Hacker News Reader

Comment by the_watcher | Hacker News Reader

> This matters to investors because, although bear markets are only fair predictors of recessions (seven of 13 postwar bear markets were followed by economic downturns)

Roughly the same predictive power as flipping a coin isn't what I'd call a "fair predictor".

View on HN

The yield inversion predictor is 100% over the past 40 years with no false positives. The only problem with yield inversion is that it may be 2-years or 3-years early. (2005 yield inversion was followed up by recession in 2007).

Here is a graph of 10-year minus 2-year, a yield inversion is whenever the graph dips below 0%.

https://fred.stlouisfed.org/series/T10Y2Y

The yield inversion predictor is incredibly powerful.

----------

EDIT: Part of the reason for yield inversions to happen is that a large number of people are buying long-term bonds, because they think a bear market will exist. Its better to hold onto a 10-year bond through a recession, because short-term rates will drop during a recession.

We don't quite have a 10-year inversion yet, but we have a 7-year inversion. 5-year yields less than 1-year at the moment. So investors prefer making less money on a 5-year (to guarantee a stable interest rate), rather than 1-year.

That means a large number of people are predicting a recession.

The problem is there will be a downturn at some point. In some of those cases on your chart the market went up for a while after the negative yield event then turned down. So event y (Recession) is going to happen at some point after event x (Yield inversion).

If some one keeps claiming recession they happen will be right at some point.

Do you have a better indicator of recession?

The yield inversion is ALWAYS within 3-years of a recession. That's pretty darn good in a world where candlesticks and dead cat bounces are your "technical" indicators...

There's even good logic for why the yield inversion is predictive of recessions: because when people buy 10-years (even if the 10-year is a lower-yield than 2-years), it means that people expect the next 2-years to suck.

So its better to lock into a long-term 10-year yield, than to hold onto a better yield for only 2-years.

--------------

Note: The 10-year has NOT inverted yet. But it is only 0.10% away from inverting.

Typically business cycles of the us economy last about 38 months. So your "predictor" that comes two years early isn't predicting anything. An analogy would be saying that you can predict 50/100 Blackjack games with 90% accuracy. Sure- you can, but it's a meaningless prediction that has no correlation to any individual game.

Example: https://awealthofcommonsense.com/wp-content/uploads/2015/02/...

I'm not sure if you've studied other "indicators".

But the point I'm trying to make is that we ALL know that indicators in economics are kinda bad. But of all the bad indicators economists have, the Yield Inversion is one of the most powerful, predictive, and correct indicators in existance.

That indicator alone should give people pause. Well, when it happens. The 10-year has NOT inverted yet, but its darn close to inverting.

Hence the good old phrase: "The Stock Market Has Predicted Nine Of The Past Five Recessions"

Yeah, but that doesn't apply to the yield inversion indicator. The yield inversion doesn't have a false positive over the past 40 years.

The 5 times the 10-year yield inverted, were followed by 5 recessions: 1980, 1984, 1991, 2001, and 2008. That's 5 for 5.

When your sample size is 5, claiming your algorithm has a 100% success rate and no false positives isn't saying much at all.

MY sample size is 40 years.

And there are other studies that go back 100+ years with similar conclusions.

https://faculty.fuqua.duke.edu/~charvey/Research/Thesis/Thes...

So it isn't a perfect indicator over 100+ years, but its still a damned good one. Furthermore, there are solid fundamental economic principles and calculations which back up the thesis. Yield Curve seems to predict GDP (or more specifically: the market participants predict a recession, which causes the market participants to buy/sell bonds in a specific way to make an inverted yield curve).

This isn't some hokey theory positioned by bear websites like zerohedge. This yield-curve inversion is a fundamental attribute taught in major schools of modern economics.

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.519...

> The historical record does not show this connection to be only a post-WWII phenomenon. The yield curve inverted between June 1920 and March 1921 and again between January 1928 and November 1929.7 Data from the 19th century are incomplete and do not easily lend themselves to analysis.8 Nevertheless, support for the thesis of the yield curve as a predictor of business cycles can be traced as far back as the mid-1800s.9

We're looking at a pattern that has been relatively reliable since the mid-1800s.

Inverted yield curves STRONGLY (but not perfectly) predicts recessions over a period of ~150 years, and perfectly predicted recessions of the last 40 years. This is one of the best indicators of all of economic theory.

> The only problem with yield inversion is that it may be 2-years or 3-years early.

The problem in finance is that early (or late) is often no better than wrong.

Depends on the goal. If you plan to make money off of indicators, well... you can't. Any powerful indicator will be gamed by other investors.

If you are setting policy, it takes years for tax-policies to have an effect on the economy. We can change the tax law today, but it won't really come into effect until next year, or the year afterwards.

On the scale of policy changes, a worst-case 2-year lag on an indicator is perfectly acceptable.

Predicting something 2-3 years early is not a "prediction", it's a failure to predict.

I'll suggest better predictors of the health of the economy would be 1) Record domestic manufacturing output, 2) Most energy security in history (US is now a net exporter of oil), and 3) lowest unemployment since 1969. The US Economy is firing on all cylinders and I wouldn't be surprised if a strong economy runs for another 2-5 years. This is not all that unusual, since the US is returning to a historical average GDP growth rate of 3-4% after an abysmal decade of 1-2% growth.

