The slope of the yield curve is likely to become an increasingly bearish indicator as this year progresses, and recent changes in the calculation of the index of leading economic indicators should not be interpreted as in any way denying that fact.
One of the best-established predictive relations in macroeconomics is the observation that when
the difference in yields between long- and short-term interest rates is low or negative, future GDP growth
tends to be slow or negative. Pretty much any measure of the size of the gap seems to support that
conclusion. Hibiki Ichiue, a student of mine who successfully defended his Ph.D. dissertation
yesterday morning (congratulations!) found that even the 3-year minus 2-year rate does quite well. Most
academic studies have used the 10-year minus 3-month Treasury rates plotted in the top graph on the
right. This spread has averaged 140 basis points over the last 50 years, but it narrowed to a
value much less than that prior to each of the last eight recessions (shown as shaded regions on
the graph), and actually became negative in six episodes, five of which led to recessions.
The spread has dropped dramatically over the last year and is now less than 90 basis points.
The natural question to ask in this situation is, how low does the spread have to go before we
start to get worried about where the economy is headed?
One way you might try to answer that question is with statistical regression analysis, in which
you try to forecast economic growth rates using lagged growth rates and lagged values for the
spread, and see if there’s some cutoff point for the spread above which movements in the spread
don’t seem to make much difference for your forecast. Time-series specifications that allow for
such effects are called threshold autoregressions or smooth transition autoregressions. Empirical
analyses using these methods have found that there does indeed seem to be such a cutoff, but it is
pretty high. John Galbraith and Greg Tkacz, in a study of GDP growth rates published in the Journal of International Money and Finance in
2000, estimated the threshold to be a little under 200 basis points. Ivan Paya, Ioannis
Venetis, and David Peel, in a study of industrial production growth published in the Journal of Forecasting in
2004, found a threshold of around 140 basis points. In other words, according to these
studies, whenever the 10-year minus 3-month spread is below 140 basis points, the lower it gets, the more pessimistic your forecast should be.
The usefulness of the yield spread for economic forecasting led the Conference Board to include it in their popular index
of leading economic indicators in 1996– they use the gap between the 10-year rate and the
overnight fed funds rate. However, beginning with this July’s release, the Conference Board
revised the method by which the index of leading economic indicators is calculated. Previously, whenever the yield spread became smaller, it contributed negatively to the index of leading indicators. Beginning with the July release, the spread will contribute negatively to the index only if the spread itself is actually negative.
The reasons for this change can be found here and here.
The key issue has to do with the way in which the Conference Board combines the information from a
variety of different indicators to form a single index that is growing over time. Some of the individual indicators that make up the index are themselves usually growing (such as stock prices) and others (like the spread) have no clear trend. The predictive relation referred to in the academic studies above is one between the level of the spread and the growth of output. Because the Conference Board relies on changes in the index as a predictor of future output growth, its previous method amounted to imposing a relation between the change in the spread and the growth of output. The Conference Board decided to fix that problem by in effect using the level of the spread itself as something that goes into determining the change in the index. Thus, under the new method, when the spread is positive, that tends to make the index go up, and when the spread is negative, that tends to make the index go down. Because the Conference Board interprets an increase in the index as a favorable signal, this has the practical effect that low but positive values of the spread no longer register as a bearish indicator.
Although Barry
Ritholtz was unreserved in his criticism of the Conference Board for this adjustment, I can see what the Board was trying to do, given the mechanics of how they are trying to combine the information from variables that are trending and nontrending. But none
of that should obscure the reality that, as a statistical proposition, the lower the spread becomes at this
point, the more pessimistic your forecast of output growth should be. This pessimism will not be reflected in the index of leading economic indicators until the spread turns negative, but that does not in any way imply that we should be unconcerned about the implications of low but positive values for the spread.
So where does that leave us right now? The Fed’s expected rate hike next week will narrow the
spread further. My nervousness about that is allayed somewhat by last week’s favorable
economic data. I’ll be more nervous if the Fed raises rates again at the following FOMC
meeting, and even more nervous if they raise rates again in November.
Looking at the
Macroblog projections of future fed funds target changes, I pretty much expect to be nervous
for the rest of the year.
