Scott Sumner  

Did the Fed cause the sub-prime boom?

Substitution in Nicotine Consu... Mankiw's Textbook on Determina...

George Selgin has a new post discussing a paper he wrote with David Beckworth and Berrak Bahadir. In the paper they argue that the Fed set its policy rate too low during the early 2000s, and that this decision contributed to the subprime boom:

It is widely believed that, in the wake of the crash, the Fed kept the federal funds target rate too low for too long, inadvertently contributing to the subprime boom. We attribute this and other Fed departures from a "neutral" policy stance to the Fed's failure to respond appropriately to exceptional rates of total factor productivity growth. We then show how the Fed, by adhering to a nominal GDP growth rate target, might have succeeded in maintaining such a neutral stance.
This is certainly a defensible claim, and I agree with the policy implications they draw from their analysis. Nonetheless, I am going to offer a few words of caution:

1. We should never leave the impression (even unintentionally) that sound monetary policy will prevent severe sectoral imbalances. It's quite possible that a housing boom of almost equal size would have occurred during the early 2000s, even under a perfect NGDP targeting policy.

2. It's dangerous to focus on interest rates, even if everyone else in the world insists on talking about monetary policy in terms of interest rates.

Here are some other claims about the housing "bubble" period, that I would argue push a little bit back against the view that the Fed played an important role:

1. Low interest rates don't imply easy money, unless it can be established that they were below the Wicksellian equilibrium rate. And this can only be determined by looking at NGDP growth rates.

2. NGDP growth during the 2001:4 to 2007:4 expansion averaged well under 6%. At the time it was the lowest growth rate during any economic expansion since NGDP data collection began in the 1940s. In retrospect, NGDP growth during the early 2000s expansion probably should have been even less. However I'm reluctant to attribute major economic events to excessive NGDP growth in an economic expansion, when NGDP growth during that expansion was unusually slow.

3. Monetary policy is not a surgical tool that can be used to affect specific sectors of the economy. It can only affect any given sector by affecting the overall economy (NGDP.) Thus in 1928 and 1929 when the Fed tried to use higher interest rates to slow the stock market boom, and the policy was a complete failure until rates were raised so high that NGDP began plunging. Then stocks crashed. It wasn't the high interest rates, it was the falling NGDP. If Fed policy contributed to the subprime boom, it was not from low interest rates, but rather excessive NGDP growth.

4. If the Fed's policy had aimed at slightly slower NGDP growth during 2001-07, then interest rates in the counterfactual policy would have been a bit higher in 2001-02, but substantially lower in 2003-07. In the long run rates tend to follow the economy, and that is why slower NGDP growth in 2001-07 would have implied lower interest rates by 2003-07. That's something to think about if you believe low rates fed the housing boom by making mortgages "affordable."

5. Even under an ideal monetary policy, the subprime boom might have occurred due to Federal government policies aimed at boosting mortgage lending (FDIC, GSEs, TBTF, CRA, etc.) and/or mistakes made by private sector lenders. There doesn't seem to be much correlation between easy money and what are viewed as "bubbles," indeed bubbles were noticeably absent in the Great Inflation of 1966-81, and the 1928-29 stock boom occurred during a period of mild deflation and extremely high real interest rates. (In fairness, NGDP growth was probably reasonably good in 1928-29.)

None of this should be viewed as criticism of the key analytical points in their article. Fed policy is too expansionary during periods of rapid productivity growth because the Fed focuses on inflation, rather than NGDP growth. With a policy of NGDP targeting, level targeting, the past 15 years would have been far more stable. However in my view that extra stability would come more from easier money after 2007 than tighter money before 2007.

PS. I put the term "bubble" in quotation marks because I don't really believe in bubbles. I'm referring to periods that other people view as bubbles.

Comments and Sharing

CATEGORIES: Monetary Policy

COMMENTS (22 to date)
E. Harding writes:

Not all credit creation is money creation. There is four times as much credit as money.
Loose credit and loose money are very different things.
As Mish Shedlock points out, there is a big difference between a debt/asset bubble and strong inflation.
The question is, how can the debt-to-NGDP ratio be kept from skyrocketing, as during the U.S. housing boom and late 1920s stock bubble?

James writes:

"Monetary policy is not a surgical tool that can be used to affect specific sectors of the economy. It can only affect any given sector by affecting the overall economy (NGDP.)"

Is this a general claim, or is it conditional on the way open market activity is conducted?

