mean 7.73 5.22 6.19 0.58 3.28 35.17
Median 7.71 5.20 6.25 1.25 3.22 37.93
Maximum 9.26 10.70 10.01 19.91 6.74 48.00
Minimum 6.97 2.00 4.94 −53.50 2.34 20.17
SD 0.45 1.64 0.84 7.80 0.80 6.71
Skewness 0.56 0.81 1.96 −2.69 1.60 −0.82
Kurtosis 3.16 4.32 9.67 21.12 7.41 2.36
Jarque–Bera 6.36 21.83 299.11 1786.20 148.54 15.75
Probability 0.04 0.00 0.00 0.00 0.00 0.00
Sum 928.15 626.00 743.39 69.93 393.18 4,220.84
Sum Sq. Dev. 24.36 319.39 83.78 7,231.63 75.98 5,371.77
Observations 120 120 120 120 120 120
TABLE 4 Correlation results

Correlation SP INF TB M1G CR DR
SP 1
INF 0.47801 1
TB 0.49601 0.450785 1
M1G −0.06294 −0.01896 −0.05621 1
CR −0.55238 −0.1364 −0.37659 0.098568 1
DR 0.252787 0.369549 0.892561 −0.04921 −0.30944 1
an important part of banks' business. The growth in nar- 5.2 | Unit root test
row money averaged 0.58% with a negative of 53%19 and
a maximum of 19.91%. Unlike other econometric techniques, the ARDL
We tested for the correlation between the variables. approach does not require pre-testing for unit roots.
The results are presented in Table 4 and offer a prelimi- Although the technique works well when variables are
nary assessment of the a priori expectations discussed integrated of different order, I(0), I(1) or a combination
under Table 2. In general, the a priori expectations of of both, unit root testing is conducted to ensure that
the relationship between interest margins and inflation the variables are not integrated of order two, that is I
as well as between interest margins and Treasury bill (2), since the ARDL technique crashes in the presence
rates are supported by the correlation coefficient results. of I(2) variables (Damane et al., 2018; Nkoro &
On the other hand, the correlation coefficient between Uko, 2016; Obeng & Sakyi, 2017; Sheriff &
interest rate spreads and the growth in narrow money, Amoako, 2014). The study made use of the Augmented
credit risk and the deposit rate, respectively, go against Dickey–Fuller (ADF) test (Dickey & Fuller, 1981) and
a priori expectations. The inconsistencies could be the Phillips and Perron (PP) test (Phillips &
explained by idiosyncrasies in Lesotho's economy that Perron, 1988) to determine the order of integration of
include the fixed exchange rate regime or the fact that the variables. The results of the ADF and PP unit root
the lion's share of banks' household credit goes to public tests are presented in Table 5.
service employees, where banks are able to deduct loan The results of both the ADF and PP tests provide
premiums from the source. Moreover, according to evidence that the interest rate spread, inflation, Trea-
CBL (2019), the country's commercial banks are highly sury bill rate, growth in narrow money and the deposit
liquid (consistently outperforming prudential liquidity rate are stationary in levels. Conversely, the credit risk
thresholds) with a liquidity stricture that mostly relies variable is stationary after first difference. Given that
on wholesale funding. none of the variables are integrated of order two, the
12 DAMANE
TABLE 5 Standard unit root tests
ADF PP
Level First Difference Level First Difference
Variable Trend and Trend and Trend and Trend and

descriptor Intercept Intercept Intercept Intercept Intercept Intercept Intercept Intercept
SP 0.0187** 0.0761* — — 0.0193** 0.0778* — —
INF 0.0004*** 0.0029*** — — 0.0053*** 0.0385** — —
TB 0.0000*** 0.0001*** — — 0.0000*** 0.0000*** — —
M1G 0.0000*** 0.0000*** — — 0.0000*** 0.0000*** — —
CR 0.2337 0.7857 0.0000*** 0.0000*** 0.1988 0.6621 0.0000*** 0.0000***
DR 0.0000*** 0.0000*** — — 0.0004*** 0.0000*** — —
Note: ***, ** and * denote that a series is stationary at 1%, 5% and 10% levels of significance, respectively.
TABLE 6 ARDL bounds testing to co-integration results
F-bound test Null hypothesis: no levels relationship
Test statistic Value Significance I(0) I(1) Evidence of co-integration

