Many people look at the spread between the yield on 2-year and 10-year Treasury securities as a forecasting tool for the business cycle. It is indeed a much better than average economic indicator, often “inverting” about a year before a recession. But there are also several misconceptions about the yield spread, which I’d like to discuss here:
1. You cannot reliably manipulate the economy by manipulating the yield spread. This is an example of the “Lucas Critique”. The yield spread predicts output because movements in short-term interest rates are strongly correlated with the business cycle. But changes in the yield spread do not cause changes in the business cycle. If the yield spread had a causal impact, then announcements of QE programs aimed at flattening the yield curve would have had a contractionary impact on asset prices. But they didn’t.
2. The implications of a sharp decline in the yield spread are far different from the implications of an outright inversion of the yield spread. This is important, as the following graph demonstrates that the yield spread has recently declined sharply, but remains positive:
There were similar sharp declines in the yield spread in 1977, 1984, and 1994, and yet in each case there was no recession in the near future.
3. The yield spread is not a particularly good indicator of whether money is too tight. Consider the following two yield spreads:
February 2006: minus 0.14%
December 2007: positive 0.99%
The February 2006 inversion did not indicate that money was too tight at that point in time. In contrast, money was clearly too tight in December 2007, and yet the yield spread was about normal, which is roughly 1%.
The yield spread also inverted in June 1998, and yet money was not too tight at that time. Nor was money too tight in January 1989, when the yield curve inverted. And obviously money was not too tight in the late 1970s. Don’t make the mistake of equating “yield curve is inverted” with “money is currently too tight”.
The best measure of whether money is too tight is the Hypermind NGDP prediction market price. And that’s a really, really sad comment on the failure of our policymakers, as well as the economics profession as a whole.
PS. David Beckworth has a somewhat different perspective on the yield spread.
READER COMMENTS
Thomas
Aug 1 2018 at 2:24pm
This raises a question I’ve often wondered: Why does the yield curve often invert before a recession? The textbook explanation I learned is that investors, anticipating a downturn, flee to safety. But an inverting yield curve indicates they’re fleeing to long-term government bonds (10-year or longer). Yet the Federal Reserve has proven to be a credible recession-fighter; since the Great Contraction of 1929-1933, the longest contraction has been only 18 months, and most have been less than a year. If that’s the case, shouldn’t investors be fleeing to 2- or 3-year bonds? And thus, shouldn’t we expect that an increase in the slope of the yield curve, rather than an inversion, would be the signal for a looming recession?
(And yes, this sort of thinking is why I stay away from macro–my instincts are almost always the opposite of the textbooks.)
John Hall
Aug 1 2018 at 3:40pm
Scott, you mentioned a lack of comments on moneyillusion, so here goes…
On 1: I’m still solidly behind the idea that it is monetary policy that is the driver of most of the U.S. business cycle. The yield curve is only relevant to the extent that it reveals the stance of monetary policy, even if it is only imperfect.
Anyway, ignore the yield curve for a moment and consider just short-term rates being near zero. This causes a liquidity trap (people don’t want to put money in banks, hold greater cash balances) and can short-circuit traditional monetary policy. If monetary policy fails to adjust, then it can cause problems for the economy. By contrast, if the yield curve inverts, then banks have less of an incentive to lend. This also disrupts monetary policy, though through a different channel. In the liquidity trap, the issue is people increasing cash balances. In the yield curve situation, it is a collapse of the bank lending channel. However, the central bank does have the ability to offset this (by increasing money supply and driving down rates) in contrast to the liquidity trap scenario. Nevertheless, I suspect that both of these effects are non-linear in some fashion, which is why they are so important.
On 2: From a statistical perspective, you might find it interesting that regressing the investment quantity index and its lags on the 10-3 month spread and lagged changes does not give statistically significant coefficients for the spread and its lagged changes. However, there does seem to be some modest evidence of a error correction process between the de-trended investment quantity index and the yield spread.