Specifically, suppose the central bank were to increase base money by purchasing oil futures at several basis points over their current price. I doubt anyone would claim this wouldn't disproportinately affect different industries depending on their sensitivity to oil.

So what if the open market desk bought government debt? Wouldn't this disproportionately affect industries more sensitive to interest rates?

Kevin Erdmann writes:

A good rule of thumb is, when in doubt, defer to George Selgin, and I don't feel qualified to comment on this paper, even though that's what I'm doing. But, I do wonder how he would explain the fact that NGDP growth and mortgage growth were pretty similar in the 1980s and the 2000s, during periods where this model suggests very different monetary policies.

Also, I'm not sure how much this matters, but in terms of thinking of the 2000s as the "subprime boom", I think we generally assume that the origination of mortgages with teaser rates caused the Fed Funds rate to be a more important factor in housing funding than it might normally be. But, I'm not sure how explanatory this is. FICO scores and average incomes for borrowers apparently didn't fall during the period. And, through 2003, prime mortgages were the overwhelming source of growth. It was only in the 2004-2006 period that sub-prime and Alt-A mortgages pushed to a very high proportion of originations. This was after rates had begun to climb back up. Most of the nonprime mortgages were originated after the Fed Funds rate had risen quite a bit above the low level of 2001-2003, when homeownership had already leveled out and begun to slightly decline. An increasing amount of real estate investment by then appears to have been from older, relatively high income households buying second homes for income in a low interest rate environment. And, as you point out, low real long term interest rates aren't a particular sign of loose monetary policy.

Anyway, food for thought.

Brian writes:

and WHY did the Fed keep the rate low? How about 9-11. On his flight back from Europe agree spans military plane flew past the NY site. With the economy weakening, an exogenous shock is not what was needed.

But simply, pump the economy so the terrorist wouldn't win.

Can't blame him. I'd have done the same.

George Selgin writes:

Scott, thanks for drawing attention to the article. Concerning your remarks, remember that we don't claim at all that the Fed "caused" the boom: we claim (as you say in your text) merely that it "contributed." That leaves plenty of scope for other causes, and for recognition that those must have played a part in the sectoral distortions to which you refer. In short, were your title accurate, you would have us on point 3, and also on point 5. But it isn't, so you don't. I'm sure I speak for my co-authors as well as myself in saying that there is plenty of blame to go around for the crisis. The Fed added fuel to the fire, that's all.

On that particular point, you refer to NGDP figures to suggest that the Fed cannot have contributed much. But our own paper actually argues that it would have made a big difference had the Fed targeted NGDP growth instead of following the policy it actually followed. (Have a look at figure 10 and the surrounding text.) Evidently our analysis are causing us to reach different conclusions than yours on this issue. In any case the paper's policy punchline is in fact one favoring an NGDP rule.

Toby writes:


whether NGDP was slow or fast when compared to the early data in the 1940s seems like the wrong benchmark. Shouldn't it be whether it was slow or fast compared to what was expected? For example, whether it was slow or fast compared to the previous seven years? Or perhaps whether it exceeded what the market seems to have predicted NGDP to be the next 5 years?

2. NGDP growth during the 2001:4 to 2007:4 expansion averaged well under 6%. At the time it was the lowest growth rate during any economic expansion since NGDP data collection began in the 1940s. In retrospect, NGDP growth during the early 2000s expansion probably should have been even less. However I'm reluctant to attribute major economic events to excessive NGDP growth in an economic expansion, when NGDP growth during that expansion was unusually slow.
Mike Hammock writes:

What should we make of the fact that some countries experienced housing booms without their central banks greatly expanding credit/increasing the money supply/pushing interest rates down (insert your preferred view of monetary policy)?

Scott Sumner writes:

E. Harding, You said:

"Loose credit and loose money are very different things."

I strongly agree. Credit problems need to be addressed through regulation (or arguably deregulation, as when we remove tax subsidies to borrowing, or moral hazard from GSEs.)

James, I don't think it matters whether they buy government debt. If they instead paid the wages of public employees with new money, the effect of the total stock of public debt outstanding would be exactly the same. If they bought a commodity like oil, then that would have some impact on the specific (oil) market.

