What Financial Intermediation Looks Like on the Ground
This chapter stays with Mauritania’s banking system and asks how to read the gap between what banks earn and what the financial system delivers. The aim is practical: by separating access, depth, pricing, liquidity, profitability, and concentration, the reader can see why the country is a useful setting for studying whether a few banks’ decisions shape aggregate credit.
EADG asks whether earnings are outrunning service
The paper’s Efficiency–Access and Depth Gap, or EADG, is best understood as a diagnostic contrast. On one side is the banking system’s efficiency dimension, which reflects profitability-related performance. On the other are access and depth: who can use formal financial services and how much financial activity the system provides relative to the economy. A high EADG means that the first side is running ahead of the second. It does not mean that banks are inefficient in every operational sense. It means that the returns being recorded are not matched by the reach or volume of intermediation.
That makes EADG useful precisely because it refuses to let one number stand in for the whole banking system. Consider two systems with equally profitable banks. In the first, the profits accompany a broad borrower base, deep private credit, competitive pricing, and many channels through which savings reach firms and households. In the second, profits come from wide spreads in a concentrated market, while credit remains shallow and difficult to obtain. The income statement looks similar; the economic service is not. EADG directs attention to that difference.
The measure is not a verdict on why the gap exists. The paper discusses several possible contributors: market power, operating costs, macroeconomic and credit risk, and the limited scale of a young financial sector. A wide margin can compensate a bank for expensive operations or expected losses. It can also reflect a lack of competitive pressure. The EADG cannot by itself distinguish those explanations. It identifies a relationship that needs a mechanism behind it.
The distinction is also important for interpreting “efficiency.” In everyday language, efficiency often means producing more with fewer resources or serving customers at lower cost. In a financial-development index, a strong efficiency score can coexist with high lending rates and limited outreach because the score is built from financial-sector outcomes rather than from a direct audit of every borrower’s experience. A bank can be financially successful while the system remains hard to enter for a small firm or a rural customer. That is why the paper treats the EADG as a starting point for investigation, not as a final diagnosis.
The other terms give the diagnosis operational meaning. Access concerns the breadth of participation: whether people and firms can obtain accounts, payments, savings products, and credit. Depth concerns the scale of financial activity relative to the economy, including the volume of private credit. Profitability describes the returns captured by banks; in the study’s bank-level data, it is measured as net income relative to total bank capital. Liquidity describes the liquid resources available to absorb pressure or meet withdrawals; the study uses cash and cash equivalents relative to total assets. Lending growth describes the change in total loan volume from one month to the next. Each measure answers a different question. None can safely replace the others.
This separation prevents two opposite mistakes. The first is to assume that high profits prove that the banking system is serving the economy efficiently. The second is to assume that a gap between profits and depth proves that every bank is extracting an unjustified rent. Mauritania’s evidence supports the narrower claim that the outcomes diverge and that the divergence deserves an explanation. The rest of the study asks whether concentration and bank-specific lending movements supply part of that explanation.
A useful way to read the system is to follow the interfaces rather than search for a single winning metric. Depositors provide funding, but concentrated deposits can make a bank sensitive to a few decisions about withdrawal or renewal. Banks transform that funding into loans, but concentrated borrowers can make the loan book sensitive to a few repayment or investment events. Profitability records the return produced by those activities and risks, while liquidity records some of the capacity to absorb pressure. Aggregate credit is the system-level result of many such balance-sheet choices. A gap between profit and depth can appear at any interface: funding may not be broad, eligible borrowers may be scarce, risk limits may be tight, or earnings may be retained instead of lent.
The distinction also changes how comparison should work. Comparing one bank’s profit with another bank’s profit says little about access unless the banks face similar borrowers, funding structures, risk, and scale. Comparing system-wide profit with system-wide credit is more informative, but it can still conceal whether the result comes from a broad expansion or a few large institutions. The Mauritanian study therefore keeps the bank-month observation and the aggregate outcome in view at the same time. That is what allows concentration to become an explanatory feature rather than a background adjective.
The paper’s indicators are consequently pieces of a map. Lending growth is closer to the flow of new intermediation than a level of assets alone. Liquidity describes room to handle funding pressure, not willingness to lend. Borrower and depositor concentration describe exposure, not the direction of the next month’s credit. A lending–deposit spread describes pricing, but not automatically the quality or inclusiveness of the loans being priced. Treating these measures as interfaces makes the central question more disciplined: which movement crosses from one part of the system to another, and which movement stops inside the bank?