Calculated F-statistic 7.955213* 10% 2.26 3.35 Yes
K 5 5% 2.62 3.79
Actual sample size 111 1% 3.41 4.68
Note: Critical values were generated by Eviews 11 and they are in line with those extracted from Narayan (2005); K is the number of regressors.
*Denotes the level of statistical significance at 1%.
study proceeded to carry-out ARDL bounds testing Using the Akaike information criteria (AIC), the follow-
approach to co-integration. ing specification of the model was chosen; ARDL (1, 0, 9, 0,
0, 8). Consistent with the findings of Demirgüç-Kunt and
Huizinga (1999), Brock and Suarez (2000), Chirwa and
5.3 | Results of the bounds test for Mlachila (2004), Crowley (2007), Folawewo and Tennant
co-integration (2008), Ahokpossi (2013), Sheriff and Amoako (2014),
Obeng and Sakyi (2017) and Khan and Jalil (2020), the
After confirming that none of our model variables are inte- macroeconomic indicators included in our study conform
grated of order two, the next step is to investigate for the exis- to a priori expectations. Our results indicate a positive and
tence of long-run co-integration between model variables highly statistically significant relationship between
under the null hypothesis of no co-integration using the bou- inflation (INF) and interest rate spreads in the long-run.
nds test approach. The results are displayed in Table 6. This confirms that banks in Lesotho perceive the infla-
The calculated F-statistic of 7.96 is greater than the tion rate to be an indicator of the cost of doing business
upper bound critical values at all levels of significance. in the country and thus price it into their interest mar-
The findings offer overwhelming evidence of the exis- gins. Although the coefficient of inflation is positive and
tence of a long-run steady relationship between the inter- highly statistically significant in the long-run, it is con-
est rate spread and chosen predictor variables in Lesotho. siderably less than one. This is a reflection that devia-
This finding is in harmony with the findings of Sheriff tions from average inflation rates are not fully passed
and Amoako (2014) and Obeng and Sakyi (2017). through into lending rates. The finding is consistent
with results discovered by Jamaludin et al. (2015), Sher-
iff and Amoako (2014) together with Obeng and
5.4 | Long-run coefficient estimation Sakyi (2017). The implication is that deviations in infla-
tion affect the interest rate spread with only a secondary
Since co-integration has been established between the effect owing to the possibility that banks expect them to
model variables, next, the study estimated the long-run be temporary.
coefficients using the ARDL framework. The results of We also find, consistent with Tarus and Man-
the long-run model are presented in Table 7. yala (2018), that the coefficient of the 91-day T-Bill rate is
DAMANE 13
T A B L E 7 ARDL results of the

Explanatory variable Coefficients T-statistics (p Value)
long-run relationship estimation
INF 0.193620*** 7.896859 (.0000)
TB 0.996124*** 9.246192 (.0000)
M1g 0.002266 0.588332 (.5578)
CR 0.002363 0.356614 (.7222)
DR −0.816829*** −8.001820 (.0000)
C 3.099173*** 5.886072 (.0000)
Note: Dependent variable = SP. *** denote the level of statistical significance at 1% respectively. The values
in parentheses are the probability values. The selected model is ARDL (1, 0, 9, 0, 0, 8).
positive and statistically significant in the long-run. A 1 % 5.5 | Short-run dynamics estimation
increase in the T-Bill rate is expected to lead to an based on ARDL
increase of 0.99% increase in the interest margin in the
long-run, holding all else constant. The finding confirms The short-run dynamics of the relationship between
that commercial banks view the T-Bill rate as an indica- interest rate spreads, macroeconomic factors and bank-
tor of government interest rate policy with the expecta- specific factors were investigated by following the ECM-
tion that positive changes in the T-Bill rate will translate ARDL model. Table 8 reports the results of estimated
into an increase in their cost of borrowing and as such short-run coefficients together with the associated model
they price it into the services they offer. Even though the diagnostic tests. The coefficient of the lagged error correc-
growth of narrow money is positive as per our a priori tion term (ECT), which represents the speed of adjust-
expectations, it is not statistically significant in the long- ment to the long-run equilibrium following a shock in
run. A similar finding was obtained by Folawewo and the previous period, is negative and statistically signifi-
Tennant (2008) for countries in sub-Saharan Africa and cant at 1% level of significance. It shows that any shock
by Obeng and Sakyi (2017) in Ghana. The implication of in the previous period associated with the model is
this finding is that banks appear not to consider the adjusted back to equilibrium in the long-run with at an
growth in money supply to result in any permanent devi- approximate speed of 32%.
ations of inflation from its average. As a result, they do From Table 8, the coefficients of inflation and the
not price-in any risk premium into their lending rate as a deposit rate in the short-run behave in a similar way to
result of money supply growth. how they behaved in the long-run. The impact of infla-
Evidence from the bank-specific factors reveals that tion on interest rate spread is still positive and highly sta-
the credit risk variable as measured by the ratio of total tistically significant. However, the magnitude is much
loans by banks to total bank assets is positive as per a smaller than it is in the long-run results. These findings
priori expectations but is not statistically significant. A are in line with those obtained by Sheriff and
possible explanation lies in the consistently low levels of Amoako (2014) coupled with Obeng and Sakyi (2017)
non-performing loans that averaged around 3.04% over and suggest that banks price-in inflation to a lesser extent
the period from January 2009 to December 2018. In addi- in the short-run than they do in the long-run. They also
tion, according to the IMF (2018), Lesotho's banks are rely less on deposit liabilities to finance credit extension
quite well capitalized such that even in the face of in the short-run than in the long-run.
extreme fiscal shocks such as expenditure cuts, salary Model diagnostic tests for serial correlation,
freezes, retrenchment of personnel and increasing arrears heteroscedasticity, the Ramsey-RESET and the long-run
to service providers, banks would easily be able to cush- normality are presented in Tables A1–A4. These diagnos-
ion against the first round impact of the shock. We also tic test results indicate that the model residuals are nor-
find, consistent with Jamaludin et al. (2015) and in a sim- mally distributed in the long-run and that they have no
ilar way to the coefficient of inflation discussed earlier, serial correlation and no heteroscedasticity. Moreover,
that the coefficient on the deposit rate is highly statisti- the results of the Ramsey-RESET show no evidence of
cally significant, but considerably smaller than one. The model misspecification. Figures A2 and A3 report the
less than one-for-one pass-through from deposit to lend- cumulative sum (CUSUM) and cumulative sum of
ing rates is a reflection that deposits, while being an squares (CUSUMSQ) tests, respectively. The tests reflect
important source of bank funding, are not the only one. that the model is robust and stable since test recursive
14 DAMANE
T A B L E 8 Results of the short-run

