Playing with Callan's periodic tables of investment returns

[siamond]

Quoting the Callan.com Web site:
Callan's Periodic Tables of Investment Returns graphically depict annual returns for various asset classes, ranked from best to worst.

This representation is pretty cool, neatly illustrating the value of diversification, and I always had in mind to assemble a spreadsheet allowing to generate such 'quilts' in an automated manner. I finally got to it, then started to play around with a few data sets.

The spreadsheet is customizable and available for broader use. It can be downloaded following instructions in this post.

Here are the annual (nominal) returns of the past 20 years of the US total market (TSM), split along the usual 3x3 matrix (Large Cap Value, Large Cap Blend, Large Cap Growth, Mid Cap Value, etc.). Data source: Simba. Click to see a larger image.



EDIT: added link to post providing instructions to download the spreadsheet.

[siamond]

Instead of focusing on annual returns, hence a limited period of time, we can also use annualized returns for time periods of 5 years. Here are the same asset classes (US 3x3 + TSM) going further back in time.

[siamond]

Why not take a look to stocks in other countries? Let's come back to the past 20 years, and compare some of the largest stock markets. Annual returns are nominal, expressed in the local currency. Data source: MSCI.

[siamond]

One more. Instead of looking at returns, why not look at premiums (over a benchmark)? Here is the same data as the previous post, but illustrating the relative premium over the returns of the US stock market.



Ideas for further exploring welcome...

[FIREchief]

I've always found these tables to be excellent food for thought. After awhile, I came to a conclusion that a person needs to ask themselves which of the following describes their approach:

a) I want to be in everything so that I'm guaranteed to have some in the "winning" bucket each year (it also guarantees a person will have some in the "losing" bucket each year)
b) I want an asset class that is never or rarely at the bottom (which also will likely never or rarely be at the top)

A US Total Market investor should be comfortable with the second statement. A total world investor, the first.

[Caduceus]

These kinds of tables are meaningless because two finance folks looking at the same company may well assign them to different categories.

I was reading the financial statements of a small-cap firm the other day. By any metric, it would be a "value" stock - low p/b ratio, low p/e ratio. But because the pundits are also predicting really aggressive earnings growth, it was slotted into the "growth" category.

It's also silly because, as Buffett has pointed out on many occasions, if a stock is really a growth stock - huge potential earnings relative to its current price - then surely it must be undervalued and hence a "value" stock as well. Conversely, a stock with low p/b ratio and low p/e ratios with zero or declining earnings surely shouldn't count as a "value" stock - since you aren't getting value for the price you are paying at all.

[Taylor Larimore]

siamond:

Thank you for posting The Callan Periodic Tables of Investment Returns. It is the best representation I know that demonstrates the benefits of diversification and the futility of trying to pick winning fund categories in advance.

Mr. Bogle is right: Don't look for the needle. Buy the haystack.

Best wishes.
Taylor

[siamond]

Something I also had in mind was to compare portfolios (as opposed to asset classes) using such representation. Here I took 10 US-only portfolios, ranging from 10% stocks/90% bonds to 100% stocks/0% bonds. And I looked at annualized returns for time periods of 5 years, in nominal terms.

All those Callan charts really looked like random quilts so far. This one, not quite. For most periods, it remains perfectly ordered by % of stocks. And every now and then, a deep crisis struck, and the order is perfectly reversed. I wasn't expecting something THAT regular, but this makes sense, either stocks returned more than bonds or bonds returned more, and then there are only two possible orderings for the various asset allocations. Now, if 2017+ could look like 1942+ or 1977+, this would be really cool. I am not holding my breath, though...

[siamond]

The previous result was based on nominal returns. Then I wondered about inflation-adjusted (real) returns. Similar pattern, of course.

Now, look at the 4 periods starting in 1962. Whether you were a bonds believer or a stocks believer or in-between, this was a VERY painful draught.

[aj76er]

Why not take a look to stocks in other countries? Let's come back to the past 20 years, and compare some of the largest stock markets. Annual returns are nominal, expressed in the local currency. Data source: MSCI.

Very nice work!

Would it be possible to also include the cap-weighted global market, much like you did with TSM on the previous table?

I found that it was kind of interesting (and expected, of course) that TSM bounced around in between of all the individual factors (i.e. both never winning and never losing). Exactly what one would expect from diversification.

