In Getting it Wrong1 William Barnett, an aggregation and index number theorist, makes a few important points about monetary policy in the United States, along with a few questionable claims for how to fix it, and some fairly bazaar speculations about why his proposals have received so little traction.
Barnett’s important points are that:
- “The misperception of permanent decrease in volatility [the great moderation] was at the core of the financial crisis and recession.”2 Clearly Wall Street and homeowners took excessive and underpriced risks and became over leveraged.
- “Their decisions [were not] dominated by greed and self-destructive, foolish behaviour.”3 “Were people less ‘greedy’ in the 1920s?”4 “During the 1920s, with less regulation, lower margins requirements, and no shortage of ‘greed’, Wall Street leverage never reached the levels attained prior to the recent crisis.”5 There is nothing “wrong with mainstream economic theory [which says that people make rational self interested choices given the information available to them]. On the contrary, a message of this book is that not enough mainstream economic theory was being used in policy design and data reporting in recent years….”6 “The primary objective of this book is to suggest that poor Federal Reserve data unnecessarily complicated private decision-making and interfered with the ability of the private sector to recognise the extent of the systemic risk existing at the time.”7
- “Once monetary assets began paying different interest rates, simple-sum monetary aggregation [currency + demand deposits + time and savings deposits, etc] became obsolete, and more complicated formulas became valid.”8 However, the Federal Reserve continued and still continues to report monetary aggregates (M1, M2, M3) that are simple sums of their components. “Increasing financial complexity with decreasing data quality is a toxic mix.”9
Getting it Wrong is devoted largely to establishing and elaborating these points. The book is divided into two parts. The first part is written in English for non-specialists and the second part provides the mathematical formula and proofs for the approach to aggregation Barnett is promoting.
A meaningful aggregate for money or anything else cannot be obtained by simply added together the components making up the aggregate if they are not perfect substitutes in their use as money. Barnett repeats a bit too often that proper and well established aggregation and index number theory as practiced, for example by the Bureau of Labor Statistics when producing the Consumer Price Index, or the Federal Reserve when producing its Industrial Production Index, is not being applied by the Federal Reserve when producing its monetary aggregates.
Barnett has developed and champions an approach to aggregation meant to capture the extent to which components of monetary aggregates provided monetary or liquidity services. Some components, such as currency are almost “pure” money (means of payment) but even currency can also be a store of value. A savings deposits with a bank is more obviously a mix of the two. Barnett argues that the extent to which a component of money provides monetary/liquidity services can be measured by the opportunity cost of holding it, its so called user cost (the difference between its own rate of interest and the rate of interest on a non monetary assets that could be held instead).
On that basis, but applying a more complex non linear aggregation methodology, a monetary aggregate of currency and bank deposits, etc, applies a heavier weight to the amount of currency than, for example, to the amount of time deposits paying a rate of interest only two percentage points less than its alternative non monetary benchmark asset.
Barnett’s aggregation methodology, with which he constructions what he calls Divisia aggregates and the Federal Reserves calls MSIs (monetary services indexes), has intuitive appeal, but the proof of the pudding is in the eating. Whether MSIs provide a better basis for monetary policy depends on whether they have a more stable, ie predictable, relationship to the price level – and hence whether controlling the growth rate of money so defined better controls inflation. Barnett complains throughout the book that the Federal Reserve has failed to use his measure. But more on this later, as he might say.
To make his case, ie to provide evidence that MSIs (he has versions using the same components as M2 called MSI2, the now discontinued M3 called MSI3, and the also now discontinued L called MSI4) would have been a better basis for monetary policy, Barnett examines two monetary episodes in some detail and makes a convincing case that his MSIs would have produced better policy.
The first was the period of 1973-74, which captured the attention of economists when the demand for money appeared to shift – the so-called “case of the missing money”. Replacing M2 or M3 with Barnett’s Divisia aggregates (MSI2 or MSI3) largely removed the shift leaving no missing money to explain.
The most dramatic episode examined was the period from October 1979 to September 1982 during which the Fed under Paul Volcker rung inflation that had reach almost 15 per cent per annum out of the economy with a medium-term reduction in average monetary growth. M2 growth dropped from around 12 to 13 per cent per year in 1976-8 to around 7 to 8 per cent per year in 1980-1.
