It’s Harder than It Looks

Let me begin with Professor Sumner’s opening assertion that “the sub-prime crisis that began in late 2007 was probably just fluke, and has few important implications for either financial economics or macroeconomics.” I maintain instead that problem loans are indeed the principal cause of our present difficulties.  Ashcraft and Schuermann (2008) analyzed a pool of about 4000 mortgages originated in 2006 by the now-bankrupt New Century Financial Corporation. All the loans in the pool were subprime, meaning one would have significant concerns about the borrowers’ ability to repay on the basis of their credit history. Furthermore, almost all of the loans called for the monthly interest payments to increase between 25 percent and 45 percent within 2-1/2 years, even if there was no change in short-term interest rates.  Yet somehow 79 percent (by dollar value) of the MBS tranches created from this pool were rated AAA by both Standard & Poor’s and Moody’s, and 95 percent were rated at least A. Something was going seriously, seriously wrong in the allocation of credit in 2006.

The only way that a significant fraction of these loans would be repaid would be if house prices continued to rise at the rates of the early 2000s. With rapidly rising house prices, the borrower has every incentive not to default, but instead should refinance and pocket the capital gain. As long as that continued, the low default rates and low correlations among default rates that went into these risk assessments might seem justified. But, when you look at the fundamentals, it was impossible for the house price appreciation to continue.

This pool of New Century loans appears representative of the $1.7 trillion in new subprime loans that were initiated in the United States between 2004 and 2006; (Ashcraft and Schuermann, 2008). Over this period there was an additional trillion dollars in alt-A loans, for which inadequate documentation of borrowers’ incomes or high loan-to-value ratio would also lead one to anticipate significant problems with repayment in an environment of falling house prices. These years also saw $3.6 trillion in new mortgages purchased or “guaranteed” by the government-sponsored enterprises such as Fannie Mae and Freddie Mac. These enterprises did not have anywhere near the capital to make such guarantees credible, and the ultimate perception of the safety of such loans by investors resulted from the assumption that the U.S. government itself would back up the loans if the GSEs could not.

This huge misdirection of capital into the U.S. housing market caused household mortgage debt to more than triple between 1995 and 2007. [1] House prices in the major U.S. metropolitan areas doubled between 2000 and 2005. [2] A downward correction in house prices and significant wave of defaults was inevitable. Credit-default swaps, complicated collateralized debt obligations constructed from underlying troubled mortgages, and off-balance sheet entities such as structured investment vehicles left many key financial institutions with leveraged exposure to this downturn. Fears of their failure crippled lending, sending economic activity into a nosedive in the fall of 2008. There is a serious indictment of policy to be made here, for which the Federal Reserve should share some of the blame. But I only attribute a small part of this failure to the excessively low interest rates of 2003-2005, and see the primary policy errors as a problem of inadequate regulatory supervision (Hamilton, 2007).

To be sure, these problems were greatly aggravated by the economic downturn itself. Rising unemployment rates and falling income exacerbated default rates. We can quite legitimately ask whether there were options open to the Federal Reserve in the fall of 2008 that might have mitigated the damage. I agree with Professor Sumner that, if the Federal Reserve had been able to achieve a five percent annual growth rate for nominal GDP over the last four quarters instead of the ‑2.5 percent actually achieved, the average American would have been better off. But I disagree on the mechanisms and tools that the Fed could have used to try to steer us toward such an outcome.

Growth of nominal GDP at an annual rate, 2000: Q1 to 2009:Q2Growth of nominal GDP at an annual rate, 2000:Q1 to 2009:Q2

Sumner appeals to the equation of exchange,

M x V = P x Y

where M is a measure of the money supply, V its velocity, and nominal GDP is written as the product of the overall price level (P) with real GDP (Y). Sumner reminds us of Hume’s notion that velocity V in part depends on the extent to which households decide keep their coins locked in chests. If we thought of V in the above equation as determined by institutional details of how often people get paid or visit the grocery store, then we’re tempted to conclude that by choosing the appropriate value for the money supply M, the Fed could deliver a desired target for nominal GDP.

