What US Bank Results Tell Us About the State of the Economy

Plus: Swedish Banks, Digitalisation of Financial Services, Blackrock

Issue #9 of Net Interest, my weekly newsletter on finance industry themes. Thanks to all of you who have forwarded and shared—keep doing what you’re doing! We’re now at nearly 2,300 readers, comprising investors, corporate executives, policymakers, students and more. I’d love that growth to continue; any feedback please do let me know.

Forwarded this? Subscribe here

What US Bank Results Tell Us About the State of the Economy

Since the onset of the virus, many datasets have become publicly available allowing people to track the economy in real time. Back in March OpenTable released its daily restaurant booking data by city; it provided a graphical illustration of how lockdowns were affecting the hospitality industry. Likewise, traffic congestion data, movie box-office revenue, daily property sales, passenger load factors on domestic flights, AirBnB searches, Google mobility reports have all emerged as gauges of consumer economic activity. This week Lael Brainard, a member of the Federal Reserve board of governors, referenced them when she said, “some high-frequency indicators…suggest that the strong pace of improvement in May and the first half of June may not be sustained.” (She actually said, “some high-frequency indicators tracked by Federal Reserve Board staff” but these days everyone has the same data.)

Her comment came at the same time as banks were reporting results for the quarter ended June. Although not as high frequency, their results nevertheless shine a light on the state of the economy. This piece looks at what that light reveals and what it means for banks. 

On the consumer side, things are looking OK. At JPMorgan, 30-day delinquency rates on home loans, credit cards and auto loans were all better than any quarter in the last five. People are also taking out more loans—retail mortgage applications were higher in June this year than in the same month last year, and June was a record month for auto originations. Spending, too, improved, with card volumes seeing a steady improvement over the quarter, flattening off towards the end of June at levels not dissimilar to last year. Bank of America added that in the first couple of weeks of July spending is higher than where it was last year.

Of course, a lot of this is driven by stimulus. Jamie Dimon of JPMorgan said:

“In the normal recession, unemployment goes up, delinquencies go up, charge-offs go up, home prices go down, none of that's true here… Savings are up, incomes are up, home prices are up. So you will see the effect of this recession, you're just not going to see it right away because of all the stimulus, and in fact, 60% or 70% of the unemployed are making more money than they were making when they were working.”

Bill Demchak of PNC agreed:

“And so Jamie went on this rant, and he's exactly right. We see consumers flush with cash. We see no delinquencies. We see consumer spending increasing. And it's all at the moment based on Government writing a check. And I just don't know how this plays out.”

These remarks highlight the positive impact government assistance is having on household cashflows. Assistance is also available from the banks themselves through loan forbearance programmes. However, compared with other countries, the take-up of these programmes is quite low. The percentage of mortgage loans currently in forbearance in the US is just over 8%. That compares with 17% in the UK, 15% in Canada and 14% in Australia.

The rate at which new borrowers are asking for forbearance is also coming down. In JPMorgan’s case, payment deferral requests were down 95% from their peak. Additionally, many borrowers have continued to pay even through their deferral period. Across its card and home lending portfolios, over half of JPMorgan customers in a deferral programme have anyway made at least one payment. The very first cohorts of borrowers that went into deferral are now coming out and their credit performance is beating expectations. JPMorgan highlighted that 80% of card customers coming off deferral are falling straight back into line; only 20% are requesting further assistance. Bank of America provided some additional colour, disclosing that without its deferral programme, write-offs in its credit card portfolio would have been only ~6% higher, not a huge number. 

To the extent that loan deferrals are a signal of distress, their lower take-up in the US may give American banks a wider firewall against losses relative to banks in other countries. However, that firewall is funded by government stimulus and so it is sensitive to stimulus being maintained. The politics of that are beyond the scope of this note, but it’s something Nathan Tankus writes extensively about.

Whatever happens, banks are getting prepared. Mike Corbat of Citigroup said, “The pandemic has a grip on the economy, and it doesn't seem likely to loosen until vaccines are widely available.” 

Helping banks get ahead of that curve is an obscure new accounting standard that deserves a shout-out. A lot has been written about new capital rules introduced in the aftermath of the financial crisis. Thanks to these rules banks today have a lot more capital than they otherwise might have. But while the Federal Reserve and others were busy drafting those, accounting bodies got busy too, and in 2016 they unveiled Accounting Standards Update 2016-13, Topic 326, Financial Instruments—Credit Losses. Also known as CECL (pronounced see-sull) it concerns itself with how banks account for bad loans. And it provides another important firewall around the banks.  

A brief background on CECL

For years banks and their regulators grappled over how best to account for loans that are going bad. Typically banks take provisions against bad loans when they start to whiff rather than waiting until they are covered in mould. This is consistent with the basic accounting principle of prudence. How strong the whiff has to be, though, and how much to provision, had remained thorny questions.