[siamond]

Would it be possible to also include the cap-weighted global market, much like you did with TSM on the previous table?

I removed Australia and Canada to make room for EM (Emerging Markets) and WORLD (Developed Markets). And since I had to choose a currency for EM and WORLD, I unified everything on USD (remember, the previous chart was expressed in local currencies). Here it is.

[abuss368]

Mr. Bogle is right: Don't look for the needle. Buy the haystack.

Best wishes.
Taylor

One of the most important investing lessons I have ever learned!

[abuss368]

Thank you for posting the table. I have always found this table very interesting and insightful. My takeaway has always been I am glad to be diversified rather than try to guess which sector will outperform (or underperform) next.

[heyyou]

Can't find the link now, but someone in the past, posted a Callan Table with zero % as the reference line for stacking the colored blocks, more clearly showing the differences between the + and - returns for each year. Low returns don't feel as bad as the negative ones. On a paper Callan Table, we can draw a reference line between each annual divide of the + and - returns.

[siamond]

Can't find the link now, but someone in the past, posted a Callan Table with zero % as the reference line for stacking the colored blocks, more clearly showing the differences between the + and - returns for each year. Low returns don't feel as bad as the negative ones. On a paper Callan Table, we can draw a reference line between each annual divide of the + and - returns.

Ah, thank you, I also had a vague recollection of an interesting format suggestion, but I couldn't remember it... Yes, that's it. Something like that?



Hm. I'd have to scratch my head a little more to fully automate this format... Maybe I should aim at doing something fully proportional to returns though, a stacked column kind of chart... Or would this detract too much from the original idea?

[heyyou]

Bea-u-tiful.

[dekecarver]

What I found interesting was playing around with Simba's spread sheet and comparing models, one of which was taking an asset class e.g. MidCap Index, MidCap Growth and MidCap Value ~33.3% each fund and comparing that over the long haul against other models. Did this after looking a the Callan's table in terms of Growth, Value... Why not go all in on one bucket and split it across those metrics?

[siamond]

What I found interesting was playing around with Simba's spread sheet and comparing models, one of which was taking an asset class e.g. MidCap Index, MidCap Growth and MidCap Value ~33.3% each fund and comparing that over the long haul against other models. Did this after looking a the Callan's table in terms of Growth, Value... Why not go all in on one bucket and split it across those metrics?

Once I'm fully set on a couple of formats, if there is enough interest, I'll probably share a spreadsheet with instructions on how to generate such 'Callan' charts by yourself.

This seems a little too specialized to add to Simba itself, so I'll keep it as a separate spreadsheet. I have in mind to allow the definition of 10 portfolios in a very similar way as in Simba, define a couple of parameters (e.g. time period, rolling returns periodicity, premium vs absolute number) and magic, you have a nice quilt. Or something like that... Ideas welcome.

[foodhype]

One more. Instead of looking at returns, why not look at premiums (over a benchmark)? Here is the same data as the previous post, but illustrating the relative premium over the returns of the US stock market.



Ideas for further exploring welcome...

This one is the most interesting.

[DoTheMath]

Once I'm fully set on a couple of formats, if there is enough interest, I'll probably share a spreadsheet with instructions on how to generate such 'Callan' charts by yourself.

Please do.

[Top99%]

Siamond,

Thanks for putting a lot of effort in creating what I find some really useful charts. Looking back though this data is especially enlightening as the current bull market in US stocks continues to extend its duration lead over all previous US bull markets.

[goGators]

How do you calculate "relative premium over the returns of the US stock market"? An example using numbers from the above tables would be great. Thank you.

[siamond]

How do you calculate "relative premium over the returns of the US stock market"? An example using numbers from the above tables would be great. Thank you.

Take the absolute return of Italy in 1998 in this post. It was 42.88%. While the US was 30.72%.

An approximate way of computing the premium would be: 42.88% minus 30.72% = 12.16%. This isn't mathematically correct though, when applied to quantities like returns which behave in a geometric manner.

The proper way of computing the premium is: (1+42.88%) / (1+30.72%) - 1 = 9.30%. Which is indeed the number you find for Italy in this post.

[RadAudit]

OP, thanks for this post.

[nedsaid]

These kinds of tables are meaningless because two finance folks looking at the same company may well assign them to different categories.