Part A: Error correction model
relationship estimation
Explanatory variable Coefficients T-statistics p Value
D(INF) 0.062126*** 5.187288 (.0000)
D(TB) 0.052423 0.742109 (.4600)
D(M1g) 0.000727 0.582531 (.5617)
D(CR) 0.000758 0.364668 (.7162)
D(TB[−8]) −0.0411** −2.628 (.0142)
D(DR) −0.738038*** −9.536436 (.0000)
D(DR[−2]) 0.202966* 1.846192 (.0683)
D(DR[−5]) −0.202519* −1.860308 (.0662)
D(DR[−7]) 0.132923* 1.824816 (.0715)
ECT(−1) −0.320865*** −6.088737 (.0000)
Part B: Diagnostic tests
Test Test statistic Probability value
R-squared 0.957231 —
Adjusted R-squared 0.945924 —
Bai and Ng long-run normality test 2.591408 .2737
Breusch–Godfrey-LM test 0.370862 .8307
BPG heteroscedasticity test 21.92859 .5228
Ramsey-RESET test 0.136441 .7128
Note: Dependent variable = SP. ***, **, * denote the level of statistical significance at 1%, 5% and 10%,
respectively. The values in parentheses are the probability values. The selected model is ARDL (1, 0, 9, 0,
0, 8).
TABLE 9 NARDL bounds testing to co-integration results
F-bound test Null hypothesis: No levels relationship

K 8 5% 2.11 3.15
residuals remain within the boundaries of the 5% critical the NARDL bounds test for co-integration. In the same
lines. way as the ARDL model, the F-statistic for testing the
existence of long-run co-integration within the NARDL
framework provides evidence of the existence of a long-
5.6 | Results of the NARDL bounds test run steady-state relationship between model variables at
for co-integration all levels of significance.
The possibility of the existence of an asymmetric impact

of macroeconomic and bank-specific factors on interest 5.7 | NARDL long-run coefficient
margins in Lesotho was further examined. The NARDL estimation
model is used to capture positive and negative shocks in
inflation (INF), the deposit rate (DR) and the Treasury The establishment of the presence of co-integration
bill rate (TB). Table 9 reports the empirical findings of between the model variables allows for the estimation of
DAMANE 15
TABLE 10 Long-run NARDL

results
INF_POS 0.179074 6.069676 (.0000)
INF_NEG −0.128463 −3.077055 (.0031)
TB_POS 0.153496 2.051129 (.0446)
TB_NEG 0.875928 5.588960 (.0000)
DR_POS −1.259538 −7.891793 (.0000)
DR_NEG −0.377585 −1.270050 (.2089)
C 10.588110 16.926584 (.0000)
Diagnostic tests
Bai and Ng long-run normality test 2.375209 .3049
Breusch–Godfrey-LM test 3.707731 .1156
BPG heteroscedasticity test 48.77145 .4418
INF
W LR 0.050023 .8230
WTBLR 4.138859 .0419
Note: Dependent variable = SP. The values in parentheses are the probability values. The selected model is
ARDL (1, 9, 1, 1, 9, 3, 7, 1, 8). WINFLR and WTBLR refer to the Wald tests of long-run and short-run symmetry
in inflation and the Treasury bill rate, respectively.
the long-run coefficients associated with the NARDL with no signs of serial correlation and heteroscedasticity.
model. Table 10 presents the long-run estimates of the Moreover, there is no evidence of model misspecification,
NARDL model. and as far as the CUSUM and CUSUMSQ stability tests
Positive and negative shocks in inflation lead to are concerned, the model is found to be robust and sta-
highly statistically significant positive and negative devel- ble. It is also important to note that from the Wald test,
opments in interest margins in the long-run, respectively, the null hypothesis of long-run symmetry is only rejected
with differing magnitudes. In each case, just as in the for the Treasury bill rate at the 5 % level. This offers proof
ARDL long-run results, the coefficients are considerably of the long-run asymmetric effects of the Treasury bill
less than one. A 1 % increase in the level of inflation rate on interest rate spreads in Lesotho.
leads to a 0.18% increase in the interest rate spread while The dynamic multipliers (i.e., the asymmetric adjust-
a 1 % decline leads to only a 0.13% decline. This shows ments) of inflation, the Treasury bill rate and the deposit
that positive shocks to inflation have more impact on the rate from an initial long-run equilibrium to a new long-
interest margin than the negative shocks. Interestingly, run equilibrium following a 1 % positive and negative
positive and negative shocks to the Treasury bill rate lead shock are presented in Figures 3–5, respectively. The
increases to the interest margin. However, the incremen- results echo those discussed in Table 10. From the fig-
tal effect of the negative shock outweighs a similar effect ures, each variable attains a new equilibrium after
from the positive shock. According to Demirgüç-Kunt approximately 15 months after each positive or negative
and Huizinga (1999) and Brock and Suarez (2000), this unitary shock.
could be a reflection of banks' hedging activities against
reinvestment risk. Although a negative shock to the
deposit rate is statistically insignificant, a 1 % increase in 6 | ROBUS TNESS CHECKS
the deposit rate under the NARDL results in a negative
and highly statistically significant decline of 1.26% in the The robustness of our findings is examined by varying
interest margin. This magnitude is greater than it was in covariates and evaluating the long-run relationship.
the ARDL long-run results. Instead of credit risk (as defined by ratio of total loans by
The NARDL model diagnostic test results are pres- banks to total bank assets), we consider credit risk as
ented in Tables A5–A7 and Figures A4 and A5. They defined in two ways. First, as the ratio of non-performing
show that the model residuals are normally distributed loans to total assets, and, secondly, as the ratio of total
16 DAMANE
FIGURE 3 Long-run dynamic multiplier of INF on SP. INF F I G U R E 5 Long-run dynamic multiplier of DR on SP. DR
denotes inflation. The horizontal axis indicates months taken to denotes inflation. The horizontal axis indicates months taken to
achieve the long-term equilibrium association. The vertical axis achieve the long-term equilibrium association. The vertical axis
indicates the magnitude of negative and positive shocks in INF indicates the magnitude of negative and positive shocks in DR
[Colour figure can be viewed at wileyonlinelibrary.com] [Colour figure can be viewed at wileyonlinelibrary.com]
provision to total assets) are presented in Tables 13 and

