Did an experiment “deepen and prolong” the Great Recession?

(Long and fairly geeky)

That’s the claim emblazoned on the cover of US academic George Selgin’s 2018 book Floored.   It is a big claim by a smart author, who has written many years (often quite sceptically, to say the least) about aspects of central banking.  And, for once, I won’t bury my conclusion: I wasn’t convinced.

The “experiment” Selgin is writing about is the decision, implemented at the start of October 2008, to pay (effectively a full market) interest rate on excess reserves (over and above the regulatory minimum the Fed persists in requiring) held by banks in their accounts at the Federal Reserve.    The Fed was late to the business of paying interest on these deposit balances (other countries, including New Zealand, had done so earlier).  It required Congressional authorisation –  itself a somewhat unusual feature –  and even having obtained that approval the new regime wasn’t supposed to be implemented until 2011.   And when the legislation was passed the focus had been on required reserve balances (not paying interest on those just represented a federal tax).   But with Fed liquidity operations during the 2008 financial crisis adding lots of excess reserves, the new interest on reserves policy was rushed into effect from 1 October 2008, with the aim of supporting demand for those reserves, and stopping market interest rates falling further than the Fed wanted.

That might seem a bit odd.  But remember that although the Federal Reserve was pretty responsive to liquidity stresses and frozen financial markets, they were slow to grasp the severity of the economic downturn.  This is from an earlier post

For example, the FOMC met two days after the Lehmans failure [in mid-Sept].  Had the Fed thought the Lehmans failure would prove “catastrophic”, or even just aggravating the severity of the recession, a cut to the Fed funds rate would surely have been in order.  There wasn’t one.  And the published records of the meeting show no sign of any heightened concern or anxiety about the financial system or spillover effects to the economy. 

It wasn’t until mid-December 2008 that the Fed funds target range was lowered to 0 to 0.25 per cent (with excess reserves now being remunerated at 0.25 per cent).

Congress had actually specified that any interest paid on reserves was to be at a rate that did not “exceed the general level of short-term interest rates”.   And yet, as various charts in the book demonstrate, the rate paid on excess reserves was to consistently exceed other short-term rates (including LIBOR, GC repo rates, and Treasury bill rates).   Notwithstanding the Congressional mandate, this wasn’t an accidental outcome, but a matter of deliberate design, since the Fed’s explicit aim –  outlined in various policy documents Selgin cites –  was to make excess reserves “attractive relative to alternative short-term assets”.    For that to happen, the interest rate had to be attractive.

Selgin’s argument is that this choice was at the root of much evil.    Specifically, by making excess reserves attractive –  so that banks didn’t want to get rid of them –  the historical link between excess reserves and broader monetary aggregate measures was broken.   Had banks been encouraged/incentivised to use, and (individually) try to get rid of, excess reserves, access to credit would have freed up much more quickly, demand would have expanded, and the economic downturn would have been (a) less deep, and (b) less prolonged.   As far as I can tell, the main channel seems to be a demand one, but he also argues that the productivity growth slowdown would also have been less severe.

Here’s why I’m not convinced.

First, take the counterfactual in which interest had not been introduced on excess reserves.  By 8 October 2008, the Fed funds target rate was still 1.5 per cent.  Without any remuneration on excess reserves, short-term market rates –  at least those that didn’t involve pricing any bank credit risk – would have been heading towards zero pretty quickly.    But by 16 December, the Fed funds target (now a range) was reduced to 0 to 0.25 per cent.   With no interest on reserves, the increasing volume of excess reserves would have reduced more market rates to zero.  But that is a difference of (a) two months and then (b) 25 basis points.    25 basis points rarely makes that much difference.

The argument is that credit conditions would have eased up more quickly.  Well, maybe, but in late 2008 and early 2009, not only was there extreme uncertainty about the creditworthiness of many marginal borrowers, but there was a great deal of reluctance among borrowers to take on new credit (who knew when demand and activity would recover?).  For entirely understandable reasons, most people were in batten-down-the-hatches mode.

I’m quite prepared to concede that lower interest rates could/would have made some difference – by then not so much in altering the depth of the trough, but in supporting the recovery that got underway from about mid-2009.  But (a) the near-zero effective lower bound on nominal interest rates prevented short-term falling to anything like the extent (to say -5 or -6 per cent) that analysts estimated (using Taylor rule frameworks) might, in the abstract have been desirable, and (b) the Fed didn’t want to lower interest rates further.  How do we know that?  Because they made no effort to utilise all the leeway they did have (various other central banks later cut their policy rates as low as -0.75 per cent), or to take emergency steps to ease the effective lower bound.   (In fact, had they been able to take their policy target rate materially negative, introducing interest on effective reserves would have been a prerequisite –  if you could earn zero on funds in the Fed accounts, while all around you rates were negative, the option of just leaving your money at the Fed would have been exceedingly attractive.)

