The Excel file did not go through (maybe PRC security settings). Attache is a text file. To use it and separate the columns convert it first to Excel (copy – paste).
I would like to compare the results of this table with my tables, but cannot figure out how to download the .TXT file. Anyone able to tell me how?
@hecht790 Thanks!
I was logged in, and it did open in my browser, but I didn't realize I could just CTRL-S to save the text file on my computer. 😬 I was looking for a "download" option for the file. Appreciate your help!
@hecht790 Seeing all those for the purposes of Pralana makes me very nervous, I'll admit. People who do monte carlo with so many assets usually use correlation matrix tables, to reflect each assets correlation with each other. Out of curiosity, do you also use all ten classes for monte carlo too or just historical? How do you find predicted returns for so many types of stocks? Do you have any bonds too? If so, if you don't lump together the stocks you'll run into trouble with either "correlated" check-box (since it will mess up on the bonds) or "uncorrelated" (which will mess up on the misc stocks). Given all the types of stock are so correlated with each other, have you considered for the monte carlo using a weighted average of all your stocks expected return and sd in a single 'stock' category? That would preserve the independence of stocks and bonds if you use "uncorrelated". I guess if you are 100% equity this is moot cause you could keep all ten but have them correlated in the monte carlo.
Once I added the data and ran the historical analysis, I got extreme results, and I am not sure they are realistic. I need to play more with the options. I ran Monte Carlo (and Historical - best year 1985, worst year 1962) with all the assets, with and without correlation. I used Simba data for returns. All my 10 equities are the same size. each 6.6% of the total assets. The remaining 33% are bonds (short, intermediate and TIPS) which are under a single asset in Pralana. I do not know if a single Stocks asset will give me a more realistic result (small, large, US, international, RIET and EM are not 100% corelated). S&P500 has different historical characteristics than my 10 assets. I used to have 2 assets, Stocks and Bonds. I split the Stocks trying to use PRC to optimize Asset Location.
Out of curiosity I ran these assets through Portfolio visualizer to find the 10y correlation matrix, which I upload here. (The pv link I generated this with is this.) It's only looking at 10 years; but you could make your own correlation matrix based on simba is you really wanted to use all the data.
My conclusion is: 1) All the stock categories are for practical purposes quite correlated with each other. Emerging the least, but even there never less than 0.71 with SV. 2) ST and IT are close to 0 correlated with all the stock categories. 3) REIT is its own animal, kind of half stock-half bond. 4) TIPs seem a bit different too, but there is much less data available so I'd ignore them, and they're still much closer to bonds then stocks.
So if I were modeling a monte carlo, I'd pick "uncorrelated" (0% correlation) but lump all the equities as "stocks" and use a weighted predicted return, and I'd use a sd based on the weighted historical average from simba. Maybe keep REITs separate. I'd keep bonds separate, lumped, to capture the diversification benefit of the 33%.
In short, three asset classes, uncorrelated (stocks/bonds/reits) .
You would know that in reality you would get some diversification benefit within the stocks that would not be evident in the monte carlo, but it would be very small. And "uncorrelated" REITS as an asset "uncorrelated" with everything else would over-estimate their diversification benefit with stocks (given it's closer to 1 than 0) but again not enough to ruin a model. But you'd capture the very potent -.10 correlation between bonds and stocks as a whole.
Historical analysis should in in theory work with all the assets without lumping. Did it?
Just my two cents that you didn't ask for. :).
@jkandell thank you,
You are correct, stocks can be bundled into a single asset. I split the Stocks trying to use PRC to optimize Asset Location. International stocks have higher dividends, lower qualified dividends and foreign tax. So, location is not obvious. Yes, historical analysis works fine. BTW Simba itself has a correlation table.
@hecht790 What did you discover about location optimization? An irritating fact I've encountered is that foreign equity used to be best in taxable cause of the credit; but for the last several years, in my case the tax from dividends has overpowered the foreign tax credit.
I discovered that few improvements would help to get a more reliable result.
- Correct the Dividends calculation. Currently the Dividends are % of growth. So, if the market did not grow (or grow negatively) you get “0” dividends. This is wrong. A better way is to calculate the dividends based on % of the asset value. If an asset loses 20% the dividends will lose 20%. This improvement is #2 on the PRC voting list.
- Add “Foreign tax paid” to each international asset. This tax is a tax credit in taxable account and lost in IRA or Roth. (it also will be nice to use the 1099 numbers instead of % of growth numbers).
- Enable dragging the assets’ names up and down (easily reorder the names) in the assets list. The order of the names affects the Asset Location algorithm. Playing with names’ order manually will help optimize the location.
- Add Asset Location Optimizer to the tool. This is probably not simple to implement and less important if the user only has US assets. However, it may provide a unique tool that also helps select specific assets.
(1) and (2) above will also improve the PRC tax report accuracy which is a great feature of PRC.
The result that I got was to locate US stocks in Taxable and International stocks in IRA/Roth. However, as I explained this result is inaccurate.
I have 3 data links that others may find useful:
1. The link below has annual total returns from 1928- for SP500, US small caps, 3 month TBill, 10yr Tbond, Baa Corporate, Real Estate, Gold, and inflation.
pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/histretSP.html
2. Shiller publishes his CAPE ratio data which has total returns for SP500, 10yr TBonds and infaltion monthly back to 1870. Note that the SP500 values are average value for the month. I'm not sure why.
3. The Early Retirement Now website has a link to their "EarlyRetirementNow SWR Toolbox v2.0". This is a google sheet for calculating safe withdrawal rates (SWR). On the Asset Returns tab it provides monthly total return data back to 1870. This has more than schillers spreadsheet and includes: inflation, SPX, 10yr BM, 30yr BM, Cash, World Ex US Equities, Commodities and Gold.
The Safe Withdrawal Rate Series - Early Retirement Now
@zuiker01gmail-com Hey, Chris, thanks for sharing this. Sterns has always been Pralana's source for this info but until now they haven't had any historical data other than the S&P, bills and bonds. We'll try to get this new data incorporated into Pralana's historical database sometime soon.
Stuart
@zuiker01gmail-com Hey, Chris, thanks for sharing this. Sterns has always been Pralana's source for this info but until now they haven't had any historical data other than the S&P, bills and bonds. We'll try to get this new data incorporated into Pralana's historical database sometime soon.
@Smatthews51, since all of the returns in Pralana from Stern (and ERN and Shiller for that matter) are from indexes, it would make sense for Pralana to request a user's average fund ER to take off the top in historical simulations. (No index fund makes the actual index returns, after all.)