14. Tables 11 and 13 provide evidence of a long-run rela-
tionship between interest rate spreads and the indepen-
dent variables. The results of the ARDL models in
Tables 12 and 14 confirm the findings of our earlier
model specification (i.e., Table 7). We find that the signs
of the covariates remain the same. Moreover, the effect of
narrow money (now, M1 plus) and credit risk is still
broadly insignificant.
7 | C O N C L U S I O N S AN D
RECOMMENDATIONS
FIGURE 4 Long-run dynamic multiplier of TB on SP. TB The aim of this paper was to investigate the macroeco-
denotes inflation. The horizontal axis indicates months taken to nomic and bank-specific determinants of interest rate
achieve the long-term equilibrium association. The vertical axis margins in Lesotho with a focus on the short- and long-
indicates the magnitude of negative and positive shocks in TB run symmetric and asymmetric nature of this relation-
[Colour figure can be viewed at wileyonlinelibrary.com] ship. To achieve this, the study used monthly time series
data from January 2009 to December 2018 together with
the ARDL bounds testing approach proposed by Pesaran
loan loss provision to total assets. Furthermore, instead et al. (2001) to test the existence of long-run co-integra-
of the growth of M1 (as defined by growth in currency tion. In addition, the existence of non-linearity in the
outside banks and transferable deposits), we consider the relationship between interest margins and the predictor
growth of the so-called M1 plus (as defined by growth in variables was evaluated using the NARDL model
currency outside banks, transferable deposits and 1-year pioneered by Shin et al. (2014). The empirical findings
call deposits). Owing to the unavailability of data, the associated with both the ARDL and NARDL models rev-
narrow definition of the interest rate spread is ealed the existence of a steady long-run relationship
maintained. Tables 11 and 12 show results with respect between the variables considered in the study. In consis-
to our model variation with growth in M1 plus and CR2 tency with the literature, the Treasury bill rate and infla-
(defined as the ratio of non-performing loans to total tion were found to have a positive and highly statistically
assets). The results of our model variation with growth in significant impact on interest margins in Lesotho in the
M1 plus and CR3 (defined as the ratio of total loan loss short and long-run. In addition, the deposit rate had a
DAMANE 17
T A B L E 1 1 ARDL bounds testing to co-integration results—model with growth in M1 plus and CR2 (defined as the ratio of non-
performing loans to total assets)

K 5 5% 2.55 3.606
T A B L E 1 2 ARDL results of the Explanatory variable Coefficients T-statistics p Value

long-run relationship estimation—
model with growth in M1 plus and CR2 INF 0.180233*** 7.329058 (.0000)
(defined as the ratio of non-performing TB 1.042218*** 9.634709 (.0000)
loans to total assets) M1g 0.002774 0.766816 (.4453)
CR2 −0.085807 −1.219752 (.2259)
DR −0.778938*** −7.70594 (.0000)
C 3.124818*** 0.415044 (.0000)
Bai and Ng long-run normality test 2.149871 0.3413
Breusch–Godfrey-LM test 0.539678 0.5850
BPG heteroscedasticity test 22.79673 0.5894
Ramsey-RESET test 0.127899 0.7215
Note: Dependent variable = SP. *** denote the level of statistical significance at 1% respectively. The values
in parentheses are the probability values. The selected model is ARDL (1, 0, 9, 0, 2, 8).
T A B L E 1 3 ARDL bounds testing to co-integration results—with growth in M1 plus and CR3 (defined as the ratio of total loan loss
provision to total assets)

K 5 5% 2.55 3.606
negative and highly statistically significant effect on the effect. Second, deposits are not the only source of credit
interest rate spread in the short and long-run. Further financing for banks in Lesotho. The magnitude of the
evidence suggests that the pass-through of inflation and respective impact of inflation and deposit rates on the
the deposit rate to the interest rate spread was less than interest rate spread was found to be smaller in the short-
one in both in the short- and long-run, signalling an run than it is in the long-run. This suggests that banks
incomplete pass-through. This confirms two things. First, price-in inflation to a lesser extent in the short-run than
inflation affects banks' lending rates with a second-round they do in the long-run. They also rely less on deposit
18 DAMANE
T A B L E 1 4 ARDL results of the

long-run relationship estimation—with
INF 0.171418*** 6.670016 (.0000) growth in M1 plus and CR3 (defined as
TB 0.870229*** 7.398233 (.0000) the ratio of total loan loss provision to
M1g 0.001963 0.537462 (.5923) total assets)
CR3 −0.165697* −1.862493 (.0659)

