Alt-M http://www.alt-m.org Ideas for an Alternative Monetary Future Tue, 24 May 2016 13:18:19 +0000 en-US hourly 1 http://wordpress.org/?v=4.1.11 On "Shadow Money" http://www.alt-m.org/2016/05/24/on-shadow-money/ http://www.alt-m.org/2016/05/24/on-shadow-money/#comments Tue, 24 May 2016 13:18:19 +0000 http://www.alt-m.org/?p=42922 The shadow banking literature has vastly and rapidly expanded since the financial crisis, and has produced some interesting pieces, as well as some exaggerated claims, in my view.  While I am not writing today to address those claims, I still wish to question a closely linked concept that has simultaneously...

The post On "Shadow Money" appeared first on Alt-M.

]]>
TMoneyPyramidhe shadow banking literature has vastly and rapidly expanded since the financial crisis, and has produced some interesting pieces, as well as some exaggerated claims, in my view.  While I am not writing today to address those claims, I still wish to question a closely linked concept that has simultaneously sprung up in the literature and in particular in the post-Keynesian one: shadow money.

One of the most elaborated and comprehensive academic research papers on this particular topic is the recently published Gabor’s and Vestergaard’s "Towards a theory of shadow money."  It’s an interesting and recommended piece.  But while I agree with some of their writings, I have to find myself in disagreement with a number of their points and examples* and in particular their central claim: that repurchase agreements ("repos" thereafter) are shadow money; that is, a type of monetary instrument used within the shadow banking system.

For some readers that might not know how a repo works, below is a concise definition provided by the IMF:

Repo agreements are contracts in which one party agrees to sell securities to another party and buy them back at a specified date and repurchase price.  The transaction is effectively a collateralized loan with the difference between the repurchase and sale price representing interest. The borrower typically posts excess collateral (the “haircut”).  Dealers use repos to borrow from MMFs and other cash lenders to finance their own securities holdings and to make loans to hedge funds and other clients seeking to leverage their investments.  Lenders typically rehypothecate repo collateral, that is, they reuse it in other repo transactions with cash borrowers.

Given repos’ (and their asset counterpart: reverse repos) properties, my view is that repos aren’t shadow money but a shadow funding instrument.  While it might not sound such a big issue, I believe the distinction is important from an analytical perspective as well as to avoid confusion.  Let me elaborate.

Gabor and Vestergaard define shadow money as “repo liabilities, promises backed by tradable collateral.”  According to them, shadow money has four key characteristics:

a) In modern money hierarchies, repo claims are nearest to settlement money, stronger in their "moneyness" than ABCPs or MMF shares.

b) Banks issue shadow money.  The incentives to issue repos are incentives to economize on bank deposits and bank reserves.

c) Shadow money, like bank money, relies on sovereign structures of authority and creditworthiness.  The state offers a tradable claim that constitutes the base asset supporting the issuance of shadow claims.

d) Repos create (and destroy) liquidity at lower levels in the hierarchy of credit claims.

They offer this chart of "modern" money hierarchy:

Shadow Money

I have to object to repos being classified as "money."

Money, as typically defined by economists, has three characteristics: it is a medium of exchange, a unit of account and a store of value.  High-powered money (the "outside money" of the financial system) currently fits this definition, as a final settlement medium.

The "moneyness" concept, a term now popularised by JP Koning’s excellent blog, asserts that various types of assets have various degrees of money-like properties.  In this quite old but classic post, JP argues that anything from beers and cattle to deposits benefits from some degree of moneyness.  In another old post, Cullen Roche provided the following good "money spectrum" chart (although I’d disagree with his outside money/deposit ranking):

scale of moneyness

Therefore, most goods and assets have some monetary properties: some can be used as media of exchange or store of value.  All represent a claim of some sort on money proper.  As a general (but inaccurate) rule, the more their price in terms of outside money fluctuate, and hence their conversion risk raises, the further away they are on the moneyness scale.  But conversion (almost) on demand also implies that, in order to have some money characteristics, a good or asset needs to be tradable.

Now let’s get back to repos as shadow money.

Repos are a liability issued by the debtor in exchange for high-powered money, of which reverse repos are the asset counterpart held by the creditor (and hence the claim on the high-powered money originally transferred, plus interest).  The debtor also transfers an extra asset (i.e. the collateral) to the creditor for security purposes at a pre-agreed haircut depending on its credit quality and market risk sensitivity.

We get here to the main point of this post: repos have little money-like property due to their non-tradability and lack of on-demand convertibility.

Indeed, a repo liability is of course non-tradable, in the same way that any debt that one owes cannot be traded for another type of liability.  It can only be refinanced and/or extinguished.  A reverse repo (or repo claim) however, could potentially have tradable properties, allowing a creditor to exchange his claim almost instantaneously on the market.  Problem is: this does not happen.  Unlike bonds or other assets, and due to the very specific features of such private agreements, there is no secondary market for repo claims.  Once a repo has been agreed upon, the contract is fixed between the two parties until maturity (or default).  Consequently, repo claims can effectively be assumed to have no liquidity.

Seen this way, it is hard to classify repos as "money," and they certainly do not deserve their third place in the moneyness hierarchy above.  So what are repos?  As I previously said, they are a funding instrument. Given that the shadow banking system makes use of repos on a large scale, we can potentially call them a "short-term secured shadow funding instrument."  And please note that repo issuance isn’t limited to banks and broker-dealers; other institutions also use them.

You’ll be tempted to reply: “what about deposits?  They have no secondary market and are not tradable either.”  This isn’t strictly accurate.  While they are both promises to pay a certain amount of money proper at a certain date, there is a very specific difference between deposit liabilities ("on demand" ones especially) and repo liabilities.  Banks themselves are deposits’ secondary market: deposits can be ‘traded’ within the bank’s own balance sheet and swapped for cash on demand.  And when dealing with a counterparty that does not hold an account with the same bank, banks take over the responsibility of transferring the underlying funds (i.e. high-powered money).

If repos aren’t "money", what else could be considered "shadow money?"  Well, assets provided as collateral do have liquidity, tradability, and therefore some "moneyness."  Those assets can sometimes be used in further transactions.  This is why I am wondering whether or not there isn’t some confusion with "shadow money" proponents’ terminology.  While their writings clearly emphasise the "shadow money" nature of repos themselves (and Poszar seems to be using the same definition here), many other academic authors have instead referred to the most commonly-used types of repo collateral (high quality and highly liquid sovereign and corporate bonds) as "shadow money" (which indeed makes more sense to me, although I do not fully adhere to this concept either).

There are plenty of things worth discussing regarding this theory of shadow money and the use of repos in general, but the money-like properties of repurchase agreements isn’t one of them.  Let’s focus on their funding properties instead.

______________

*I also believe that their shadow money expansion theory is subject to the same critique as other endogenous outside money theories, such as MMT’s.

PS: The fact that repos are backed by marketable collateral does not confer any specific monetary property to repo claims.  Marketable collateral is used in many other types of lending transactions, in particular in private banking-type lending.  Also, repos and any other collateralised lending are expected to be repaid at par, independently of the valuation fluctuations of their underlying collateral.

PPS: Baker and Murphy build on Gabor’s and Vestergaard’s piece and just published a blog post that argues for a new "investment state," in a typical post-Keynesian interventionist fashion.

The post On "Shadow Money" appeared first on Alt-M.

]]>
http://www.alt-m.org/2016/05/24/on-shadow-money/feed/ 5
Dodd-Frank, Economic Policy Uncertainty, and Bank Lending http://www.alt-m.org/2016/05/21/dodd-frank-economic-policy-uncertainty-and-bank-lending/ http://www.alt-m.org/2016/05/21/dodd-frank-economic-policy-uncertainty-and-bank-lending/#comments Sat, 21 May 2016 15:18:47 +0000 http://www.alt-m.org/?p=42544 Michael Bordo, John Duca, and Christoffer Koch recently produced a Cato Institute Research Brief on how policy uncertainty affects bank lending.  The Cato Research Briefs in Economic Policy series allows economic researchers to write an accessible 2,000 word summary of an academic article that cuts through the required literature review,...

The post Dodd-Frank, Economic Policy Uncertainty, and Bank Lending appeared first on Alt-M.

]]>
Research-Briefs RotatedMichael Bordo, John Duca, and Christoffer Koch recently produced a Cato Institute Research Brief on how policy uncertainty affects bank lending.  The Cato Research Briefs in Economic Policy series allows economic researchers to write an accessible 2,000 word summary of an academic article that cuts through the required literature review, regression explanations, etc., required in the full version.  Bordo, Duca, and Koch based their Research Brief on an NBER article they published in February, entitled, “Economic Policy Uncertainty and the Credit Channel: Aggregate and Bank Level U.S. Evidence over Several Decades.”

The authors find that, holding other macroeconomic and regulatory factors constant, policy uncertainty “significantly slows U.S. bank credit growth.”  To measure uncertainty, Bordo, Duca, and Koch use an economic policy uncertainty (EPU) index based on newspaper article wording.  The EPU index shows consistently higher uncertainty in the four years after the Great Recession compared to four years after the five prior recessions.  What has caused elevated uncertainty post ’08 relative to comparable time periods?  Bordo et al. argue that Dodd-Frank accounts for at least part of this relatively high uncertainty.  Dodd Frank not only vastly overhauled financial regulation, but did so in a way that left a uniquely large and substantive amount of rulemaking authority to administrative agencies.  Three years after Dodd-Frank’s passage only half of the rules the law mandates had been finalized.  Even now, regulators have still not implemented twenty-five percent of the rules required by the law.

Just what has the increased uncertainty, caused at least in part by Dodd-Frank, meant for bank lending?  Examining data from 1961-2014, Bordo, Duca, and Koch show that uncertainty has a highly significant effect on real per capita bank loan growth.  This significance is robust across various macroeconomic and regulatory variables, like changes in GDP growth, the fed funds rate, and bank capital rules.  The authors also delineate their results by bank size, noting that uncertainty has a greater adverse impact on smaller banks.

The connection that Bordo, Duca, and Koch show between policy uncertainty and depressed bank lending is surely an important consideration for policymakers charged either with evaluating the legacy of Dodd-Frank, or with thinking about the reasons for slow post recession growth in general.

The post Dodd-Frank, Economic Policy Uncertainty, and Bank Lending appeared first on Alt-M.

]]>
http://www.alt-m.org/2016/05/21/dodd-frank-economic-policy-uncertainty-and-bank-lending/feed/ 4
Posner on the Legality of the Fed's Last-Resort Lending http://www.alt-m.org/2016/05/19/posner-legality-feds-last-resort-lending/ http://www.alt-m.org/2016/05/19/posner-legality-feds-last-resort-lending/#comments Thu, 19 May 2016 13:12:23 +0000 http://www.alt-m.org/?p=40635 A recent Marginal Revolution post has alerted me to Eric Posner's January 2016 working paper, "What Legal Authority Does the Fed Need During a Financial Crisis?"  Posner's paper is remarkable, both for its assessment of the legality of the Fed's emergency lending operations during the recent crisis, and for the...

The post Posner on the Legality of the Fed's Last-Resort Lending appeared first on Alt-M.

]]>
EricPosnerA recent Marginal Revolution post has alerted me to Eric Posner's January 2016 working paper, "What Legal Authority Does the Fed Need During a Financial Crisis?"  Posner's paper is remarkable, both for its assessment of the legality of the Fed's emergency lending operations during the recent crisis, and for the policy recommendations Posner offers based on that assessment.[1]

Posner's account of the Fed's actions reads like a long bill of indictment.  The Fed's Bear Stearns rescue, for starters, "was legally questionable."  The Fed couldn't legally purchase Bear's toxic assets, and it knew it.  Instead it created a "Special Purpose Vehicle" (SPV), named it Maiden Lane, and lent Maiden Lane $28.82 billion so that it could buy Bear's toxic assets.  Voila!  What would have been an illegal Fed purchase of toxic assets was  transformed into a Fed loan "secured" by the very same assets.

