If we extend the time frame out to look at monthly or annual returns, the picture is clearer and the distribution looks log normal even if the tails are still fat.
Are programs that use a log-normal distribution superior to those that use a normal distribution? Possibly. Does it make a big difference? In many cases no. It doesn’t. Moreover, the difference shouldn’t matter if your expectations for what MCS can and cannot do are reasonable. More on that in a minute.
The second criticism is that with many believing future returns will be lower than historical averages, using historical data overstates a client’s chances of success.
To this, I say, “duh.” If you think returns will be lower in the future than they have been in the past, of course projecting those lower returns will suggest weaker results. Do we really need software to make that case?
Don’t get me wrong. Arguments that returns are likely to be lower going forward than what the historical record would suggest, at least for a while, do not strike me as outlandish by any means.
There are few certainties with respect to markets, but with bond yields so low, there is a mathematical certainty that many of the good quality bonds on the parts of the yield curve we would consider will lag the returns of similar bonds in the past. How long this will persist is unknown.
It is also true that by many valuation measures, stocks are not cheap. How “not cheap” they are is subject to debate, of course, but regardless of what you think about valuation, it would be a stretch to assume current valuations are permanent.
Nonetheless, are the critics correct in saying that if you use historical parameters with MCS, you are misleading clients? That depends on how you use MCS.
If you are using it as a crystal ball and telling clients definitively “you are going to be fine,” you might be encouraging an exaggerated sense of security. But using weaker than historic assumptions and definitively stating that clients are up a creek could cause an unneeded level of fear too. It can also cause them to work longer or spend less in retirement unnecessarily.
If you use return assumptions that are too gloomy for too long, you will be simulating an environment that has never existed. There may be value in that.