Unemployment Rate over 12 Month MA Strategy
This thread discusses using the monthly Unemployment Rate (UR) as a long term means to time the market. Tom4Jean posted the original article,Using economic indicators to time the market | Investing For A Living, over in FogSailing's thread.
The basic idea is to track the UR versus its 12 month Moving Average. As long as the UR is below the MA you stay invested in the market. As soon as it rises above you get out. The idea is that this method can be used as a modification to Buy-N-Hold by getting you out of the market during economic downturns and remaining invested (holding) otherwise. The article claims great results, but I thought we should see for ourselves and thus this thread was created.
The first step was to get the data which I got from the Bureau of Labor Statistics here: Bureau of Labor Statistics Data
Starting in Jan 2000 I plotted this data is a spreadsheet and computed its 12 month MA. Using this system I came up with the following IFTs:
Jan 2000 |
IN |
|
Aug 2000 |
OUT |
Whipsaw |
Sep 2000 |
IN |
|
Jan 2001 |
OUT |
|
Oct 2002 |
IN |
Whipsaw |
Nov 2002 |
OUT |
|
Oct 2003 |
IN |
|
Jun 2007 |
OUT |
|
May 2010 |
IN |
|
Nov 2010 |
OUT |
Whipsaw |
Dec 2010 |
IN |
|
This gives us only 10 IFTs in 16 years so it does appear to be long term Buy-N-Hold with a few IFTs.
Looking at the dates we see that this system did keep us out of the 2008 market crash. Maybe it kept us out too long, but hey, at least you didn't lose your shirt following this thing. That is the kind of protection this system is supposed to offer. We also see that it kept us out of a good chunk of the 2000 Dot Com Bust but not all of it.
This shows us our first problem with this system, monthly whipsaws. There is one going in to the economic downturn and one coming out. There is one more in our data set at Nov 2010. Do these make a difference? I don't know. Only 3 whipsaw months in 16 years doesn't sound like a lot. I mean you get more than that in a typical year with the LMBF systems. Still, we really don't know until we test this data and compute some backtested results.
That points us to another problem -- data latency. You can't make the trade for the month on the month because the data doesn't come out until the following month. So we are already behind a month plus some days and that isn't even taking into account the fact that the government also likes to revise the numbers after that. Well those are the kinds of problems we are only going to see if we follow these numbers monthly in real time. That's the purpose of this thread, to see if we can make this thing work.
In the meantime how do we compute backtested results to see if we even want to follow this system? To be realistic I think at a minimum we need to build in a two month lag on the trigger. Why? Look at last month's (April) UR. It came out the 1st week of May. That means you can't use April's data until you are already into May. Since we are working with monthly data that means a decision on April data won't go into effect until June. That's a lag of 2 month's. If you have any other ideas on that left me know. In the mean time I'll try to come up with some data using a two month latency and get back to you. :)
Re: Unemployment Rate over 12 Month MA Strategy
OK, so per the above discussion I decided to try this method with a 2 month lag on the IN/OUT trigger. I chose to test two methods. The first one I called "UR C&F" and it goes 100% into the C Fund while we are IN and 100% into the F Fund while we are OUT. The second one I called "UR S&F" as it does the same thing with the S & F Funds. Here are the annual results together with our 5 funds:
Year |
G Fund |
F Fund |
C Fund |
S Fund |
I Fund |
UR C&F |
UR S&F |
2000 |
6.42% |
11.67% |
(9.14%) |
|
|
(8.17%) |
|
2001 |
5.39% |
8.61% |
(11.94%) |
(9.04%) |
(21.94%) |
(0.35%) |
(1.91%) |