In 1998 SunTrust bank of Atlanta (now part of Truist) was forced to restate its earnings after the US Securities and Exchange Commission deemed it had taken too much in the way of provisions. When the financial crisis came along in 2007 banks were criticised for taking too little. Many banks were considered well capitalised moments before they failed because they had not properly accounted for losses building on their balance sheets. The stress test of 2009 (a subject of last week’s Net Interest) was an attempt to reverse this to see what banks would look like if forced to provision their loans for an adverse economic environment. 

The new CECL accounting standard addresses the “too little, too late” problem that handicapped banks during the financial crisis. Previous standards only allowed provisions to be taken if losses were “probable”. That meant there had to be some evidence of loan deterioration. CECL lowers the bar, allowing provisions to be based on estimates of expected losses under future economic conditions. So even if the loan looks OK right now, if there’s any expectation that it may deteriorate, then a provision can be booked regardless. Put simply, whatever’s happening on the ground, if the bank has a “reasonable and supportable” forecast that things are going to get worse, it can take action. [1]

CECL was introduced at the beginning of the year. At the time there was some controversy that it would lead to ‘procyclicality’—that banks would be forced to take sharply higher provisions at the first sign of a recession, before its impact could be properly felt. Back then no-one had any idea that one of the worst recessions in history was just around the corner. 

But from a CECL perspective, this recession has one favourable characteristic. It was pretty well telegraphed. Forecasts are usually an extrapolation of current trends—expectations typically move slowly via a steady drip of downgrades one way and a steady drip of upgrades on the way back. Because it was so widely telegraphed this recession was discounted very quickly. That’s why the market fell at its fastest rate in history back in March; that’s what allowed central banks to respond so forcefully. 

What’s in the numbers

Back in March banks didn’t have much time to mobilise. Lockdowns were enforced days before books were closed on the quarter and economic forecasts had to be finalised. At the beginning of April consensus was that unemployment would peak out at 9.6%; by the end of April consensus had shifted up to 17.5%. No surprise that when banks reported their first quarter earnings in the middle of that month, they were already warning that provisions in the second quarter would be higher.

Since then, expectations have converged at a more stable level. Banks use these expectations to inform their provisioning. They typically take a weighted view across a range of four or five scenarios. Those banks that have disclosed their base case scenario point towards unemployment of around 10% at the end of 2020, declining to around 7.5% at the end of 2021. JPMorgan is a bit more bearish; Wells Fargo is a bit more bullish. Their forecasts are laid out in the table below. As well as tilting its base case more recessionary this quarter, JPMorgan also increased the weighting on its downside scenario, a whopper that incorporates 22% unemployment at year end. According to Jamie Dimon, “if the base case happens, we may be over-reserved. I hope the base case happens.”

Some banks without the wherewithal to do all that economic forecasting outsource the job to Moody’s. Moody’s is one of those coveted companies you don’t get fired for using, and it has brought that to bear via its ImpairmentStudio platform, part of Moody’s Analytics. (The same principles apply to CECL outsourcing as they do to credit ratings—each incremental user reduces the risk for other users, creating a feedback loop which entrenches Moody’s market position.)

These economic forecasts led to some pretty punchy loan loss provisions being taken. In aggregate the five biggest banks increased their provisions by US$28 billion in the quarter. Those with capital markets businesses had the fortune of generating bumper trading revenues with which to absorb these provisions. Sometimes challenged, the strategic rationale of bolting investment banking on to retail banking has paid off.

The key point here is that as long as the economy does not deteriorate more than currently expected (and loan performance tracks economic performance as it historically has), banks shouldn’t have to take more provisions. In the financial crisis, investors knew earnings were going to get worse through time as provisions were slowly ramped up. (How much of that was in the price was a different matter.) In this crisis, it is much less obvious that earnings have to get worse. Investors need to believe not that the economy will worsen but that expectations of it will. 

To put it another way, the top five banks currently have sufficient provisions to absorb losses across 2-4% of their loan books. Within credit cards specifically JPMorgan is now reserved at 12.6%; assuming all its 30+ day delinquencies go bad (1.7%) and all its forbearance balances follow (~4%) that’s ~6% of the book gone, which its reserve covers twice over.

The table below captures the five biggest banks’ economic assumptions and current loss reserve standings. The table also shows Goldman Sachs’ economic forecasts and those the Fed embedded into its stress tests and U,V and W coronavirus scenarios. Another way to benchmark banks’ reserves is to look at how they stack up against the losses the Federal Reserve throws at them in its stress test. They are roughly 50% there. That’s a pretty bearish scenario, though. As Jamie Dimon said, “You all are doing estimates that show DFAST [Dodd-Frank Act Stress Tests] and Fed adverse. We will not lose that kind of money on credit, okay?”

