Overview

In Mauritania’s 48-month aggregate sample, the macro-only model explains little short-run credit movement: R-squared is 0.110. Adding any of eight size-weighted lending shocks yields positive, 1%-significant coefficients of 0.634–0.923 and R-squared of 0.872–0.981, while the equal-weight placebo adds little. The result ties explanatory power to uneven bank size, but the contemporaneous relationship is partly built into related lending data and is not a universal causal share.

Learning outcomes

Macro controls alone explain little of monthly aggregate credit growth in this 48-observation sample (R-squared 0.110).
Size-weighted lending shocks raise R-squared to 0.872–0.981 across eight specifications, with coefficients of 0.634–0.923.
The equal-weight placebo adds little, linking the result to bank-size concentration—not proving a causal supply share.

Keywords

Mauritanialendingshockscreditconcentrationplacebo

When Lending Shocks Become Aggregate Shocks

This chapter tests the central empirical claim about Mauritania’s dominant banks: do size-weighted, bank-specific lending movements explain the monthly swings in aggregate credit that macroeconomic variables alone leave unexplained? The reader should come away able to read the coefficient, fit, and placebo evidence together—recognizing both the strength of the lending relationship and the reasons it is not a universal causal share of national credit.

The macro-only model leaves the main swings unexplained

The paper begins the aggregate exercise with a deliberately limited benchmark. Aggregate credit growth is regressed on inflation, the policy rate, M3 growth, and exchange-rate changes, without a granular bank-level term. This baseline has an R-squared of 0.110 and an adjusted R-squared of 0.028 across 48 monthly observations. The macro variables alone therefore organize only a small part of the month-to-month movement in system credit in this sample.

That result does not mean that monetary policy, inflation, liquidity, or the exchange rate are irrelevant to lending. It means that the common variables, as entered in this short aggregate regression, do not account for the pronounced swings in credit growth. A macro-only model sees the background conditions shared by the banking system. It does not yet see whether a few dominant institutions are making bank-specific decisions that move the aggregate at the same time.

This is the problem the granular term is meant to address. The term is not raw lending growth. It is a size-weighted sum of bank-level residuals produced after the first-stage controls and fixed effects are applied. Adding it asks whether the part of bank movement not represented by the macro benchmark is aligned with aggregate credit. That is an incremental question: how much does the system learn when the concentrated banking structure is represented explicitly?

A useful way to read the baseline is as a missing-component diagnosis. Suppose aggregate credit slows in a month when the policy rate and inflation barely change. A macro-only regression has limited information about whether demand softened across the economy, whether a few banks tightened standards, whether a large project was delayed, or whether a funding event affected one dominant lender. The low baseline fit does not distinguish those possibilities. It tells us that common conditions by themselves are insufficient to describe the observed high-frequency movement.

The baseline also protects the main result from a misleading comparison. A high R-squared after adding the lending shock would be less informative if there were no low-information benchmark. By keeping the macro-only specification visible, the paper asks what the bank-level term adds rather than presenting a large fit statistic without a counterfactual model.

Adding the lending shock changes the picture

Across all eight constructions of the lending residual, the size-weighted shock enters positively and is reported as statistically significant at the 1 percent level. The coefficients range from 0.634 to 0.923. The mapping is economically close to proportional: the paper interprets a 0.01 increase in the granular lending shock as associated with roughly a 0.65–0.92 percentage-point increase in same-month aggregate credit growth. The exact coefficient changes as the first-stage controls change, but the direction and broad magnitude remain stable.

Published by DeepMox

FAQ

How much did the macro-only model explain about Mauritania’s credit growth?

In Mauritania’s 48 monthly observations, the macro-only model had R-squared 0.110 and adjusted R-squared 0.028, so it explained only a small share of month-to-month credit movement.

What does a 0.01 increase in Mauritania’s granular lending shock mean?

It is associated with roughly a 0.65–0.92 percentage-point increase in same-month aggregate credit growth, using the reported 0.634–0.923 coefficients.

What did the equal-weight placebo find in Mauritania?

It found little aggregate information: coefficients were about −0.003 with standard errors near 0.007, and R-squared rose only to at most 0.115.

Previous11 minMeasuring Mauritania’s Bank Shocks

Next11 minMauritania’s Profitability Shocks and Credit

Mauritania’s Lending Shocks and Aggregate Credit

When Lending Shocks Become Aggregate Shocks

The macro-only model leaves the main swings unexplained

Adding the lending shock changes the picture

FAQ