The key to my unreserved criticism is that the Conf Board’s change simply runs against experience. A widening yield curve is stimulative, and therefore should suggest exapnading ecoonmic activity going forward — certainly, a year hence is plnety of lead time.
Conversely, a narrowing spread reflects a decrease of stimulus. That suggests less Eco-activity.
If I am reading the Conf Bd notes correctly, the new methodology has a narrowing spread not as a neutral, but a positive.
Quite frankly, that’s where the line gets crossed for me, from legitimate economic analysis to silly cheerleading.
Can you imagine: “Yeah! The yield curve is flattening!”
Me neither . . .
Your definition of SLOPE is totally wrong. You determine the SLOPE by dividing, not subtracting or adding. A 3-month Treasury bill at 3% combined with a 10-year T note at 4%, with a difference of 100 basis points, produces a SLOPE of 1.33. It is not the same as a 3-month Treasury at 10% and a 10-year T yield at 11%. Even though the difference is exactly the very same 100 basis points, the SLOPE of the latter is 1.10. As it relates to monetary policy, the former is fairly accommodative, and the latter is restrictive.
Prof. Hamilton:
I understand that you are worried about the yield curve inversion. However, I am not sure if your fear is warranted at this point in time. Although yield curve inversions have portended recessions in the past, there were also other signs that accompanied them.
One such sign, which I think is important, is the spread between corporate bonds and treasury. In the past, when yield curves have inverted, the spreads between corporate bonds and treasury have also widened. This implies that bond investors were taking their money out of the “more risky” corporate bonds and investing them in the “less risky” treasury bonds (of equal duration), thereby pushing the treasury bond yield down (in that part of the curve), hence the yield curve inversion. For example, BBB-rated corporate bond spread, in March 2001 (as the economy was weakening), was about 250 basis points (with the 10-yr LIBOR spread at 90 basis points). Today, the BBB-rated corporate bond spread is about 126 basis points (with 10-yr LIBOR at 45 basis points). Looks like the investors are not pursuing the course of “flight for safety” yet. Hence, financing costs for the corporations are still at non-recession levels.
Another sign is the stock market: The NASDAQ and NYSE, as of yesterday, have more advancing issues than declining issues, once again indicating a healthy market for equities. Here too, investors are not “fleeing for safety”.
Based on these “signs”, I am tempted to think that the almost-flat yield curve is more due to some non-economy-related technical pressure on the 10-yr part of the curve than an economy-related technical pressure on the 10-yr part of the yield curve.
What do you think? As always, if I have erred in my analysis, please feel free to correct me.
I enjoy the pedagogical approach you take in your blogs. This approach helps non-economists like me better understand and appreciate the finer concepts of economics. Thank you for your efforts, and I hope you continue blogging for a long time to come.
Hank, it’s very curious to me that (1) you think my post offered a definition of the slope, (2) the definition that you think you saw strikes you as totally wrong, (3) you give examples of “correctly” calculating the slope as Y2/Y1, and (4) you suppose that you know, and I do not, the methodology and findings of the many dozens of academic studies on this topic.
But, since you ask, the slope is defined as dY/dX, for which the usual finite approximation is (Y2 – Y1)/(X2 – X1), and the difference between this magnitude and (Y2 – Y1) is simply a matter of the units in which X is measured. The typical study estimates a different coefficient bj for each spread, so these units are implicit in the coefficients.
I think the valid point you’re trying to make is that the level of the short rate in addition to the spread between the long and short rate can be informative. There is a separate literature on this question as well.
I agree, the steepness of the curve is usually a good indicator of whether we are headed into a recession. Note, however, that the average fed funds rate at the time the curve inverts is close to 10%, not 3.5%. There’s a simple explanation for why the curve inverts, and it has less to do with recessions and more to do with the front end being more sensitive to fed funds movements than the long end. Whenever the fed feels it needs to put a clamp on rampant inflation, so they tighten, which restricts growth. However, currently, inflation is by no means rampant.
Great points, Madan. Yes, there are many other favorable indicators, though I continue to view oil as a negative.
