Santander Paid a Dividend by Accident
Here is a dumb story. Banco Santander is a bank in Spain. It's also got quite a lot of bank in the U.S. So much so that its U.S. bank holding company is subject to the Federal Reserve's annual stress test. It took the stress test. It failed, because the Fed found that it was inept at capital planning:
The Federal Reserve objected to the capital plan from Santander due to widespread and significant deficiencies across the BHC’s capital planning processes. Specific deficiencies were identified in several areas, including governance, internal controls, risk-identification and risk-management, management information system (MIS), and assumptions and analysis that support the BHC’s capital planning processes.
Santander apparently got the results and said "Deficiencies across the capital planning processes? We'll show them deficiencies across the capital planning processes!" Because the punishment for failing a stress test is that you're not allowed to pay dividends, and about a month after failing the stress test, Santander went out and declared a dividend. That's some bad capital planning!
It's sort of excusable. Santander's U.S. bank holding company is called Santander Holdings USA. That's what took the stress test, and failed the stress test and wasn't allowed to pay dividends. So Santander sort of ticked off, OK, we can't pay dividends from Santander Holdings to our Spanish parent company, got it, check. The idea is, you can move money around within the Santander U.S. entity -- the subsidiaries can pay dividends up to the holding company -- but you can't move it out of the U.S. system. Santander Holdings, which is regulated by the Fed, can't move money up to Banco Santander SA, which isn't. Fair enough. And Santander Holdings didn't pay a dividend. Fine.
What happened instead is that Santander Consumer USA Holdings Inc. paid a dividend. Santander Consumer is a non-bank financial company that makes car loans. It is 60.5 percent owned by Santander Holdings, but is a public company, and the rest of its shareholders are just regular public shareholders. And as a public company with a public company board of directors and a fiduciary responsibility to its shareholders, it declared a quarterly dividend of $0.15 per share, or $52.3 million total, out of quarterly earnings of $81.5 million. Of that money, $31.6 million went to Santander Holdings, and so stayed within the Fed-regulated system, but $20.7 million left the system to Santander Consumer's public shareholders.
Santander knew that it wasn't supposed to let money leave the system by a dividend from Santander Holdings to the Spanish parent. And it knew that it was allowed to let money move within the system, for instance by a dividend from a wholly owned subsidiary to Santander Holdings. But its mistake was letting a partially owned subsidiary declare a dividend, and letting $20.7 million leak out of the system.
The Fed looks pretty good in all of this. For one thing, good call on the planning-process deficiencies! But also, the Fed noticed the problem on May 22, after the dividend was declared but before it was paid on May 30. It generously let Santander Consumer pay the dividend anyway -- otherwise it'd have been a bit of a mess for shareholders -- so long as Banco Santander contributed an offsetting $20.9 million of capital to Santander Holdings, so that no money would leave the U.S. system. But that was the last time; Santander Consumer did not declare a dividend for the second quarter, and explained that it won't be paying dividends until the parent fixes its stress-test problems.
Which probably weren't helped by this. Now the Fed has further punished Santander by ... entering a Written Agreement saying that it won't do this any more. Will it do another dumb thing? Maybe! But it will be a different dumb thing at least.
Dan Davies wrote a good post yesterday about the intersection of bank regulation and technology:
[I]n the old days, if the regulator asked for something, the banks were very much allowed to give them what data they could scrounge up without too much inconvenience, and then tell them to go and whistle up a rope if they wanted any more. As a result, a culture grew up under which banks tended to assume that if the IT systems could process a transaction and deliver the quarterly results, then they were integrated. The failings reported in the Fed letter to Deutsche Bank -- Excel spreadsheets updated by hand, databases which systematically got the sign wrong on long and short positions, all of that sort of thing -- are the natural consequence of an environment in which innovation in product design was praised, but in which nobody wanted to spend money on a sufficiently flexible architecture to integrate all the new products into the existing risk management systems.
Nowadays, of course, regulatory requests for data are not like that. They come down with the force of a royal command performance, with the threat of (at the very least) not approving the year’s dividend payout. And the regulators want the data in the format they specify, without errors and on time.
Yep! It's easy to describe what you'd want to run a bank: You have a system, and it pulls in all your data about all of your positions, and it describes and aggregates all of those positions in all of the ways you might need. For every position and every business and every decision you're about to make, it can tell you the economic and tax and accounting and regulator-capital and stress-test and securities-regulation consequences, and then you can think about those all in an integrated way.
So you decide you want to declare a dividend, and to model the consequences you put "$0.15 per share" in your dividend calculator, and it tells you that your capital ratios will go down by X and your liquidity will go down by Y and your GAAP equity will go down by Z but also -- and this is crucial -- it pops up a big red box saying "NO YOU IDIOT YOU CAN'T PAY A DIVIDEND REMEMBER THAT FED THING??" And there's a siren or whatever in case you're not looking.
It is sort of astounding how far the banks are from that; Davies says "it seems very unlikely to me that the cost of catching up with the years of underinvestment will be less than $1 billion per bank." I don't say it's astounding because it's easy to build that system; it's easy to describe generically, but the amount of information and judgment you'd need to integrate seems endless. Banks are very complicated, both in the sense that they have lots of positions and in the sense that everything is subject to multiple overlapping and qualitative and occasionally contradictory regulatory regimes. But it's astounding because it's just so much harder for humans to do this than it is for computers. No human both sees the whole story and knows the details. The Santander Holdings person who did the stress tests, and the Santander Consumer board members who approved the dividend, probably never met. If your computers can't give you an integrated view of what you're doing, you're left relying on your (impersonal, distributed) gut, and your gut's algorithm is probably worse than your computer's.
"Dear Santander, don't pay any dividends, love, the Fed" is actually a pretty simple set of instructions as bank regulation goes. And Santander still messed it up. And not even in a super-dumb or shady way! You can understand how it happened. Imagine how often less obvious mistakes happen.
Here's the earnings release with that information.
Who had been told, in Santander Consumer's public offering last year, to expect quarterly dividends. The IPO prospectus doesn't mention the stress-test risk but does have some foreshadowing. E.g.:
Because of our relationship with Santander, we are subject to the regulatory oversight of the Federal Reserve System (the “Federal Reserve”). This oversight has led us to develop and maintain extensive risk management and reporting procedures and has helped us to continually adapt to the evolving regulatory requirements for consumer finance in the United States.
Not extensive enough!
I don't know why it's $20.7 million of dividend but $20.9 million of contribution. The Fed's order today says $21 million.
See slide 6 here. The May dividend was on first-quarter earnings; the second-quarter dividend would have been paid in August I guess.
Incidentally my vague assumption is that banks are relatively bad at this, both because they are so big and complicated and because they prioritize information systems less than certain other nimbler financial competitors do. Like, my wild guess would be that the big quanty hedge funds and high-frequency trading firms are usually better-integrated and have clearer views of themselves in their totality than the big banks do. On the other hand! That Latour case yesterday is at least as dumb as this Santander case, in a different way. Latour really did integrate its economic and regulatory systems so they could seamlessly use the same data. It just did it by ignoring the actual regulatory requirements.
This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.
To contact the author on this story:
Matthew S Levine at email@example.com
To contact the editor on this story:
Toby Harshaw at firstname.lastname@example.org