Patrick Harker on Real-Time Payments and the Future of the Federal Reserve

Patrick Harker · The Wharton School, University of Pennsylvania

Fed back-office consolidation has little savings left; payments and most cost already consolidated.

Harker responds to Governor Chris Waller's call to consolidate Fed back-office functions. The payment rails are already consolidated under a single payments executive, and IT is moving toward consolidation. Supervision is a delegated function of the Board of Governors, so more consolidation is possible there. Local bank examiners are still needed, so not everything can be centralised. The large cost of the Fed is its staff, and those staff are already consolidated or on the way to it. He says HR is one area where consolidation would save little. He is therefore skeptical that much more savings remain.

“They've already either been consolidated or on their way to being consolidated. So I understand what Chris is saying, but at the same time, I'm not sure how much more juice there is to squeeze.”at 33:45

Patrick Harker · The Wharton School, University of Pennsylvania

Balance sheet is an operational tool, not core monetary policy, reserved for exigent circumstances.

Harker says the balance sheet is an operational issue, not a major tool of monetary policy, and should be used only in exigent circumstances. He describes the 2019 runoff: the Fed had estimates of appropriate reserves ranging from about $500 billion to $1 trillion and settled on a $500 billion target, but reserves ended up near $1.5 trillion. He says the level of ample reserves cannot be known a priori; the only signal is market reaction, which made the episode humbling. He says the Fed moved to an ample reserves regime with trade-offs compared with the old system.

“It is not the major tool of monetary policy. Okay. I think we gotta take that off the table because it should only be used in exigent circumstances.”at 41:44

Patrick Harker · The Wharton School, University of Pennsylvania

Fed crisis interventions were overdone; it should adopt time limits like the Bank of England.

Harker says that in retrospect the Fed probably overdid its response to the Silicon Valley Bank episode, because it was very worried about contagion, and contagion may not have been as bad as feared. He calls this a moral hazard problem that is hard for officials to handle under pressure from markets and Congress. He says the Bank Term Funding Program, offered with no haircut on collateral and no stigma, was probably not well thought out and was done in haste. He argues the Fed should borrow the Bank of England's practice of putting time limits on its crisis measures, and should focus on avoiding the cliff in the first place.

“Or at least the Fed should borrow a page from the Bank of England and put time limits on what we do.”at 42:59

Patrick Harker · The Wharton School, University of Pennsylvania

Fed should treat inflation as job one, not trade it off against employment; Harker has turned hawkish.

Harker says the dual mandate is in law, but trying to trade off inflation against unemployment led the Fed into trouble. He argues the Fed should adopt a hierarchy: inflation is job one, and only when inflation is near target should officials worry about unemployment. He says current labor market developments, such as AI and immigration, have little to do with what the Fed controls, so the Fed should focus on inflation. He says he has become surprisingly more hawkish, and he was also one of the early voices calling for rate increases coming out of the pandemic.

“the idea that we have a dual mandate, it's in law, it's what we do. But this is what leads us down a bad path. I think we should actually start to think more in a hierarchical fashion. ... Inflation is job one.”at 47:39

Patrick Harker · The Wharton School, University of Pennsylvania

Community banks can thrive, not just survive, by deeply understanding a niche customer base.

Many people argue that community banks cannot compete because technology rewards scale. Harker's counterexample is a small Pennsylvania bank, transcribed as the Bank of Burdenham, which he says is the first de novo bank since the financial crisis. It serves the Amish community heavily. The bank worked with regulators, including the FDIC, to take an alternative approach to identity. It accepted genealogical records and attestation in place of IDs and credit scores. It runs mobile branches that the FDIC treats as branches. The community insures barns and rebuilds them after fires. The bank has had very few write-offs. Harker also cites a Wyoming bank that serves ranchers. He concludes that deep customer knowledge is the alternative to scale.

“But if you really understand your customer, I mean, really deeply understand your customer, you cannot just survive, but you can thrive.”at 14:42

Patrick Harker · The Wharton School, University of Pennsylvania

US countercyclical capital buffer should be raised from zero; release it for all banks at once.

The US countercyclical capital buffer has been zero since its inception, while the UK runs a 2% buffer. Harker says raising it does not force banks to hold more capital, because regulators can lower other capital requirements to offset it. He argues a buffer built up in good times and released for everyone at once carries no stigma, and it keeps the system away from the edge of a crisis, where the Fed otherwise has to reach for QE. He admits there is some friction and that banks pushed back when he argued for raising it in Fed debates.

“In The UK, it's 2% capital. Now people are saying that's forcing banks to hold more capital. No. No. You just... You lower the other capital requirements.”at 44:01

Patrick Harker · The Wharton School, University of Pennsylvania

Discount window use gets publicly disclosed and read as trouble, so stigma limits its usefulness.

Harker responds to proposals to make the discount window more accessible, such as counting discount window collateral toward liquidity coverage ratios or reviving term auction facilities. He argues that the core problem is stigma: a bank's discount window borrowing is publicly disclosed, and the market reads it as a sign the bank is in trouble. He contrasts this with the Bank Term Funding Program, which everyone used without stigma because all banks took it. The stigma is hard to engineer around, so he points to a countercyclical buffer released for everyone at once as a better approach.

“This gets publicly disclosed that you went to the discount window, which the market's gonna read as that bank's in trouble.”at 46:31

Patrick Harker · The Wharton School, University of Pennsylvania

The Fed should have skipped FedNow and gone straight to tokenized money; private RTP could serve.

Harker was against starting FedNow. He thought the private clearinghouse's RTP network could already provide the service. He conceded there was a real case for FedNow anyway, because small banks may lack access to the private network. His preference was to leapfrog to the next technology, a tokenized dollar, tokenized deposit or central bank digital currency. He lost that internal argument, and as an institution the Fed then committed to building FedNow. He says he was then fully behind the decision, and he still has design quibbles.