As for other causes of the flattening yield curve, the Fed has caused a very rapid increase in short rates over the last year. I see that Fed tightening as clearly the most important reason for the narrowing spread.
On balance, I conclude that it’s OK for the Fed to go ahead with next week’s rate hike, and probably the following expected hike as well. But let’s take a careful new evaluation of the landscape before pushing above 3.75.
There’s a nice animated graphic of the UK and US yield curves at:
http://www.yieldcurve.com/marketyieldcurve.asp
Some similar arguments to Madan’s have been made for the UK, where the curve inverted some time ago. Recently however UK GDP growth has stalled quite suddenly as consumer demand has slowed (coincident with the stagnation of the real estate market), so it may be that the yield curve has been flashing a red light after all.
The recent strong performance of the UK stock market would seem to contradict any recessionary signal from the yield curve, but the forecasting ability of equities cannot always be relied upon (e.g. see Estrella and Mishkin, http://www.ic.sunysb.edu/Class/eco360/articles/Estrella%20Mishkin%20Yield.pdf).
Maybe it’s different this time, but presumably there were equally seductive arguments to dismiss the inversion in the past, since markets didn’t react until much later, when recession was knocking at the door.
Why is it that the inverted yield curve predicts slower growth? Is it that the market, seeing slower growth ahead, expects the Fed to lower rates, hence wants to buy long bonds so as to profit when rates come down, and this lowers today’s yield on longer bonds?
Or is the other direction, that the lowered long term interest rates somehow redirect economic activity such that growth is inhibited?
Sometimes I read commentary suggesting that the Fed has been trying (without much success) to get long term rates up, presumably in order to reduce the chance of the economy falling into recession. I’m not sure how to interpret this without knowing the direction of causation between the yield spread and future growth.
I have an explanation for the discrepancy between the yield curve’s weak signal and those stronger indicators cited by Madan Manoharan. Corporate spreads and stock prices reflect expectations of profitability, whereas Treasury yields ultimately reflect expectations of inflation and of what the Fed will eventually have to do to maximize growth without sparking inflation. I suggest that the labor market (over several decades, but particularly over the past decade) has changed in such a way as to allow firms a large cushion of profitability without producing inflationary strains. Under these circumstances, the expectation is that the Fed will eventually find that it can and should reverse its tightening policy without risking inflation and that it will do so even as profits remain adequate.
In the late 1990s we experienced a booming economy but without much apparent strain on the labor market. In late 2003, we experienced rapid growth and booming profitability, yet the labor market remained weak. The labor market has since begun to recover, but by most indications (with the noted exception of the unemployment rate) it has a long way to go. Between the “strong but not stressed” labor market of the late 1990s and the “weak but not recessionary” labor market of late 2003, there is a lot of “neutral territory” which could allow for an economy that is “strong” in terms of profitability (hence low corporate spreads) but “weak” in terms of labor demand (hence low Treasury yields).
Hal, the natural tendency is for the long rate to exceed the short rate. There are two interpretations of the factors that can overturn that normal tendency. The first view emphasizes that the Fed controls the short rate directly but influences the long rate only indirectly. When the Fed tightens quickly, it will drive the short rate up relative to the long rate. Moreover, when the Fed tightens quickly, economic activity slows down. So, according to this theory, the predictive power of the yield spread has to do with the ability of the Federal Reserve to affect the economy and their often manifest tendency to go too far.
The second view of this relation thinks of the spread between long and short yields as reflecting the implicit forward interest rate (what you think at date t the cost would be of borrowing money at date t + 1 in order to repay at date t + 2). If period t + 2 is going to be characterized by poor investment opportunities (a low physical marginal product of capital) and low consumption levels (so that people would want to have funds at that time), there would be extra saving at date t + 1 and less investment at date t + 1 causing the interest rate between t + 1 and t + 2 to go down. According to this theory, forces outside the control of the Federal Reserve are the cause of both the narrowing spread and the subsequent economic downturn. Under this interpretation, although the spread moves prior to the downturn, it should not be thought of the cause of the downturn, it’s just reflecting the market’s expectation of that eventual downturn.