The system is concentrated in more than one place
Mauritania’s banking structure gives the gap a concrete shape. The paper compares system-wide lending with the lending of the five largest banks and the remaining institutions, and finds that aggregate credit is overwhelmingly driven by the five largest. This is not the same as saying that smaller banks do nothing. It means that their movements are less likely to determine the direction of the system when the dominant lenders move differently or at a much larger scale.
The data show concentration inside banks as well as concentration among banks. The study’s confidential supervisory panel covers 15 banks over 49 monthly observations from June 2020 through June 2024, producing 735 bank-month observations before variables that require lags reduce the usable sample. Two banks were excluded: one began operations in November 2023, and another had inconsistencies and a high incidence of outliers. The aggregate regressions reported later use 48 observations. These details matter because the study is a high-frequency case analysis, not a timeless description of every financial system.
The balance sheets are visibly uneven. Median total assets are about 6.55 billion Mauritanian ouguiya, compared with a mean of 8.76 billion and a maximum above 33.6 billion. Median loans are about 4.20 billion, while the mean is about 4.85 billion. Deposits show a similar dispersion: the median is about 4.38 billion and the mean about 5.85 billion. A mean that is pulled upward by a few large institutions can make the average bank look more influential than a typical bank is—or make the typical bank look more capable of offsetting a dominant one than it really is.
The funding and lending relationships are concentrated too. The five largest depositors account for about 15 percent of deposits on average, reaching as high as 50 percent in some bank-months. The five largest borrowers account for a median 21 percent of a bank’s loans, with the upper quartile at 31 percent. These figures do not prove that concentration is harmful. They do show why a bank may be sensitive to the decision of a small number of depositors or borrowers. A withdrawal, refinancing need, repayment problem, or large project can matter disproportionately when the balance sheet is not diversified.
Pricing provides another piece of context. Average loan rates in the sample cluster around 11 percent, while average deposit rates cluster around 6 percent, producing a lending–deposit spread near 600 basis points. Such a margin is economically meaningful, but it is not self-interpreting. It may compensate for credit and operating risk, reflect a constrained funding market, or indicate limited competition and pricing power. The paper’s data do not settle those alternatives. They make clear, however, why a bank can generate substantial earnings without automatically facing the competitive pressure that would turn those earnings into cheaper or more widely available credit.
This is the system the paper measures: dominant banks connected to concentrated borrowers and depositors, operating under common macroeconomic conditions, with profitability and lending growth as separate outcomes. The diagram is not a claim that every relationship is causal. It is a map of the interfaces at which the outcomes can diverge. Macro conditions can affect all banks, but a bank-specific decision can still matter if the bank is large. Depositor and borrower concentration can shape a bank’s room to expand, but their presence does not tell us whether a particular month’s profit was retained or lent.
The same caution applies to the study’s use of the five largest banks. Their importance is a structural fact in the paper’s comparison, not proof that they are the only source of credit or that size alone causes weak access. Large institutions can have scale advantages, wider distribution, and stronger buffers. They can also become points of systemic dependence. The relevant question is how their size interacts with the timing of lending decisions and with the ability of smaller institutions to compensate when a dominant bank changes course.
This is where the distinction between level and movement becomes important. A large bank may hold a large stock of loans and still reduce new lending in a particular month. A smaller bank may grow rapidly from a low base without changing system-wide credit very much. The paper’s later granular analysis is designed around that difference: it asks not only who is large, but whether unexpected changes at large institutions line up with the monthly movement of the aggregate. The structural description supplies the reason to test the relationship; it does not itself establish the result.
There is also a selection boundary. The panel begins with supervisory information on active banks, but two institutions are excluded for specific data or timing problems, and the remaining observations are confidential. This means that the reported concentration and relationships describe the analytical sample the authors use. They should not be casually converted into a claim about every financial provider, every informal lender, or every form of Islamic and conventional finance in Mauritania. The study explicitly says that the loan-type breakdown was not reliable enough for a separate decomposition.
That is why the data provide context rather than a complete causal explanation. They establish a concentrated, high-margin, unevenly distributed banking system in which profits, lending, funding, and borrower exposure can move on different tracks. They also define what the study cannot observe directly. The records do not reveal every rejected loan, every credit-committee decision, or the full use made of retained earnings. The breakdown between Islamic and conventional lending was not reliable enough for the authors to analyze. And because the data are confidential, readers cannot independently reconstruct the bank-level panel from the paper alone.
The result is a sharper question, not a finished answer. In Mauritania, if a few banks dominate credit and their balance sheets contain concentrated relationships, then their unexpected lending movements may carry more weight than an average-bank model assumes. The next step is to explain how the paper isolates those movements and tests whether they survive aggregation into system-wide credit.