But clever as the Fed's gambit was, it  wasn't so clever as to render it entirely innocent of legal hanky-panky.  "The problem," Posner observes,

is that the transaction provided that the value of the Fed's interest would be tightly connected to the value of the underlying assets.  If the assets fell in value by as little as 4%, the Fed would lose money… By contrast, in a [properly] secured loan…the lender bears very little to no risk from the fluctuation of asset values.  Functionally, the Maiden Lane transaction was a sale of assets, not a secured loan.

In rescuing AIG, the Fed resorted to the same "legally dubious" bag of tricks it employed in saving Bear, creating two more Maiden Lane vehicles, and again assuming considerable downside risk.  The Fed also grabbed a 79.9 percent equity stake in AIG, which it placed in a trust established for the sole benefit of the U.S. Treasury.  That transaction was later held by the Court of Federal Claims to have been been unauthorized by the Federal Reserve Act, and therefore illegal.

In fact, according to Posner, all of the Fed's emergency lending programs, the TALF alone excepted, "raised legal problems."  In each case the Fed violated the spirit of 13(3), which requires that its loans be "secured to the satisfaction of the Federal Reserve bank."  Posner is especially good at explaining the speciousness of  Fed lawyers' claims that the Fed's loans were indeed secured:

Imagine, for example that the Fed would like to make an unsecured loan to Joe Shmo, who has no assets.  Following the legal division's advice, the Fed could create an SPV called Shmo LC.  Shmo LC would then lend money to Joe, and in return receive an unsecured note from him, that is, an IOU.  Shmo LC would get its money from the Fed, which would make a section 13(3) loan to Shmo LC secured by Shmo's note.

All of which would be just dandy, were it not for the inconvenient fact, pointed out by Posner, "that the Fed, Congress, and all other relevant actors have always understood section 13(3) to [require] 'real' security — in the sense of collateral that would render the loan riskless or close to that."

Suppose, on the other hand, that the Fed's lawyers were in fact correct.  Suppose that it was perfectly legal for the Fed to have "secured" many of its 13(3) loans using assets of doubtful value.  In that case, the Fed's claim that it was unable legally to rescue Lehman Brothers was itself a sham, for a Fed that might have legally lent to Joe Shmo could certainly have lent legally to what was at the time the United States fourth-largest investment bank.  Fed officials can't have it both ways: either they lied about Lehman, or they broke the law left-and-right with the 13(3) loans they did make.

Posner himself believes that the Fed let Lehman fail for political and "operational" reasons rather than legal ones, and that the questionable legality of yet another Joe Shmo-type operation — and an especially blatant one at that — merely provided it with "a convenient excuse" for avoiding political backlash.  I'm pretty sure he's right.  But although he recognizes the capricious nature of the Fed's decision to let Lehman fail, neither that awareness nor his understanding that the Fed rode roughshod over existing laws causes Posner to see any need to limit the Fed's last-resort lending powers.

Quite the contrary: so far as Posner is concerned, the fact that the Fed has been inclined to bend or break the law, or to invoke it only when it found doing so convenient, is reason, not for strengthening the rules governing the Fed's last-resort lending, but for getting rid of them altogether!  According to him, the problem isn't that the Fed thumbed its nose at existing laws.  It's that "gaps in the government's powers" made all that nose-thumbing necessary.  What we need to do, Posner says, is fill those "gaps."

For Posner, filling the power gaps means, first of all, consolidating within a single "Financial Crisis Response Authority" (FCRA), and preferably within the Fed itself, all of the crisis-response powers presently divided among it, the Treasury, and the FDIC.  It also means granting to that authority the power to:

  • buy assets, including equity.
  • make unsecured loans to non-bank financial institutions.
  • control non-bank financial institutions…in order to force them to pay off counterparties, lend money, and so on.
  • wind up insolvent non-bank financial institutions….
  • force non-bank financial institutions to raise capital.
  • dictate terms of transactions, control the behavior of firms (for example, forcing them to lend), or acquire them where necessary.

The recent crisis shows, Posner insists, that "all these powers are necessary":

Because of the fear of stigma, even liquidity-constrained financial institutions will be inclined to delay before borrowing from emergency credit facilities.  The FCRA needs the authority to force those firms to borrow, and also to force healthy firms to borrow at the same time in order to prevent the market from picking off the weakest firm.  Moreover, the crisis showed that when financial institutions accept emergency loans, they have strong incentives to hoard cash when the system as a whole benefits only if they lend into the market a portion of the money they borrow.  For this reason, the FCRA needs the authority to order firms to enter financial transactions.  Finally, the crisis showed that financial institutions that should be given emergency money may not be able to offer collateral for a loan, and it may be very difficult to value the collateral in any event.  The FCRA needs the authority to make capital injections, unsecured loans, and partially secured loans; and to buy assets

Posner fails to recognize that the stigma problem to which he refers can be solved, without forcing anyone to borrow, by substituting an auction-style lending facility for the Fed's discount window, as was in fact done during the crisis when the TAF was established; and his suggestion that financial firms' hoarding of cash was something the Fed would have prevented had it had the power to do so, rather than something the Fed deliberately encouraged, doesn't square with the facts.  But these are secondary points.  What's most troublesome about Posner's proposal is that it fills the "gaps" in the Fed's power so generously as to do away with practically all limits upon the Fed's ability to meddle with people's property.  His suggestion that a FCRA should be perfectly free to make unsecured loans and commandeer equity, for example, would allow it to lend to Joe Shmo on whatever terms it likes, and to nationalize private firms at will, with utter impunity.

Posner insists nonetheless that his proposed FCRA would not command unlimited power.  Instead, its power would be checked by means of "a robust legal regime" that would "correct abuses after [a] crisis."  Specifically, firm shareholders would "be entitled to sue" the FCRA "and receive a remedy if they can show that the FCRA's actions were unreasonable":

The usual post-crisis analyses by independent government agencies with the power to compel testimony and discover documents from the FCRA will facility the litigation by collecting facts and making them publicly available.

With all due respect to Professor Posner, he seems here to be putting a great deal of weight on an awfully thin reed, if not a broken one.  If it was far from easy for Starr International to convince a court that the New York Fed acted illegally, and if Starr received no "remedy" even despite doing so, how much harder would it be for a plaintiff to establish that the actions of a much more powerful Fed (or FCRA), though perfectly legal, were nevertheless both harmful and "unreasonable"?  And suppose such a plaintiff somehow managed to prevail.  Might Professor Posner, or some like-minded legal scholar, not be inclined in that case to regret the discovery of still another "gap" in the Fed's power, and to recommend, on the basis of the very same arguments Posner offers for filling already apparent gaps, further reforms aimed at ruling-out such lawsuits, to guard against the possibility, however remote, that the threat of them might discourage some desirable (if not obviously "reasonable") anti-crisis measure?

Yet it would be unfair to accuse Posner of depending entirely on lawsuits to constrain his proposed FCRA.  For it's clear that Posner's case for an FCRA wielding vast powers mainly rests, not on the naive belief that such an authority could be adequately constrained by the threat of successful litigation alone, but on the assumption that it will never (or hardly ever) abuse its powers  in the first place.

What's the basis for that bold assumption?  Posner certainly can't claim that it's difficult to conceive of ways in which his proposed FCRA might behave badly.  The rescuing of firms that might safely be allowed to fail, including ones that richly deserve to fail, is only one obvious example.[2]

Instead, Posner's postulate of an infallible Fed appears to take shape as a mutation of his much less heroic (if still doubtful) claim that the Fed did nothing wrong during the recent crisis.  Early in his paper he explains that he plans "to assume that the mainstream view that the Fed acted properly during the financial crisis by lending widely is correct."  As a means for determining what reforms would have allowed the Fed to avoid errors of omission (but not ones of commission) during the recent crisis, the assumption makes perfect sense, even if it happens to be false.  But Posner isn't content to limit himself to such a counterfactual exercise: he wants to draw conclusions concerning "what reforms are necessary to supply [the Fed] with the proper legal authority" going forward.  To do so he segues, perhaps unconsciously, from assuming, arguendo, that the Fed acted correctly this last time around, to assuming, implicitly, not only that it will act correctly in any future crisis, but that it will do so even if it wields much vaster powers than before.

Is it being uncharitable to suggest that, once we grant the assumption that the only errors that a government agency is ever likely to commit are errors of omission, we do not really need a legal scholar, or any other sort of expert, to tell us how to make that agency function ideally?  The granting of unlimited power is in that case a no-brainer.  To anyone who isn't prepared to altogether set-aside the possibility of errors of commission, on the other hand, the sort of reforms Posner proposes must seem naive — and dangerous — in the extreme.

And that's what troubles me most about Posner's proposal.  It's not that his logic is bad.  It's seeing a legal scholar, and a very intelligent one, blithely cast aside the very idea of the rule of law, while championing its opposite: the arbitrary decisions of (practically) omnipotent bureaucrats.

Nor does Posner not realize what he's doing.  On the contrary: he's aware of the complaint of other scholars that the Fed already demonstrates the dangers of substituting  the rule of men for the rule of law, even citing a recent Cato Journal article by CMFA Adjunct Scholar Lawrence White to that effect.  But having recognized White's complaint, Posner dismisses it summarily, on the technical grounds that

the constitutional limitations on delegation of power to agencies — embodied in the nondelegation doctrine – are effectively nil.  The requirement that the LLR use its powers to unfreeze the financial system would supply the intelligible principle required by the nondelegation doctrine under recent precedents.

But White isn't arguing a point of U.S. Constitutional Law.  He's appealing to a fundamental legal principle that's older than the U.S. Constitution, and one that transcends the laws of any particular nation.  Posner's response therefore misses the point entirely.  The question isn't whether or not Congress may grant Fed officials unlimited power.  It's whether it ought to grant them so much power.  Posner thinks it should.  I don't know about you, but I hope Congress puts more weight on the advice of John Locke, David Hume, and John Adams.

_______________________

[1] Although I concern myself here only with his account of the Fed's emergency lending, Posner also assesses the legality of the crisis-related actions of the Treasury and the FDIC.

[2] I dare readers to peruse the aforementioned list of proposed FCRA powers, and to submit comments — under their real names — to the effect that they are unable to imagine others.

Posner himself (p. 30) recognizes that Congress had reasons for not awarding the Fed unlimited last-resort lending powers.  However, he notes that the reasons have been "mostly political, including distrust of the Fed, and popular resentment of the bailouts of Wall Street firms," and appears to dismiss them simply for that reason.  As for the moral hazard problem — the sole non-"political" reason he recognizes for limits upon the Fed's last-resort lending power —  Posner dismisses it as well, claiming that it can be adequately contained by means of "ex ante regulation such as capital requirements, which are independent of the LLR's power."  But as Richard Fisher has argued,

Requiring additional capital against risk sounds like a good idea but is difficult to implement.  What should count as capital? How does one measure risk before an accident occurs?  And how does one counteract the strong impulse of the regulated to minimize required capital in highly complex ways?  History has shown these issues to be quite difficult.  While we do not have many examples of effective regulation of large, complex banks operating in competitive markets, we have numerous examples of regulatory failure with large, complex banks.

Nor is it clear that even the strictest capital requirements can suffice to rule out excessive risk taking when its the case, as it would be under Posner's proposed regime, that regulatory authorities need not hesitate to bail out, not only firms' uninsured creditors, but their shareholders.