It’s not entirely clear the implications of this new accounting standard have filtered down into commentators’ thinking. CECL changes the complexion of a bank’s income statement. Part of it reflects what genuinely happened in the quarter—fees that came in, interest income that accrued etc. Part of it reflects the current moment in time—unrealised trading profits marked to market as at the reporting date. And part of it now reflects the future—losses that haven’t been incurred yet but might if the bank’s latest economic forecasts come to pass. Past, present and future; bank earnings are a kind of reflection of A Christmas Carol

Valuations aside, this means that US bank earnings already reflect a future that other companies’ earnings don’t. There’s a well-documented disconnect out there across segments of the market—an historic skew between optimists on the one hand and pessimists on the other. Tesla’s market cap is now higher than any US bank bar one; it’s bigger than Wells Fargo and Citigroup combined, yet generates a fraction of the profits.

Of course there are various reasons to hate on banks. As detailed in the first issue of Net Interest, they are susceptible to being hijacked by Government as a tool of policy; they are subject to the deleterious impact of zero and negative interest rates; they don’t offer much growth. But unlike in past crises, a credit-induced earnings collapse is not one of those reasons.

[1] European accounting standard setters did something similar, although not quite. CECL loan impairment is based on lifetime expected credit loss. The European equivalent (IFRS 9) looks out only a year for loans performing as expected. IFRS also assigns a higher probability to longer-term economic forecasts to avoid a premature reaction to short-term stress. All of which means that like-for-like, provisioning rates should be higher in the US.

Share Net Interest

More Net Interest

Swedish Banks

Swedish banks reported earnings for the quarter this week, too. They’re an interesting group to look at because they operate in an economy that was not shut down as extensively as others. Based on Google data, the drop-off in trips to retail and recreation sites was only 20% in April, versus 75% in the UK. Retail consumption was correspondingly down only 1% between February and April, versus 18% in the UK. Of course, Sweden is an open, export-oriented economy and so it was unlikely to remain immune from the collapse in economic activity overseas; industrial production was down 16% in April, not that far off the UK (down 24%). SEB economists are forecasting a 5% decline in GDP for the year overall, better than the 9% decline they are forecasting for the Euro Area. 

Against this, banks’ loan books are performing quite well.

  • In the case of Svenska Handelsbanken, “You need to look 30 years back in time to find a lower loan loss number than the one we have posted in the second quarter.”

  • In the case of SEB, provisions were higher but it was all oil: “It is very difficult to see corporates with problems outside of oil.”

  • Nordea reported higher provisions, but it operates across multiple Nordic geographies. Within Sweden, its loan book contributing just 5% to group impaired loans and 10% to group provisions, despite making up a quarter of total group loans. 

Svenska Handelsbanken admits the picture looks weird: “This may come as a surprise to some, having such a low level in the middle of a crisis.” There’s some debate about whether the source of the surprise is better underwriting or under-provisioning, but the Swedish coronavirus response could have something to do with it. 

Digitalisation of Financial Services

A favourite quote of the Coronavirus era is one attributed to Lenin: “There are decades where nothing happens, and there are weeks where decades happen.” Satya Nadella of Microsoft has his own version: “We have seen two years’ worth of digital transformation in two months.”

It’s the same at banks. For years banks have tried to incentivise customers to stay out of branches and use digital channels instead. For various reasons—demographics, compliance—they weren’t 100% successful and had to run two full-cost channels in parallel. Now, customers are shifting their usage patterns. Wells Fargo said that on a broadly stable number of active digital users, log-ins were up 10% versus the first quarter and branch visits were down 19%.

Perhaps the most striking datapoint comes from Bank of America which tracks sales closed via digital channels. Having hovered at around 25-30% of total consumer banking sales historically, they rose to 47% in the second quarter. In the past BBVA has identified 50% as a kind of tipping point beyond which legacy costs can begin to be brought down. On that, Brian Moynihan of Bank America said, “Yes. I think the time to figure that out will be a little bit later.”


I once saw David Harding, founder of the Winton hedge fund complex, present an ecological model of asset management. He ran a simulation of 100 funds forward using historical market data, with one simple rule laid on top: that if a fund fails, its assets are redistributed across the remaining funds. Given how difficult it is to launch a new fund, the model has merit. Its punchline is that there will always and only ever be a single survivor, who eventually gobbles up all the assets. 

I was reminded of the model when I looked at Blackrock’s earnings this morning. The firm reported US$7.32 trillion in assets under management. That’s a lot. It’s bigger than the GDP of any country bar China and the US. And the firm is still gobbling up assets—in the second quarter alone it captured US$100 billion of inflows, picking up share from others across most asset classes. There’s still plenty of time for the simulation to run, though—the firm’s market share is still <10% among the top 500 firms globally.