story

How do robo-advisers invest, and what can we learn from them?

by Vincent Skiera
Posted on September 07, 2020

A robo-adviser is a fully algorithmic investment strategy, aimed at giving each individual a personalized portfolio that optimally matches his or her personal preferences. Robo-advisers execute investments and customer on-boarding automatically, without human involvement. This crucial feature empowers them to offer services for much lower investment amounts than human advisers typically require. Where human advisers often require investments of at least €50,000 - rather unattainable for the majority of the population - some of their AI counterparts have a minimum investment amount of only €10 (Finanztest, 2018).

My research explores the net benefits or net costs from the perspective of various stakeholders. The data I use observes one particular robo-adviser introduced by a bank in 2017 that I call “a typical robo-adviser” going forward, representative for robo-advisers in general and behaving similarly to other robo-advisers. In my previous blog post, I showed how a robo-adviser can benefit society at large by increasing participation in financial markets.1 In this post, I explain how exactly robo-advisers invest. What can retail investors learn from their investment behaviour?


How does a robo-adviser tick?

Most robo-advisers’ investment strategies are based on value-at-risk models. Personalization enters the investment strategy by taking into account information that the specific customer - a retail investor - provided during the on-boarding process, usually through a questionnaire (think of factors such as investment horizon, risk tolerance, financial literacy, income, and wealth). The robo-adviser's main aim is to construct a portfolio that provides a high return but is unlikely to exceed the loss that the customer considers to be tolerable.

A customer’s on-boarding questionnaire, for instance, may indicate that (s)he wants a portfolio in which a yearly loss that exceeds 25% of the portfolio’s value is expected to happen only once every 20 years. After identifying this level of risk tolerance, the robo-adviser calculates which portfolio with such loss characteristics achieves the highest expected return. It then allocates the customer’s money accordingly, typically across many asset classes, ranging from stocks and corporate bonds to sovereign bonds and commodities, and that cover all regions worldwide.

As the prices of a portfolio’s holdings change over time (with stocks increasing and corporate bonds decreasing in value) the allocation of a customer’s funds across asset classes and regions also changes. This change in allocation may lead to a situation in which the portfolio is no longer optimal, at least not according to the customer’s criteria. As stocks are riskier than bonds, for example, an increase in the amount of money invested in stocks may mean that the risk of the portfolio exceeds what the customer is willing to tolerate.

Similarly, expected returns associated with various asset classes and the risks that these asset classes have in common may also change over time. In other words, a robo-adviser buys and sells assets to re-achieve the intended allocation, given the current expected returns and common risks, a process known as rebalancing. In my example, rebalancing would include selling stocks and buying bonds.

A robo-adviser covers each asset class and region with exchange traded funds (ETFs). An ETF is a set of securities, typically replicating an index such as the S&P500, whose ownership is bought and sold on an exchange. The popularity of ETFs has greatly increased among both retail and institutional investors, as they are easy to trade and give great diversification at low cost.


Skiera-fig-1.jpg#asset:15264

Figure 1 depicts the make-up of the average portfolio managed by retail investors, versus the portfolio managed by a typical robo-adviser at the end of 2018. Importantly, the label of ETF encompasses many different assets. For example, one ETF may contain stocks in the US, replicating the S&P500, whereas another ETF may contain corporate bonds in Europe.

How does a robo-adviser invest?

Figure 1 shows the make-up of an average retail investor's portfolio, in a sample of over 15,000 retail investors based in Germany, and that of a typical robo-adviser. As you can see right away, the robo-adviser exclusively invests in ETFs, while retail investors invest in a whole range of different assets, their main holdings being stocks (54% of assets). Diversified assets, such as mutual funds and ETFs, make up 38% of their portfolios.

So why do robo-advisers only invest in ETFs? They want a diversified portfolio, and ETFs allow them much diversification at low cost and with low complexity. Instead of holding all 500 stocks of the S&P500, for instance, a robo-adviser can hold one ETF replicating the S&P500 instead.


Skiera-fig-2.jpg#asset:15267

Figure 2 shows the distribution of the number of trades in 2018, for portfolios managed by retail investors and by a typical robo-adviser. For retail investors the unit of observations is the number of trades a single retail investor does in 2018, while for a typical robo-adviser the unit of observation is the number of trades in a particular customer’s portfolio in 2018. Trades above 200 have been grouped in the rightmost bar.