I was reading the financial statements of a small-cap firm the other day. By any metric, it would be a "value" stock - low p/b ratio, low p/e ratio. But because the pundits are also predicting really aggressive earnings growth, it was slotted into the "growth" category.

It's also silly because, as Buffett has pointed out on many occasions, if a stock is really a growth stock - huge potential earnings relative to its current price - then surely it must be undervalued and hence a "value" stock as well. Conversely, a stock with low p/b ratio and low p/e ratios with zero or declining earnings surely shouldn't count as a "value" stock - since you aren't getting value for the price you are paying at all.

The Callan periodic tables have limits to their usefulness but it doesn't mean they are meaningless either. What it shows is that asset classes change places in rank of best performance and worst performance. As the old saying goes, every dog has their day. Some asset classes are consistently more near the top than others. Other asset classes are consistently near the bottom than others. This is not meaningless or useless information.

One thing is that financial data is "slippery". That is numbers are not precise to begin with. I have discussed how the numbers provided by accountants involve judgment and accruals. So we see some educated guesses as part of the numbers. Also definitions of things like "growth" and "value" vary to some degree. We know the market works from earnings estimates, if the earnings numbers from history are not precise, certainly earnings estimates are less so. But that is what we have. To a degree, all of this is in the eye of the beholder.

To say that our tools are imprecise isn't the same thing as saying they are useless.

[Dottie57]

What I found interesting was playing around with Simba's spread sheet and comparing models, one of which was taking an asset class e.g. MidCap Index, MidCap Growth and MidCap Value ~33.3% each fund and comparing that over the long haul against other models. Did this after looking a the Callan's table in terms of Growth, Value... Why not go all in on one bucket and split it across those metrics?

Once I'm fully set on a couple of formats, if there is enough interest, I'll probably share a spreadsheet with instructions on how to generate such 'Callan' charts by yourself.

This seems a little too specialized to add to Simba itself, so I'll keep it as a separate spreadsheet. I have in mind to allow the definition of 10 portfolios in a very similar way as in Simba, define a couple of parameters (e.g. time period, rolling returns periodicity, premium vs absolute number) and magic, you have a nice quilt. Or something like that... Ideas welcome.

A spreadsheet would be wonderful.

[MJW]

The Callan tables show me that I need to get in and out of US small value and emerging markets at the right times.

[Ricola]

Much Thanks!

[MIretired]

How do you calculate "relative premium over the returns of the US stock market"? An example using numbers from the above tables would be great. Thank you.

Take the absolute return of Italy in 1998 in this post. It was 42.88%. While the US was 30.72%.

An approximate way of computing the premium would be: 42.88% minus 30.72% = 12.16%. This isn't mathematically correct though, when applied to quantities like returns which behave in a geometric manner.

The proper way of computing the premium is: (1+42.88%) / (1+30.72%) - 1 = 9.30%. Which is indeed the number you find for Italy in this post.

Thanks for charts. I've been missing this point all along. I think we get sidetracked into other areas, and lose sight of things.

So, the Callan charts primarily show the idea to diversify similarly risked assets.
But I am looking at the stock/bond chart you made(actually thought it was a bar chart; but must be a copy of spreadsheet cells) with real returns. And how it shows for 5 yr. independent times, stocks won 78% since 1927, and bonds won the other 22%. (I gather you used the formula of (1+Er1)/(1+Er2) weighted by % holdings for these charts. E is the asset, r is it's return. Also: these assets were rebalanced annually?)
2 points: notice the order of stk/bnd weighting in returns was not disrupted. ie: 20% stocks never beat 10% while losing to 30%.
And I was interested in this chart being made with 10 yr independent returns since 1927. But with 2 rows of cells for each row: one starting in 1927 and every 10 yrs., and one under it starting in 1932 and every 10 years. A rolling mix looking like a brick background. One particular note: I think a 50/50 stk/bnd beat both 100/0 and 0/100 at some point during the '00's.
PS: also adding in that 0/100 stk/bnd AA.
I am somewhat trying to create this sheet myself, and am seeing the amount of time and work going into it. Not to mention I got the weighted return wrong--didn't consider W1(1+Er1)/W2(1+Er2). It might not be worth it because a multiple efficient frontier would show it better. In fact I think the EF of TSM/BND for 2000-2009 is a straight horizontal line.