The post Posner on the Legality of the Fed's Last-Resort Lending appeared first on Alt-M.

]]>
http://www.alt-m.org/2016/05/19/posner-legality-feds-last-resort-lending/feed/ 16
Some Simple Monetarist Arithmetic of the Great Recession and Recovery http://www.alt-m.org/2016/05/18/some-simple-monetarist-arithmetic-of-the-great-recession-and-recovery/ http://www.alt-m.org/2016/05/18/some-simple-monetarist-arithmetic-of-the-great-recession-and-recovery/#comments Wed, 18 May 2016 13:07:22 +0000 http://www.alt-m.org/?p=41493   The familiar chart below illustrates the depth of the decline in real output during the 2007-09 Great Recession (the shaded period), and the failure of the recovery to return real output to its “potential” path (in other words, to eliminate the estimated “output gap”) during the subsequent years up...

The post Some Simple Monetarist Arithmetic of the Great Recession and Recovery appeared first on Alt-M.

]]>
 

Nominal GDP, Output Gap, productiviy, resource misallocation, velocity of moneyThe familiar chart below illustrates the depth of the decline in real output during the 2007-09 Great Recession (the shaded period), and the failure of the recovery to return real output to its “potential” path (in other words, to eliminate the estimated “output gap”) during the subsequent years up to the present day.  The second chart puts the same 2007-16 period in the context of the previous decades, showing how exceptionally prolonged the current below-potential period is by contrast to previous postwar recessions and recoveries.

White Graph 1, cropped

WhiteArithmeticFigure2

It is instructive to decompose the path of real GDP into its components, nominal GDP and the price deflator. Here are the natural logs of nominal GDP (call it Y) and real GDP (call it y).  From the definition y = Y/P, it follows that ln y = ln Y – ln P, so the growing vertical difference between the two series reflects the rising price level.

White Graph 3

Here is the path of the natural log of the GDP deflator, which is the difference between the real and nominal GDP series. (Note: FRED uses 2009=100 as the index base; I have re-normalized to 2004=1 by dividing by 90).

White Graph 4

Monetarist theory sharply distinguishes real from nominal variables.  Nominal shocks (changes in the path of the money stock or its velocity) have only transitory effects on real variables like real GDP. Accordingly an account of the path of real GDP in the long run (and 6+ years of recovery should be enough time to reach the long run) must be explained by real and not merely by nominal factors.  An account of the path of nominal GDP, by contrast, cannot avoid reference to nominal factors.  So we need distinct (but compatible) accounts of the two paths.

To account for the path of nominal GDP we can use the simple accounting decompositions M = φY and ln M = ln φ + ln Y, where φ (“phi”), following the notation of Michael Darby’s canonical monetarist textbook, is a mnemonic for fluidity, defined as M/Y, and thus the inverse of velocity. Before the Great Recession, the velocity of M2 was on a gradual downward path, falling around 1% per year.  During the Recession it fell much more steeply.  Since the Recession it has been falling around 2% per year.  Correspondingly, φ was on the rise, but its path shifted upward and become slightly steeper.  Meanwhile, the path of M2 has hardly budged, with M2 rising at an annualized 5.9% rate between the midpoint of the previous recession and the midpoint of the Great Recession, and at a 6.7% rate since.  Here are the observed paths of log M2 and log nominal GDP on the same scale, and in the next chart the path of log M2 fluidity:

White Graph 5

White Graph 6

As a stylized approximation, let us treat the path of log M2 as a smooth line without variation.  Then all the variation in nominal GDP is explained by the variation in fluidity.  The downward displacement of the path of nominal GDP during the Great Recession, and its slower growth afterward, corresponds to the upward displacement of fluidity (drop in velocity) during the Recession and its more rapid growth since the recession.  Further simplifying, let us treat the shift in fluidity as due to an upward shift in desired fluidity at a single date t*.  We can explain the downward shift in the steady-state path of the log of nominal Y as the result of an exogenous upward shift in the steady-state path of the log of fluidity resulting from the public’s demand to hold larger M2 balances relative to income Y, in the face of an unvarying path of M2.  The next diagram shows the shifts in steady-state paths, constrained by the accounting identity that ln M = ln φ + ln Y.

White Graph 7

Assuming gradual adjustment of actual to desired fluidity, we can add hand-drawn transitory adjustment paths.  This yields a stylized representation of the actual time series seen above.

White Graph 8

A downward displacement of the steady-state path of nominal GDP, due to a contractionary money supply or velocity shock, can bring a transitory decline in real GDP.  As is familiar, people try to remedy a felt deficiency in money balances by reducing their spending.  In the face of sticky prices, reduced spending generates unsold inventories and hence cutbacks in production and layoffs until prices fully adjust.  As Market Monetarists have long argued, the Federal Reserve could have avoided the downward displacement of the path of nominal GDP if policy-makers had immediately recognized and met the rise in fluidity (the drop in velocity) with appropriately sized expansionary monetary policy.  (I leave aside the Traditional Monetarist critique of the track record of stabilization policy, which argues that central bankers cannot be expected to get the timing and magnitude right in a world where to do so they would have to forecast velocity shocks better than market price-setters.)

Although a one-time un-countered rise in desired fluidity can temporarily knock real GDP below course, such a nominal shock should not persistently displace its growth path, as the first chart above indicates has happened.  We can expect monetary equilibrium to be roughly restored by appropriate adjustment in the price level relative to the nominal money stock, bringing real balances up to the newly desired level, within three or four years at most.  (In the data plot above, we see the GDP deflator shift to a lower path already in 2009.)  Real GDP should around the same time return to its steady-state path as determined by non-monetary factors (labor force size and skills, capital stock, total factor productivity as governed by technologic improvements, policy distortions, and so on).  To explain the continued low level of real GDP relative to estimated potential since 2011 or so, we need a persistent shock to the path of real GDP.

I suggest that real GDP has shifted to a lower path because of a shrinkage in the economy’s productive capital stock — a problem that better monetary policy (not feeding the boom) could have helped to avoid, but cannot now fix.  During the housing boom, investible resources that could have gone into augmenting human capital, building useful machines and sustainable enterprises, and conducting commercial research and development, were instead diverted to housing construction. In the crisis it became evident that the housing built was not worth the opportunity cost of the resources allocated to it.  That major misallocation of resources has lowered the path of the capital stock below its previous trend.  I do not know precisely how the contribution of capital input is measured when the CBO estimates potential output, but I hypothesize that potential output is currently overestimated because capital wastage has not been fully recognized.  I welcome comments and evidence on this hypothesis.

The post Some Simple Monetarist Arithmetic of the Great Recession and Recovery appeared first on Alt-M.

]]>
http://www.alt-m.org/2016/05/18/some-simple-monetarist-arithmetic-of-the-great-recession-and-recovery/feed/ 20
Interest on Reserves and the Fed's Balance Sheet http://www.alt-m.org/2016/05/17/interest-on-reserves-fed-balance-sheet/ http://www.alt-m.org/2016/05/17/interest-on-reserves-fed-balance-sheet/#comments Tue, 17 May 2016 20:47:34 +0000 http://www.alt-m.org/?p=41492 (Recently I was invited to testify on the title subject, along with Robert  Eisenbeis, Todd Keister, and John Taylor, before the House Committee on Financial Services Subcommittee on Monetary Policy and Trade.  The topic is, as loyal Alt-M readers know, one that I have plenty to say about. Here's my written...

The post Interest on Reserves and the Fed's Balance Sheet appeared first on Alt-M.

]]>
SelginTestimony(Recently I was invited to testify on the title subject, along with Robert  EisenbeisTodd Keister, and John Taylor, before the House Committee on Financial Services Subcommittee on Monetary Policy and Trade.  The topic is, as loyal Alt-M readers know, one that I have plenty to say about. Here's my written testimony.)

May 17, 2016

Chairman Huizenga, Ranking Member Gwen Moore, and distinguished members of the Committee on Financial Services Monetary Policy and Trade Subcommittee, my name is George Selgin, and I am the Director of the Cato Institute’s Center for Monetary and Financial Alternatives.  I am also an adjunct professor of economics at George Mason University, and Professor Emeritus of Economics at the University of Georgia.  I am grateful to all of you for having granted me this opportunity to testify before you on the subject of “Interest on Reserves and the Fed’s Balance Sheet.”

The Federal Reserve was originally given the authority to pay interest on bank reserves effective October 1, 2011 by the Financial Services Regulatory Relief Act of 2006.  The intent of that step was to increase commercial banks’ efficiency by reducing the opportunity cost they incurred in being required to hold reserves that bore no interest.

The Emergency Economic Stabilization Act of 2008 subsequently accelerated the effective date upon which the Fed might begin paying interest on reserves to October 1, 2008.  The Fed in turn actually began paying banks interest on both required reserves and excess reserves on October 9, 2008.

The rationale behind the early deployment of the Fed’s authority to pay interest on reserves was entirely different from that behind the original, 2006 measure. Interest on reserves was to be relied upon, not as a means for improving banks’ efficiency, but as a new Federal Reserve instrument of monetary control.  Specifically, it was resorted to as a contractionary monetary measure, meant to prevent monetary expansion that would otherwise have taken place as a consequence of the Fed’s post-Lehman emergency lending operations.  As Chairman Ben Bernanke explained at the time:

our liquidity provision had begun to run ahead of our ability to absorb excess reserves held by the banking system, leading the effective funds rate, on many days, to fall below the target set by the Federal Open Market Committee.  … Paying interest on reserves should allow us to better control the federal funds rate, as banks are unlikely to lend overnight balances at a rate lower than they can receive from the Fed; thus, the payment of interest on reserves should set a floor for the funds rate over the day. With this step, our lending facilities may be more easily expanded as necessary.[1]

In his memoir Chairman Bernanke says that “by setting the interest rate we paid on reserves high enough, we could prevent the federal funds rate from falling too low, no matter how much [emergency] lending we did."[2]

According to Richmond Fed economists John R. Walter and Renee Courtois, Fed officials were concerned at the time that, in pushing the fed funds rate below its target, the Fed's emergency credit injections might end up “increasing the overall supply of credit to the economy beyond a level consistent with the Fed’s macroeconomic policy goals, particularly concerning price stability…. Once banks began earning interest on the excess reserves they held, they would be more willing to hold on to excess reserves instead of attempting to purge them from their balance sheets via loans made in the fed funds market, which would drive the fed funds rate below the Fed’s target for that rate.”[3]

The Fed’s decision had reflected the FOMC’s belief in the days immediately following Lehman’s failure that the inflation outlook was highly uncertain, and that, in the absence of interest payments on reserves, continued emergency lending could well push inflation above the Fed’s 2% target.  In retrospect, the Fed’s fears were tragically misplaced. Instead of assisting it in achieving either its federal funds rate or its inflation target, the Fed’s decision to begin paying interest on bank reserves contributed to a collapse in nominal spending that was already in progress, helping thereby to turn the subprime crisis into a more general macroeconomic downturn.  That the Fed realized that the macroeconomic situation was rapidly worsening even before it actually began paying interest on reserves was reflected in its decision to further reduce its federal funds target, from 2% to 1.5%, on October 8, 2008.  The Fed chose not to reconsider its decision to commence paying banks to hold reserves a day later.

The rapid decline in the growth rate of nominal GDP, from about 3.5% at the start of 2007 to minus 3.3% by the second quarter of 2009, is shown in Figure 1, which also shows the progress of adjustments to the fed funds target and the “effective” federal funds rate, which is the average rate of interest paid on actual overnight loans.  The collapse in spending is ipso-facto evidence that the Fed’s stand was overly tight.  The figure shows that the Fed’s rate target had become more-or-less irrelevant by the third quarter of 2008, and that this continued to be the case after it began paying interest on bank reserves.  The latter policy did, however, reduce the volume of interbank lending and overall credit expansion, contributing thereby to the collapse of nominal GDP.