How does a robo-adviser's investment behaviour differ from a retail investor’s?

Robo-advisers also trade much more frequently than retail investors do. Figure 2 shows that the median number of trades per retail investor was 6 in 2018, whereas a typical robo-adviser executed a median of 81 trades per customer—more than 13 times as many trades.2

Why do robo-advisers trade so often? As discussed above, they trade to rebalance their customer's portfolio, often buying multiple ETFs for a few hundred euros each, and selling multiple other ETFs, also for a few hundred euros. This rebalancing leads to many trades, but each trade is small.


What can retail investors learn from robo-advisers?

Given the differences between robo-advisers and retail investors, it is natural to wonder whether human investors should behave more like their algorithmic counterparts. A robo-adviser differs in two main points from a retail investor: (1) it holds only diversified assets, generally in the form of ETFs, and (2) it trades much more often.

Let's first consider asset composition. As noted above, robo-advisers strive for diversification. In general, an undiversified portfolio— holding only a very small number of stocks, perhaps even fewer than five —is problematic: the average stock has the same expected return as its index, but an individual stock often has 2 to 3 times as much risk as its index. Thus, an individual stock exposes the investor to unrewarded risk.3

“Many retail investors take on much more risk than necessary to achieve their expected return.”

The fact that the portfolios of robo-advisers and retail investors look very different (Figure 1) does not necessarily imply that investments made by robo-advisers are more diversified. For example, a retail investment portfolio that includes a large combination of stocks may be just as diversified as a robo-adviser’s portfolio that includes stock ETFs. Moreover, ETFs and mutual funds make up a large portion (38%) of retail investors’ invested assets, hinting at substantial diversification. Indeed, many of the mutual funds that retail investors select target indices similar to those of robo-advisers’ ETFs, achieving similar diversification.

It seems that retail investors would benefit from adopting robo-advisers’ prioritization of diversification. In our sample, 2,208 out of 15,159 retail investors (14.5%) had an undiversified portfolio4, meaning many retail investors take on much more risk than necessary to achieve their expected return. It may come as no surprise that many of the undiversified retail investors in my sample actually started use a typical robo-adviser later on, increasing their diversification.

What about trading frequency? Should retail investors also imitate the robo-adviser’s tendency to trade often? Not quite. The downside of trading at such high frequency (75 additional trades for the median retail investor!) is that each trade is costly for the retail investor. A typical bank might charge a retail investor at least €5 for each trade. If a retail investor were to make 81 trades per year (like a robo-adviser) instead of 6, this would translate into additional costs of at least €3755per year—and probably even more. Moreover, studies have shown that when retail investors trade very frequently, their return before fees is largely unchanged, but the trading costs greatly reduce the return after fees (Barber & Odean, 2000).

Given these fees, why does a robo-adviser trade so often? First of all, a robo-adviser has much lower fees than a retail investor. It can cancel out trades, avoiding trading fees when it buys an ETF for one customer and sells the same ETF for another customer. Furthermore, its large size allows it to negotiate better prices than retail investors, reducing the fees from trading even further.

To sum up, retail investors can learn a great deal from robo-advisers, particularly with regard to the value of diversification, but they should be cautious about imitating them in all respects, particularly with regard to trading frequency.

Vincent Skiera is a Doctoral Candidate of Finance at Haas School of Business at UC Berkeley.

Footnotes

  1. We even extended the analysis of the previous blog post, showing that our Robo-adviser’s customers who are new to financial markets would have otherwise not participated in financial markets, meaning that the number of new active investors did not decline with the introduction of our Robo-adviser.
  2. The mean for an investor is 22 trades due to some outliers with very high numbers of trades, while our Robo-adviser has a mean of 86 trades for each customer in 2018, still almost 4 times as much, and more than 60 trades more.
  3. The risk is unrewarded, as the investor does not gain any additional expected return from taking the additional risk in the single stock. Holding many stocks allows the investor to reduce the unrewarded risk from holding individual stocks.
  4. I classify a portfolio as undiversified if it holds fewer than 5 assets, e.g. stocks or bonds, and no ETFs or mutual funds.
  5. €375 =(81-6)× €5

References

  • Barber, B., & Odean, T. (2000). Trading Is Hazardous to Your Wealth:The Common Stock Investment Performance of Individual Investors. Journal of Finance, 55(2), 773-806.
  • Finanztest. (2018). Robo-adviser: Die Maschine Machts. Finanztest, 42-47.