This is 2002-2011:

[siamond]

Can't find the link now, but someone in the past, posted a Callan Table with zero % as the reference line for stacking the colored blocks, more clearly showing the differences between the + and - returns for each year. Low returns don't feel as bad as the negative ones. On a paper Callan Table, we can draw a reference line between each annual divide of the + and - returns.

I actually found an easy way to automate this formatting variant. Here is the full chart with the international data for the past 20 years.

[LadyGeek]

siamond:

Thank you for posting The Callan Periodic Tables of Investment Returns. It is the best representation I know that demonstrates the benefits of diversification and the futility of trying to pick winning fund categories in advance.

Mr. Bogle is right: Don't look for the needle. Buy the haystack.

Best wishes.
Taylor

New investors should read this wiki article: Callan periodic table of investment returns

The Callan Table is the best visual information showing the importance of diversification, reversion-to-the-mean, and the impossibility of forecasting asset-class returns. It is a primary reason Bogleheads favor total market index funds.

[siamond]

I am looking at the stock/bond chart you made(actually thought it was a bar chart; but must be a copy of spreadsheet cells) with real returns.

Yes, those are spreadsheet cells with a good dose of conditional formatting.

And how it shows for 5 yr. independent times, stocks won 78% since 1927, and bonds won the other 22%. (I gather you used the formula of (1+Er1)/(1+Er2) weighted by % holdings for these charts. E is the asset, r is it's return. Also: these assets were rebalanced annually?)

Yes, annual rebalancing is implicitly assumed. And yes, the portfolio return is weighted by the portfolio components. I am not sure to understand your formula in this case though. If Rs = stocks return, Rb=bonds return; Ws and Wb = respective portfolio weights, the portfolio return is Ws(1+Rs)+Wb(1+Rb)-1. Or more simply Ws*Es+Wb*Eb.

2 points: notice the order of stk/bnd weighting in returns was not disrupted. ie: 20% stocks never beat 10% while losing to 30%. [...]One particular note: I think a 50/50 stk/bnd beat both 100/0 and 0/100 at some point during the '00's.

This regularity surprised me a bit to begin with, but then I realized that if stock returns are higher than bonds returns, then it stands to reason that the more stocks you have in the portfolio, the higher the return. And then if bonds returns are higher, then it's the very exact reverse ordering. So, no, there is no way a 50/50 stk/bnd can beat both 100/0 and 0/100 at some point in time.

And I was interested in this chart being made with 10 yr independent returns since 1927. But with 2 rows of cells for each row: one starting in 1927 and every 10 yrs., and one under it starting in 1932 and every 10 years. A rolling mix looking like a brick background. PS: also adding in that 0/100 stk/bnd AA.

With a bit of copying and pasting, I created a chart similar to what you suggested, here is the link. I am not too sure what this is telling you though... Let me know if this is of interest.

[CedarWaxWing]

Once I'm fully set on a couple of formats, if there is enough interest, I'll probably share a spreadsheet with instructions on how to generate such 'Callan' charts by yourself.

Please do.

1+

[MIretired]

I am looking at the stock/bond chart you made(actually thought it was a bar chart; but must be a copy of spreadsheet cells) with real returns.

Yes, those are spreadsheet cells with a good dose of conditional formatting.

And how it shows for 5 yr. independent times, stocks won 78% since 1927, and bonds won the other 22%. (I gather you used the formula of (1+Er1)/(1+Er2) weighted by % holdings for these charts. E is the asset, r is it's return. Also: these assets were rebalanced annually?)

Yes, annual rebalancing is implicitly assumed. And yes, the portfolio return is weighted by the portfolio components. I am not sure to understand your formula in this case though. If Rs = stocks return, Rb=bonds return; Ws and Wb = respective portfolio weights, the portfolio return is Ws(1+Rs)+Wb(1+Rb)-1. Or more simply Ws*Es+Wb*Eb.
Well. Right about the not dividing the weighted returns. It's an addition of weights. Or called weighted return. But I've sensed something else: but in a sense you answered. To get the 5 yr CAGR, you must have run the annual rebalancing returns of each AA independently.You just take the sets of 5 each annual returns. Then compound the desired weighting of each 5 times(or 4.)