The reason I tend to favor the first of these views is that it seems to me that right now the Fed is very much in a position where they could choose to continue to raise short rates quickly or choose not to do so. I think it’s largely up to them what happens from here.
Morevoer, even if you accept the second view, to the extent that the Fed has some discretion here, it ought not to be slamming on the brakes if the economy is about to slow down significantly on its own.
But let me repeat, the current indicators (including today’s employment figures) otherwise look pretty good, so I’m seeing this primarily at this point as a caution flag for how quickly the Fed should think about raising rates beyond the next two presumed hikes. I think they should pause at 3.75.
Dr. Hamilton:
I hope that you are well and thanks for commenting on the Conference Board’s recent change in their composition of their Leading Index of economic indicators.
Back in 2000, while speaking with Victor Zarnowitz at the Conference Board, who I believe is responsible for helping assemble that index, he remarked to me that what one should focus on is the ‘relative’ changes in the spread and there they are taking place on the curve. That is, is the short-rate increasing relatvie to the long-term rate, or is the long rate falling relative to the short-rate? Both have different economic implications.
In the current context, I do share you view that should the Fed continue with its aggressive rate hikes at the short-end the lagged effects on real gdp growth several quarters out from now may impact output in a not too friendly manner. What’s more, with Chinese demand for oil slowing, in addition to a potential slowdown in US real gdp growth 2 – 5-quarters out from now, the energy market could get hit with an unexpected decline in the demand for crude, which has the potential to result right back into the cyclical behavior for crude oil and therefore drop prices considerably.
Thus, the words of the late MIT Professor Rudi Dornbusch that once stated, ‘the Fed has murdered every post World War 2 expansion’, may ultimately ring true once again despite the Fed being too busy worried about excess stimulus while at the same time contractionary effects from past rate hikes will begin to impact certain areas of US output in the near future….
If I might ask a question, don’t the GDP and other economic indicators of today represent the results of decisions made 6mos to a year or more ago, and in that context, given the amount of stimulus generated by the long period of low rates, shouldn’t things be much MORE inflationary and with much better growth and wage numbers than they currently are? Aren’t these generally moderate numbers more of a warning sign that in an environment of now rapidly rising rates, we should be genuinely concerned about what conditions will be like 6 mos to a year FROM now and not so sanguine about today’s numbers?
Under the “we gave it our best monetary shot, and THIS is all we got out of it?” theory…
Thanks for your thoughts.
Yes RM, I think this is the question we’re all trying to sort out. As I’ve mentioned in some previous posts, the inventory component of the latest GDP figures makes me a little less pessimistic. But I share your concerns.
Interesting. Sounds like it’s still very much an open question.
Thanks.
Isn’t the yield curve better defined by the 3 month and the 30 yr?
It’s pretty flat now, but another month or so of higher-than-expected job gains and wage increases might set off a grease fire under those long-term rates.
My guess about the Fed is that they’d rather over-react than under-react. If they over-react now and things reverse then they can over-react in the other direction later. But if they under-react now then things can get really out of hand and they will have to super-over-react later, which would be worse.
Carnival of the Capitalists
Welcome to this week’s Carnival, sponsored by MarketingLinx, the best marketing site on the web, as determined by you. Contribute a link today to get your reward! This week’s contributions were fantastic. Hosting really is a different experience from r…
You might want to put the earlier comments on quality spreads in the context of the economic cycle. Quality spreads tend to move in lock step with capacity utilization, so changes in the quality spread have an extremely strong cyclical pattern. Consequently, a widening of the quality spread is forecast a slower economy when the default risk from more risky investments rises. The quality spread tend to be as good a leading indicator as the yield spread.
For Brian Wesbury’s always excellent comments on this, see his Monday Morning Outlook (http://research.claymore.com/docs/w080805%20-%20The%20Fed,%20the%20Economy%20and%20Stocks%20.pdf):
“Over the past 15 years, inversions of the yield curve have only occurred when the Fed pushed the real (or inflation-adjusted) federal funds rate above 3.9%. Today, the real federal funds rate is 1.33%, well below a neutral level. The narrowing yield curve has been caused by the carry trade, a strategy of borrowing at the short end of the curve and investing at the long end, not excessively tight money. We forecast that the Fed will lift rates on Tuesday and then at each of its three remaining meetings this year. But, even a year-end rate of 4.25% will not cause economic damage. We continue to forecast 4.0% real growth into 2006, rising inflationary pressures, strong profit growth, buoyant equity markets, and much higher long-term bond yields.”