But one runs into an immediate practical problem in the bewildering variety of different magnitudes that might be thought to correspond to the money supply M, such as M1 (checkable deposits plus currency held by the public) or the monetary base (currency plus reserves, the latter being the electronic credits that private banks could use to turn into currency if they wished). Obviously two different M’s must imply two different V’s for the above equation, and we can’t think of both V’s as being entirely unaffected by actions taken by the Fed. For that matter, there are different concepts one could use for P x Y on the right-hand side of the above equation. Is it the dollar value of sales of all final goods and services (that is, nominal GDP, as my discussion above assumed), the dollar value of all transactions (as the earlier monetary theorists supposed and the notion of a dollar physically changing hands invites), the dollar value of consumption expenditures, or something else? It is clear that there is a long menu of different values we might refer to when we talk about the “velocity of money,” and at most one of these can actually determine the level of nominal GDP.

When you dive into such details, you are led to the conclusion that the above equation is not a theory of income determination, but instead is a definition of V. If we use a different measure of the money supply or a different measure of nominal transactions, then we must be talking about a different number V. What the equation really does is define a value of V for which the resulting expression is true by definition.

As Milton Friedman himself was quite clear, it is ultimately an empirical question as to whether a given candidate V so defined simply changes passively in response to the Fed implementing a change in the favored measure of M. The diagram below plots the growth rate of M1 along with the negative of the growth rate of the velocity implied when M1 is our measure of the money supply and nominal GDP is the measure of transactions. The strong impression is that in quarters in which M1 grew a lot, the M1-velocity shrank by an offsetting amount, leaving the quarter-to-quarter correlation between M1 growth and nominal GDP growth quite weak.

Top panel. Annual growth rate of M1, 1980:Q1 to 2009:Q2. Bottom panel Annual growth rate of the ratio of M1 to nominal GDP.

The same conclusion emerges if you prefer to use the monetary base as your measure of the money supply.

It’s necessary to spell out a mechanism other than the equation of exchange by which the Federal Reserve is asserted to have the power to achieve a particular target for nominal GDP. One mechanism that we all agree is relevant in normal times is the Fed’s control of the short-term interest rate. Lowering this will usually stimulate demand and eventually lead to faster nominal GDP growth. However, there are two problems with advocating this tool in the fall of 2008. First, most of us are persuaded that the stimulus from such a policy takes some time to affect the economy, so that, if implemented in October 2008, it is not a realistic vehicle for preventing a decline in 2008:Q4 nominal GDP.  Second, we quickly reached a point at which the fed funds rate fell to its nominal floor, an essentially zero percent interest rate, from which there is no more down to go.

Now, I am in agreement with Professor Sumner that this does not mean that monetary policy is completely ineffective in such a situation. I do not endorse the “liquidity trap” scenario, though my reasons are different from those given by Sumner. I believe that once the nominal T-bill rate falls to zero, not much is achieved by further open market purchases of T-bills by the Federal Reserve. However, there is still an opportunity for monetary stimulus in such a situation through purchase of assets other than T-bills, and I believe that this was something the Fed should have tried in the fall of 2008.

Unfortunately, until the beginning of 2009, the Federal Reserve was doing everything it could to prevent its actions from stimulating the economy in the usual fashion. It was viewing the slowly unfolding credit problems as primarily a crisis in lending, in which the Fed felt it needed to step in as lender of last resort on what ultimately proved to be a massive scale. [3] The Fed wanted to lend extensively, but did not want to see currency held by the public increase. For this reason, it sold off a significant portion of its holdings of T-bills through September 2008, in a paired set of actions, lending with one hand and selling T-bills with the other, that might be described as “sterilizing” the lending operations so as to prevent them from having an effect on the money supply. [4] When the Fed ran out of T-bills to sell, it asked the Treasury to create some more for the Fed to use just for sterilization. More importantly, In October the Fed began paying interest on reserves, in effect borrowing directly from banks, and creating an incentive for banks to hold the newly created deposits as a staggering burgeoning of excess reserves, again preventing its actions from increasing the value of M1. Excess reserves amounted to $833 billion by August 2009, or more than the sum of all the currency issued by the Federal Reserve between its creation in 1913 and 2008.