Figure 1

SelginIORTestimonyFigure1

The Fed’s policy of paying interest on excess reserves, combined with the substantial scale of its post-Lehman emergency lending and the even greater scale of later rounds of Quantitative Easing, led to a massive accumulation of banking system excess reserves.  As Figure 2 shows, excess reserves, which between 2002 and 2008 had seldom exceeded $1.8 billion, had risen to almost $2.7 trillion in August 2014, and as of this April still exceeded $2.33 trillion.

Figure 2

SelginIORTestimonyFigure2

Although some authorities[4] have claimed that the scale of the Fed’s reserve creation alone made a corresponding increase in bank holdings of excess reserves inevitable, that is not correct.  Although the total quantity of bank reserves is largely determined by the Fed’s rather than commercial bankers’ decisions, banks are always capable in principle of reducing their holdings of excess reserves by swapping them for other assets.  Although the swapping does not destroy reserves, it does result in overall growth in the quantity of bank deposits, together with a corresponding increase in required reserves and a like reduction in excess reserves.  Until the third quarter of 2008 this process kept bank excess reserves roughly constant despite steady growth in total Federal Reserve Bank assets and the monetary base; and it might have done the same afterwards had circumstances not been such as to encourage banks to accumulate excess reserves.  Nor did the tremendous scale of the Fed’s asset purchases itself matter: During the notorious Weimar hyperinflation, for example, the growth in total bank reserves far exceeded that witnessed in the U.S. since Lehman’s bankruptcy.  Yet Germany’s banks, instead of accumulating excess reserve, increased their lending and deposit creation proportionately, and eventually more than proportionately, with terrible consequences.

Nor is U.S. banks’ decision to accumulate excess reserves attributable to the panic that followed the Fed’s decision to allow Lehman Brothers to go bankrupt.[5]  Although banks’ fear that their counterparties might be allowed to go bankrupt would make them reluctant to lend to other banks, it alone would not necessarily cause them to decisively favor reserves over low-risk Treasury securities.  Furthermore, as Figure 3 shows, although the TED spread — a widely-used measure of the perceived risk of bank failures, equal the difference between the interest rate on short-term interbank lending and the interest rate on Treasury securities — spiked not long after Lehman's failure, the spread returned to normal levels afterwards, mainly in response to the Fed’s decision to rescue AIG, while banks’ excess reserve holdings did not.  The persistent increase in bank holdings of excess reserves suggest that the payment of interest on such reserves, rather than banks’ reassessment of the risk of counterparty failures, is behind the increase.

Figure 3

SelginIORTestimonyFigure3

Finally, the timing of the substantial rise in banks’ excess reserve holdings, as shown in the next chart, is also consistent with the view that the Fed’s policy of paying interest on excess reserves contributed more to the increase than Lehman’s failure did.  As Figure 4 shows, although banks accumulated excess reserves immediately following Lehman’s failure, most of the increase in excess reserves occurred after the Fed began paying interest on reserves.

Figure 4

SelginIORTestimonyFigure4

Some experts doubt that the very modest return on excess reserves — for most of the period between October 2008 and December 2015 the rate of interest on excess reserves was fixed at just 25 basis points — can have sufficed to induce banks to hoard reserves.  However, banks’ willingness to hold excess reserves depends, not on the absolute return on such reserves, but on how that return compares to the return on alternative liquid and risk-free assets, such as Treasury bills.  As Figure 5 shows, the interest rate on excess reserves has generally exceeded the yield on Treasury bills.  The same figure shows how the volume of interbank loans has tended to vary according to the difference between the rate of interest on excess reserves and the yield on Treasury securities, which can be regarded here as a proxy for market rates more generally.

Figure 5

SelginIORTestimonyFigure5

Because reserves began to bear a higher return than safe governments securities, the demand for those securities did not increase substantially after Lehman’s failure (Figure 6).

Figure 6

SelginIORTestimonyFigure6

As I’ve noted, a desire to prevent its emergency lending from contributing to the availability of federal funds supplied the original inspiration for the Fed’s decision to begin paying interest on bank reserves, so it is no surprise that the policy should have been responsible for the actual decline in interbank lending that took place after Lehman’s failure.  Once they were able to earn interest on their excess reserves exceeding the effective federal funds rate, banks (mainly smaller ones) that until the crisis had generally been net interbank lenders, withdrew from that market, while those (mainly larger ones) that had previously tended to participate as borrowers found it both necessary and no longer onerous to hold substantial quantities of excess reserves instead.

Besides contributing to the collapse in interbank lending, the Fed’s decision to reward banks for holding excess reserves prevented the creation of additional reserves from giving rise to corresponding growth in other kinds of bank credit by short-circuiting of the base-money “multiplier” that normally connects growth in bank reserves to more substantial growth in bank deposits.  As the Figure 7 shows, the M1 multiplier, the ratio of M1 (currency in circulation plus demand deposits) to the monetary base (currency in circulation plus total bank reserves) fell from 1.617 on September 10th to half that value by the beginning of 2010.

Figure 7

SelginIORTestimonyFigure7

The collapse of the money multiplier was in turn responsible for the failure of the Fed’s large-scale asset purchases to give rise to any corresponding increase in bank deposits, bank credit, and nominal GDP.  Instead, banks’ holdings of excess reserves grew almost in lock-step with the Fed’s creation of new base money.  Had banks not been rewarded for holding excess reserves, a much smaller program of Quantitative Easing might have given rise to a much more substantial increase in bank deposits, bank lending, and nominal GDP.

Partly owing to the repressive effect of interest on reserves on bank deposit creation, most forms of bank lending, instead of being revived by the Fed’s creation of fresh bank reserves, remained stagnant or (in the case of Commercial and Industrial Loans) continued to decline long after Lehman’s failure.  Commercial and Financial Lending declined until the third quarter of 2010, as seen in Figure 8.  And although it has made up for lost ground since, it remains well below the level consistent with its pre-boom trend.  Moreover, because the crisis resulted in a large and lasting decline in net “shadow” bank lending to non-financial firms,[6] especially by Money Market Mutual Funds, much of the revival in commercial bank lending has  consisted of lending to corporate borrowers that had previously relied upon funding from shadow banks.  Lending to small businesses has suffered correspondingly.

Figure 8

SelginIORTestimonyFigure8

Banks’ unprecedented accumulation of excess reserves has as its counterpart a very large Fed balance sheet relative to both overall economic activity and private lending.  As Figure 9 shows, the increase relative to GDP is the largest since the World War II era, when the Fed was committed to setting a floor on the governments’ wartime borrowing costs by serving as a “last resort” purchaser of its bonds.[7]  That commitment finally ended with the so-called “Treasury Accord” of 1951.[8]  Although the Fed’s balance sheet reached its highest historical level relative to GDP during the Great Depression, that record mainly reflected that era’s extreme drop in GDP, as opposed to growth in the absolute size of the Fed’s balance sheet.

Figure 9

SelginIORTestimonyFigure9

Substantial growth in the Fed’s balance sheet, combined with the incomplete revival of bank lending since the crisis, has caused the Fed’s overall share of bank-based financial intermediation to triple, as seen in Figure 10:

Figure 10

SelginIORTestimonyFigure10

Such a large increase in the Fed’s role in the allocation of scarce savings is much to be regretted, as it almost certainly means that those savings are not being devoted to their most productive or welfare-enhancing uses.  At best central banks are inefficient financial intermediaries, not the least because efficient intermediation forms no part of their official responsibilities. Instead, their acquisition of interest-earning assets is supposed to be incidental to their tasks of regulating overall monetary conditions and serving as lenders of last resort.  They are, furthermore, generally supposed to avoid exposing themselves — and, indirectly, taxpayers — to loss, and are for that reason expected to fully secure their last-resort loans and to limit their outright asset purchases to safe government securities.  Commercial banks, in contrast, are not similarly constrained, and cannot be if they are to take full advantage of opportunities for productive lending.

Until the recent crisis, the Fed was no exception to the general rules governing central banks.  Before early 2008 Fed assets consisted overwhelmingly of U.S. Treasury bills, notes, and bonds. Since the crisis, however, the Fed’s asset holdings have changed considerably, in ways that generally involve still greater departures from any efficient use of scarce funds, including a substantial increase in MBS holdings and long-term Treasury securities acquired during several rounds of Quantitative Easing (Figure 11):

Figure 11

SelginIORTestimonyFigure11

Although the Fed’s crisis-related asset purchases may have been instrumental in combating the panic and subsequent recession, its continued holding of non-traditional assets long afterwards constitutes a serious distortion in the allocation of scarce capital, including a perpetuation of the very misallocations of which irresponsible private lenders (encouraged in many cases by government policies[9]) were guilty in the years leading to the crisis.

Despite the counterproductive consequences of the Fed’s original decision to employ interest payments on bank reserves as an instrument of monetary control, and the inefficient allocation of savings to which banks’ hoarding of excess reserves contributes, the Fed continues, seven and a half years since the crisis, not only to rely on that new instrument, but to rely on it and changes in the interest rate it offers in its overnight reverse repurchase agreements (ON RRPs) exclusively for monetary control purposes, while dispensing entirely with traditional open market operations.  Its decision to do so, and more specifically, to maintain a positive rate of interest on excess reserves, and even to increase that rate (as it did in mid-December 2015), is to be regretted.

The December rate hike itself appears in retrospect to replicate the Fed’s error of October 2008, when it employed interest on reserves to avoid an unwanted loosening of credit, on the grounds that such a loosening might prevent it from achieving its policy targets.  In electing last December to raise the interest rate paid on excess reserves from 25 to 50 basis points, the FOMC pointed to a “considerable improvement in labor market conditions,” while declaring that it was “reasonably confident that inflation will rise, over the medium term, to its 2 percent objective.”[10]  As of this writing, both core and headline PCE inflation remain below the Fed’s 2% target, while the unemployment rate is again at 5%, its level in October 2015. Many observers have since concluded that the December rate hike was a mistake.

However, it would be more accurate to claim that, while the December doubling of the rate of interest paid on excess reserves was a mistake, the decision to pay 25 basis points on those reserves was a mistake as well.  As David Beckworth has put it in a blogpost on the topic, “The Fed…got ahead of the recovery well before December.”[11]  According to the Fed’s own estimates, as seen in Figure 12 below, the “natural” fed funds rate, which is the rate consistent with a stable level of spending growth and inflation, has been persistently negative since Lehman went bankrupt.  Consequently, in setting a positive funds rate target band, the upper bound of which was determined by the interest rate on excess reserves, the Fed maintained an excessively tight policy.

Figure 12

SelginIORTestimonyFigure12

It is owing to the perception that natural rates in their own struggling economies are also negative that several foreign central banks, including the ECB and the central banks of Denmark, Sweden, Switzerland, and, starting in January this year, Japan, have turned to charging rather than paying interest on bank excess reserve holdings.  The step has been controversial, and its consequences have not clearly fulfilled the hopes of those central bankers that have resorted to it.  However, regardless of its merits the policy turn raises obvious questions concerning the Fed’s decision to continue pursuing its opposite strategy.

Besides contributing to what may have been an excessively tight policy stance, the continuation of interest payments on excess reserves also serves to perpetuate the Fed’s unusually heavy involvement in the allocation of savings, and the consequent mal-investment of those savings.