Not quite sure what you’re suggesting Tim. Are you saying that, because real rates aren’t that high, the yield curve won’t invert, or that, even if the yield curve does invert, because real rates are below 4%, it won’t matter?
The point of the studies I was drawing on was that the yield curve doesn’t have to go all the way to inversion before it should start to factor into your forecast.
1. How mechanically does the Fed drive interest rates, how does this create money supply and who gets the inserted money supply?
2. Why do I read a lot of things that don’t make intuitive sense. The Fed raising rates in response to high growth (surely increased production would need increased money supply, no, to keep value of currency same?
3. Why did the WSJ have a story about the Fed trying to drive long term rates down by raising rates. Then to drive them up by…raising rates. Especially when long term rates mostly are driven by expectations of future inflation of the money supply not present, anyhow?
Yikes, TCO! Answering those three little questions (and all their subplots) seems like a job for a full college course, not my little blog!
Here’s a very nice free online book from the Fed:
http://www.ny.frb.org/education/addpub/monpol/
Maybe if some other readers have some similar requests, I can take a stab at some of these at some point.
Yeah. Ok. I will buckle down and look at it. IT seems like at the end of the day, everything is finance (NPV) and supply and demand curves. I don’t understand how it all works, but I know enough to see things in the press (and even from economists) that are not self-consistent or that seem to discuss things without really thinking it back to basic supply and demand.
Is there a good textbook of the level of popular intuition as Brealey and Myers Corporate Finance or Copeland’s Valuation, but that discusses monetary policy?
Walking the tightrope
So here we are again on the day of an open market committee meeting. Not much to talk about in the way of what they will do, but plenty to talk about in the way of what they should do….
A good part of the discussion above seems to be about whether to make a recession forecast out of the currently narrow spread whilst short rates are still low (though rising). Here is a recent academic study shedding light on this:
Andrew Ang and Monika Piazzesi from Columbia/Chicago have a paper forthcoming in the JoEconometrics (http://gsbwww.uchicago.edu/fac/monika.piazzesi/research/GDP.pdf) where they estimate the relationship between GDP growth and interest rates using the entire termstructure. Their technical “trick” is to impose no-arbitrage between the various yields. If they did that in the right way, their estimates gain efficiency. The results say that the short-rate is a more reliable predictor (see the paper why think so). Since current short rates are historically (still) low, their results would imply that there is no recession expected to be around the corner.
Of course this runs against all the spread-based evidence quoted above. Prof. Hamilton: Do you have any views on whether the No-Arbitrage structure imposed in that paper is more help or hindrance for inference?
btw: Thanks to all for the interesting discussion above!
emt
Elmar, I’m a big fan of the Ang, Piazzessi, and Wei paper and general research program. I see that paper as saying that the short rate has information in addition to that contained in the spread, not that the spread has no predictive power.
One way to get more accurate estimates is to impose restrictions, which is the Ang, Piazzessi and Wei approach. Another alternative is to get more data. The experience from looking at a large number of different countries is that the slope of the yield curve has important predictive power.
Looking at the two diagrams why not to say, for sake of profane simplicity,” ooh, if the interest differential becomes less than 1%, we are going into recession” ? (no phd granted).
macro sux. micro rulz.
Recessions in the American economy are brought on by such a complex web of domestic and international factors — social, political, and economic — that I wonder how much fed tightening really contributes to the problem.
Perhaps we give the Fed too much credit.
If the American economy is inherently cyclical in nature, if it is prone to periodic booms and busts, then to what extent can the Fed really avoid recessions and guarantee a soft landing?
Even assuming Fed influence over short term rates is the primary cause of recessions or contributes significantly to recessions, it is probably impossible for the Fed or any economist to predict the optimal measured pace at which short term rates should be raised.