DR −0.557167*** −3.415071 (.0000)
C 3.846553*** 7.244080 (.0000)
Bai and Ng long-run normality test 2.453905 0.2931
Breusch–Godfrey-LM test 0.745139 0.6890
BPG heteroscedasticity test 19.78216 0.6550
Ramsey-RESET test 1.171921 0.2820
Note: Dependent variable = SP. ***, * denote the level of statistical significance at 1% and 10%, respectively.
The values in parentheses are the probability values. The selected model is ARDL (1, 0, 9, 0, 0, 8).
Lesotho's currency (Loti) and the South African Rand. As a result,

liabilities to finance credit extension in the short-run
financial markets and the public are assured that every Loti in cir-
than in the long-run. The empirical findings based on the
culation is backed by an equivalent amount of foreign currency
NARDL model rejected the null hypothesis of symmetric (Damane, 2020).
relationship between interest margins and the Treasury 3
Other CMA member countries are Namibia and Swaziland.
bill rate. Positive and negative shocks to the Treasury bill 4
Other SACU member countries are Botswana, South Africa and
rate lead to increases in the interest rate spread with the
Swaziland, Namibia.
positive shocks having less of an impact than the negative 5
These are Standard Lesotho Bank, Nedbank and First
shocks. Inflation's impact on the interest margin con-
National Bank.
forms with theory. However, positive shocks to inflation 6
The Central Bank of Lesotho (the Bank) was first established as
have more of an impact on the interest margin than do
the Lesotho Monetary Authority in 1978, under the Lesotho Mon-
negative shocks. etary Authority Act of 1978 starting operations in 1980. In 1982,
In the main, the study's findings reflect that macro- through the Act of Parliament, the name Lesotho Monetary
economic instabilities and bank-specific factors such as Authority was changed to the Central Bank of Lesotho and the
high and volatile prices and deposit rates sustained at Bank was conferred additional functions and responsibilities. In
low levels are more likely to create a widening of interest August 2000, the Central Bank of Lesotho Act 2000 (the Act)
rate spreads in Lesotho. In light of these results, authori- came into force and bestowed a fair amount of autonomy on the
Bank, and defined the primary objective of the Bank as achieve-
ties and policymakers are advised to ensure price and
ment and maintenance of price stability.
general macroeconomic stability while also pursuing pol- 7
A 2011 FinScope Survey on financial inclusion among consumers
icies aimed at maximizing savings. Failure to do so is
in Lesotho revealed that access to both formal and informal finan-
likely to result in elevated financial sector risk and uncer- cial products stood at 80.9% of the population (Sekantsi, 2019).
tainty that can lead to wider interest rate margins. Similarly, in the 2014 FinScope consumer survey report for Zim-
babwe pointed out that financial inclusion in Zimbabwe increased
E N D N O T ES from 38% in 2011 to 69% in 2014 on account of mobile money.
1 8
Lesotho is a member of the Common Monetary Area (CMA) of Core banking entails the improvement of back-end system
Southern Africa along with South Africa, Namibia and Eswatini. processing of all banking transactions including, but not limited
Further discussion on the CMA in the context of the paper is to, deposit, loan and credit processing together with teller ser-
offered in Section 2. vices, sales, loans orientation, payments systems, automatic teller
2
The Central Bank of Lesotho's (CBL's) primary mandate, as machines (ATMs) and electronic services (Central
spelled out in section 5 of the Central Bank of Lesotho Act of Banking, 2017).
9
2000, is to achieve and maintain price stability in Lesotho. This Commercial banks' loans to GDP ratio and commercial banks'
objective is achieved through the CBL's fixed exchange rate mone- deposits to GDP ratio.
tary policy regime that ensures the one-to-one peg between 10
Interest rate spread.
DAMANE 19
11
Ratio of broad money to GDP. Crowley, J. (2007). Interest rate spreads in English-speaking African
12
The ratio of commercial bank branches per 100,000 adults. countries (No. 7-101). International Monetary Fund.
13 Damane, M. (2019a). The effects of regional cross listing on firm
Growth rates in the interest rate spread are calculated in 2017
value and financial performance: Drawing lessons for Lesotho.
with 2007 as a base year.
African Journal of Accounting, Auditing and Finance, 6(3),
14
The list of Caribbean countries includes Barbados, a country that 201–235.
operates under a fixed exchange rate regime where its currency Damane, M. (2019b). An analysis and comparison of banking sector
the Barbados dollar is pegged with the United States dollar. development in the CMA. Central Bank of Lesotho Occasional Ana-
15
The maximum lag lengths are normally selected by means of lytical Note. May 2019, Issue 3. https://www.centralbank.org.ls/