The alternative to continuing the present policy is, of course, to dispense with interest payments on excess reserves while restoring conventional open market operations as the Fed’s primary instrument of monetary control.  Restoring efficient credit allocation in turn means reducing the size of the Federal Reserve’s balance sheet both absolutely and relative to that of private intermediaries.

For the Fed to do all of these things while maintaining a proper monetary policy will be challenging.  But for it to avoid taking these steps is for it to continue to contribute to the economic malaise that has made for a slow and still unsatisfactory recovery from the 2008 crisis.  And although the task of normalizing monetary policy may be difficult, it is hardly impossible.  The phasing-out of interest on excess reserves, together with the lowering of interest payments on ON RPPs, will help to revive the money multiplier, thereby not just allowing but necessitating a compensating unwinding of the Fed’s post-crisis balance sheet.  If it isn’t to disrupt markets the unwinding must be both gradual and anticipated: one proposal would have the Fed begin by committing to sell $4-$5 billion in short-term Treasuries each week.[12]  Such a sale would, incidentally, more than make up for the reduction in Fed interest payments to Money Market Funds, by returning to the marketplace securities that such funds have long been craving.

Having the Fed return to its pre-crisis policy of zero interest on excess reserves does not mean forgetting the arguments that supported the 2006 legislation that originally granted the Fed the right to pay interest on reserves.  However, meeting the spirit of those arguments requires only that the Fed be able to pay interest on banks’ required, as opposed to their excess, reserves.  So long as excess reserves bear no interest, banks have little reason to accumulate them, and would therefore suffer little from the inefficiency connected to their slight holdings.  Economic efficiency is in any case better enhanced by encouraging banks to put excess reserves to use, than by paying them to hoard such reserves.

___________________

[1] (http://www.federalreserve.gov/newsevents/speech/bernanke20081007a.htm)

[2] (Courage to Act, pp. 325-6).

[3]https://www.richmondfed.org/~/media/richmondfedorg/publications/research/economic_brief/2009/pdf/eb_09-12.pdf

[4] See, for example, Todd Keister and Gaetano Antinolfi, http://libertystreeteconomics.newyorkfed.org/2012/08/interest-on-excess-reserves-and-cash-parked-at-the-fed.html#.Vzog6ORGR2A

[5] This claim has been put forward by Alex Cukierman, among others.  See “U.S. Banks’ Behavior since Lehman’s Collapse, Bailout Uncertainly and the Timing of Exit Strategies.” Working paper, August 30, 2014.

[6] See Joshua Gallin, “Shadow Banking and the Funding of the Nonfinancial Sector.” Working paper 2013:50, Federal Reserve Board.

[7] The chart comes from Lowell R. Ricketts and Christopher J. Waller, “The Rise and (Eventual) Fall in the Fed’s Balance Sheet.” The Regional Economist, January 2014, Federal Reserve Bank of St. Louis.

[8] A still larger ratio during the Great Depression mainly reflected the tremendous GDP collapse of that episode rather than of absolute growth in the Fed’s size.

[9] See Peter Wallison, Hidden in Plain Sight: What Really Caused the World’s Worst Financial Crisis and Why It Could Happen Again. New York: Encounter Books, 2015.

[10] https://www.federalreserve.gov/newsevents/press/monetary/20151216a.htm

[11] http://macromarketmusings.blogspot.com/2016/02/the-fed-did-not-make-mistake-in-december.html

[12] http://www.ft.com/cms/s/0/520377e8-037e-11e5-b55e-00144feabdc0.html#axzz48p3YArDG.  See also Norbert Michel (http://www.heritage.org/research/reports/2014/08/quantitative-easing-the-feds-balance-sheet-and-central-bank-insolvency), who proposes that the Fed take until 2020 to sell 75% of its long-term securities and MBS, at a rate of $45 billion each month, while holding the other 25% until they mature.

The post Interest on Reserves and the Fed's Balance Sheet appeared first on Alt-M.

]]>
http://www.alt-m.org/2016/05/17/interest-on-reserves-fed-balance-sheet/feed/ 57
A Monetary Policy Primer, Part 4: Stable Prices or Stable Spending? http://www.alt-m.org/2016/05/16/monetary-policy-primer-part-4-stable-prices-vs-stable-spending/ http://www.alt-m.org/2016/05/16/monetary-policy-primer-part-4-stable-prices-vs-stable-spending/#comments Mon, 16 May 2016 13:17:45 +0000 http://www.alt-m.org/?p=39917 Changes in the general level of prices are capable, as we've seen, of eliminating shortages or surpluses of money, by adding to or subtracting from the purchasing power of existing money holdings.  But because such changes place an extra burden on the price system, increasing the likelihood that individual prices...

The post A Monetary Policy Primer, Part 4: Stable Prices or Stable Spending? appeared first on Alt-M.

]]>
NoisyPrices2Changes in the general level of prices are capable, as we've seen, of eliminating shortages or surpluses of money, by adding to or subtracting from the purchasing power of existing money holdings.  But because such changes place an extra burden on the price system, increasing the likelihood that individual prices will fail to accurately reflect the true scarcity of different goods and services at any moment, the less they have to be relied upon, the better.  A better alternative, if only it can somehow be achieved, or at least approximated, is a monetary system that adjusts the stock of money in response to changes in the demand for money balances, thereby reducing the need for changes in the general level of prices.

Please note that saying this is not saying that we need to have a centrally-planned money supply, let alone one that's managed by a committee that's unconstrained by any explicit rules or commitments.  Whether such a committee would in fact come closer to the ideal I'm defending than some alternative arrangement is a crucial question we must come to later on.  For now I will merely observe that, although it's true that unconstrained central monetary planners might manage the money stock according to some ideal, that's only so because there's nothing that such planners might not do.

The claim that an ideal monetary regime is one that reduces the extent to which changes in the general level of prices are required to keep the quantity of money supplied in agreement with the quantity demanded might be understood to imply that what's needed to avoid monetary troubles is a monetary system that avoids all changes to the general level of prices, or one that allows that level to change only at a steady and predictable rate.  We might trust a committee of central bankers to adopt such a policy.  But then again, we could also insist on it, by eliminating their discretionary powers in favor of having them abide by a strict stable price level (or inflation rate) mandate.

Monetary and Non-Monetary Causes of Price-Level Movements

But things aren't quite so simple.  For while changes in the general price level are often both unnecessary and undesirable, they aren't always so.  Whether they're desirable or not depends on the reason for the change.

This often overlooked point is best brought home with the help of the famous "equation of exchange," MV = Py.   Here, M is the money stock, V is its "velocity" of circulation, P is the price level, and y is the economy's real output of goods and services.  Since output is a flow, the equation necessarily refers to an interval of time.  Velocity can then be understood as representing how often a typical unit of money is traded for output during that interval.  If the interval is a year, then both Py and MV stand for the money value of output produced during that year or, alternatively, for that years' total spending.

From this equation, it's apparent that changes in the general price level may be due to any one of three underlying causes: a change in the money stock, a change in money's velocity, or a change in real output.

Once upon a time, economists (or some of them, at least) distinguished between changes in the price level made necessary by developments in the "goods" side of the economy, that is, by changes in real output that occur independently of changes in the flow of spending, and those made necessary by changes in that flow, that is, in either the stock of money or its velocity.   Deflation — a decline in the equilibrium price level — might, for example, be due to a decline in the stock of money, or in its velocity, either of which would mean less spending on output.  But it could also be due to a greater abundance of goods that, with spending unchanged, must command lower prices.  It turns out that, while the first sort of deflation is something to be regretted, and therefore something that an ideal monetary system should avoid, the second isn't.  What's more, attempts to avoid the second, supply-driven sort of deflation can actually end up doing harm.  The same goes for attempts to keep prices from rising when the underlying cause is, not increased spending, but reduced real output of goods and services.  In short, what a good monetary system ought to avoid is, not fluctuations in the general price level or inflation rate per se, but fluctuations in the level or growth rate of total spending.

Prices Adjust Readily to Changes in Costs

But what about those "sticky" prices?  Aren't they a reason to avoid any need for changes in the price level, and not just those changes made necessary by underlying changes in spending?  It turns out that they aren't, for a number of reasons.[1]

First of all, whether a price is "sticky" or not depends on why it has to adjust.  When, for example, there's a general decline in spending, sellers have all sorts of reasons to resist lowering their prices.  If the decline might be temporary, sellers would be wise to wait and see before incurring price-adjustment costs.  Also, sellers will generally not profit by lowering their prices until their own costs have also been lowered, creating what Leland Yeager calls a "who goes first" problem.  Because the costs that must themselves adjust downwards in order for sellers to have a strong motive to follow suit include very sticky labor costs, the general price level may take a long time "groping" its way (another Yeager expression) to its new, equilibrium level.  In the meantime, goods and services, being overpriced, go unsold.

When downward pressure on prices comes from an increase in the supply of goods, and especially when the increase reflects productivity gains, the situation is utterly different.  For gains in productivity are another name for falling unit costs of production; and for competing sellers to reduce their products' prices in response to reduced costs is relatively easy.  It is, indeed, something of a no-brainer, because it promises to bring a greater market share, with no loss in per-unit profits.  Heck, companies devote all sorts of effort to being able to lower their costs precisely so that they can take advantage of such opportunities to profitably lower their prices.  By the same token, there is little reason for sellers to resist raising prices in response to adverse supply shocks. The widespread practice of "mark-up pricing" supplies ample proof of these claims.  Macroeconomic theories and models (and there are plenty of them, alas) that simply assign a certain "stickiness" parameter to prices, without allowing for the possibility that they respond more readily to some underlying changes than to others, lead policymakers astray by overlooking this important fact.

A Changing Price Level May be Less "Noisy" Than a Constant One

Because prices tend to respond relatively quickly to productivity gains and setbacks, there's little to be gained by employing monetary policy to prevent their movements related to such gains or setbacks.  On the contrary: there's much to lose, because productivity gains and losses tend to be uneven across firms and industries, making any resulting change to the general price level a mere average of quite different changes to different equilibrium prices.  Economists' tendency — and it hard to avoid — to conflate a "general" movement in prices, in the sense of a change in their average level, with a general movement in the across-the-board sense, is in this regard a source of great mischief.  A policy aimed at avoiding what is merely a change in the average, stemming from productivity innovations, increases instead of reducing the overall burden of price adjustment, introducing that much more "noise" into the price system.

Nor is it the case that a general decline or increase in prices stemming from productivity gains or setbacks itself conveys a noisy signal.  On the contrary: if things are generally getting cheaper to produce, a falling price level conveys that fact of reality in the most straightforward manner possible.  Likewise, if productivity suffers — if there is a war or a harvest failure or OPEC-inspired restriction in oil output or some other calamity — what better way to let people know, and to encourage them to act economically, than by letting prices generally go up?  Would it really help matters if, instead of doing that, the monetary powers-that-be decided to shrink the money stock, and thereby MV, for the sake of keeping the price level constant?  Yet that is what a policy of strict price-level stability would require.

Reflection on such scenarios ought to be enough to make even the most die-hard champion of price-level or inflation targeting reconsider.  But in case it isn't, allow me to take still another tack, by observing that, when policymakers speak of stabilizing the price level or the rate of inflation, they mean stabilizing some measure of the level of output prices, such as the Consumer Price Index, or the GDP deflator, or the current Fed favorite, the PCE ("Personal Consumption Expenditure") price-index.  So long as changes in total spending ("aggregate demand") are the only source of changes in the overall level of prices,  those changes will tend to affect input as well as output prices, so policies that stabilize output prices will also tend to stabilize input prices.   General changes in productivity, in contrast, necessarily imply changes in the relation of input to output prices: general productivity gains (meaning gains in numerous industries that outweigh setbacks in others) mean that output prices must decline relative to input prices; while general productivity setbacks mean that output prices must increase relative to input prices.  In such cases, to stabilize output prices is to destabilize input prices, and vice versa.

