© 2016-2019 Fowler Drew Limited

  • Black Facebook Icon
  • Black Twitter Icon
  • Black LinkedIn Icon

Authorised and regulated by the

Financial Conduct Authority 

  • Stuart Fowler

Robo advisers

Publishing, music, photography have all succumbed. In finance, electronic trading and peer-to-peer lending are disrupting existing business models. Could investment management be next, asked FT’s Pauline Skypala in her Smart Money column: ‘Robo advisers’ arrive to pose tech challenge to investment managers‘.

Beta harvesters

Prompted by a Forbes Magazine article that coined the phrase ‘robo adviser’, Pauline refers to two new US businesses that rely on quantitative techniques (computer-resident, mathematical, decision rules) to build and then maintain asset-allocation driven portfolios assembled using low-cost trackers (ETFs in these cases). ‘Total costs’, observes Pauline, ‘come in at 0.5% or less – much lower than the 1.5-2% typical of standard investment advice and management for retail customers’.

They do not try to time markets but rebalance the weights to prevent excessive drift from target weights arising from differential returns. Even these rules will have a performance impact however. In other words, these investors are low-cost ‘beta harvesters’, rejecting the ‘selection alpha’ or outperformance that is the labour-intensive activity that explains so much of the high industry cost of so-called active management. With beta harvesting, everyone can be a winner by holding ‘efficient’ risk exposures at the asset allocation level. Alpha is by definition a zero-sum game, with a loser for every winner.

The American market is much more attuned than we are to the idea that alpha hunting brings problems as well as costs: selection decisions that have to be made and lived with by the client, not just professional advisers and managers (see my post Six degrees of implausibility). One of the best-selling investment books in the States is Charlie Ellis’ Winning the loser’s game. Its simple sporting analogies provided the light-bulb moment for many private clients who opted out of alpha hunting.

What they typically retain, however, is diversified asset allocation, optimised by reference to volatility, ie the short-term movements in portfolio level or what we term ‘path risk’. It’s a standardised solution, blind to different investors’ own single or multiple time horizons, to the nature of their own liabilities (needs or wants) and to the benefits and consequences resulting from better or worse outcomes than they need or want.

Portfolio asset allocation optimised on common-period volatility was one of the first applications of computer-resident decision processes as far back as the 1980s when increases in computing power allowed portfolio construction theory developed in the 1950s to be put into practice.

Quant alpha hunters

Ironically, robots are also perfectly capable of selecting skilled managers, where there is sufficient data on the past performance of the funds they manage. Quantitative rules apply probability theory and appropriate statistical tests to identify funds whose performance is unlikely to have been achieved by chance alone. They can do this, but the results may not make it worth doing. Alpha hunters typically reject a quantitative approach because they believe that the soft factors they think are so important will not show up in the data. But if they don’t, they aren’t worth knowing because they don’t predict performance.

Quantitative techniques are widely used in many hedge fund strategies where either extreme speed or ruthless objectivity are seen as critical sources of an edge. The purpose of the edge is to pile alpha on top of returns that, like the underlying balanced portfolio of conventional wealth management, are optimised on short-term volatility. But this application of technology is probably best viewed by most private investors as as a component of a portfolio rather than a solution for all their investment needs. It is also increasingly being seen as being replicated more cheaply by a combination of cash (with no industry charges) and a much smaller risky portfolio (which the industry cannot make so much out of).

The use of quantitative techniques, both in the institutional and retail markets, has itself been handicapped by the shortcomings of the volatility-optimised approach to risk control. In the USA, the active ‘quant’ share appears to have peaked in 2006 at about $800 bn (compared with $2,500 bn in passive quant). It survived the 2000 bear market but not the 2008/9 bear market, because the underlying hope of diversification as a risk control was that asset correlations would not increase in a bad market. They did. Bye bye risk control. The active quant share dropped to around $300 bn and is only slowly recovering.

Replicate or replace?

So there is no question that robots can do what expensive humans currently do. But is that the right question? Surely what we should be asking is whether technology can help us do something different that would be not only cheaper but also more useful for clients.