information criteria such as Akaike information criterion (AIC), images/Publications/Research/Analytical_Note_Issue_3.pdf
Schwarz information criterion (SIC) and Hannan-Quinn infor- Damane, M. (2020). Investigating the structural stability of the the-
mation criterion (HQ). oretical money multiplier model in Lesotho: A flexible least
16
That is, Equation (1) assumed the existence of a symmetric rela- squares approach. International Journal of Statistics and Eco-
tionship between the variables. nomics, 21(2), 70–89.
17 Damane, M., & Sekantsi, P. S. (2020). Nexus between financial
The dynamic multipliers show the traverse of the interest rate
innovation and economic growth in Lesotho: Evidence from
spread variable from an initial equilibrium to a new equilibrium
ARDL and nonlinear ARDL approaches. International Journal
given positive and negative shocks to the inflation variable.
of Financial Innovation in Banking (in press). https://www.
18
This falls within the upper bound of the 3–6% inflation target band inderscience.com/info/ingeneral/forthcoming.php?jcode=ijfib
of Lesotho's fixed exchange rate anchor economy, South Africa. Damane, M., Sekantsi, P. S., & Molapo, S. S. (2018). Testing the sta-
19 bility of money demand function in Lesotho. International
The minimum corresponds to a drastic fall in currency in circula-
tion in 2011. Journal of Sustainable Economy, 10(4), 383–404.
Demirgüç-Kunt, A., & Huizinga, H. (1999). Determinants of com-
mercial bank interest margins and profitability: Some interna-
DATA AVAILABILITY STATEMENT tional evidence. The World Bank Economic Review, 13(2),
The data that support the findings of this study are 379–408.
openly available in at https://www.mediafire.com/file/ Dickey, D. A., & Fuller, W. A. (1981). Likelihood ratio statistics for
r5447h6a6ngf1cl/Bank_Spreads_-_Model_Data.xlsx/file autoregressive time series with a unit root. Econometrica: Jour-
nal of the Econometric Society, 1, 1057–1072.
Folawewo, A. O., & Tennant, D. (2008, July). Determinants of inter-
ORCID
est rate spreads in Sub-Saharan African countries: A dynamic
Moeti Damane https://orcid.org/0000-0002-4139-4638 panel analysis. In A paper prepared for the 13th Annual African
Econometrics Society Conference (pp. 9–11).
R EF E RE N C E S Ho, T. S., & Saunders, A. (1981). The determinants of bank interest
Ahokpossi, M. C. (2013). Determinants of bank interest margins in margins: Theory and empirical evidence. Journal of Financial
Sub-Saharan Africa (No. 13-34). International Monetary Fund. and Quantitative Analysis, 16(4), 581–600.
Allen, L. (1988). The determinants of bank interest margins: A note. IMF (International Monetary Fund). (2018). IMF Executive Board
Journal of Financial and Quantitative Analysis, 23(2), 231–235. Concludes 2017 Article IV Consultation with the Kingdom
Anchang, M. T. (2016). The institutional determinants of financial of Lesotho. Available Online: https://www.imf.org/en/News/
development: Evidence from the Southern African Development Articles/2018/02/23/pr1863-imf-executive-board-concludes-2017-
Community (SADC) (Doctoral dissertation). article-iv-consultation-with-the-kingdom-of-lesotho (Accessed on
Ayaya, O. O. (1997). Financial reform and credit extension in Leso- 3rd December 2020).
tho. Development Southern Africa, 14(4), 547–560. Jamaludin, M. F., Klyuev, M. V., & Serechetapongse, A. (2015).
Bai, J., & Ng, S. (2005). Tests for skewness, kurtosis, and normality What drives interest rate spreads in Pacific Island countries? An
for time series data. Journal of Business & Economic Statistics, empirical investigation (No. 15-96). International Monetary
23(1), 49–60. Fund.
Bhatt, V. V. (1988). On financial innovations and credit market evo- Johansen, S. (1988). Statistical analysis of cointegration vectors.
lution. World Development, 16(2), 281-292. Journal of Economic Dynamics and Control, 12(2–3), 231–254.
Brock, P. L., & Suarez, L. R. (2000). Understanding the behavior of Johansen, S., & Juselius, K. (1990). Maximum likelihood estimation
bank spreads in Latin America. Journal of Development Eco- and inference on cointegration—With applications to the
nomics, 63(1), 113–134. demand for money. Oxford Bulletin of Economics and Statistics,
Central Bank of Lesotho (2019). Supervision Annual Report 2018. 52(2), 169–210.
Central Bank of Lesotho. Karantininis, K., Katrakylidis, K., & Persson, M. (2011). Price trans-
Central Banking (2017). Lesotho launches new core banking system mission in the Swedish pork chain: Asymmetric nonlinear
without a hitch. https://www.centralbanking.com/central- ARDL (No. 726-2016-49995, pp. 1–11).
banks/financial-stability/fmi/3229711/lesotho-launches-new- Khan, M., & Jalil, A. (2020). Determinants of interest margin in
core-banking-system-without-a-hitch Pakistan: A panel data analysis. Economies, 8(2), 25.
Chirwa, E. W., & Mlachila, M. (2004). Financial reforms and inter- Khoabane, S. (2018). A Comparative Analysis of Banking Services