So, which?  Appeal to menu costs supplies a ready answer: if a burden of price adjustment there must be, let the burden fall on the least sticky prices.  Since "input" prices include wages and salaries, that alone makes a policy that would impose the burden on them a poor choice, and a dangerous one at that.  As we've seen, it means adding insult to injury during productivity setbacks, when wage earners would have to take cuts (or settle for smaller or less frequent raises).  It also increases the risk of productivity gains being associated with asset-price bubbles, because those gains will inspire corresponding boosts to aggregate demand which, in the presence of sticky input prices, can cause profits to swell temporarily.  Unless the temporary nature of the extraordinary profits is recognized, asset prices will be bid up, but only for as long as it takes for costs to clamber their way upwards in response to the overall increase in spending.

What About Debtor-Creditor Transfers?

But if the price level is allowed to vary, and to vary unexpectedly, doesn't that mean that the terms of fixed-interest rate contracts will be distorted, with creditors gaining at debtors expense when prices decline, and the opposite happening when they rise?

Usually it does; but, when price-level movements reflect underlying changes in productivity, it doesn't.  That's because productivity changes tend to be associated with like changes in  "neutral" or "full information" interest rates.  Suppose that, with each of us anticipating a real return on capital of four percent, and zero inflation, I'd happily lend you, and you'd happily borrow, $1000 at four percent interest.  The anticipated real interest rate is of course also four percent.  Now suppose that productivity rises unexpectedly, raising the actual real return on capital by two percentage points, to six percent rather than four percent.  In that case, other things equal, were I able to go back and renegotiate the contract, I'd want to earn a real rate of six percent, to reflect the higher opportunity cost of lending.  You, on the other hand, can also employ your borrowings more productively, or are otherwise going to be able (as one of the beneficiaries of the all-around gain in productivity) to bear a greater real interest-rate burden, other things equal, and so should be willing to pay the higher rate.

Of course, we can't go back in time and renegotiate the loan.  So what's the next best thing?  It is to let the productivity gains be reflected in proportionately lower output prices — that is, in a two percent decline in the the price level over the course of the loan period — and thus in an increase, of two percentage points, in the real interest rate corresponding to the four percent nominal rate we negotiated.

The same reasoning applies, mutatis mutandis, to the case of unexpected, adverse changes in productivity.  Only the argument for letting the price level change in this case, so that an unexpected increase in prices itself compensates for the unexpected decline in productivity, is even more compelling.  Why is that?  Because, as we've seen, to keep the price level from rising when productivity declines, the authorities would have to shrink the flow of spending.  Ask yourself whether doing that will make life easier or harder for debtors with fixed nominal debt contracts, and you'll see my point.

Next: The Supply of Money

_____________________________

[1] What follows is a brief summary of arguments I develop at greater length in my 1997 IEA pamphlet, Less Than Zero.  In that pamphlet I specifically make the case for a rate of deflation equal to an economies (varying) rate of total factor productivity growth.  But the arguments may just as well be read as supplying grounds for preferring a varying yet generally positive inflation rate to a constant rate.

The post A Monetary Policy Primer, Part 4: Stable Prices or Stable Spending? appeared first on Alt-M.

]]>
http://www.alt-m.org/2016/05/16/monetary-policy-primer-part-4-stable-prices-vs-stable-spending/feed/ 8
Two Cheers for the Leverage Ratio http://www.alt-m.org/2016/05/12/two-cheers-for-the-leverage-ratio/ http://www.alt-m.org/2016/05/12/two-cheers-for-the-leverage-ratio/#comments Thu, 12 May 2016 13:14:08 +0000 http://www.alt-m.org/?p=40641 In a previous blog posting, I suggested that there is no case for capital adequacy regulation in an unregulated banking system.  In this ‘first-best’ environment, a bank’s capital policy would be just another aspect of its business model, comparable to its lending or reserving policies, say.  Banks’ capital adequacy standards...

The post Two Cheers for the Leverage Ratio appeared first on Alt-M.

]]>
BeautyContestIn a previous blog posting, I suggested that there is no case for capital adequacy regulation in an unregulated banking system.  In this ‘first-best’ environment, a bank’s capital policy would be just another aspect of its business model, comparable to its lending or reserving policies, say.  Banks’ capital adequacy standards would then be determined by competition and banks with inadequate capital would be driven out of business.

Nonetheless, it does not follow that there is no case for capital adequacy regulation in a ‘second-best’ world in which pre-existing state interventions — such as deposit insurance, the lender of last resort and Too-Big-to-Fail — create incentives for banks to take excessive risks.  By excessive risks, I refer to the risks that banks take but would not take if they had to bear the downsides of those risks themselves.

My point is that in this ‘second-best’ world there is a ‘second-best’ case for capital adequacy regulation to offset the incentives toward excessive risk-taking created by deposit insurance and so forth.  This posting examines what form such capital adequacy regulation might take.

At the heart of any system of capital adequacy regulation is a set of minimum required capital ratios, which were traditionally taken to be the ratios of core capital[1] to some measure of bank assets.

Under the international Basel capital regime, the centerpiece capital ratios involve a denominator measure known as Risk-Weighted Assets (RWAs).  The RWA approach gives each asset an arbitrary fixed weight between 0 percent and 100 percent, with OECD government debt given a weight of a zero.  The RWA measure itself is then the sum of the individual risk-weighted assets on a bank’s balance sheet.

The incentives created by the RWA approach turned Basel into a game in which the banks loaded up on low risk-weighted assets and most of the risks they took became invisible to the Basel risk measurement system.

The unreliability of the RWA measure is apparent from the following chart due to Andy Haldane:

Figure 1: Average Risk Weights and Leverage

Average RWAs

This chart shows average Basel risk weights and leverage for a sample of international banks over the period 1994–2011.  Over this period, average risk weights show a clear downward trend, falling from just over 70 percent to about 40 percent.  Over the same period, bank leverage or assets divided by capital — a simple measure of bank riskiness — moved in the opposite direction, rising from about 20 to well over 30 at the start of the crisis.  The only difference is that while the latter then reversed itself, the average risk weight continued to fall during the crisis, continuing its earlier trend.  “While the risk traffic lights were flashing bright red for leverage [as the crisis approached], for risk weights they were signaling ever-deeper green,” as Haldane put it: the risk weights were a contrarian indicator for risk, indicating that risk was falling when it was, in fact, increasing sharply.[2]  The implication is that the RWA is a highly unreliable risk measure.[3]

Long before Basel, the preferred capital ratio was core capital to total assets, with no adjustment in the denominator for any risk-weights.  The inverse of this ratio, the bank leverage measure mentioned earlier, was regarded as the best available indicator of bank riskiness: the higher the leverage, the riskier the bank.

These older metrics then went out of fashion.  Over 30 years ago, it became fashionable to base regulatory capital ratios on RWAs because of their supposedly greater ‘risk sensitivity.’  Later the risk models came along, which were believed to provide even greater risk sensitivity.  The old capital/assets ratio was now passé, dismissed as primitive because of its risk insensitivity.  However, as RWAs and risk models have themselves become discredited, this risk insensitivity is no longer the disadvantage it once seemed to be.

On the contrary.

The old capital to assets ratio is making a comeback under a new name, the leverage ratio:[4] what is old is new again.  The introduction of a minimum leverage ratio is one of the key principles of the Basel III international capital regime.  Under this regime, there is to be a minimum required leverage ratio of 3 percent to supplement the various RWA-based capital requirements that are, unfortunately, its centerpieces.

The banking lobby hate the leverage ratio because it is less easy to game than RWA-based or model-based capital rules.  They and their Basel allies then argue that we all know that the RWA measure is flawed, but we shouldn’t throw out the baby with the bathwater.  (What baby? I ask. RWA is a pretend number and it’s as simple as that.)  They then assert that the leverage ratio is also flawed and conclude that we need the RWA to offset the flaws in the leverage ratio.

The flaw they now emphasize is the following: a minimum required leverage ratio would encourage banks to load up on the riskiest assets because the leverage ratio ignores the riskiness of individual assets.  This argument is commonly made and one could give many examples.  To give just one, a Financial Times editorial — ironically entitled “In praise of bank leverage ratios” — published on July 10, 2013 stated flatly:

Leverage ratios …  encourage lenders to load up on the riskiest assets available, which offer higher returns for the same capital.

Hold on right there!  Those who make such claims should think them through: if the banks were to load up on the riskiest assets, we first need to consider who would bear those higher risks.

The FT statement is not true as a general proposition and it is false in the circumstances that matter, i.e., where what is being proposed is a high minimum leverage ratio that would internalize the consequences of bank risk-taking.  And it is false in those circumstances precisely because it would internalize such risk-taking.

Consider the following cases:

In the first, imagine a bank with an infinitesimal capital ratio.  This bank benefits from the upside of its risk-taking but does not bear the downside.  If the risks pay off, it gets the profit; but if it makes a loss, it goes bankrupt and the loss is passed to its creditors.  Because the bank does not bear the downside, it has an incentive to load up on the riskiest assets available in order to maximize its expected profit.  In this case, the FT statement is correct.

In the second case, imagine a bank with a high capital-to-assets ratio.  This bank benefits from the upside of its risk-taking but also bears the downside if it makes a loss.  Because the bank bears the downside, it no longer has an incentive to load up on the riskiest assets.  Instead, it would select a mix of low-risk and high-risk assets that reflected its own risk appetite, i.e., its preferred trade-off between risk and expected return.  In this case, the FT statement is false.

My point is that the impact of a minimum required leverage ratio on bank risk-taking depends on the leverage ratio itself, and that it is only in the case of a very low leverage ratio that banks will load up on the riskiest assets.  However, if a bank is very thinly capitalized then it shouldn’t operate at all.  In a free-banking system, such a bank would lose creditors’ confidence and be run out of business.  Even in the contemporary United States, such a bank would fall foul of the Prompt Corrective Action statutes and the relevant authorities would be required to close it down.

In short, far from encouraging excessive risk-taking as is widely believed, a high minimum leverage ratio would internalize risk-taking incentives and lead to healthy rather than excessive risk-taking.

Then there is the question of how high ‘high’ should be.  There is of course no single magic number, but there is a remarkable degree of expert consensus on the broad order of magnitude involved.  For example, in an important 2010 letter to the Financial Times drafted by Anat Admati, she and 19 other renowned experts suggested a minimum required leverage ratio of at least 15 percent — at least five times greater than under Basel III — and some advocate much higher minima.  Independently, John Allison, Martin Hutchinson, Allan Meltzer and yours truly have also advocated minimum leverage ratios of at least 15 percent.  By a curious coincidence, 15 percent is about the average leverage ratio of U.S. banks at the time the Fed was founded.

There is one further and much under-appreciated benefit from a leverage ratio.  Suppose we had a leverage ratio whose denominator was not total assets or some similar measure.  Suppose instead that its denominator was the total amount at risk: one would take each position, establish the potential maximum loss on that position, and take the denominator to be the sum of these potential losses.  A leverage-ratio capital requirement based on a total-amount-at-risk denominator would give each position a capital requirement that was proportional to its riskiness, where its riskiness would be measured by its potential maximum loss.

Now consider any two positions with the same fair value.  With a total asset denominator, they would attract the same capital requirement, independently of their riskiness.  But now suppose that one position is a conventional bank asset such as a commercial loan, where the most that could be lost is the value of the loan itself.  The other position is a long position in a Credit Default Swap (i.e., a position in which the bank sells credit insurance).  If the reference credit in the CDS should sharply deteriorate, the long position could lose much more than its current value.  Remember AIG! Therefore, the CDS position is much riskier and would attract a much greater capital requirement under a total-amount-at-risk denominator.