However, by Barnett’s measure policy was much tighter. MSI2 grew on average at 4.5 per cent per annum over this period and actually decline over part of 1981-2. “The simple-sum aggregates’ growth rates were at the intended levels, but the Divisia growth rates were half as large, producing an unintended negative shock of substantially greater magnitude than intended. A deep recession resulted.”10
Having presented his evident, Barnett asks: “Are there any shortcomings in measurement to prevent a central bank from using Divisia monetary aggregates? Is there any reason at all to prefer the disreputable simple-sum monetary aggregates to the state-of-the-art Divisia monetary aggregates? The answer to both questions is one simple unequivocal word – no! In measurement, central banks should do the best they can, not the worst they can. It doesn’t get any worse than simple-sum aggregation.”11
“Do the economists on the Federal Reserve Board’s staff not know they are ‘doing it wrong’? Of course, they know…. So why are those economist producing such poor financial data in flagrant violation of basic principles of index number theory, aggregation theory and elementary accounting? They are rationally pursuing their bureaucracies’ self-interest.”12
When asked why they chose not to adopt his index, Barnett claims that he “never provide[s] a direct answer to that question”.13
He then proceeds to offer a rather lame public choice explanation: “The Fed itself measures and produces the financial and monetary data relevant to monitoring the product that the Fed produces, the liquidity of the economy. Is there a conflict of interests?”14
“If the Fed produces data of poor quality, it accomplishes two things: (1) it cannot be held accountable for policy mistakes, because the world is more mysterious and hence more decisions must be left to the Fed’s ‘discretion….’.”15
These arguments are silly, though Barnett’s recommendation that the Fed set up an independent data unit deserves some consideration. The Fed is accountable for and judged by the extent to which it achieves price stability and full employment (its unfortunate dual mandate). If it is using an inferior indicator of policy or policy target, it is reducing its ability to fulfil its mandates. It has no public choice incentive to do that.
The reason the Fed continues to rely on simple-sum aggregates (M1 and M2), though the Federal Reserve Bank of St Louis computes and publishes MSI1 and MSI2, is that it has carefully examined Barnett’s index, as well as others, and failed to find them empirically superior.16
The primary study found that: “There is little clear improvement in terms of either demand equation or reduced form equation performance of the experimental monetary aggregates as compared to the conventional measures.”17
This result may reflect technical measurement problems with determining component user costs during periods of an inverted yield curve or a rapid change in interest rates, which may subsequently have been over-come by the version of MSI aggregates now published by the St Louis Fed. These revised indexes, however, do not perform much better than conventional simple-sum M1 or M2 during the “missing money” episode in the mid 70s, though their superiority during the Volcker “Practical Monetarist Experiment” at the beginning of the 1980s remains the same. Barnett is fully aware of this and other studies at the Fed but makes no mention of, nor reference to, them. What is his motive for this, as he might ask?
Barnett can be quirky in other ways as well. He seems to think that the Fed’s non-borrowed reserves (NBR), used as a policy target during the “Monetarist Experiment”, are “those bank reserves held by banks, but not borrowed by banks from the Federal Reserve”.18 “It follows immediately from elementary logic… that non-borrowed reserves cannot be negative,”19 he insists. Money is fungible. There is no way of determining what part of bank reserves are borrowed. NBR are properly defined as bank reserves less borrowed reserves, which can be and have been negative (borrowing to finance cash withdrawals from banks). Barnett insists on multiple occasions that negative NBR “is an oxymoron, making the Federal Reserve look silly.”20
In fact, Barnett’s insistence on this point makes him look silly.
On other themes, Barnett provides an interesting history of the main players in the development of index number theory, especially those relevant for monetary policy (Irving Fisher, François Divisia, Henri Theil, Dale Jorgenson, Milton Friedman, W Erwin Diewert and others). He provides good explanations for the basis for his Divisia indices understandable by laymen. His criticisms of calls for policy audits of the Federal Reserve (they are already subject to operational audits) are well founded. “However, expanded audit would be an inferior solution to the creation of an independent data institute.” 21
His inside professional gossip is interesting to me at least having been at many of the same places (UC Berkeley, U of Chicago, Board of Governors of the Federal Reserve) at the same time, though I only met Barnett once at the board.
Above all, Barnett has firmly established the inappropriateness of simple-sum aggregation of the components of “money” after the liberalisation of interest rates and convincingly documented the harm it can do when formulating and implementing monetary policy. Research to find better aggregates should continue.
- William A. Barnett, “Getting It Wrong: How Faulty Monetary Statistics Undermine the Fed, the Financial System, and the Economy,” MIT Press, Cambridge MA, 2012
- Op cit, page 3.
- Op cit, page 5.
- Op cit, page xxv.
- Op cit, page 10.
- Op cit, page 150.
- Op cit, page 133.
- Op cit, page xxviii.
- Op cit, page xxxi.
- Op cit, page 104.
- Op cit, page 51.
- Op cit, page 153.
- Op cit, page 147.
- Op cit, page 35.
- Op cit, page 148.
- David E. Lindsey and Paul Spindt, “An Evaluation of Monetary Indexes,” Special Studies Paper 195, Board of Governors of the Federal Reserve System, March, 1986.
- Op cit, Appendix D-14
- Barnett, page 102, footnote 6
- Op cit, page 130.
- Op cit, page 153.
- Op cit, page 81.