### What’s Wrong with Robo 1.0?

Please forgive me if this sounds like a rant against simplistic approaches to projections, but I guess that is what it basically is. I’m going to ignore all the customer-focused aspects of this for now and simply focus on the anchoring bias I see in a lot of implementations of the calculations. Bad assumptions = bad outcomes.

I'm getting a little concerned about the prevalence of robos that advertise their "Nobel Prize-winning technology" and simplistically project a fixed real return to the point of retirement (occasionally with some standard error bands) to tell someone their projected $1m will fund the lifestyle of their dreams. There are lots of problems with this entire process and I want to address a few of them briefly.

#### Real Returns

What about the correlation of inflation with financial assets? While equities have had a relatively stable long-run real return of about 6%pa over the past 200+ years, the long run is rolling 30-year periods! And even at that interval, there is some substantial variation. But how about bonds? Just taking the last 20 odd years shows that the relationship with real bond yields and inflation is quite meaningful. Property is hardly immune to a correlation with inflation either. To think about all assets only in real return space (and then adding the token 2.5% inflation value on) misses the incredible short-term correlation that fears of deflation have on financial assets in the shorter term. Which leads me to my next point.

#### Sequencing Risk

Simply extrapolating any annual value over decades and adding in a standard error band misses the massive consequence of real-world market gyrations and the effects it has on the ability of an individual to accumulate wealth over time. The "short-term" fall in 2008 was catastrophic for anyone retiring any time around that period. Much less so for someone who had just started their working life and had most of their capital locked up in their future earnings (Human Capital). The ratio of Investable Capital to Human Capital forms a nice proxy for how serious a short-term fall can be to an individual.

#### Asset Mix and Distance to Liability Value

This follows on from sequencing risk - the financial aspects of retirement planning mostly comprises an Asset Liability Modelling problem. The challenge is estimating the value of the liability (which comprises two broad estimates – how much spending will there be per annum in retirement and how many years the retirement will last – but also includes binding constraints such as the in/ability to go back to work for health reasons, ageism etc) as well as projecting the value of assets. Simply projecting the current risk profile (don't get me started on how badly that is determined) to the day of retirement can massively overexpose a pre-retiree to unnecessary market risk. If they need the proverbial "$1m" to retire and the forward projection of their assets at the date of retirement using a zero-coupon bond value is $1.2m, why should they take as much risk as someone who's forward projection is $500k? Just because they "like risk"? If the liability has been accurately valued, the risk in the investment portfolio should reflect both the desires of the individual as well as the practical consideration of how close they are to their goal.

#### Expected Returns

Embedded in a return prediction – especially one over decades – is a view of the world. Will it be like the average of the past 10 years, 20 years, 50 years? How confident am I in that prediction? What if we get a world that isn't in a historic data set? Simply extrapolating historic returns from a period into the future is an appalling way to avoid having to think about what might actually turn out and how confident we can be in that expectation. Expected returns are not easy to predict, but the longer the time horizon (up to a point) the better chance we have of being approximately right if we can understand what is driving returns on that time frame. The key here is "approximately right". These return expectations will still reflect a view of the world, but by combining different views of the world (either through scenario planning or Monte Carlo techniques) at least there is an acceptance that more than one future is possible.

#### Expected Asset Correlations

In a similar way that extrapolating the past into the future for expected returns implies a view of the world, so does taking a historical covariance matrix and assuming that holds into the future. Should equities and bonds be positively correlated? They were in the 1970s and both went down (in price). They were in the 1990s and both went up. That correlation was being driven by an underlying force – inflation moving higher and then lower was the central theme shifting the way assets behaved in relation to each other. If I'm fixing my inflation at 2.5% forever, does that mean I need to find a period where inflation was stable historically and use that correlation matrix to describe the future? And I haven't even started on the correlation between investible assets and Human Capital - do wages go up in line with inflation?

Enough rant. Each of these areas requires thousands of words in their right to explain in better detail. My next piece will be some of the thinking we have at OnTrack as to how to better tackle these issues.