Test out the Claudia Sahm rule as a timing strategy

Research
1

I see there are different versions of using unemployment stats as a timing indicator to stay out of the market.
I also understand one of those who have done well is the Claudia Sahm signal.

Outlined here;
https://www.forbes.com/sites/randybrown/2020/01/09/this-new-rule-to-identify-recessions-could-give-investors-an-edge/?sh=3b1dff4255be
https://alfred.stlouisfed.org/series?seid=SAHMREALTIME&utm_source=series_page&utm_medium=related_content&utm_term=related_resources&utm_campaign=alfred

And has been tested on CXO, with the following approach:

"stock market timing strategy that combines the market 10-month simple moving average (SMA10) and the Sahm Recession Indicator (Sahm), which signals the start of a recession when the 3-month SMA of the U.S. unemployment rate is at least 0.5% higher than its low during the last 12 months. Specifically, the strategy:

Holds the S&P 500 Index (SP500) unless it is below its SMA10 and Sahm first signals a recession.
Subsequently holds cash until SP500 crosses above its SMA10."

Is it possible to investigate such a strategy on this platform?

2

Here’s a start:

https://www.portfolio123.com/port_summary.jsp?portid=1657701

3

Thank you so much.

It comes as no surprise that you have already tested the strategy:)

Is it possible to make it in screens? It’s easy enough to get the “ticker” and the condition of purchase “Close (0, $ SP500)> SMA (210, 0, $ SP500)”

But is it possible to enter the condition for sale?

ticker("spy")
Close (0, $SP500) > SMA (210, 0, $SP500)
SMA (3, 0, ##UNRATE) > 0.5 + LowVal (12, 0, ##UNRATE) > 0 and Close (0, $SP500) < SMA (210, 0, $SP500)
4

No, screens have opposite buy/sell rules. In other words, the sell rule is implicitly the opposite of the buy rule. If you want the buy and sell conditions to be different, then you have to use a simulation.

5

Whycliffes -

The Unemployment Rate has historically been a highly accurate indicator for the start of bear markets. However, there is a crucial problem with using it as a timing indicator.

Logically, when small businesses (collectively, the US’s largest employer) begin to see a significant decline in traffic/sales, they usually cut the cost of their traditionally most significant expense first. They can’t do anything about their sunk cost for plant/equipment or commitments like rent (think of a restaurant or retail business), but they will lay off employees who aren’t needed when the customer base shrinks. Employees are usually the first thing to go. Therefore, historically, a rise in the unemployment rate has been highly correlated to declines in GDP - i.e., the start of a recession.

Likewise, equity investors are especially sensitive to economic contractions, which have historically caused a correlated earnings and stock price contraction. Hence, a significant rise in the Unemployment Rate is highly correlated to stock-price declines. An increase of 0.5% above a low is used because it helps investors discriminate between random noise and meaningful signals.

HOWEVER, for the first time since Unemployment statistics began being tracked in 1948, the Unemployment Rate in February-March 2020 WAS NOT an accurate signal for the beginning of a bear market. It failed in 2020 partly because of the idiosyncratic nature of the Covid Pandemic and partly because the Unemployment series lagged the market so much last year.

Regarding the Unemployment Rate being an effective timing tool, it’s clear it has been effective historically because economic contractions have always been a progressive, cumulative series of events that results in the deterioration of the economy and corporate earnings. One factor after another begins to turn south, we can usually see the progression of negative influences building up. Finally, there is a point of support that cracks and the dam breaks.

For example, in the ‘Financial Crisis’ recession, we saw negative market factors growing through all of 2007 as the pressures in the housing market piled up. Breadth measures such as the Advance-Decline Line turned down months before the market did. Moreover, the S&P 500 high occurred on October 10, 2007, but the actual decline didn’t start for two more months, and it took 18 months for stocks to find a bottom.

In the Covid Crash, the market’s decline began on February 24, 2020, but the Unemployment Rate was still at 4.40% (considered ‘Full Employment’) at the March 6, 2020 labor report for February. Since we don’t get updates of the prior month’s Unemployment Rate until the first Friday of every month, that data series is seriously lagging in an event like the pandemic. Worse, we then apply a threshold (based on past prices) to generate a signal, making it lag even more. By the April 3 report, Unemployment in March had skyrocketed to 14.8%, and more than 20 million Americans were out of work.

As you can see from this chart, stocks had already begun to climb higher and were in a powerful rally when the Unemployment Sets turned downward:

As a result, the Claudia Sahm timing you highlighted was a terrible indicator in 2020. You can see from this Custom Series that the market was already in a powerful uptrend when the timing system went to cash in 2020:

You would have been required to absorb ALL of the -34% selloff in March 2020, and exited your positions while in the rally off the bottom. Also, the signal performed in a less than stellar manner in 2001, exiting the market when the dot-com crash was more than halfway over.