2 points: notice the order of stk/bnd weighting in returns was not disrupted. ie: 20% stocks never beat 10% while losing to 30%. [...]One particular note: I think a 50/50 stk/bnd beat both 100/0 and 0/100 at some point during the '00's.

This regularity surprised me a bit to begin with, but then I realized that if stock returns are higher than bonds returns, then it stands to reason that the more stocks you have in the portfolio, the higher the return. And then if bonds returns are higher, then it's the very exact reverse ordering. So, no, there is no way a 50/50 stk/bnd can beat both 100/0 and 0/100 at some point in time.
Well.I mis-wrote. Meant 20% didn't win over 10% AND 30%. This because of minvariance. Part of why I wanted to see 0% stocks. Wrong on this, too. The 50/50 beating 100 and 0 I've done backtest on. But hard to remember since I didn't save it. I'm thinking it was LongT, now, though I'd thought it was IntermT.Nope. Can't make sense of it. Might have been an annual WD test. I don't know where I got any of this paragraph from. Or the prior paragraph, either.(During the 2000's with VFITX. ??

And I was interested in this chart being made with 10 yr independent returns since 1927. But with 2 rows of cells for each row: one starting in 1927 and every 10 yrs., and one under it starting in 1932 and every 10 years. A rolling mix looking like a brick background. PS: also adding in that 0/100 stk/bnd AA.

With a bit of copying and pasting, I created a chart similar to what you suggested, here is the link. I am not too sure what this is telling you though... Let me know if this is of interest.

[MIretired]

With a bit of copying and pasting, I created a chart similar to what you suggested, here is the link. I am not too sure what this is telling you though... Let me know if this is of interest.

That's mainly what I meant. But didn't expect the mis-ordering now. I really wanted to see whether stocks won 25 or 35 yrs. in a row 2 times using 10 yr holding periods, overlapping by 5 yrs.This chart is confusing looking, though. Guess I don't need to try and make that chart now. But it did show me that. It could've been made using just the 2 weights of 100/0 and 0/100 stks/bnds, too.

Edit to add

[siamond]

Yes, annual rebalancing is implicitly assumed. And yes, the portfolio return is weighted by the portfolio components. I am not sure to understand your formula in this case though. If Rs = stocks return, Rb=bonds return; Ws and Wb = respective portfolio weights, the portfolio return is Ws(1+Rs)+Wb(1+Rb)-1. Or more simply Ws*Es+Wb*Eb.

Well. Right about the not dividing the weighted returns. It's an addition of weights. Or called weighted return. But I've sensed something else: but in a sense you answered. To get the 5 yr CAGR, you must have run the annual rebalancing returns of each AA independently.You just take the sets of 5 each annual returns. Then compound the desired weighting of each 5 times(or 4.)

Ok, let me take this step by step. Each cell of the table includes annualized returns for time intervals of 5 (or 10) years, inflation-adjusted, for a given Asset Allocation made of US stocks and bonds. Annualized returns mean the (geometric) average return per year that would produce the same compound growth as the actual returns over this time interval.

First thing in a spreadsheet is to use the nominal annual returns of the individual asset classes (here US stocks and bonds) to infer the portfolio annual returns, as we just discussed. If Rs = stocks return (nominal), Rb=bonds return (nominal); Ws and Wb = respective portfolio weights; then NRp (nominal portfolio return for this year) is: NRp = Ws*Es+Wb*Eb.

Then we need to adjust for inflation, to get from a nominal number to a real number. To do this, we need to geometrically subtract the effect of inflation for this year. If Ius = inflation in the US, then the RRp (real portfolio return for this year) is: RRp = (1+NRp)/(1+Ius)-1.

Now let's take those numbers for 5 years in a row: RRp1, RRp2, RRp3, RRp4, RRp5. Now the question is, what the corresponding annualized (average) return, also known as Compound Annual Growth Rate (CAGR) for this time interval? You can look up the exact mathematical formula on Wikipedia, but spreadsheet software like Excel provides a very handy function for this, and the way I do it is the following: GEOMEAN(1+RRp1, 1+RRp2, 1+RRp3, 1+RRp4, 1+RRp5)-1.

I hope this helps... Maybe some of us should create a few wiki entries about the math of investments with spreadsheets...

[siamond]

With a bit of copying and pasting, I created a chart similar to what you suggested, here is the link. I am not too sure what this is telling you though... Let me know if this is of interest.

