How much capital is enough capital?  Is there such a thing as too much capital?  What kind of capital is the right kind of capital?  One would think that almost a decade after the financial crisis, we would know at least the official answer to these critical questions.  But we don’t.  That’s why a new FedFin study explores not only regulatory capital for banks as it has been constructed since the crisis, but also all the changes to it still under way and what the impact of all this uncertainty means for monetary-policy transmission, financial-stability, and the future of banking.  We’ve raised more of our own questions than we’ve answered in this study, but I’ve one clear conclusion for your consideration:  regulators need to answer these questions at least to their own satisfaction before making still more changes to rules with increasingly obvious and often unintended costs.

As we worked through our new study, I came to see U.S. capital regulation as a sort of self-driving car.  Think of the bank as a machine that needs to get somewhere working through a complex landscape according to a set of sophisticated internal controls.  After being directed by the board, the bank’s CEO climbs into the driver’s seat to set the destination – say the grocery store (i.e., a ROE of ten percent and 55 mph).  Now, though, a power from on high – Google for the car, the feds for the bank – intervenes.  This omniscient authority decides the driver is too fat for her own good and – after months of disagreement and despite remaining differences – it redirects the destination from the grocery store to the gym.  Genuflecting as she goes, the CEO then bravely sets out to this new destination – a 7.5 ROE at 30 mph.

Can the car get to the gym – that is, can the bank achieve the objective set for it by the regulators and still have lunch – i.e., make investors happy?  Here’s where the self-driving part really kicks into gear – the driver can only achieve both regulatory and profitability goals if all of the assumptions made by the auto-pilot work under real-world conditions. 

This might be relatively easy if the bank were on a nice road with guardrails, clear lane markings, no intersections, and most importantly no semi-trailer trucks, wandering two-year olds, or texting drivers gunning obliviously for our sedate sedan.  And, the sun has to be out – our car still can’t quite work in the dark or rain.

Business conditions have to mimic self-driving ones with regard to simplicity, stability, and certainty for our car to work according to rules dictated from on high according to controls built into the dash.  Like our car, the self-driving bank has a set of internal controls hard-wired by the federal regulators to meet a series of often-incompatible prudential objectives.  For capital, there are as many as the dozen dials the ABA has calculated – Collins Amendment baselines, various stress-test thresholds, internal capital allocations, minimum risk-based and leverage thresholds, and resolution buffers just to name a few.  GSIBs have still more, making their dashboards even more incoherent, but all banks have a lot of capital speed limits to meet all at once all the time – rain, shine, competitors, the odd financial crisis, whenever and however.

All of these capital pedometers aren’t the only internal controls banks must meet all of the time.  There’s a tachometer demanding liquidity compliance that must also be met even though the better the bank meets liquidity requirements, the more dangerous seem its capital requirements, especially the leverage one, because of the capital requirements applicable to even no-risk assets booked to comply with the liquidity rules.  Each time the bank revs up its liquidity compliance, it stalls out on capital. 

My bank car also has additional internal controls – think of them as blind-spot checks and braking controls.  However, set again from on high, they watch out only for what the powers that be think might be on the road, not what actually shows up – the deer they forgot to program in based on expectations that our bank would drive only in the city.  

In short, our bank stalls out and runs into the unexpected unless it creeps along at ten mph in the right-hand lane on a sunny day.  This is safety, but not the soundness of business purpose and economic function banks must have in order to meet shareholder demands and provide the financial intermediation on which the economy relies. 

There’s one other problem – and it’s a big one – with this self-driving bank – the Fed is trying to put its foot on the pedal even though the self-driving car no longer has one.  That is, the FRB’s monetary policy is aimed at increasing bank lending to power up the recovery, but banks can’t lend because the car’s controls won’t let them.

Our new paper, along with an earlier one focusing in particular on monetary policy go into this in detail.  We demonstrate that the Board will need to change its policy tools and rely far more on its portfolio and non-banks if it wants to keep banks as self-driven vehicles – that is, as utilities.  Non-banks are of course still purring along in their Ferraris, making the roads still more challenging not just to banks, but also the Fed.

I probably wish more than most that there were really self-driving cars.  Now that would get us all where we want to go in ways that are good for us and everyone around us.  Wanting it won’t, though, make it any more so than wanting big banks to be safe can make them safer.  These are complex problems that require up-front recognition of complexity risk not just at big banks, but also in regulatory policy-making and the financial markets in which it must work.