One concern I have is that long-term rates are determined by the market place. And while the market place is often efficient, it is not always efficient. And while economic actors are often rational, at times they act irrationally.
Is BAIDU really worth $100 a share in the long-term? Was YAHOO really worth say $1500 a share at a PE ratio of 300 in the year 2000 (I don’t know the exact figures). Were Japanese investors who overpaid for Hawaii real estate in the 1970s and 1980s correct? They determined the inflated market price for land by overpaying. Were the prices set by the market rational?
Why assume that buyers and sellers of long-term bonds are paying rational prices? They are assuming a 2% inflation rate long-term. Why should their assumption be accurate?
What alternative does the Fed have. If the Fed does not raise rates, we may face an overheated economy and hyper-inflation.
If economies are cyclical in nature, a recession must occur sooner or later.
Shouldn’t we be more concerned about fundamental economic factors? The Arab oil embargo in the 1970s and the high price of funding the Vietnam War certainly hurt the American economy in the 1970s.
Shouldn’t we be more worried about $65 a barrel oil prices and the cost of funding the Iraq invasion? Isn’t it mainly those fundamental economic factors which bring on recessions, rather than the Fed’s efforts?
The main problem is that this inquiry lacks any meaningful control.
If the Fed were to pause, will that really prevent a recession, or will it merely delay the onset of an inevitable recession by 3 months?
The Fed should just keep current inflation under wraps. They shouldn’t care about recession or especially about long term investor decisions (equity or long term bond rates). Just keep the ship in the channel. don’t overdrive it. Wathc a couple bouys down. But don’t worry about the rest.
I don’t actually know anything. that is just my intuition after reading all the weird stuff from maco people in the newspaper.
ENJOY THE BUSH MICRO-BOOM WHILE IT LASTS
What is the yield curve, and why should you care? It has nothing to do with Katherine Harris. Professor Hamilton (Econobrowser) conducts a useful little seminar, and one of his commenters reminds us of the Curse of (Rudiger) Dornbusch: “[T]he…
just get rid of the Fed. their political manipulation of the FF rate, not to mention their constant over- and under-shooting, shd be replaced by a simple number (function of trend productivity and population growth), call it 3%, and just not tampered with. over time things will work themselves out, both busts and booms.
It seems like the EU’s version of a Fed does less manipulation/gaming/fiddling with the system. Rationally this will do less damage and the expectation of less damage should help those making long term capital decisions. Unfortunately the structurally greater dynamism of the US economy (more capitalistic, growing population, etc.) makes us usually do better. So people get confused and think our Fed deserves credit.
(Mis) leading indicators?
This is a belated post, inspired by a 17 August blog by Mark Thoma on The Use of Leading Economic Indicators in Economic Forecasting. Mark cites a paper by Stock and Watson and a Bloomberg column by Caroline Baum, which asked why economic forecasters …
Could someone explain what are the factors to determine slope of yield curve? Thanks.
Steve, the slope of the yield curve is determined by whether short-term interest rates are expected to rise or fall (rapidly rising rates would mean a steeper yield curve) and the extent to which investors place a premium on short-term relative to long-term bonds. The second effect is inherently harder to explain but it is very clear from numerous academic studies that it is there and it matters.
My own thoughts about why the yield curve helps predict economic downturns were described in this earlier post.
i want to know the fact that change the slope of the yield curve.
Like so many things in Economics, an inversion is more complex than a 30 second tv snippet allows for.
Yield curve inversions have invariably led to economic wobbles and slowdowns; More severe inversions (duration and depth) have led to recessions.
That given is what made the LEI change (and its tortured methodology) so absurd. Essentially, the CObnference Board replaced a reliable economic indicator with a goofy one — and then added a methodology that defied common sense. Yes, unreserved criticism is what the cheerleaders got and deserved.
I expect that they will eventually beat a (not so) hasty retreat, and reverse themselves.
So far, we have seen only a mild and shallow inversion. That’s a warning; If it deepens and continues, it will have more ominous implications
The best explanation I’ve read to date comes from former Dallas Fed Economist Lacy Hunt:
http://bigpicture.typepad.com/comments/2005/12/explaining_yile.html