I agree with Sumner that the Fed made an error in abandoning its traditional goal of monetary expansion. The sterilization efforts ultimately made it much harder for the Fed to do what I think they now understand would be desirable, namely, provide a more traditional stimulus to aggregate nominal GDP. In my opinion, the preferred policy in the fall of 2008 would have been to acknowledge more aggressively the losses financial institutions had absorbed on existing loans, impose those losses on stockholders, creditors, and taxpayers, and retain as the Fed’s first priority the stimulus of nominal GDP rather than trying to lend to everybody.

Let me mention one other real-world complication that has made it difficult for the Fed to achieve a faster growth of nominal GDP. Macroeconomists often like to think of inflation in terms of an aggregate price index, the variable P in the equation of exchange. But, particularly in the current environment, aggressive stimulus by the Fed is unlikely to show up as higher wages or the prices of most services, but instead would raise relative commodity prices and could in a worst case scenario precipitate a currency crisis, both of which would be highly destabilizing in their own right. I agree with Sumner that the Fed could and should have done more, but would caution that it is also possible for the Fed to try to do too much. I come back to the perspective with which I opened — given the earlier regulatory lapses, significant economic losses could not have been prevented by any monetary policy that could have been implemented in the fall of 2008.

Finally, I would like to comment on Sumner’s intriguing suggestion that we might be able to sidestep all of the complications as to how the Fed achieves a particular target for nominal GDP growth by having it fix the price of a futures contract settled on the basis of nominal GDP growth. Perhaps there are some more details of what Sumner has in mind that I am missing, but I don’t really understand how it could work. The essence of any Fed operation is an exchange of assets of equivalent value. Traditionally, the Fed uses the money it creates to purchase a T-bill, thereby increasing the quantity of money in circulation. However, a futures contract is not an asset, but instead is an agreement between two parties for which neither party initially compensates the other. The market value of that agreement at inception is, by definition, zero.  The Fed may participate in the market with an infinitely large position, say, and thereby cause the contract to price in five percent nominal GDP growth. But the initial act of taking one side or the other of such contracts will not create or destroy any money.

For example, you and I might agree today that if fourth-quarter GDP growth is above five percent, I pay you a sum based on the gap, and if it is below five percent, you pay me. Let’s say we make exactly that agreement, so that the futures price starts out precisely where the Fed wants it to be. If at expiry nominal GDP growth equals five percent, the value of the contract to either party is, again by definition, zero, because neither of us compensates the other. Hence, if the contracts when issued were priced for a five percent growth (as I understand they’re supposed to be, under this system), and the economy experienced that hoped-for 5 percent growth, there is no avenue whereby the money supply could increase or decrease over the period of time covered by the contract as a result of Federal Reserve participation in the market.

If, for sake of argument, we ignore the issues discussed above and think of V as institutionally fixed and view M x V = P x Y as determining nominal GDP, it would be necessary to have 5 percent growth of M to achieve 5 percent growth of nominal GDP. I do not see how the Fed “fixing” the price of a futures contract based on nominal GDP growth could achieve that objective. This is simply a manifestation of the broader point I’ve been making above. If the Fed does not have the economic ability to achieve a particular target for nominal GDP growth, then it is not feasible for it to fix the expiry value of a futures contract settled on the basis of nominal GDP growth.

In conclusion, Sumner is raising some important issues and I agree with him on some of the key points. But while I agree that monetary policy in the fall of 2008 could have been improved upon, doing so was much trickier than might appear.

James D. Hamilton is professor of economics at the University of California, San Diego


[1] Federal Reserve Board Flow of Funds, Table L.2.

[2] Case-Shiller/S&P Home Price Index for 20 metropolitan areas,….

[3] Keister and McAndrews (2009) provide an exposition of this view.

