That’s a part of the big picture I’m trying to understand. What is Dozens’s target market?
I’m fully aware that at the moment it’s focused on the spender-saver-investor, but what’s the mid to long-term plan? Do they want to target at experienced investors, or a step further, day-traders? It’s worth pointing out that the spender-saver-investors will gradually become more experienced, but unlikely to turn themselves into day-traders.
For those soon-to-be novice investors, I’d argue that there may be enough warnings on the app, but the recommendations/guidelines given to them are way too misleading, and may have dire consequences. See my comments on the irresponsible goal setter above.
Ideas to fix and improve the goal setter:
- Ask the user when do they need the money, instead of calculating it based on the risk level the user choose. If they need the money in 5 years or less, recommend them to use cash savings products instead.
- Ask the users whether they have expensive (as defined by APR) debts, and make it clear that this doesn’t include the student loan.
- Check (and ask if they don’t have enough on Dozens, because they might have it elsewhere) the user have enough emergency funds.
- Change the predicted annual return to a more conservative and realistic number. More on this points below.
- If the investment fall short of the expected return, what can the user do in order to meet their goal? (e.g.: delay the purchase, draw savings from elsewhere, or purchase something less ideal but cheaper instead, depending on what the goal is)
- Have some pre-build portfolios for each risk group. As a starting point, those multi-asset funds (perhaps the world equity fund too) can form a portfolio on their own. I’d argue that most if not all other funds are not suitable to be hold alone.
Let’s discuss the expected annual return. My best guess is, right now it’s using the annual return (CAGR) over a long period of time. This might be okay if your are investing for 30 years or longer, but we rarely do. For a shorter time horizon, it is a fairly misleading number. Invest for one year in the highest risk group, the reality will fall short of it more than half of the time. I.e. the user will have less than 50% chance to reach their goal. Even hold the EM fund for 5 years, there’s still 35% chance to loss money. The chance does improve if the user chooses a lower risk fund, but that’s not the whole story either. It becomes virtually impossible to reach their goal if the user chooses the lowest risk fund - short duration gilts. (this is exactly the reason why I said those funds are not suitable to be hold alone)
So, what number should you use in the app? I personally would recommend use the 85th percentile annualized real return (CAGR) over rolling N-years periods from a few decades worth of data, where N is the number years the user intended to stay invested or 10, whichever is smaller. For the examples in the post below, I use individual fund for the sake of simplicity. But the returns should be measured on portfolio, not individual funds. And the portfolio needs annual rebalancing.
For example, invest 10 years in the emerging market equity fund, the recommended method estimated annual real return is about -2.3% (using last 50 years historical data, before fund & platform costs and tracking error). That’s not a typo, it is negative. Because if you invest in EM equity alone for 10 years, there’s a non-negligible chance (more than 15%) that you will end up losing more than 20% (equivalent to 2.3% per year compound) of the initially invested money.
Note: if insufficient data is available (e.g.: the robotic fund, which tracks the STOXX Global Automation & Robotics Index, and historical data is only available since 20 June 2011), this becomes very tricky, and I don’t really have any good recommendation at the moment. Monte Carlo simulations may help, but I don’t like it because it’s hard to take the order of events into consideration without creating any bias. Anyone who can think of something that may work, please feel free to write down your ideas.
I’m gonna explain the methodology a bit.
Why percentile? CAGR is good for measuring past performances, but it’s biased toward to the extreme but rare values, so it’s not a good measurement for managing expectations. If a fund reliably returns 1% annually but had an one-off event doubled it’s value in one of the year over the last 50 years, it will have an average return of 2.39%. But can you expect that you will get a 2.39% annual return by investing in it for the next 10 years? I highly doubt about it. Anyone reasonable would expect a 1% annual return. A quantile number solves this problem nicely. You will get exactly 1% annual return for this fund, whichever percentile number you choose, except the 1st and 2nd percentiles.
Why 85th percentile? It result in 15%, or about 1 in 6.67 chance to fall short of the expected return. This is a small enough probability and should work for most people. You could pick 80th (20%, one-fifth chance) or 90th (10%, one-tens chance) if you like, but I wouldn’t go further than those two numbers. Go below 90th will include way too much noise and make the expected return unrealistically low. Go above 80th will introduce significantly higher chance to fall short of the expectation of the investors. You may also add an additional “average luck” scenario, which would use the median (50th percentile) number, this should give the user a good sense of what to expect from an average market condition. But if you do provide it, please be responsible and warn the user that they will only have about 50% chance to achieve this number.
Why annualized real return? Annualized percentage rate makes it easier to compare different products. You’d expect a bank to advertise a 2% AER 5 years fixed interest savings account, not 10.4% 5 years fixed interest savings account for the same reason.
Why annualized real return? For those who aren’t familiar with the terms, real return means inflation-adjusted return. The opposite is nominal return, which is not adjusted for inflation. The reason to use real return is because investment often involves a fairly long time horizon, and inflation cannot be ignored. Invest £100,000 at 8% can make you a millionaire in 30 years, how exciting? But by that time £1,000,000 would only worth a bit over £414,000 in today’s money if the inflation was 3%, not so exciting any more, right? If someone is saving for an university fund for their new-born baby, they will almost certainly use the cost of university in today’s money as their goal, unless they are financially well-educated, but then they should be able to read and understand the small prints on the app, and pick the correct numbers to use anyway.
Why over rolling N-years periods? For high volatility investments, such as EM equity, annual return over the entire time period (8.3% in the last 50 years) isn’t very representative and will give you a false sense of high return and low risk. The discrete returns is even worse - too many extreme values. You need one representative number that doesn’t over or under-estimate the returns, and doesn’t create a false sense of security, that’s why I recommend the percentile return over rolling N-years periods. When the N is close to the user’s time horizon (but capped at certain level, to avoid being too optimistic), this will give a realistic range of returns the user can expect, and probabilities to achieve the returns.
Why measure the portfolio, not individual funds? When you save £10,000 in a 5 years fixed rate cash savings account, do you care if the bank split the money into 5 different pots, some gains money and some losses money? No you don’t. You only care that you will get your £10,000 plus interest back at the end of the 5 years. The same applies to investments. You invest £10,000 in 5 funds for 10 years, the only thing really matters is what you get back in total at the end of the 10 years, not how these 5 funds individually performed.
The advantage of this method for measuring long-term returns are:
- Being a percentile number, it’s less affected by very rare and extreme events. These events will have bigger impact on the CAGR over the entire period.
- It still takes considerations of the very rare and extreme events, and both the frequency of them and the movement ranges matter too.
- It takes consideration of the sequence of events, which is something often missing from Monte Carlo simulations based on standard derivation and average returns. E.g.: the market tends to bounce back up after a crash, but it’s not guaranteed (see Japan).
- It also takes other indirect and often delayed factors into consideration, such as currency movements caused by the not always timely fiscal and monetary policies responding to the market movements, and the effect of the eventual unwinding of the stimulus.
Honestly, the expected returns with probabilities will benefit experienced investors too, and help everyone makes informed choice about the risk they take and the returns they can expect. For an example, even you are a very adventurous and risk-seeking person, if you can achieve your long-term financial goal with 99% chance by investing in a lower risk portfolio, why would you choose to invest in a higher risk portfolio and gives you only 80% chance to meet your goal?
Plus, I haven’t seen any platform offering this kind of tools. You will have a selling point if you made this.