Misunderstanding the net interest margin
Lately, there have been a lot of discussions in the media and in the academic sphere surrounding banks’ net interest margin in the low (or negative) interest rate environment. I have explained before how lowering interest rates below a certain threshold led to ‘margin compression’ (see here), which in turned depressed banks’ profitability and hence their internal capital generation, solidity and ability to lend.
The net interest margin (NIM thereafter) is roughly the difference between the average interest rate earned on assets and the average interest rate paid on funding, and is usually defined as
Net interest income / average earning assets, with NII being the difference between interest income (from loans and securities mostly) and interest expense (on deposits and other types of debt/funding instruments)
We now see conflicting articles and research pieces on the effects of low rates on banks’ NIM (see two of the most recent ones by the St Louis Fed here and other Fed researchers here). But, to my knowledge, most, if not all of those pieces make the same fundamental mistake: they do not look at risk-adjusted NIMs.
‘Risk adjustment’ is a critical concept but sadly often overlooked in the literature. I once defined the interest rate on a loan as the following:
LR = RFR + IP + CRP – C,
where LR is the loan rate, RFR is the applicable, same maturity, risk-free rate, IP the expected inflation premium, CRP the credit risk premium that applies to that particular customer and C the protection provided by the collateral (which can be zero).
As I explained elsewhere, margin compression occurs when the risk-free rate declines so much that interest rates banks pay on their funding reaches the zero lower bound while their interest income continues to decline (which led me to hypothesise that the zero-lower bound was actually a ‘2%-lower bound’ in the case of the banking/credit channel of monetary policy). This however assumes no fundamental change in the rest of the economy’s credit (or default) risk.
Indeed, in bad economic times, the CRP usually increases for most borrowers, partially offsetting the effects of the decline in the risk-free rate on new lending. Moreover, bankers can easily boost their NIM by lending relatively more to higher-risk customers or investing in higher-risk projects, even in good economic times. Consequently, it looks like the headline NIM isn’t suffering or declining that much. It can sometimes even improve, in particular when economic conditions are benign. For instance, emerging market banks often boast high NIMs, but also high default rates (and high ‘losses given default’). In such cases, margin compression seems not to be occurring. But this is just an accounting illusion.
See the example in the chart below, which represents the hypothetical evolution of the different components of a given unsecured loan rate throughout a long recession:
Once you adjust the NIM for the loan book’s underlying risk, the story is different. Banks’ interest income can rise but the risk of default on new lending, as well as that of their legacy loan portfolio, also rises. Because the CRP is often fixed at inception, legacy lending now underpays relative to its risk profile, potentially implying economic losses down the line.
Most studies don’t factor this phenomenon in. They look at unadjusted NIMs, which in many cases do not provide any useful information.
A very good and quite recent paper on banking mechanics by Claudio Borio and his team (The influence of monetary policy on bank profitability), which looks at the impact of the shape of the yield curve on margin compression and banks’ profitability, does understand that accounting plays a significant role:
The second form [of dynamic effects in the transmission of the level of interest rates to net interest income], which is more relevant, relates to accounting practices. Any interest margin on new loans also covers expected losses. But provisions in the period we examine follow the “incurred loss model”, so that, in contrast to interest rates, they are not forward-looking. As a result, extending new loans raises profitability temporarily, since losses normally materialise only a few years later at which point loans also become non-performing, eroding the interest margin. This also means that if lower market rates induce more lending, they will temporarily boost net interest margins. The strength of this effect will depend on background economic conditions. For instance, it is likely to be weak precisely when interest rates are unusually low and the demand for loans anaemic.
However, they stop short of providing a solution, or a correction, to this effect. To be fair, risk-adjusted NIMs are not directly observable and very difficult to estimate, given that disclosures about banks’ loan portfolio are very limited and that only some of their customers (i.e. large corporates) have bonds or credit default swaps traded on the secondary market. Therefore, some analysts use the following ex-post adjusted NIM ratio:
(Net interest income – loan impairment charges) / average earning assets
Default risk, expressed in the income statement by loan impairment charges (LICs – also called loan-loss provisions), is directly deducted from net interest income, making the NIM easier to compare across banks or countries. But even this version can be highly inaccurate, as LICs are backward-looking and depend on each bank’s accounting policies. In the short-run, some banks tend to over-provision, others to under-provision.
You’ve reached the end of this post perhaps wondering whether I had a solution to this problem. Unfortunately no, I don’t. But I believed that a clarification was in order. In finance, or economics in general, any decision involves risk-taking, and studies that do not take risk into account must be taken with a pinch of salt.
PS: The inflation premium is stripped out of the risk-free rate in this post, but in practice benchmark market rates such as Treasuries already factor in inflation expectations.
This post was re-published on Alt-M.