On 3: I agree.
stoney batter
Aug 1 2018 at 3:57pm
Hi Scott. Longtime reader but I’ve never commented. Since you noted the lack of comments on this particular post, I’ll wade in:
You said: “changes in the yield spread do not cause changes in the business cycle. If the yield spread had a causal impact, then announcements of QE programs aimed at flattening the yield curve would have had a contractionary impact on asset prices. But they didn’t.”
There are probably some confounding variables that drive both the yield curve and the business cycle. Absent a policy shock (like QE), the combination of factors that flatten the yield curve may still signal a recession: e.g. risk aversion, expectations for less near-term policy accommodation, and reduced expectations for longer-term nominal GDP growth.
Usually when sales of ice cream rise, the murder rate also rises. The former does not cause the latter, but both are caused by hot weather. If the Fed holds an ice cream social and boosts the sales of ice cream on a given day, the murder rate would not change. This does not mean that next week, when the ice cream social is over and it’s very hot again, the two variables won’t continue to move in unison again.
Rajat
Aug 1 2018 at 5:25pm
Agree with (1) and (2). I think (3) is also right, but the yield curve inversion signal tends to occur 1-2 years out from a recession. Probably rate cuts within the year prior to a recession cause the inversion to disappear. This is consistent with what you’ve said before, that central banks like the Fed are slow to shift into gear and like to move ‘deliberatively’. This means that while they cut rates as they notice the economy slowing, they cut too late and too little to prevent recessions – hence no ‘mini-recessions’.
This leaves the question of why the Hypermind market is not giving any warning that the Fed is close to triggering a recession. Maybe it needs to look further out in advance? Maybe it just doesn’t have enough participation to be efficient? Tyler Cowan’s conversation with Vitalik Buterin was interesting in this regard.
Whatever the reason, I think we have to be very careful in dismissing a metric that has been a useful and correct early-warning signal over the last 40 years or so.
Scott Sumner
Aug 2 2018 at 1:02pm
Thomas, The standard explanation is that rates fall sharply during recessions, and since long rates represent the average of expected future short rates, they fall when a recession is expected.
John, I’m surprised that investment is not correlated with lags of the spread.
Stoney, Good ice cream analogy.
Rajat, Just to be clear, I’m not “dismissing” this indicator, I’m agreeing with it. The indicator does not forecast a recession, at the moment. It forecasts slower growth. And I’m on record predicting slower growth ahead.
As far as Hypermind, I’d point out that the stock market also does not currently forecast a recession, and it has a LOT more liquidity than Hypermind.
Good point about 1 to 2 years ahead, but this is actually a problem for the yield spread. The original models suggested a recession within 12 months, which was often the case in earlier decades (1920-70). In more recent decades, there seemed to be a longer lag, which led to the 1 to 2 year window. But since expansions only tend to last for 5 years on average, we need to be careful not to engage in confirmation bias. Changing the window is a bit of a red flag.
All that said, I still regard the yield spread as one of our best indicators, subject to the three caveats in this post.
Willy2
Aug 2 2018 at 4:50pm
“The yield spread predicts output because movements in short-term interest rates are strongly correlated with the business cycle.” ??
The US was already in a (shallow) recession in 2005 & 2006. But the yield curve inverted in very late 2005/very early 2006. But the yield curve started to steepen again in the 2nd quarter of 2007, when the US recession deepened. So, interest rates are not correlated to the business cycle.
Ricardo
Aug 2 2018 at 8:51pm
Here’s my naive view: the (2yr) shorter-end rates are mostly Fed oriented, while the (10yr) longer rates are more free market oriented (r*, longer run productivity, desired savings vs desired investment, and all that). So, the case of an inversion means the market expects the Fed will (over?) tighten more in the short-run than the expected longer run balance of desired savings vs. desired investment.
Andrew_FL
Aug 6 2018 at 10:52am
If there’s nothing more to the business cycle than an economy that just hums right along until it gets knee-capped out of nowhere by tight money, how can the yield curve inversion be a pretty good predictor of a recession in the future if it’s almost never inverted when money is actually tight? How does the yield curve somehow “know” that policy makers at the fed will in a year or so suddenly lose their minds and strangle an innocent boom in the crib?
Comments are closed.