est rate spreads in the commercial banking system in Malawi. Fees in the Southern African Customs Union (SACU) Countries.
IMF Staff Papers, 51(1), 96–12. https://www.centralbank.org.ls/index.php/component/k2/
20 DAMANE
item/10-a-comparative-analysis-of-banking-services-fees-in-the- Phillips, P. C., & Perron, P. (1988). Testing for a unit root in time
southern-african-customs-union-sacu-countries series regression. Biometrika, 75(2), 335–346.
Kwasi Obeng, C. (2018). Is the effect of exchange rate volatility on Qamruzzaman, M., & Jianguo, W. (2018). Nexus between financial
export diversification symmetric or asymmetric? Evidence from innovation and economic growth in South Asia: Evidence from
Ghana. Cogent Economics & Finance, 6(1), 1460027. ARDL and nonlinear ARDL approaches. Financial Innovation,
Lacheheb, M., & Sirag, A. (2016). Oil price and inflation in Algeria: 4(1), 20.
A nonlinear ARDL approach. Topics in Middle Eastern and Saunders, A., & Schumacher, L. (2000). The determinants of bank
African Economies, 18(2), 45–60. interest rate margins: An international study. Journal of Inter-
López-Espinosa, G., Moreno, A., & de Gracia, F. P. (2011). Banks' national Money and Finance, 19(6), 813–832.
net interest margin in the 2000s: A macro-accounting interna- Sekantsi, L. P. (2019). The nature of financial inclusion in Lesotho:
tional perspective. Journal of International Money and Finance, Challenges and main policy considerations to increase access to
30(6), 1214–1233. and use of quality financial services. Journal of Digital Banking,
Männasoo, K. (2013). Determinants of bank interest spreads in 3(4), 330–345.
Estonia. Eastern European Economics, 51(1), 36–60. Sheriff, I., & Amoako, G. (2014). Macroeconomic determinants of
Maudos, J., & De Guevara, J. F. (2004). Factors explaining the inter- interest rate spread in Ghana: Evidence from ARDL modelling
est margin in the banking sectors of the European Union. Jour- approach. Journal of Finance and Bank Management, 2(2),
nal of Banking & Finance, 28(9), 2259–2281. 115–132.
McShane, R. W., & Sharpe, I. G. (1985). A time series/cross Shin, Y., Yu, B., & Greenwood-Nimmo, M. (2014). Modelling asym-
section analysis of the determinants of Australian trading bank metric cointegration and dynamic multipliers in a nonlinear
loan/deposit interest margins: 1962–1981. Journal of Banking & ARDL framework. In Festschrift in honor of Peter Schmidt
Finance, 9(1), 115–136. (pp. 281–314). New York, NY: Springer.
Michis, A. A. (2016). Market concentration and nonlinear pricing in Tahir, S. H., Shah, S., Arif, F., Ahmad, G., Aziz, Q., & Ullah, M. R.
European banking. Journal of Economics and Business, 85, 1–12. (2018). Does financial innovation improve performance? An
Molapo, S., & Damane, M. (2016). An econometric approach to pri- analysis of process innovation used in Pakistan. Journal of
vate sector credit in Lesotho. Journal of Business in Developing Innovation Economics Management, (3), 195–214.
Nations, 15(1), 31–45. Tarus, D. K., Chekol, Y. B., & Mutwol, M. (2012). Determinants of
Motelle, S. I., & Biekpe, N. (2014). Financial intermediation spread net interest margins of commercial banks in Kenya: A panel
and stability of the banking system in the Southern Africa Cus- study. Procedia Economics and Finance, 2, 199–208.
toms Union. Managerial Finance, 40, 276–299. Tarus, D. K., & Manyala, P. O. (2018). What determines bank inter-
Mowatt, R. (2001). Prospects for financial sector reform in the context est rate spread? Evidence from Sub-Saharan Africa. African
of regional integration in SADC? Johannesburg: University of Journal of Economic and Management Studies, 9, 335–348.
the Witwatersrand. Zhang, Z., Tsai, S. L., & Chang, T. (2017). New evidence of interest
Narayan, P. K. (2005). The saving and investment nexus for China: rate pass-through in Taiwan: A nonlinear autoregressive dis-
Evidence from cointegration tests. Applied Economics, 37(17), tributed lag model. Global Economic Review, 46(2), 129–142.
1979–1990.
Nkoro, E., & Uko, A. K. (2016). Autoregressive Distributed Lag
(ARDL) cointegration technique: Application and interpretation.
Journal of Statistical and Econometric Methods, 5(4), 63–91.
Obeng, S. K., & Sakyi, D. (2017). Macroeconomic determinants of
interest rate spreads in Ghana. African Journal of Economic How to cite this article: Damane M.
and Management Studies, 8(1), 76–88. Investigating the determinants of commercial bank
Pesaran, M. H., & Shin, Y. (1999). An autoregressive distributed-led
interest rate spreads in Lesotho: Evidence from
modelling approach to cointegration analysis. Cambridge: Cam-
autoregressive distributed lag (ARDL) and non-
bridge University Press.
Pesaran, M. H., Shin, Y., & Smith, R. J. (2001). Bounds testing linear ARDL approaches. Int J Fin Econ. 2020;1–23.
approaches to the analysis of level relationships. Journal of https://doi.org/10.1002/ijfe.2370
Applied Econometrics, 16(3), 289–326.
DAMANE 21
A P P EN D I X A TABLE A5 Breusch–Godfrey serial correlation LM test for