George, The title was aimed at a broader set of ideas than your specific paper. Indeed the comment "even unintentionally" alluded to my worry that many readers would draw the conclusion that we are claiming NGDP targeting would prevent events like the subprime boom. I tried to write the post in such a way that indicated I accepted the basic qualitative arguments you put forward, but thought the share of Fed responsibility in the subprime boom was rather low, and perhaps (I'm not certain) lower than you three estimated. Perhaps I should have worded it more carefully.

Notice that my list of 5 factors was directed against a general argument, not (specifically) your paper.

Scott Sumner writes:

Mike, That suggests that monetary policy is not always the cause of housing booms.

George Selgin writes:

Mike Hammock, none of the metrics you refer (monetary aggregates or credit measures or absolute interest rates) has much to do with the metrics of our paper, which refer to a rough estimate of the "natural" rate and comparison of that with the Fed's rate targets. So what people should "make" of booms in other countries is that it might be worth applying the same metrics we apply for the U.S. to those cases.

In any case, the fact that housing booms occurred in so many places under so many different circumstances poses a challenge to any uni-causal theory--which is a good reason for not relying on such a theory, but not a good reason for concluding than the single causes in question are all irrelevant. (Try this: "Oh! Spain has no GSEs. So, the GSEs cannot have played any role in the U.S. housing boom." Keep arguing that way and see whether you don't end up convincing yourself that the boom was caused by...nothing!)

That said, it would be nice to see some critically commentary directly addressing the arguments my co-authors and I actually offer in claiming that Fed policy was too easy. So permit me to make clear what we do and do not do: We never use absolute interest rate levels as a measure of monetary easiness; on the contrary, we adhere to an explicitly Wicksellian framework. We reach our conclusions about the Fed's stance on the basis of empirics, which we check against a number of alternative indicators. We also go into the FOMC's deliberations to show how that Committee itself understood that it was setting rate targets at below "equilibrium" (natural) levels. We consider both Taylor-Rule and NGDP-targeting counterfactuals, concluding that either would have resulted in a less severe boom, but that the latter would have come closest to keeping rates near their "natural" levels. We never suggest, finally, that monetary policy alone was responsible for the housing booms experienced either in the U.S. or elsewhere.

ThomasH writes:

Maybe this is just paraphrasing Scot's saying:

However in my view that extra stability [from NDGP targeting] would come more from easier money after 2007 than tighter money before 2007.

But would we be worrying much about monetary policy 2000-2008 if the Fed had maintained NGDP in 2008-15?

Nick writes:

The main problem with the paper is that it assumes that the housing sector was dependent on interest rates during the boom. But the main driver of the boom was the "innovation" in "affordable mortgage products", where the focus was on short term low (dollar) mortgage payments. The mortgage market became focused on initial payment affordability at the expense of ignoring ultimate affordability. Interest rates became only secondary importance (if at all) in determining a home buyers decision about how much borrowing to take on.

Thus, the argument describing Fig 5 of the paper is unsound (which is the lynchpin that connects the interest rate gap and the housing market), and needs to incorporate this fundamental shift in how mortgages were underwritten and how borrowers decided how much they would borrow during this period. Without incorporating those important variables the correlation is spurious. It would be like finding a correlation between the interest rate gap and the "SUV boom" of the 90's without taking into account the low oil price during that period.

Although one might argue that products like hybrid ARM's were dependent on (artificially) low short term rates, the large expansion of products like teaser rates at round numbers (e.g. 1.9%, 0.9%), interest-only loans, 40 year mortgages, and neg-am loans (which all lower initial payments at even high interest rates) illustrates that no matter what the rates actually were, lenders would find a way to make initial payments low enough for borrowers to qualify. And it is that focus on the borrower's ability to budget for the initial cash payments that drove borrower behavior, not the interest rate at which borrowers could obtain financing.

Jose Romeu Robazzi writes:

@E Harding, Prof Sumner
"Loose credit and loose money are very different things."
I agree with this too. The great recession might be recognized as a period of "tightening regulation" and "temptative loose money".

I agree with Selgin in the sense that the monetary policy stance, empirically evaluated, might have contributed to the event that some now call "housing bubble".

Easy money may be a necessary condition, but no sufficient condition, during the several periods before the bust, for the bubble formation process.

But what would be the other conditions? In my view, faulty finance structures, with either or both term mismatch / low equity levels (high leverage). Therefore, the focus of financial regulation should be prevent term mismatch between assets and liabilities in any financing structure, also, prevent very high leverage structures.