When you combine BOTH the Unemployment Indicator AND the 200-day Moving Average (which is also a very lagging indicator), you have the ultimate system to make you regret ever hearing the name Claudia Sahn! haha

There are far better risk-assessment and avoidance signals that are available to you on P123. For example, here are the signals for one of many indicators that I use:

If you followed this signal alone, you would exit in a timely manner just before this risk indicator increased over a certain threshold and re-enter the market just after a rally begins. It has some whipsaws, but when combined with several other uncorrelated indicators from the [color=blue]50+ Indicators[/color] I use on my site, those whipsaws are virtually eliminated.

It will take time, but you’ll get the hang of it after a lot of trial and error – and continue doing what you’re doing: Think outside the box and ask questions. Best of luck!

6

The indication cannot be generated here?: https://www.portfolio123.com/indicators?uiu=1737

7

That looks very good at first glance. I haven’t used the Macro Charts much myself . . .

8

Please accept my apologies for my poor English. That was a question, not a statement. Is it possible to implement the Claudia Sahm rule in a macro chart?

9

ETFOptimize,
Above statement is not quite correct. My Unemployment Model did signal recession end of March 2020.

It also signaled on July-3-2020 that the recession had ended.

EDIT The title on the Figure should read July UER= 5.4% (NOT September)

Fig-8.-9-26-2021.png
Fig-8.-9-26-2021.png1364×943 113 KB

10

I think so–or at least you can come very close. See https://www.portfolio123.com/doc/side_help_item.jsp?id=34 and if you have specific questions, contact Paul, who knows more about these charts than I do.

11

Hi Georg,

I think we are talking about two different things. My point was that an increasing Unemployment Rate has historically (as you said, since 1948) been a surprisingly accurate indicator of a BEAR MARKET. However, my point was that 2020 was different, and unemployment DID NOT coincide with the Bear Market. I said:

I didn’t reference the RECESSION. I’m interested in timing the market and have no interest in when a recession occurs. It’s pretty well known that the market leads the economy and the S&P 500 is actually used as a leading indicator in the LEI. I seem to recall that you use the SP500 in your recession indicator also. Maybe I’m wrong on that…

What I said, and what my chart showed, was that the UER (Unemployment Rate) - for the first time - was NOT an accurate coincident indicator for the start (and end) of the Bear-Market downturn. So my statement was 100% correct based on what I said.

I think you misunderstood me. In fact, you said:

Exactly! But at the end of March 2020, the stock market had already bottomed (March 23) and was headed sharply higher!

And…

Yes, the recession may have ended in July, but that fact provided no benefit for investors (imo). For me and my clients, we are interested in market-timing signals, not recession-timing signals.

By July-3-2020, which you mark as the end of the recession, the S&P 500 had already gained a phenomenal 42% and reached 3,179.72 from its low of 2,237.40 on March 23. You mark the Covid-related economic recession in the summer of 2020 with dates (late March to July 3) that coincided almost exactly with the initial, turbocharged phase of the BULL MARKET, rather than marking a Bear Market.

Am I wrong about these dates? Georg, your work is extraordinary (imho), but as was my original point, the Covid Crash was different than any other Bear Market we have ever seen in stock market history. Check the releases - at the time the stock market (SP500) had bottomed on March 23, weekly Unemployment was just beginning to go up.

Moreover, we’re still seeing bizarre market impacts this week – 18 months later! —All created by an out-of-control Federal Reserve Board with bankers that have gotten in WAY over their heads in trying to manipulate markets (in my opinion).

12

Hi Chris,
I agree with your assessment that the market did not respond, as it did in the past, to the 2020 recession. Apologies that I misread your post.

Like you, my interest is also to time the market, problem is that over shorter periods it is impossible to do.

13

I agree with you Georg, over short-term periods (daily and shorter periods, imo), it is virtually impossible to generate highly accurate timing signals.

The primary reason for this fact is that there is too much noise in daily prices. Price can be very volatile during short-term (daily) periods because experienced traders and institutions are trying to push prices to extremes to attain their most profitable trade targets. Inexperienced investors don’t know what they are doing and will buy or sell at extreme prices as a reaction to conscious or subconscious emotions, affecting prices at the margins.

There is just too much noise in short-term prices for a timing system to be accurate. On the other end, I find that monthly pricing is too long and will miss out on critical market infection points.

For these reasons, I generally use weekly timing signals, which corresponds nicely with P123’s weekly updates of fundamentals and many macroeconomic factors. Weekly signals are far more accurate because a large percentage of timers don’t want to have over-the-weekend exposure to impactful events. An obvious example is the Lehman moment in 2008 when the firm’s bankruptcy was announced on Sunday evening.

With so many investors eliminating positions on Friday close to avoid weekend exposure, the weekly price tends to be far more accurate from a ‘fairly valued’ perspective. Obviously, the longer the span, the more aligned with the fair value an equity’s or ETF’s price will be. Moreover, ETF prices are much more accurate because ETFs hold dozens, hundreds, or even thousands of individual equities, and weekly timing offers a high probability of more accurate signals. Therefore, I invest with ETFs on a weekly basis.