And, of course, it wasn’t just the Fed that didn’t want (or believe it necessary for) short-term interest rates to be lower.    Bond yields for a long time were pricing a pretty quick rebound in short-term interest rates, and it wasn’t until 2012 –  well after the trough of the recession –  that US 10 year Treasury yields got down to around 1.5 per cent.   The Fed and markets were mostly, and repeatedly, focused on the first tightening (which didn’t actually take place until 2015).   That was a mistake, of course, but presumably involved the best expert judgement of the FOMC about where short-term rates neeeded to be.  If they’d read the economy correctly, official short-term rates would have been set lower, regardless of the interest on reserves policies.

The other main reason I’m sceptical is that all of this is a US-specific story, and yet the US experience of the recession and recovery wasn’t unusually bad, even though the US was itself the epicentre of the crisis.    If Selgin’s story was correct, the combination of being the crisis epicentre and adopting the IOR policy in the middle of it all, should have left the US economic performance looking pretty poor relative to, say, (a) other big countries (G7) with their own currencies, and (b) non-crisis advanced countries.  That should be so whether we focus on either real GDP per capita or real GDP per hour worked.

There are three other G7 floating exchange rate countries.  Two –  Japan and Canada –  didn’t have a homegrown financial crisis at all, while the UK was caught up in the US crisis.  Of those countries, all three had slower higher productivity growth than the US over the decade after 2007, and the US also had higher growth in real GDP per capita (although the differences with Japan and Canada are small).    Comparing the US with the smaller floaters who didn’t have a crisis (Australia, Norway, Israel, New Zealand) the US looks to have been pretty much in the middle of the pack.  Precise comparisons depend on which periods and which variables you focus on, but the US experience really doesn’t stand out in the way the Selgin hypothesis appears to require.  Of course, one never knows the (economic) counterfactual, but much as he criticises the IOR policy, Selgin doesn’t (that I noticed) set out one either.

In his book Selgin cites another short piece he wrote last year specifically about New Zealand’s experience with interest on reserves.  He claims our experience supports his case.   I’m also not convinced about that.

Some history.  When the OCR system was introduced in 1999, we designed it as (what is known as) a channel system.  The OCR itself was set half-way between the interest rate paid of deposits (25 basis points below OCR) and the rate at which banks could borrow (collateralised) from the Reserve Bank (25 points above OCR).  Actual settlement cash balances were tiny (of the order of $20 million in total).   Banks were required to keep their accounts in credit at the end of each day, but otherwise they had no real demand for positive balances. $1 each was, in principle, sufficient.    Banks lent and borrowed among themselves to manage the impact of net customer flows between them.

That system worked only because of a strange bifurcation we had (consciously and deliberately) introduced a few years earlier.  Until the late 1990s, interbank settlements were done only once at day (leaving lots of intraday credit exposures which could have led to havoc etc if ever there was a bank failure).  Like most other countries, we replaced that with a real-time gross settlement (RTGS) system, under which large wholesale transactions (mostly fx, but also fixed interest securities) were settled individually during the course of the day.  In a system with perhaps $30-40 billion of daily transactions, $20 million of total reserves wasn’t going to be enough.

To meet these liquidity needs, we set up a system of collateralised intra-day credit (“autorepo”), in which we lent banks billions of dollars during the day and took their securities as (in economic terms) collateral. At the end of the day, all those intraday transactions were unwound, and the system ended each day still with only $20 million or so of aggregate sewttlement balances.  By the mid 2000s, however, there was impetus for change from several sources.  Maintaining the separate software (the “autorepo module” in Austraclear) seemed cumbersome, probably expensive, and unnecessary.  And the stock of government bonds was steadily diminishing (lots of fiscal surpluses, and high offshore demand for NZ government bonds) and the Bank’s appetite for lending on paper issued by banks themselves (mostly bank bills) –  which had never been high – was diminishing.

And so we adopted a radically simpler system.  Instead of lots of lots of intraday repos, and the $20m balances at the end of the day, we just combined the two.  The Reserve Bank injected additional liquidity (billions of dollars of it) and bought various financial assets with the proceeds.  Bank wanted –  or needed – substantial balances to cope with the ups and downs of intra-day settlement flows.    They held more settlement cash balances and fewer securities, and we issued more settlement cash balances and held more securities.  This chart from Selgin’s New Zealand paper illustrates the magnitude of the change.


Since the change was not intended to have any macroeconomic consequences, unsurprisingly there was a considerable change in the ratio of (say) broad money to settlement cash balances.  But for big picture purposes, that change itself was inconsequential.