Also, law-makers should aim at a run free banking system, eliminating moral hazard, which in my view are incentives to leverage and term mismatch in financial structures

Mike Hammock writes:

Scott and George, thanks for the responses.

George Selgin writes:

Nick's remark reminds me of similar claims made by Peter Wallison is his otherwise excellent work on the role of the GSEs in the boom. The problem, though, is that having new credit "products" to offer, or being willing to lend at lower standards, are not the same thing as having more $$$ to lend. All the lowering of credit standards one can imagine alone doesn't add up to a single risky mortgage unless the funds are there to be lent in the first place.

In short, while other factors were to blame for such a large share of credit being directed toward risky mortgages, the extent of the boom depended on the total nominal supply of mortgage credit, and that total in turn depended to an important extent on Fed policy.

Lawrence H. White writes:
3. Monetary policy ... can only affect any given sector by affecting the overall economy (NGDP.)

Scott, are you really asserting that monetary policy never has relative-price or sectoral effects, contrary to Cantillon or Wicksell?

Nick writes:

The argument would continue by saying that the extra $$$ to lend came not from an excessively easy Fed policy, but rather at the expense of other possible lending that banks could have done. For example, consumer debt, which did not increase as much as mortgage debt, and did not experience a boom and bust.

Then the extra mortgage lending could be seen as merely a relative shifting of credit away from certain sectors and into others, rather than as a runaway expansion of credit as a whole. And the relative shift can be explained better by the idiosyncratic changes to the mortgage market at the time that were not seen in other lending markets that did not crash than it can by a macroeconomic solution that would seem to implicate them all.

Bob Murphy writes:

Larry White, if I understood him correctly at the time, Scott once settled that same dispute with me by arguing that if the Fed (say) buys my house for $10 million, and obviously enriches me, that that is fiscal policy, not monetary policy.

Kevin Erdmann writes:


Homeownership topped out in 2Q 2004, prices topped out at the end of 2005, and most of the subprime boom happened from 2004-2006. Also, during this period rent was rising, suggesting that, from a consumption perspective, homes were still undersupplied.

This suggests that the new mortgage products were being used because of demand for them from potential buyers who were trying to create housing supply where it was needed. Probably, these products were novel, in part, because they just hadn't been that useful before. In the 1970's, the last time low long term real rates caused Price/Rent ratios to rise significantly, nominal mortgage rates were in the double digits because that really was a time period with loose monetary policy. So, for marginal households, cutting their downpayment down and accepting a larger or variable monthly payment would not have been helpful.

George Selgin writes:

Agreed, Nick, that it is conceivable that riskier borrowers are bidding available funds away from others, with no general expansion of credit. But that's not what was going on in the U.S. Look at any outstanding prime debt measures you choose, and you will see substantial growth through 2006, if not beyond.

Nick writes:

@Kevin Erdmann:

I would quibble with your figures a little bit. Prices topped out during the summer of 2006, whether you use Case-Schiller or CoreLogic, especially in the biggest bubble cities

Homeownership bounced around its highest level throughout 2004-2006, and especially out west where these products were most prevalent. It's not really right to say that homeownership declined meaningfully after mid-2004, until the bubble burst after 2006.

I'm not sure what "This suggests that the new mortgage products were being used because of demand for them from potential buyers who were trying to create housing supply where it was needed." means. Nobody "needed" another PUD 30 miles from downtown Phoenix, and supply had nothing to do with the issue of a 40 year old 2 bedroom in San Bernadino tripling in value in about as many years.

Also note that another factor in the reduced underwriting was the agent/principal problem that came from securitizing so many non-agency loans. That didn't exist in the 70's, so banks wouldn't have lent under the terms they did during the bubble. To use another bad car analogy, it would be like expecting cars to have remote keyless entry before they even had power locks.

Kevin Erdmann writes:

Nick, thanks for the response. On homeownership rates, my point is that they didn't increase after that, so the bulk of subprime mortgages were not associated with attracting marginal new households into homeownership. I think homes were roughly priced in line with alternative inflation protected fixed income investments at the time. I'm not sure that anyone has presented a compelling mechanism for why credit was pushing home prices above the levels justified by alternatives, considering the evidence that the period with the most subprime mortgages was not associated with ownership growth, falling FICO scores, rising debt service ratios, or lower incomes of borrowers.

The idea that there was an oversupply needs to answer the elephant in the room: Rising rents and high rent inflation throughout the period until the crisis hit.

Comments for this entry have been closed
Return to top