That's mainly what I meant. But didn't expect the mis-ordering now. I really wanted to see whether stocks won 25 or 35 yrs. in a row 2 times. This chart is confusing looking, though. Guess I don't need to try and make that chart now.

Yes, if we step back a bit here, I don't think this kind of chart is terribly useful to compare variations of weights (%) in the composition of an Asset Allocation. The more usual representations like growth charts, drawdown charts, rolling returns charts, are more telling for this. If you're minimally spreadsheet literate, then please use the Simba backtesting spreadsheet for this type of analysis.

Callan charts are very cool when segmenting the market in various ways (e.g. by factors like size and value; by geography like countries; by industry sectors; maybe also segment the bonds market per duration and/or credit risk; etc). It then shows very well the quasi-randomness of what is winning and what is losing year over year, the foolishness of narrow bets (notably based on recent past performance), and the power of diversification. But they are not as enlightening when playing with asset allocation weights.

[LadyGeek]

This thread is now in the wiki: Callan periodic table of investment returns (External links --> Forum discussions).

Also note the wiki article now appears in the "Bogleheads® investing start-up kit" navigation menu (bottom of the page), as it teaches new investors the principle of diversification - a key concept.

Mobile users can see the navigation menu by tapping on the "External links" heading (opens up the section for viewing).

[siamond]

Once I'm fully set on a couple of formats, if there is enough interest, I'll probably share a spreadsheet with instructions on how to generate such 'Callan' charts by yourself.

This seems a little too specialized to add to Simba itself, so I'll keep it as a separate spreadsheet. I have in mind to allow the definition of 10 portfolios in a very similar way as in Simba, define a couple of parameters (e.g. time period, rolling returns periodicity, premium vs absolute number) and magic, you have a nice quilt. Or something like that... Ideas welcome.

A spreadsheet would be wonderful.

Once I'm fully set on a couple of formats, if there is enough interest, I'll probably share a spreadsheet with instructions on how to generate such 'Callan' charts by yourself.

Please do.

1+

Since you guys expressed interest, I packaged my working spreadsheet for broader use (I mean, assuming you're minimally spreadsheet literate). See next post.

[siamond]

Here is a packaging of my Callan Table Calculator spreadsheet. Please download it here, and use a reasonably recent version of Excel:
https://drive.google.com/uc?id=1-EZIwGV ... t=download

The 'README' tab provides instructions. Those of you used to the Simba backtesting spreadsheet will find a familiar structure, allowing the user to name and define 10 portfolios, using some controls to be more specific about exactly what type of return (e.g. nominal, real, premium over a benchmark, annual or time intervals) you'd like to see in the Callan's table. I provided two formats for Callan tables, the regular 'quilt' and the distribution around a 'zero' axis.

Feedback welcome...

PS. after solving a small compatibility issue, it appears to work fine in LibreOffice, although I didn't play much with it.

[Dottie57]

Here is a packaging of my Callan Table Calculator spreadsheet. Please download it here, and use a reasonably recent version of Excel:
https://drive.google.com/uc?id=1-EZIwGV ... t=download

The 'README' tab provides instructions. Those of you used to the Simba backtesting spreadsheet will find a familiar structure, allowing the user to name and define 10 portfolios, using some controls to be more specific about exactly what type of return (e.g. nominal, real, premium over a benchmark, annual or time intervals) you'd like to see in the Callan's table. I provided two formats for Callan tables, the regular 'quilt' and the distribution around a 'zero' axis.

Feedback welcome...

PS. it appears to work -almost- fine in LibreOffice, but one formula has to slightly change in the Callan table cells. Contact me by PM if interested. I'll try to find a better solution to make it work the same in both Excel and LibreOffice.

Thank you. Love the confirmation: buy the whole haystack"

[Bob]

Thank you for developing and sharing this tool. Very interesting and helps me look at my portfolio nominal and real returns in a better contexr.
Great stuff !

[Noobvestor]

Why not take a look to stocks in other countries? Let's come back to the past 20 years, and compare some of the largest stock markets. Annual returns are nominal, expressed in the local currency. Data source: MSCI.

Any given country really is all over the map (yes yes, pun intended)

[vitaflo]

One thing I'm not a fan of in these tables is it doesn't show scale. For example in the chart in the OP, the difference between Highest and Lowest in 1998 is 47%. The difference in 2017 is only 16%. But the table makes it looks like these differences are similar because it's just laid out high to low.