2002 |
5.00% |
10.27% |
(22.05%) |
(18.14%) |
(15.98%) |
1.70% |
3.36% |
2003 |
4.11% |
4.11% |
28.54% |
42.92% |
37.94% |
8.47% |
5.17% |
2004 |
4.30% |
4.30% |
10.82% |
18.03% |
20.00% |
10.82% |
18.03% |
2005 |
4.49% |
2.40% |
4.96% |
10.45% |
13.63% |
4.96% |
10.45% |
2006 |
4.93% |
4.40% |
15.79% |
15.30% |
26.32% |
15.79% |
15.30% |
2007 |
4.87% |
7.09% |
5.54% |
5.49% |
11.43% |
8.91% |
9.91% |
2008 |
3.75% |
5.45% |
(36.99%) |
(38.32%) |
(42.43%) |
5.45% |
5.45% |
2009 |
2.97% |
5.99% |
26.68% |
34.85% |
30.04% |
5.99% |
5.99% |
2010 |
2.81% |
6.71% |
15.06% |
29.06% |
7.94% |
29.91% |
37.15% |
2011 |
2.45% |
7.89% |
2.11% |
(3.38%) |
(11.81%) |
(0.12%) |
(4.45%) |
2012 |
1.47% |
4.29% |
16.07% |
18.57% |
18.62% |
16.07% |
18.57% |
2013 |
1.89% |
(1.68%) |
32.45% |
38.35% |
22.13% |
32.45% |
38.35% |
2014 |
2.31% |
6.73% |
13.78% |
7.80% |
(5.27%) |
13.78% |
7.80% |
2015 |
2.04% |
0.91% |
1.46% |
(2.92%) |
(0.51%) |
1.46% |
(2.92%) |
It does look like this system does reduce our losses during economic downturns. Unfortunately it also reduces the initial recovery. I guess you can't have one without the other. The question is do you come out ahead in the long run. We will have to see.
Also notice that we have basically been pegged in the market these last 5 years so our methods produce the same results as the individual C & S Funds. 2011 actually underperformed these funds because it suffered from the whipsaw trigger in late 2010. :( Yep, whipsaws still hurt.
Re: Unemployment Rate over 12 Month MA Strategy
OK, here are the 5, 10 & 15 year annualized returns:
5 Year Annualized Returns
|
G Fund |
F Fund |
C Fund |
S Fund |
I Fund |
UR C&F |
UR S&F |
CAGR |
2.03% |
3.57% |
12.63% |
10.65% |
3.78% |
12.13% |
10.41% |
Risk |
0.34% |
3.57% |
11.31% |
15.58% |
13.39% |
11.77% |
15.79% |
10 Year Annualized Returns
CAGR |
2.94% |
4.74% |
7.36% |
8.03% |
3.20% |
12.49% |
12.29% |
Risk |
1.14% |
2.86% |
18.03% |
21.42% |
20.67% |
10.54% |
14.01% |
15 Year Annualized Returns
CAGR |
3.51% |
5.12% |
5.03% |
7.61% |
3.56% |
9.95% |
10.43% |
Risk |
1.26% |
2.95% |
18.22% |
21.44% |
21.31% |
9.66% |
12.43% |
The 5 year returns don't quite live up to their benchmark funds, but then we have basically been in a bull market during that time and that's not where this system helps you.
When you look out over the 10 and 15 year returns you see where the strength of this system lies. It does indeed save you during the down years and that is what it is supposed to do. So if you are a Buy-N-Hold kind of investor that's worried by another economic downturn you may want to look into a system like this to tell you when to step aside for a while. According to this system we aren't there yet. :)
Re: Unemployment Rate over 12 Month MA Strategy
Thanks for the analysis Catcus, I enjoy seeing this stuff back-tested. Definitely a worthy option to buy-and-hold, and my all-time favorite strategy that I rarely use, the LMBF-1 CI->S...
At the risk of bringing up a competitor, there's another TSP site that uses the current business cycle as a guide to being invested in stocks. It's really a buy-and-hold for the most part, shifting from S to C depending on where we are in the business cycle (or at least where the person thinks we are), and if we're still growing. I bring it up because he uses unemployment as one of the factors to determine the business cycle, along with PMI, M2 money supply growth and yield spreads. Not a trading strategy at all, but I think it might have side-stepped the last two downturns moving to G if the numbers dictate.
Your numbers - wow, it really just points out how poor the I fund has performed for the associated risk. Unless you really feel it's due for a pop, I would rather stay in the F fund for the lower risk over the long term.
Thanks again. Can't ask much more from a system - it beats buy-and-hold with lower risk.