This is the question Fowler Drew sought to answer. We combined well-developed theories of consumption and investor utility to design and manage portfolios (at the asset allocation level) that would deliver the horizon-specific outcomes each client defined, in the form they needed or wanted, subject to the constraints that they specified (such as sustaining real spending within tolerable limits, based on how they viewed the benefits and consequences, and never running out of capital). And to be really useful, we knew we had to be able to make regular projections of outcomes that were always consistent with current portfolio values.

The model we use to plan a defined-outcome goal, construct the required portfolio and then manage it throughout the plan was built and parameterised in 1998/9. It was tested almost immediately by a bear market and subsequent move to exceptional peak levels of real equity prices in the dot-com boom, all this before this firm was started and before we could put the model into practice in real portfolios. Since 2005, we have relied exclusively on the model to manage goal-based portfolios. We have made only small, gradual refinements to the model and none to the underlying concepts. It has done exactly what we asked of it, including re-projecting new, goal-specific, outcome probabilities that have transformed the clients’ experience of living with volatility.

The key difference relative to models that required implausibly good short-term forecasting was that short-term liabilities sensitive to volatility are bound (by the typical outcome constraints) to be invested in risk free assets: cash for the shortest liabilities, index linked gilts and National Savings certificates for the next set of liabilities. Volatile risky asset are only held against longer-duration liabilities where time is available either to speed up or slow down the response to changing prices and changing future returns. Maybe there is an element of ‘mental accounting’ that explains the advantage of goal-based and outcomes-driven investing but if this is what helps ordinary investors to live their lives with confidence about their financial goals, regardless of what is happening to economies and markets, it is still an advantage worth having.

A disruptive robo version of what we do is perfectly possible in the mass market. The challenge is not scaling up the management of the assets. It is rather the user’s remote interaction with a model or engine to decide (in an iterative ‘conversation’) the plan parameters: resources, time horizons, target outcomes and risk approach. This might need to be both within and across different competing goals. We know none of this requires understanding of markets or investment, as long as people can articulate their needs and wants and can react to information about outcomes (such as visualising the benefits and consequences). And it doesn’t just arise at the outset: the point about outcomes-driven investing is that progress might trigger revisiting the variables, such as taking more or less risk, lowering or raising the targets or shifting the allocation of capital between two goals. As a service, it is more about journey management than portfolio management.

Pension freedoms: green light for technology

I wonder whether the spur to change has not already happened, in the shape of the pension freedoms announced in the March 2014 Budget. By undermining the insurance industry’s predictable income from annuities, the Chancellor will unleash a competitive response from these firms. What we know about the modern life insurance firm is that it is a big user of quantitative techniques to manage both its own balance sheet and the retail products created on or off its balance sheet. The kind of probabilistic modelling we do, driven by risk management more than return seeking, is amazingly rare and culturally alien and threatening in traditional wealth management firms. But it is deep in the DNA of insurance companies.

I have not forgotten that one of the first customers for our model was an insurance-company internet subsidiary: Egg. Though Prudential sold Egg just before it rolled out its pilot investment-advice service, the marketing team at Egg had understood perfectly the transformative power of journey management.

Right on cue, in an article in FTfm the same week, two senior business heads at Allianz Global Investors argued how deficient the standard solutions for retirement income are in the UK and set out Allianz’ stall for alternatives that will unquestionably be rich in quantitative processes. So too are the compelling alternatives that already exist in modern versions of with-profits like Prufunds and in drawdown with upward-only income guarantees, such as MetLife offers.

Before that, BlackRock, a massive global competitor with a strong bias to quantitative techniques, had reacted within weeks to the UK pension changes by showing its interest in providing more intelligent solutions.

The tech challenge is definitely here and conventional wealth managers I talk to know it. But having a realistic survival plan is a different matter if you’re culturally opposed to management by algorithm. If they look back at the occupational pension market in the UK, they will see that the old-fashioned balanced managers that owned that space have mostly already gone the way of the book publishers and record labels that once thought they owned their space.

#activemanagement #assetallocation #businessmodels #costs #investmentprocess