However, one consequence –  the one that is the focus of Selgin’s note –  was that we no longer had a channel system. If we were paying an interest rate on $8 billion of settlement cash balances, that was going to be the operative Reserve Bank rate (hardly anyone borrowed from us again through the standing facility, except to test that it still worked) and the OCR was redefined as the deposit rate.  We’d moved –  consciously and deliberately –  to a floor system.   Selgin makes quite a lot of the idea that a mere a 25 basis point change had materially altered demand for reserves (hence the parallel he seeks to draw with the US), but in fact what really created the new demand was the Reserve Bank’s decision to end autorepo (and special intraday credit).   Banks could simply not have settled their payments –  sometimes well over $1 billion each  –  without holding a lot more settlement balances.

The simple system was initially introduced didn’t last long, in part because the early stages of the financial crises (abroad) broke upon us relatively soon.   Initially, we’d been willing to pay the OCR rate on any balances banks had in their accounts at the end of each day.   But for a variety of reasons, many people at the Bank were not happy about the consequences of this, including the fact that if a bank wanted to hold more settlement cash balances and couldn’t find any other bank keen to lend to them, we had to respond by increasing settlement cash balances, or see market interest rates potentially move out of line with our target.    I took the view that if there was a higher demand for settlement cash balances –  and macro conditions were where we wanted them – we should simply supply them.  The taxpayer typically profited from us doing so.

From memory I was the lone dissenter on the relevant committee when it was decided to introduce “tiering”. Under these arrangements, we would tell each bank how much they were allowed to hold at a full OCR interest rate, and anything else in their accounts at the end of each day would earn a rate a lot lower.  There were earnest bureaucratic efforts to devise formulae to determine how much banks should be allowed to hold, and through my remaining years in the Bank these were updated very so often.  I could never quite reconcile myself to this sort of (perhaps largely harmless) “central planning” or rationing.

Selgin –  who has written a lot in the course of his career about free banking, and the (not free) pre-Federal Reserve regime –  seems to think that tiering (which made excess reserves quite unattractive to banks) made a material difference, including to our economic performance.  Tiering was actually introduced as part of a package and (at least in my observation) the other component – lending secured on bank bills –  was at least as important, if not more so.  Selgin argues that, in consequence, credit spreads in New Zealand never blew out to the extent in the US and Europe, while somewhat grudgingly conceding what looks to me quite an important difference:

though the difference also reflected the fact that New Zealand’s banks were not so encumbered with toxic assets as some U.S. and European banks.

As in, not at all exposed to the sort of assets creating problems then in the Northerm Hemisphere.

It is certainly true that –  as compared to the US system –  tiering did lead to some new overnight interbank lending. Selgin puts a lot of store on this (and on the death of the overnight market in Fed funds in US), but I think it is a mistaken emphasis.   First, had our banks had anything like the degree of concern about each other that US banks did, there’d have been no interbank lending during the crisis.  And second, and more important, banks have plenty of other interactions in which to monitor the creditworthiness of each other.  Overnight loans have always seemed pretty minor relative to those other exposures (eg collateralisation on net positions in derivatives markets etc).

Selgin’s bottom line about New Zealand is this

The RBNZ’s success in keeping credit flowing may have in turn contributed, if only to a modest extent, to New Zealand’s Great Recession being  both one of the first to end and one of the shallowest.

And yet, with barely any domestic financial crisis ourselves and –  on Selgin’s telling –  with superior monetary management, here is the chart of per capita GDP.

crisis costs 2019

Yes, our recession was a little shallower than that of the United States –  you’d surely expect that when we didn’t have big banks toppling over, or new ones being bailed out almost every week.  A year or so later, we actually had a relapse, and across the whole decade there was rarely more than 1 per cent difference between the total cumulative growth rate.   And this is the chart on which New Zealand looks relatively good.

Here, by contrast, from the OECD data, is real GDP per hour worked for the two countries.

us nz prod

Our underperformance isn’t necessarily much worse than it was pre-crisis, but (a) the US did have the crisis and Selgin asserts it was very costly, and (b) we have the monetary management system that he regards as superior to that in the US.  There just doesn’t seem to be anything in the data to support a story that interest on reserves really made any material difference to macro outcomes.

The bigger issue always was the over-optimism about the outlook that meant that Fed wasn’t as aggressive as it could have been (actually neither was the Reserve Bank).  That remains a concern now when the authorities in neither country have done anything to “fix” the near-zero lower bound constraint.  And, by definition, the next serious downturn is getting closer every day.

For those (geeks) interested in such things, it is an interesting and stimulating book.   And he raises some points which I found more persuasive about the reluctance of the Fed to more actively reduce the size of its balance sheet, even years after the crisis.  That has the effect of leaving the central bank nearer the centre of the credit allocation process than it really should be. In turn, that risks inviting Congressional (or industry) pressure in the next downturn for the central bank to do more of that credit allocation: if there is a role for government in such matters, it is surely one for fiscal policy and Congress itself, not for the central bank.

Those readers with institutional subscriptions can read my, considerably shorter, review of Floored on the Central Banking journal website.