> 3) lowest unemployment since 1969

What was the US's unemployment in 2007, before the crash? It was 5%.

https://www.bls.gov/spotlight/2012/recession/pdf/recession_b...

Recessions CAUSE unemployment. You've got cause-and-effect backwards. First comes recession, THEN comes unemployment. Usually, unemployment happens as the recession recovers.

In 2009, unemployment was 10%. But guess what? The recession was over, and 2010 was one of the best recoveries of all time.

> 1) Record domestic manufacturing output

That's practically a tautology. When US GDP shrinks (and a large portion of GDP is manufacturing), you have the DEFINITION of a recession.

--------------

You're looking at historical indicators of a Recession. IE: When everyone is unemployed and the factories are closed, we can look back and say "Yeah, a recession happened 6 months ago". But there's literally NO predictive power in those two attributes.

------------

> 2) Most energy security in history (US is now a net exporter of oil)

While this is likely good policy, it doesn't seem to be indicative of recessions or not. Case in point: we've grown incredibly in the 1950s when we were importing oil. The 80s and 90s were also built on top of foreign imports, but that didn't stop the economic boom times.

> Predicting something 2-3 years early is not a "prediction", it's a failure to predict.

Why do you say that?

Predicting the economy will experience a recession is like predicting a person will die: you'll be right eventually, but without knowing when, you can't do much useful with that information.

IMO, this is the time for policy makers to start to debate policy changes. It takes over a year for tax laws to have any effect for example, and maybe 2 or 3 years before tax laws really kick in and affect the economy. (First year, people will be typically surprised at their new refunds instead of really spreading it over the year)

Trump's Tax Cuts were written into law in 2017, but only these months (tax season 2019) is when people are fully understanding the scope of the tax law, as it hits our pocketbooks.

A surefire predictor within 2-years or so is good for policymakers, but probably bad for traders.

I think before you can declare any indicator a success or failure, you need to put into context what decision you are trying to make off the indicator.

If you are trying to decide wether to dump all your stocks into treasury bills in the next 90 days, but before an actual recessionary drop, then yes it's a poor indicator (and likely there is no indicator to answer this question).

But if you're looking at a long term allocation of your assets, or a corporation trying to plan investments long term then maybe it's relevant.

It's hard to look at the long run GDP rate (from the50's on) and think we're going to return to past rates until we fix our inequality problem. They've been inversely related for multiple decades now.

In December of 1994 the 10 year minus 2 year was at 0.15 (currently it is 0.18 in Feb 2018), but then we had to wait for the next recession until 2001. It could go back up like in August of 1984 or January 1995, or continue downward and put us in a recession in a couple years.

Fair, we're not in a 10-year inverted scenario yet. But we're pretty close.

> (currently it is 0.18 in Feb 2018)

Note that the 1-year is inverted with the 2-year. So while 10year-2year is 0.18, the 10year-1year is 0.1.

https://www.treasury.gov/resource-center/data-chart-center/i...

The lower-end of the curve is ALREADY inverted. But that's not "the indicator", the 10-year is the indicator that is tracked. Still, seeing the curve invert between 1-year and 5-years is worrysome to me.

This is a less reliable indicator: but when the lower end of the curve inverts, usually the higher end inverts soon afterwards as well.

I think we dipped briefly during intraday trading a few weeks ago.

On the 10-year?

We've definitely dipped on the 5-year and 7-year. But I'm not aware of any inversion on the 10-year (even an intra-day one). Or was it on the Global Bond market? (I think I recall one story about the global bond market becoming inverted...)

Hmmm... do you have a citation by any chance? I'm definitely interested if you can remember where you saw that data, for the full details of the situation.

Turns out I misremembered. I was thinking of the inversion which happened in December, which was as you said, the 3- and 5- year briefly inverted (by 1 basis point).

Everyone arguing with you seems to be fixated on how this is framed as an indicator instead of what yield curve inversion actually means and the effects it has that likely lead to a recession.

It looks like inversions are 'predicting' boom economies. Recessions follow booms. Here's the chart with real median income added. Notice that inversions coincide with local maxima of median income (stock market level would probably be a better comparison but S&P only lets FRED show the last ten years of their data)

https://fred.stlouisfed.org/graph/?g=mXhU

I don't think that's the right logic.

Say there are ten "events" that have had recessions follow them (or not). Each of these events happened 10 times.

For the first type of a event, a recession happened just once afterwards.

For the second type of event, a recession happened twice.

For the third type of event, a recession happened three times

The third, a recession happened three times.

The fourth, a recession happened four out of ten times.

The fifth, a recession happened five times.

Etc...

This has very little to do with flipping a coin, and much more to do with deciding the right time to pay attention.

Half a chance of getting hit by a car is not the same as "flipping a coin."

Generally, throughout the past hundred years or so, there has been much less than a 50% chance to enter a recession. I don't know the numbers, but for any given year, it could be 10%. If there is now a 50% chance, isn't that a 5 fold increase in risk?

Yes, this is just pretty simple Monty Hall/Bayes stuff, your example is on point.

50% is 50% better than 0%. If you knew there was a 50% chance you'd die on your way to work today, you'd probably call off work. If you know there's a 50% chance a recession is coming, as an investor, you take similar precautions.

Reply on news.ycombinator.com