Quote:
Originally Posted by
Cactus
OK, here are the 5, 10 & 15 year annualized returns:
5 Year Annualized Returns
|
G Fund |
F Fund |
C Fund |
S Fund |
I Fund |
UR C&F |
UR S&F |
CAGR |
2.03% |
3.57% |
12.63% |
10.65% |
3.78% |
12.13% |
10.41% |
Risk |
0.34% |
3.57% |
11.31% |
15.58% |
13.39% |
11.77% |
15.79% |
10 Year Annualized Returns
CAGR |
2.94% |
4.74% |
7.36% |
8.03% |
3.20% |
12.49% |
12.29% |
Risk |
1.14% |
2.86% |
18.03% |
21.42% |
20.67% |
10.54% |
14.01% |
15 Year Annualized Returns
CAGR |
3.51% |
5.12% |
5.03% |
7.61% |
3.56% |
9.95% |
10.43% |
Risk |
1.26% |
2.95% |
18.22% |
21.44% |
21.31% |
9.66% |
12.43% |
The 5 year returns don't quite live up to their benchmark funds, but then we have basically been in a bull market during that time and that's not where this system helps you.
When you look out over the 10 and 15 year returns you see where the strength of this system lies. It does indeed save you during the down years and that is what it is supposed to do. So if you are a Buy-N-Hold kind of investor that's worried by another economic downturn you may want to look into a system like this to tell you when to step aside for a while. According to this system we aren't there yet. :)
1 Attachment(s)
Re: Unemployment Rate over 12 Month MA Strategy
For anyone who want's to look at the numbers I used, here is the spreadsheet.
It's nothing fancy. The UR is in column B the MA is in col C and the indicator is in D. Let me know if you find any errors.
Re: Unemployment Rate over 12 Month MA Strategy
Tsunami posted a link to this article In Search of the Perfect Recession Indicator | PHILOSOPHICAL ECONOMICS over on his thread that discusses using this Unemployment Rate method as an economic indicator. One thing I found interesting in that article is that it indicates that this method is actually a leading indicator. It shows that since 1948 the method actually preceded the start of the recession by an average of 3.45 months. If that is the case the latency of our data should not be an issue. Now we only have to worry about revised numbers. :rolleyes:
Re: Unemployment Rate over 12 Month MA Strategy
Cactus,
I took a quick look at your spreadsheet and I think our data match perfectly. What I was saying regarding this indicator not triggering a signal to exit stocks has to do with the criteria, which are described in the article and I'll paste them below. The UE trend has to turn up (if the monthly rate spikes above the trend, that's just noise and not part of the criteria). Plugging numbers into your spreadsheet that change in trend will not occur unless the UE rate surged to 5.6% on the 6/3 jobs report. Looking at something more realistic, if I plug in 5.1% for the May report, then 5.2% for the June report, then 5.3% for the July report...at that point the 12-month trend finally flat-lines at 5.05%...then if the August report hits 5.4% it finally turns the 12-month trend up and a sell signal would be triggered at that time, but only if the S&P price trend was also down, part two of the indicator (the price trend currently is down already, so we're just waiting for confirmation by the UE rate. So, realistically, I think that as Jim Puplava has been saying on his weekend podcasts on Financialsense.com, and other gurus, a recession doesn't look likely until 2017 at this point given that this indicator typically gives a 3 or 4 month advance notice of a recession.
Here's the criteria below, and this is what I plugged into my system, but only as a signal to exit stocks and go to G or F for the duration of the bear market (or if you're very bold, you could short the market during this period with an inverse ETF outside of the TSP)...the re-entry signal for my system comes from a monthly S&P chart and just looks for two indicators, one of them is something like the fast MACD crossing above the slow MACD from a low level, below 30 I think...I forget the specifics that worked beautifully in the last two bear markets (certainly no guarantee it will work next time though), I don't have my spreadsheet at work:
Using the unemployment rate as an input, the specific trading rule for GTT would be:(1) If the unemployment rate trend is downward, i.e., not indicating an oncoming recession, then go 100% long U.S. equities.
(2) If the unemployment rate trend is upward, indicating an oncoming recession, then defer to the price trend. If the price trend is upward, then go 100% long U.S. equities. If the price trend is downward, then go to cash.
To summarize, GTT will be 100% invested in the market unless the unemployment rate trend is upward at the same time that the price trend is downward. Together, these indicators represent a double confirmation of danger that forces the strategy to take a safe position.