[4] See Hamilton (2009) for more details.


Ashcraft, Adam, and Til Schuermann. 2008. “Understanding the Securitization of Subprime Mortgage Credit,” working paper, Federal Reserve Bank of New York (

Hamilton, James D. 2007. “Commentary: Housing and the Monetary Transmission Mechanism.” In Housing, Housing Finance, and Monetary Policy, Federal Reserve Bank of Kansas City, pp. 415-422.

Hamilton, James D. 2009. “Concerns about the Fed’s New Balance Sheet,” in The Road Ahead for the Fed, edited by John D. Ciorciari and John B. Taylor, Stanford: Hoover Institution Press.

Keister, Todd, and James McAndrews. 2009. “Why Are Banks Holding So Many Excess Reserves?” Federal Reserve Bank of New York Staff Papers (

Also from this issue

Lead Essay

  • In this month’s sure-to-be controversial lead essay, Bentley University economist Scott Sumner argues that almost everything economists and economic policymakers thought they knew about the role of monetary policy in the recent recession and financial collapse is wrong. Sumner contends that the resources of monetary policy were not exhausted, as many economists believed, but were barely used. Flying in the face of conventional wisdom, Sumner maintains that monetary policy in the run-up to the finacial crisis was not highly expansionary, but was in fact disastrously contractionary. Sumner offers a short history of monetary economics to put into historical perspective the role of allegedly failed monetary policy in the financial crisis and recession. He proposes a strategy for central bankers — targeting forecasts of nominal GDP — that might help avert future crises. In conclusion, Sumner warns of the political dangers of misdiagnosing the crisis: unless the record is set straight, free markets will once again take the fall for a failure of monetary policy.

Response Essays

  • University of California, San Diego economist James D. Hamilton disputes Scott Sumner’s claim that the sub-prime crisis was a fluke with few lessons for macroeconomics. According to Hamilton, the booming U.S. housing market represented a “huge misdirection of capital,” and the overexposure of key financial institution to the housing market’s downward correction crippled lending and sent the economy into a nosedive. Hamilton agrees that the Fed might have limited the damage had it kept the growth rate for nominal GDP higher, but he disagrees with Sumner about the tools available to the Fed to achieve this. Hamilton notes that tools available to the Fed depend on which of the possible specifications of the money supply and its velocity actually determine nominal GDP. Hamilton says unconventional paths to monetary stimulus were open the Fed in late 2008 and that “the preferred policy … would have been to acknowledge more aggressively the losses financial institutions had absorbed on existing loans, impose those losses on stockholders, creditors, and taxpayers, and retain as the Fed’s first priority the stimulus of nominal GDP rather than trying to lend to everybody.” Hamilton concludes with some worries about Sumner’s favored tool for targeting nominal GDP growth.

  • University of Georgia economist George Selgin agrees with Scott Sumner that “tight money was the proximate cause of the post-September 2008 recession” and that “a policy of nominal income growth targeting might have prevented the recession.” Selgin encourages Sumner to acknowledge the role easy money played in the subprime crisis, and argues that Sumner’s five-percent nominal income growth target is “unnecessarily and perhaps dangerously high.” Selgin favors a two or three percent target, which he contends would be less likely to perpetuate boom-bust cycles.

  • San Joses State’s Jeffrey Rogers Hummel begins with a brief history of economic thought about the causes of the business cycle, which leads to a call for “a measure of epistemic humility.” Hummel signs on to much of Sumner’s story about the Fed behavior in 2008, and accepts his criticism of the widespread use of interest rates as the main indicator of monetary policy. But Hummel departs sharply from Sumner’s prescription for better monetary policy — a rule to target the forecast of nominal GDP growth. “The … critical defect of Sumner’s Rule,” Hummel argues, “is its blithe assumption that money, unlike any other good or service, requires not merely government provision but detailed, sophisticated, and flexible government management.” Hummel raises doubts that even the best such rule would be well-applied, and calls for the “abolition of the Fed, elimination of government fiat money, and complete deregulation of banks.”