The really toxic positions would be revealed to be the capital-hungry monsters that they are.  Their higher capital requirements would make many of them unattractive once the banks themselves were to made to bear the risks involved.  Much of the toxicity in banks’ positions would soon disappear.

The trick here is to get the denominator right.  Instead of measuring positions by their accounting fair values as under, e.g., U.S. Generally Accepted Accounting Principles, one should measure those positions by how much they might lose.

Nonetheless, even the best-designed leverage ratio regime can only ever be a second-best reform: it is not a panacea for all the ills that afflict the banking system.  Nor is it even clear that it would be the best ‘second-best’ reform: re-establishing some form of unlimited liability might be a better choice.

However, short of free banking, under which no capital regulation would be required in the first place, a high minimum leverage ratio would be a step in the right direction.

_____________

[1] By core capital, I refer the ‘fire-resistant’ capital available to support the bank in the heat of a crisis.  Core capital would include, e.g., tangible common equity and some retained earnings and disclosed reserves.  Core capital would exclude certain ‘softer’ capital items that cannot be relied upon in a crisis.  An example of the latter would be Deferred Tax Assets (DTAs).  DTAs allow a bank to claim back tax on previously incurred losses in the event it subsequently returns to profitability, but are useless to a bank in a solvency crisis.

[2] A. G. Haldane, “Constraining discretion in bank regulation.” Paper given at the Federal Reserve Bank of Atlanta Conference on ‘Maintaining Financial Stability: Holding a Tiger by the Tail(s)’, Federal Reserve Bank of Atlanta 9 April 2013, p. 10.

[3] The unreliability of the RWA measure is confirmed by a number of other studies.  These include, e.g.: A. Demirgüç-Kunt, E. Detragiache, and O. Merrouche, “Bank Capital: Lessons from the Financial Crisis,” World Bank Policy Research Working Paper Series No. 5473 2010); A. N. Berger and C. H. S. Bouwman, “How Does Capital Affect Bank Performance during Financial Crises?” Journal of Financial Economics 109 (2013): 146–76; A. Blundell-Wignall and C. Roulet, “Business Models of Banks, Leverage and the Distance-to-Default,” OECD Journal: Financial Market Trends 2012, no. 2 (2014); T. L. Hogan, N. Meredith and X. Pan, “Evaluating Risk-Based Capital Regulation,” Mercatus Center Working Paper Series No. 13-02 (2013); and V. V. Acharya and S. Steffen, “Falling short of expectation — stress testing the Eurozone banking system,” CEPS Policy Brief No. 315, January 2014.

[4] Strictly speaking, Basel III does not give the old capital-to-assets ratio a new name.  Instead, it creates a new leverage ratio measure in which the old denominator, total assets, is replaced by a new denominator measure called the leverage exposure.  The leverage exposure is meant to take account of the off-balance-sheet positions that the total assets measure fails to include.  However, in practice, the leverage exposure is not much different from the total assets measure, and for present purposes one can ignore the difference between the two denominators.  See Basel Committee on Banking Supervision, “Basel III: A global regulatory framework for more resilient banks and banking systems.”  Basel: Bank for International Settlements, June 2011, pp. 62-63.

The post Two Cheers for the Leverage Ratio appeared first on Alt-M.

]]>
http://www.alt-m.org/2016/05/12/two-cheers-for-the-leverage-ratio/feed/ 25
What China Can Teach Us About the Future of Banking http://www.alt-m.org/2016/05/11/what-china-can-teach-us-about-the-future-of-banking/ http://www.alt-m.org/2016/05/11/what-china-can-teach-us-about-the-future-of-banking/#comments Wed, 11 May 2016 13:07:59 +0000 http://www.alt-m.org/?p=40282 A few weeks ago, Citi published a quite fascinating 100-page report on financial innovations, from blockchain to P2P lending, in various regions of the world.  It’s a highly recommended and very comprehensive reading that I won’t be able to summarize in a short blog post. From this report, it is...

The post What China Can Teach Us About the Future of Banking appeared first on Alt-M.

]]>
traditional-piggy-banks (1)
A few weeks ago, Citi published a quite fascinating 100-page report on financial innovations, from blockchain to P2P lending, in various regions of the world.  It’s a highly recommended and very comprehensive reading that I won’t be able to summarize in a short blog post.

From this report, it is clear that China’s financial system has adopted innovations at a much faster pace than the Western world.  And if there is a defining characteristic of the Chinese system, it is its very erratic and repressive regulatory framework, which made me once call China a "spontaneous Frankenstein banking system":

Financial regulation in China is quite a mess.  China seems to be the world testing ground for some of the most ridiculous banking rules.  With all their related unexpected consequences.

In an earlier post, I also highlighted that

China is an interesting case.  Underneath its very tight government-controlled financial repression hide numerous financial experiments aimed at bypassing those very controls.  The Chinese shadow banking system is now a well-known financial Frankenstein, with multiple asset management companies, wealth management products and other off-balance sheet entities providing around half the country’s credit volume.  The more the government tries to regulate the system, the more financial innovation finds new workarounds and become increasingly more opaque.

We already knew that the Chinese financial system was completely distorted from years of regulatory repression and crony capitalism, as a whole new report on finance in China by The Economist demonstrates (see the editorial here, and the report starting here).  Echoing my worries, The Economist calls for China to "free up" its "financial jungle."  Citi’s and The Economist’s reports now allow us to quantify the effects of those distortions.  Indeed, China leads the world in fintech and digital disruption in general; it has some of the largest fintech firms and, as Citi said, it is now "past the tipping point."

While its very large e-commerce has been a strong driver of the rise of alternative payment providers in the country, Citi points at a number of other factors that have facilitated the rise of those third-party payment companies, among which an under-developed banking system viewed by the public as quite unreliable (unsurprising given how tightly controlled banking is in China, which has stifled customer-oriented innovation), and "relaxed regulation."  Citi points out that Chinese regulators have now proposed new tightened rules for the payment sector, so brace yourself for further innovations in this space.  For now, Alipay handles more than three times the volume of transaction that Paypal does, and payment firms have more retail customers than banks have and are now expanding into offline payments.

CitiDigitalDisruptionFigure35

CitiDigitalDisruptionFig39&40

China also has the largest P2P lending market in the world, four times bigger than that of the US.  Citi analysts forecast that P2P loans are going to represent a sizeable 9% of total retail lending in China by 2018.

The driver of this growth is, typically, mostly regulatory constraints on traditional banking that triggered regulatory arbitrage:

P2P lending platforms target segments that are unserved or under-served by existing banking system such as consumer credit and small and micro business lending.  Traditional banks are not particularly good at serving this customer segments due to tougher Know Your Client/Anti-Money Laundering (KYC/AML) requirements as well as tightened lending standard post global financial crisis.

And one would add that capital requirements on certain category of customers (such as SMEs) play a large role here too, as I keep pointing out on this blog (see at the end of this post).  The same reasons are behind the development of such lending platforms in Europe and the US.  And indeed, as Citi writes:

According to China MSME Finance Report 2014 by Mintai Institute of Finance and Banking, almost 80% of SMEs were not served by the banks.  The explosive growth in the P2P lending has met the needs of SMEs which cannot get formal financing.

And:

Chinese banks are under tight regulations such as reserve requirement, loan-to-deposit ratios (LDR), KYC, AML, and so on.  There was however little regulations for the P2P lending sector.  There is also no capital requirement.

Furthermore, Chinese monetary repression is also a driver, as P2P lending allows savers to earn higher returns.  Here again, Chinese regulators are looking at ways to scrutinize and more tightly control the sector.

What are the effects of all this?  As The Economist points out, China’s shadow banking sector is the largest and possibly the fastest growing in the world:

TheEconomistPeoplesRepublicDebt

TheEconomistMinnowGiant

There is a fundamental difference between the Chinese banking system and the Western one however.  Chinese banks, despite being extremely large, have historically had no ability to grow outside of the Communist party’s grip and no ability to adapt to consumer demand as a result.  Citi points out that there were only 8.1 bank branches per 100,000 adults in China, vs. around 30 in the Eurozone and the US.  With little banking presence, fintech firms have found it easy to rapidly grow.

Yet, developed economies do have a lesson to learn from the Chinese experience.  The more regulatory constraints are put in place on banks, the more innovative ways around them will spontaneously emerge and the more complex and opaque ("Frankenstein-like") the financial system will become.  And sadly, it looks like Europe and the US have decided to follow China’s footsteps.

PS: The following chart is revealing.  Most of the financial products that are at most risk of disruption (SME and personal loans, deposits…) are also those that are the most affected by regulatory requirements and low interest rates.

CitiDigitalDisruptionFig25

PPS:  A very good introduction to the Chinese financial mess is Walter’s and Howie’s Red Capitalism. However the book was "only" updated in 2012, and plenty has happened since then, in particular in the fintech portion of China’s shadow banking sector.

_______________

[This article originally appeared on Spontaneous Finance]

The post What China Can Teach Us About the Future of Banking appeared first on Alt-M.

]]>
http://www.alt-m.org/2016/05/11/what-china-can-teach-us-about-the-future-of-banking/feed/ 10
A Monetary Policy Primer, Part 3: The Price Level http://www.alt-m.org/2016/05/10/monetary-policy-primer-part-3-price-level/ http://www.alt-m.org/2016/05/10/monetary-policy-primer-part-3-price-level/#comments Tue, 10 May 2016 13:17:38 +0000 http://www.alt-m.org/?p=39188 Few people would, I think, take exception to the claim that, in a well-functioning monetary system, the quantity of money supplied should seldom differ, and should never differ very much, from the quantity demanded.  What's controversial isn't that claim itself, but the suggestion that it supplies a reason for preferring...

The post A Monetary Policy Primer, Part 3: The Price Level appeared first on Alt-M.

]]>
macdonalds with Text 2Few people would, I think, take exception to the claim that, in a well-functioning monetary system, the quantity of money supplied should seldom differ, and should never differ very much, from the quantity demanded.  What's controversial isn't that claim itself, but the suggestion that it supplies a reason for preferring some path of money supply adjustments over others, or some monetary arrangements over others.

Why the controversy?  As we saw in the last installment, the demand for money ultimately consists, not of a demand for any particular number of money units, but of a demand for a particular amount of monetary purchasing power.  Whatever amount of purchasing X units of money might accomplish, when the general level of prices given by P, ½X units might accomplish equally well, were the level of prices ½P.  It follows that changes in the general level of prices might, in theory at least, serve just as well as changes in the available quantity of money units as a means for keeping the quantity of money supplied in line with the quantity demanded.

But then it follows as well that, if our world is one in which prices are "perfectly flexible," meaning that they always adjust instantly to a level that eliminates any monetary shortage or surplus, any pattern of money supply changes will avoid money supply-demand discrepancies, or "monetary disequilibrium," as well as any other.  The goal of avoiding bouts of monetary disequilibrium would in that case supply no grounds for preferring one monetary system or policy over another, or for preferring a stable level of spending over an unstable level.  Any such preference would instead have to be justified on other grounds.

So, a decision: we can either adopt the view that prices are indeed perfectly flexible, and proceed to ponder why, despite that view, we might prefer some monetary arrangements to others; or we can subscribe to the view that prices are generally not perfectly flexible, and then proceed to assess alternative monetary arrangements according to their capacity to avoid a non-trivial risk of monetary disequilibrium.