To me it would be much more interesting to show not just high to low, but also the range from high to low graphically. I think you would get a better take on the actual differences between the asset classes each year this way.

[siamond]

One thing I'm not a fan of in these tables is it doesn't show scale. For example in the chart in the OP, the difference between Highest and Lowest in 1998 is 47%. The difference in 2017 is only 16%. But the table makes it looks like these differences are similar because it's just laid out high to low.

To me it would be much more interesting to show not just high to low, but also the range from high to low graphically. I think you would get a better take on the actual differences between the asset classes each year this way.

Yes, I agree, this is a big drawback of this representation. I've pondered about the scaling issue as well, but then the approach I took (a good dose of conditional formatting of fixed-size cells, fed by sorted tables) would no longer work. I started to play with a stacked column chart with Excel, we can make it work reasonably ok, but then we really lose the visual effect about diversification that the Callan's periodic tables illustrate so well. So I think we should take it for whatever it is, a representation of the quasi-randomness of winners vs. losers year over year, and no more.

[siamond]

Just for fun, I assembled a table for bonds. Here it is. Click to see a larger image.

LTT = Long-Term Treasuries; ITT = Intermediate Term Treasuries; STT = Short Term Treasuries; T-Bills = Treasury Bills; IT Corp = Intermediate Term Corporates; HY Corp = High Yield Term Corporates; TIPS = Treasury Inflation-Adjusted Securities; Global = Global (unhedged) bonds; Int'l = ex-US (hedged) bonds; TBM = US Total Bonds Market.

[GAAP]

New investors should read this wiki article: Callan periodic table of investment returns

wiki wrote:
The Callan Table is the best visual information showing the importance of diversification, reversion-to-the-mean, and the impossibility of forecasting asset-class returns. It is a primary reason Bogleheads favor total market index funds.

The wiki synopsis of what the table illustrates is accurate and clear, but this is a very nice synopsis of some of the implications:

I've always found these tables to be excellent food for thought. After awhile, I came to a conclusion that a person needs to ask themselves which of the following describes their approach:

a) I want to be in everything so that I'm guaranteed to have some in the "winning" bucket each year (it also guarantees a person will have some in the "losing" bucket each year)
b) I want an asset class that is never or rarely at the bottom (which also will likely never or rarely be at the top)

A US Total Market investor should be comfortable with the second statement. A total world investor, the first.

[LadyGeek]

I would agree, but should that post be in the wiki?

From the perspective of a new investor, that statement is confusing. I don't see the distinction between (a) and (b).

======================
On a more technical note, footnote 4 describes a metric known as "coefficient of variation". Would that be useful here?

[LadyGeek]

Can't find the link now, but someone in the past, posted a Callan Table with zero % as the reference line for stacking the colored blocks, more clearly showing the differences between the + and - returns for each year. Low returns don't feel as bad as the negative ones. On a paper Callan Table, we can draw a reference line between each annual divide of the + and - returns.

I actually found an easy way to automate this formatting variant. Here is the full chart with the international data for the past 20 years.

Have you experimented with a stacked bar (column) chart? It would be similar to the above, but the height of each block would be scaled relative to the total. I think this will give you a sense on the magnitude of the relative returns (larger returns have larger block sizes), which may be closer to heyyou's suggestion.

(This is just a suggestion, implement or not as you wish.)

[staythecourse]

Sorry if I am repeating as I am late to the party and too lazy to read the other replies.

Can the periodic table reflect more of all the MAJOR subasset classes and not the granular country or cap size subassets? I think a quilt of: sp500, russell 2000v, EAFE, EM, TIPS, TBM, Tinternational bonds, cash, gold, commodities (CCF), REITS, etc... would be the most useful represented over the last 20 years (for example). Then an interactive way so one can make one's on portfolio weights and see how it shows up in that same quilt over the last 20 years.

It would be a great way of seeing diversification in action where it should never be the leader or the worst each year. My guess should be in top 1/3 and out of bottom 1/3 each year IF truly diversified. Also would be a great illustration of the effect of increasing weighted average to different subasset classes (for example: equal weighted to all the sub asset classes vs. 5% to REIT and gold)

Good luck.