NARDL
F-statistic 1.029031 Prob. F(2,59) 0.3637

TABLE A1 Breusch–Godfrey serial correlation LM test
Obs*R-squared 3.707731 Prob. Chi-square(2) 0.1566
Obs*R-squared 0.370862 Prob. Chi-square(2) 0.8307
TABLE A6 Heteroscedasticity test for NARDL: Breusch–
Pagan–Godfrey
TABLE A2 Heteroscedasticity test: Breusch–Pagan–Godfrey
F-statistic 0.932833 Prob. F(23,87) 0.5567 Obs*R-squared 48.77145 Prob. Chi-square(48) 0.4418
Obs*R-squared 21.95859 Prob. Chi-square(23) 0.5228 Scaled explained SS 31.41463 Prob. Chi-square(48) 0.9692
Scaled explained SS 41.68541 Prob. Chi-square(23) 0.0099
TABLE A7 Bai and Ng (2005) long-run normality test for

TABLE A3 Ramsey-RESET test NARDL
Omitted variables: Squares of fitted values Statistic Prob.
Value df Probability Skewness −0.49317 .689054
t-statistic 0.369379 86 .7128 Skewness 3/5 2.25114 .012188
F-statistic 0.136441 (1, 86) .7128 Kurtosis 1.108916 .133733
F-test summary Normality 2.375209 .304951
Sum of Sq. df Mean Squares

Test SSR 0.001152 1 0.001152
Restricted SSR 0.727505 87 0.008362
Unrestricted SSR 0.726353 86 0.008446
TABLE A4 Bai and Ng (2005) long-run normality test
Statistic Prob.
Skewness −0.28478 .612094
Skewness 3/5 3.139727 .000846
Kurtosis 1.431219 .076184
Normality 2.591408 .273705
22 DAMANE
FIGURE A1 Graphical depiction of study variables [Colour figure can be viewed at wileyonlinelibrary.com]
F I G U R E A 2 CUSUM graph [Colour figure can be viewed at F I G U R E A 3 CUSUMSQ graph [Colour figure can be viewed
wileyonlinelibrary.com] at wileyonlinelibrary.com]
DAMANE 23
F I G U R E A 4 NARDL CUSUM graph [Colour figure can be

viewed at wileyonlinelibrary.com]
F I G U R E A 5 NARDL CUSUMSQ graph [Colour figure can be

viewed at wileyonlinelibrary.com]

Ijfe 2370

Uploaded by

Copyright:

Available Formats

Ijfe 2370

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ijfe 2370

Uploaded by

Copyright:

Available Formats

Received: 7 July 2020 Revised: 17 November 2020 Accepted: 20 November 2020

Investigating the determinants of commercial bank interest

Research Department, Central Bank of

1 | INTRODUCTION interest rate spread or interest margin is the difference

TABLE 1 Variable names and sources

Nature of variable Name of variable Abbreviation Variable state Source of data

Abbreviations: BOS, Bureau of Statistics; CBL, Central Bank of Lesotho.

The study makes use of monthly data from January 2008

TABLE 2 A priori expectations

Relationship between Expected sign Rationale Source

in the study because of the potential endogeneity of Xx

where all variables are as previously defined, Δ is the first

through is represented by decomposing the macroeco- Xw X

ΔSPt = α0 + π 1 SPt − 1 + π 2+ INFt+− 1 + π 3− INFt−− 1

TABLE 3 Descriptive statistics

TABLE 4 Correlation results

TABLE 5 Standard unit root tests

Level First Difference Level First Difference

Variable Trend and Trend and Trend and Trend and

TABLE 6 ARDL bounds testing to co-integration results

F-bound test Null hypothesis: no levels relationship

Test statistic Value Significance I(0) I(1) Evidence of co-integration

T A B L E 7 ARDL results of the

T A B L E 8 Results of the short-run

TABLE 9 NARDL bounds testing to co-integration results

F-bound test Null hypothesis: No levels relationship

Test statistic Value Significance I(0) I(1) Evidence of co-integration

The possibility of the existence of an asymmetric impact

TABLE 10 Long-run NARDL

provision to total assets) are presented in Tables 13 and

F-bound test Null hypothesis: No levels relationship

Test statistic Value Significance I(0) I(1) Evidence of co-integration

T A B L E 1 2 ARDL results of the Explanatory variable Coefficients T-statistics p Value

F-bound test Null hypothesis: No levels relationship

Test statistic Value Significance I(0) I(1) Evidence of co-integration

T A B L E 1 4 ARDL results of the

CR3 −0.165697* −1.862493 (.0659)

Lesotho's currency (Loti) and the South African Rand. As a result,

A P P EN D I X A TABLE A5 Breusch–Godfrey serial correlation LM test for

F-statistic 1.029031 Prob. F(2,59) 0.3637

TABLE A7 Bai and Ng (2005) long-run normality test for

Omitted variables: Squares of fitted values Statistic Prob.

Value df Probability Skewness −0.49317 .689054

t-statistic 0.369379 86 .7128 Skewness 3/5 2.25114 .012188

F-statistic 0.136441 (1, 86) .7128 Kurtosis 1.108916 .133733

F-test summary Normality 2.375209 .304951

Sum of Sq. df Mean Squares

TABLE A4 Bai and Ng (2005) long-run normality test

F I G U R E A 4 NARDL CUSUM graph [Colour figure can be

F I G U R E A 5 NARDL CUSUMSQ graph [Colour figure can be

You might also like