“I was, by the way, not a fan. I'd be completely honest. I was not a fan of starting FedNow... I thought that the private sector, the clearinghouse had the private sector RTP. That could provide that service... But I thought we should just leapfrog to what the new technology is gonna be”at 20:45

Patrick Harker · The Wharton School, University of Pennsylvania

Real-time payments should allow a short delay; 16 seconds cuts fraud, 30 seconds cuts bank liquidity needs.

Fraud moves quickly on fast payment rails, and it is under-reported, so the true scale is unknown but large. Harker argues that a short hold gives systems time to screen payments. In a calibrated model he ran with a Wharton finance colleague, adding 16 seconds lets the system better detect fraud. Doubling the delay to 30 seconds allows more payments to be netted, which lowers the liquidity banks must hold in the network. Harker argues that tying up less liquidity matters because banks that must hold more liquidity for a given return take more risk on what is left. He acknowledges the counterargument that card-style checkout needs milliseconds, but says the system can be engineered around that, and that for most payments a little delay has real benefits.

“If you just add sixteen seconds into the payment, you can reduce fraud because now you can better look for that fraud. Right? ... if you add just another... Say double that to thirty seconds, you can now start netting more payments, and the banks can have lower liquidity”at 24:38

Patrick Harker · The Wharton School, University of Pennsylvania

Uniform ~3% card fees make cash users, often lower-income, subsidize credit card users.

Card swipes still carry about a 3% merchant fee in the US. If there is no separate cash and credit price, cash payers are subsidizing credit users. Harker says cash payers are usually lower-income and card users are usually higher-income, so the transfer runs from poorer to richer households. Other jurisdictions, such as Europe, cap these rates at around 50 basis points. He acknowledges that caps have costs and trade-offs. He says the fee is a major frustration for merchants and for customers, especially when the surcharge appears explicitly on the bill.

“the people who pay cash, if there's not a separate price for cash or credit, the people who are paying cash are subsidizing the people who are paying for credit.”at 29:56

Patrick Harker · The Wharton School, University of Pennsylvania

AI and immigration drive labor market weakness, which lies outside the Fed's control.

Harker argues that what is happening in the labor market has little to do with what the Fed can control. He names AI and immigration as forces affecting the labor market, and says they have nothing to do with what the Fed can do positively or negatively in the short to medium run, and perhaps not in the long run. From this he concludes the Fed should not try to trade off employment against inflation.

“AI, immigration. We can go down the list of all sorts of... They have nothing to do. And they're affecting the labor market.”at 48:14

Patrick Harker · The Wharton School, University of Pennsylvania

AI can do math proofs better than humans; edge shifts to ideas, creativity and experienced judgment.

Harker says AI is becoming so good at mathematics that the old prize for a clever proof is disappearing, because AI can do proofs better than people. He argues that what will distinguish people is their ideas and creativity, the data they bring, and the slant they take on a problem. Using AI well, in his view, depends on years of experience building judgment, which is why he could use AI to replicate his speechwriting process and why Druckenmiller could, but a young person today cannot. He concludes that a one-size-fits-all approach to AI in writing and other fields is wrong.

“AI can do the proofs better than you can. So it's gonna be really your ideas and your creativity in terms of the data you bring, the slant you take on a problem”at 52:41

Patrick Harker · The Wharton School, University of Pennsylvania

Most colleges lose money on athletics; pay-for-play undermines universities' purpose.

Harker argues that most US colleges and universities lose a lot of money on athletics. Public universities often subsidise athletics through student fees, which students increasingly question. He traces the current model to NCAA governance changes that concentrated money in the Power Five conferences, which he saw coming when he sat on the Division I board. He says the purpose of athletics is leadership, grit and character, not a farm system for pros, since most athletes never turn pro. He thinks universities should ask whether making money from athletes is worth their role. He says the Ivy League's simpler model is the one to keep.

“The vast majority of colleges and universities in America lose money on college athletics.”at 8:56

Patrick Harker · The Wharton School, University of Pennsylvania

Higher-ed mergers and closures are unprecedented; flagships will do okay, regional schools struggle.

Harker says the Philadelphia Fed sees unprecedented merger activity and closures in higher education. He attributes this to two forces. One is the demographic cliff. The other is that international students are less likely to come to the US, and many smaller liberal arts colleges depended on them for enrollment. Closures hit communities where the college is the main employer. He expects the mega and flagship universities to do okay and niche liberal arts colleges to survive by offering something unique. Mid-tier state universities face consolidation, and this is already visible in Pennsylvania. He cites the Philadelphia Fed's database of economic impact from education and health institutions across 520 US regions.

“What we're seeing right now is unprecedented. We are seeing merger and acquisition activity in higher education or simple closures.”at 10:27

Patrick Harker · The Wharton School, University of Pennsylvania

Tokenized deposits should beat stablecoins because banks can pay interest on them.

Stablecoins already exist and are in legislation, and banks are responding with tokenized deposits. Harker sees one built-in advantage for tokenized deposits: they can pay interest, which stablecoins cannot. If a customer can do everything with a stablecoin and also earn interest, he expects the customer will probably choose the deposit. He qualifies this with 'we'll see how this all shakes out' and 'the market has to do what the market's gonna do'. He also says that crypto is a different case. Some treat it as an asset class, and he is not sure about that, but it is not a transaction tool.

“and they have one advantage of a tokenized deposit versus stablecoin. They can pay interest. ... if I can do everything I can do with a stablecoin and get some interest, I'm probably gonna do that.”at 28:44