Holy nosedive Batman, just noticed that the market has finally tipped it's hand and is crumbling. Now I'm definitely glad I'm following my system and am safely in the G fund!
Re: Unemployment Rate over 12 Month MA Strategy
Ah yes, thank you, Tsunami. I see the difference. The article you linked to does explicitly state that you are following the trend of the 12 month moving average. The wording of the article that Tom4Jean linked to stated: "when the unemployment rate crosses the 12 month moving average." That is what my spreadsheet uses for its signals and why I have both curves on the plot. I'm looking for the crossovers.
As you mentioned this method does give us spikes which results in whipsaws. The trend following method seams to smooth that out so you don't get whipsaws but still transition for real in about the same place.
I like that better. I think I'll rework my spreadsheet tonight using the trending method and see how much of a difference it makes. Thanks again. :)
Re: Unemployment Rate over 12 Month MA Strategy
Cactus, I still have one unresolved issue with the article that maybe you can figure out. On both your spreadsheet and mine, we have September 2007 as the month when the UE 12-month trend last turned up, warning of the last recession just 2 months before it officially started...but in the first table in that link the author shows March 2007 as the month the UE trend turned up. Going further back I have January 2001 as the previous time the UE trend turned up (which, since the price trend was also down, would have triggered a sell signal for stocks on the day that jobs report was released, which was 2/2/2001)...checking your spreadsheet, we match on that too, January 2001 was when this method warned of an impending recession....and that also matches the authors chart. I'm wondering if the author just made a mistake and the "2007.03" should be "2007.09" in that chart, but it's driving me a little nuts wondering. Let's see, my data stops in late 1998 since I only wanted to cover the last two recessions, but your spreadsheet goes back further....he's got April 1990 as the previous UE trend turn...that matches your spreadsheet too if you consider the tie of 5.30% in 4/1990 as a turn (I'll need to keep that definition in mind for my system)....I'll just assume he screwed up and go with my data...but I see he has an email address... so I just emailed him to ask, and I will post again if he replies.
Re: Unemployment Rate over 12 Month MA Strategy
Tsunami, I concur that the change in trend occurred in Sep 2007. These are the dates I have for the change in trend starting in 2000:
Jan 2000 IN
Jan 2001 OUT
Nov 2003 IN
Sep 2007 OUT
Jun 2010 IN
If you use the crossover of the Unemployment Rate over the 12MMA as mentioned in the first article that Tom4Jean linked to that only pushes the trigger back to Jun 2007. The author is clearly early with his Mar 2007 date.
One posibility is that the Mar 2007 date was computed with unrevised data. The data you and I are using is Seasonally Adjusted and by now also revised if need be. Looking at the trend numbers from March through September you will notice they don't differ from month to month by more than 2/100.
At the end of the article you linked to it said it is possible (though cumbersome) to access unrevised data through FRED. You may want to go that route if you are really curious. As for me that sounds like too much work. I'd rather go forward from here documenting the Unemployment rate as it is reported and then seeing when and how often the Government revises the numbers.
Re: Unemployment Rate over 12 Month MA Strategy
You guys are doing great work here, thank you and keep it up!
Re: Unemployment Rate over 12 Month MA Strategy
OK, Looking at the triggering dates in the previous post we see that using the change in trend trigger does eliminate the whipsaws we saw with the crossover trigger. There actually are still some whipsaws in the data in the 90s, so technically this reduces whipsaws -- it doesn't eliminate it.
You will also notice that some of the trigger dates for trending lag those for crossover. How does that effect our data? Here are the years that produced different results: Crossover / Trend
|
UR C&F |
UR S&F |
UR C&F |
UR S&F |
2000 |
(8.17%) |
|
(9.14%) |
|
2002 |
1.70% |
3.36% |
10.27% |
10.27% |
2003 |
8.47% |
5.17% |
4.11% |
4.11% |
2007 |
8.91% |
9.91% |
13.27% |
14.66% |
2010 |
29.91% |
37.15% |
22.70% |
29.54% |
2011 |
(0.12%) |
(4.45%) |
2.11% |
(3.38%) |
It looks like some years are better and some years are worse. That should all even out in the end. I'll have to compute some 5, 10 & 15 year returns of this and compare it to the previous posting