Your guide does not hesitate for a moment to recommend the latter course.  For while some prices do indeed appear to be quite flexible, even adjusting almost continually, at least during business hours (prices of goods and financial assets traded on organized exchanges come immediately to mind), in order for the general level of prices to instantly accommodate changes to either the quantity of money supplied or the quantity demanded, it must be the case, not merely that some or many prices are quite flexible, but that all of them are.  If, for example, the nominal stock of money were to double arbitrarily and independently of any change in demand, prices would generally have to double in order for equilibrium to be restored.  (Recall: twice as many units of money will command the same purchasing power as the original amount only when each unit commands half as much purchasing power as before.)  It follows that, so long as any prices are slow to adjust, the price level must be slow to adjust as well.  Put another way, an economy's price level is only as flexible as its least flexible prices.

And only a purblind observer can fail to notice that some prices are far from fully flexible. The reason for this isn't hard to grasp: changing prices is sometimes costly; and when it is, sellers have reason to avoid doing it often.  Economists use the expression "menu costs" to refer generally to the costs of changing prices, conjuring up thereby the image of a restaurateur paying a printer for a batch of new menus, for the sake of accommodating the rising costs of beef, fish, vegetables, wait staff, cooks, and so forth, or the restaurants' growing popularity, or both.  In fact both the restaurants' operating costs and the demand for its output change constantly.  Nevertheless it usually wouldn't make sense to have new menus printed every day, let alone several times a day, to reflect all these fluctuations!  Electronic menus would help, of course, and now it is easy to conceive of them (though it wasn't not long ago).  But those are costly as well, which is why (or one reason why) most restaurants don't use them.

The cost of printing menus is, however, trivial compared to that of changing many other prices. The prices paid for workers, whether wage or salaried, are notoriously difficult to change, except perhaps according to a prearranged schedule, which can't itself accommodate unexpected change.  Renegotiating wages or salaries can be an extremely costly business, as well as a time-consuming one.

"Menu costs" can account for prices being sticky even when the nature of underlying changes in supply or demand conditions is well understood.  Suppose, for example, that a restaurant's popularity is growing at a steady and known rate.  That fact still wouldn't justify having new menus printed every day, or every hour, or perhaps even every week.  But add the possibility that a perceived increase in demand may not last, and the restaurateur has that much more reason to delay ordering new menus: after all, if demand subsides again, the new menus may cost more than turning a few customers away would have.  (The menus might also annoy customers who would dislike not being able to anticipate what their meal will cost.)  Now imagine an employer asking his workers to take a wage rate cut because business was slack last quarter.  Get the idea?  If not, there's a vast body of writings you can refer to for more examples and evidence.

These days it is common for economists who insist on the "stickiness" of the price level to be referred to, or to refer to themselves, as "New Keynesians."  But the label is misleading.  Although John Maynard Keynes had plenty of innovative ideas, the idea that prices aren't perfectly flexible wasn't one of them.  Instead, by 1936, when Keynes published his General Theory, the idea that prices aren't fully flexible was old-hat: no economists worth his or her salt thought otherwise.[1]  The assumption that prices are fully flexible, or "continuously market clearing," is in contrast a relatively recent innovation, having first become prominent in the 1980s with the rise of the "New Classical" school of economists, who subscribe to it, not on empirical grounds, but because they confuse the economists' construct of an all-knowing central auctioneer, who adjusts prices costlessly and continually to their market-clearing levels, with the means by which prices are determined and changed in real economies.

Let New Classical economists ruminate on the challenge of justifying any particular monetary regime in a world of perfectly flexible prices.  The rest of us needn't bother.  Instead, we can accept the reality of "sticky" prices, and let that reality inform our conclusions concerning which sorts of monetary regimes are more likely, and which ones less likely, to avoid temporary surpluses and shortages of money and their harmful consequences.

What consequences are those?  The question is best answered by first recognizing the crucial economic insight that a shortage of money must have as its counterpart a surplus of goods and services and vice versa.  When money, the means of exchange, is in short supply, exchange itself, meaning spending of all sorts, suffers, leaving sellers disappointed.  In contrast, when money is superabundant, spending grows excessively, depleting inventories and creating shortages.

Yet these are only the most obvious consequences of monetary disequilibrium.  Other consequences follow from the fact that, owing to different prices' varying degrees of stickiness, the process of moving from a defunct level of equilibrium prices to a new one necessarily involves some temporary distortion of relative price signals, and associated economic waste.  A price system has work enough to do in coming to grips with ongoing changes in consumer tastes and technology, among many other non-monetary factors that influence supply and demand for particular goods and services, without also having to reckon with monetary disturbances that call for scaling all prices up or down.  The more it must cope with the need to re-scale prices, the less capable it becomes of fine-tuning them to reflect changing conditions within particular markets.

Hyperinflations offer an extreme case in point: during them sellers often resort to "indexing" local-currency prices to the local currency's exchange rate with respect to some relatively stable foreign currency, or to simply posting prices in foreign currency while accepting local currency in payment at the going rate of exchange.  In so doing, they largely cease referring to specific conditions in the markets for their particular goods, settling instead for keeping their prices roughly consistent with overall monetary conditions.  In light of this tendency it's hardly surprising that hyperinflations lead to all sorts of waste, if not to the utter collapse of the economies they afflict.  If relative prices can become so distorted during hyperinflations as to cease entirely to be meaningful indicators of goods' and services' relative scarcity, it's also true that the usefulness of price signals in promoting the efficient use of scarce resources declines to a more modest extent during less severe bouts of  monetary disequilibrium.

What sort of monetary policy or regime best avoids the costs of having too much or too little money?  In an earlier post I suggested that keeping the supply of money in line with the demand for it, without depending on help in the shape of adjustments to the price level,  is mainly a matter of achieving a steady and predictable overall flow of spending.  But why spending?  Why not maintain a stable price level, or a stable and predictable rate of inflation?  If, as I've claimed, changes in the general level of prices are an economy's way of coping, however imperfectly, with monetary shortages and surpluses, then surely an economy in which the price level remains constant, or roughly so, must be one in which such surpluses and shortages aren't occurring.  Right?

No, actually.  Despite everything I've said here, monetary order, instead of going hand-in-hand with a stable level of prices or rate of inflation, is sometimes  best achieved by tolerating price level or inflation rate changes.  A paradox?  Not really.  But as this post is already too long, I must put off explaining why until next time.
______________________
[1] For details see Leland Yeager's essay, "New Keynesians and Old Monetarists," reprinted in The Fluttering Veil.

The post A Monetary Policy Primer, Part 3: The Price Level appeared first on Alt-M.

]]>
http://www.alt-m.org/2016/05/10/monetary-policy-primer-part-3-price-level/feed/ 20
Competition in (British) Banking http://www.alt-m.org/2016/05/06/competition-in-british-banking/ http://www.alt-m.org/2016/05/06/competition-in-british-banking/#comments Fri, 06 May 2016 13:12:32 +0000 http://www.alt-m.org/?p=39577 Writing for FT.com’s “The Exchange” blog, economists Diane Coyle and Jonathan Haskel suggest that Britain’s regulators — namely, the Competition and Markets Authority and the Bank of England — have got it wrong on competition in banking.  The authors argue that “the CMA and the BoE” have overlooked “the ruinous effect on...

The post Competition in (British) Banking appeared first on Alt-M.

]]>
Pink Piggy BankWriting for FT.com’s “The Exchange” blog, economists Diane Coyle and Jonathan Haskel suggest that Britain’s regulators — namely, the Competition and Markets Authority and the Bank of England — have got it wrong on competition in banking.  The authors argue that “the CMA and the BoE” have overlooked “the ruinous effect on competition of the ‘too big to fail’ subsidy” in their recent reports and policy announcements, and that, if anything, it is becoming “harder than ever for new entrants to gain a foothold” in the banking market.

In my view, Coyle and Haskel are right, but their argument doesn’t go far enough.

Let’s start with the basics: what is the too-big-to-fail subsidy, and how does it affect competition in banking?  The fundamental idea is that the bigger a bank is, the more likely it is to be bailed out if it runs into trouble.  The events of the 2008 financial crisis seem to confirm this, as do the assumptions of government assistance that some rating agencies build into their “support” ratings.  And as the 2011 report of Britain’s Independent Commission on Banking points out:

If one bank is seen as more likely to receive government support than another this will give it an unwarranted competitive advantage.  As creditors are assumed to be less likely to take losses, the bank will be able to fund itself more cheaply and so will have a lower cost base than its rival for a reason nothing to do with superior underlying efficiency.

The result is that small banks struggle to compete against larger rivals, while market entrants have difficulty establishing themselves against privileged incumbents.  All of this makes the banking sector less dynamic — and more comprehensively dominated by large, established firms — than it might otherwise be.

As Coyle and Haskel see it, however, Britain’s CMA thinks the problem has already been solved: that the competitive playing field has been leveled by the Bank of England’s proposed “systemic risk buffer,” according to which larger banks must hold more equity capital against their risk-weighted assets than smaller competitors.  In consequence, the CMA’s October 2015 provisional report on Britain’s retail banking market mostly ignored the too-big-to-fail problem, focusing instead on the rather more mundane question of how consumers can be encouraged to switch bank accounts more often.

Yet the CMA’s position is mistaken, say Coyle and Haskel, for three reasons.  First, switching bank accounts doesn’t always make sense for consumers: in the UK, at least, one bank account is pretty much the same as another, so consumers’ status quo bias is often quite rational.  Second, the level of additional capital big banks must hold as a systemic risk buffer is not high enough to outweigh the funding benefits that accrue from being too-big-to-fail.  Third, the stepped schedule of systemic risk buffer requirements outlined by the Bank of England might make big banks less likely to compete with each other, by effectively creating high marginal tax rates when banks move from one “systemic risk buffer” tier to another.  As Coyle and Haskel say, “This might restrain the emergence of gargantuan banks, but the purpose of competition is to promote rivalry, not hold up expansion at arbitrary regulator-determined thresholds.”

So far, so good.  But there’s a bigger picture here that Coyle and Haskel don’t see, or at least fail to mention.  For one thing, it isn’t just lower funding costs that make too-big-to-fail such an anti-competitive doctrine.  In fact, the very act of bailing out a failing institution itself constitutes a powerful strike against market competition.  As Europe Economics’ Andrew Lilico has put it, “company failure is an essential and ineliminable part of the competition process.  One of the most important obstacles to new entry in the banking sector, impeding competition, is that failing banks are saved by the government.”  If you want smaller banks to grow, and new banks to prosper, in other words, you can’t keep saving their bigger rivals from the consequences of bad investments.

More important still are the grounds upon which banks compete.  And it’s here that our financial regulatory authorities have the most to answer for.  Yes—of course—banks should compete with one another to provide the best possible service at the best possible price.  In an ideal world, however, banks would compete on something else as well: namely, their safety, stability, and reliability.  That banks do not tend to compete on these grounds today is testament to the fact that their depositors, bondholders, and shareholders do not see the need to pay attention to such things.  “Regulatory badging,” that illusory sense that banks must be safe because they are subject to regulators’ oversight, means that people seldom ask how highly-leveraged their banks really are.  Deposit insurance means they might not care about the answer, even if they ask the question.  And too-big-to-fail compounds the problem: if your bank is going to be bailed out, why worry about its risk profile?  No amount of regulatory oversight can compensate for this loss of competitive market discipline.

Ultimately, then, Coyle and Haskel are right to stress the importance of competition: if financial stability is the goal, then competition must be central to any banking reform agenda worthy of the name.  But before regulators can be part of the solution, they must understand the ways in which they are part of the problem.  And that, alas, has yet to happen.

The post Competition in (British) Banking appeared first on Alt-M.

]]>
http://www.alt-m.org/2016/05/06/competition-in-british-banking/feed/ 1