There’s an old joke about analysts – you don’t need them in a bull market and you don’t want them in a bear market. Traditional analysts have their flaws, but a new study suggests there is an alternative – robo-analysts.
Robo-analysts refer to research firms that use technology to mass-produce recommendations with limited human involvement. Robo-analysts’ upgrades and downgrades don’t make the headlines and markets pay little heed to them, but that may be a mistake, according to a new study from researchers at Indiana University.
The study, Man Versus Machine: A Comparison of Robo-Analyst and Traditional Research Analyst Investment Recommendations, examined more than 76,000 recommendations made by seven robo-analyst firms between 2003 and 2018 and found their stock recommendations were much more profitable than those made by human analysts.
They also appear to be much more objective, issuing roughly four times as many sell recommendations as traditional analysts.
Good as the results are, the robots aren’t going to replace human analysts any time soon.
Firstly, while the robots do much better overall, they are not perfect, and their sell recommendations are no better than those made by ordinary analysts.
Secondly, while computers can quickly digest huge amounts of data, they can’t incorporate “soft” information – for example, information gleaned from conference calls or discussions with company management – into their analysis.
Thirdly, the study notes that analysts facilitate access to management and help organise specialised investor conferences, “boutique services that institutional investors increasingly value and place emphasis on”.
Still, the robots have already made inroads into the investment industry. Computerised strategies are responsible for the vast majority of stock market trading these days and investors’ exposure to quantitative strategies is expected to continue increasing in the coming years.
Ordinary investors are increasingly availing of the services of robo-advisers, online services that automate financial planning by using algorithms to construct and manage portfolios.
“Ripe for disruption”
Evidence increasingly suggests the analyst industry is “ripe for disruption”, say the researchers. Over the last two decades, structural and regulatory changes, such as Mifid II, which prevents firms from bundling the cost of analyst research into trading fees, have forced analysts to cut costs and slash budgets. Separate research, for example, shows analysts are increasingly likely to issue earnings forecasts for multiple firms on the same day.
This bundling of forecasts is a cost-saving measure, and one that has resulted in “less accurate, less bold, and less informative” reports, as one study put it. The Indiana researchers contend that the overall quality of traditional analyst reports “appears to have declined”.
Additionally, time-pressed analysts can only do so much, so they are more likely to focus on corporate news releases such as earnings reports that are “easier to process and interpret”. Some research indicates analysts are slower to respond to information contained in company regulatory filings, presumably due to the sheer volume and complexity of the information they contain.
In contrast, robo-analysts don’t need to be reliant on earnings announcements, as they can quickly incorporate the complex information contained in the aforementioned filings.
Then there is the age-old problem of the conflicts of interest faced by traditional analysts. One reason investors listen to analysts is because of their close access to company management, but analysts risk getting the cold shoulder from executives if they slap a sell rating on the company in question.
Additionally, it’s hard for an analyst to be an impartial commentator if they are reviewing a company with whom their firm has an investment banking relationship with.
In contrast, robo-analyst firm New Constructs states that its research is “uncompromised”, and the study notes that many other robo-analyst firms make similar claims.
Certainly, robo-analysts issue recommendations that are “substantially more negative” than traditional analysts’; almost a quarter of ratings (24 per cent) are sell recommendations, compared to just 6 per cent among traditional analysts.
There is also a wide gulf when it comes to buy recommendations; they account for almost half (47 per cent) of traditional analyst recommendations, compared to just 32 per cent for robo-analysts.
Economic incentives are not the only reason for this optimism gulf. The study finds robo-analyst reports are also less optimistic than those produced by independent analyst research firms which sell their reports directly to clients and which don’t have any conflicts of interest related to investment banking.
One explanation for this is that, while independent analysts may not have any motivational biases, they do have cognitive biases that can undermine their judgment.
Whatever the explanation, the reality is robo-analysts produce “significantly more revisions” to their forecasts, whereas traditional analyst recommendations “tend to be very persistent” and become “stale over time”. Standard recommendations are “sticky”, with separate research showing that analysts affiliated to investment banks are especially slow to downgrade stocks with buy or hold recommendations.
Again, limited attention as well as economic incentives is likely to play a part in such instances.
A paper recently published in The Accounting review finds analysts are “less likely to issue timely earnings forecasts” for a firm if they also have to cover another company announcing earnings on the same day. Similarly, they are “more sluggish in providing stock recommendations and less likely to ask questions in earnings conference calls as their number of concurrent announcements increases”, the paper notes.
Those conclusions echo those of University of California behavioural finance expert Prof. David Hirshleifer, whose research shows that very busy analysts are more likely to herd with the consensus forecast or to reissue previous forecasts, and are more likely to issue less accurate rounded forecasts (forecasts ending with zero or five).
Ironically, while robo-analysts appear to issue more accurate and objective recommendations than their traditional counterparts, financial markets continue to pay much more attention to the latter.
When a traditional analyst upgrades a stock, it tends to generate a pretty decent price pop of around 3 per cent.
In contrast, robo-analyst recommendations are largely ignored; “at best”, the authors note, their upgrades drive a price increase that is barely one-tenth the size of the return associated with traditional analyst upgrades.
However, while markets don’t react immediately, portfolios based on robo-analysts’ buy recommendations can generate “substantial” returns for investors over time, the study notes.
For now, this remains a niche area. If the robots are coming, however, then it’s bad news for analysts but potentially very good news for active investors seeking a more scientific investing approach.