Data platforms such as Stocktwits may hurt long-term forecasts

AllAfrica

4 days ago

Since the beginning of the century, the number of satellites orbiting Earth has increased more than 800%, from less than 1,000 to more than 9,000. This profusion has had a number of strange and disturbing repercussions. One of them is that companies are selling data from satellite images of parking lots to financial analysts. Analysts then use this information to help gauge a store's foot traffic, compare a retailer to competitors and estimate its revenue.
This is just one example of the new information, or 'alternative data', that is now available to analysts to help them make their predictions about future stock performance. In the past, analysts would make predictions based on firms' public financial statements.
According to our research, the plethora of new sources of data has improved short-term predictions but worsened long-term analysis, which could have profound consequences.
In a paper on alternative data's effect on financial forecasting, we counted more than 500 companies that sold alternative data in 2017, a number that ballooned from less than 50 in 1996. Today, the alternative data broker Datarade lists more than 3,000 alternative datasets for sale.
In addition to satellite images, sources of new information include Google, credit card statistics and social media such as X or Stocktwits, a popular X-like platform where investors share ideas about the market. For instance, Stocktwits users share charts showing the evolution of the price of a given stock (e.g. Apple stock) and explanations of why the evolution predicts a price increase or decrease. Users also mention the launch of a new product by a firm and whether it makes them bullish or bearish about the firm's stock.
Using data from the Institutional Brokers' Estimate System (I/B/E/S) and regression analyses, we measured the quality of 65 million equity analysts' forecasts from 1983 to 2017 by comparing analysts' predictions with the actual earnings per share of companies' stock.
We found, as others had, that the availability of more data explains why stock analysts have become progressively better at making short-term projections. We went further, however, by asking how this alternative data affected long-term projections. And we found that over the same period that saw a rise in accuracy of short-term projections, there was a drop in validity of long-term forecasts.
Because of its nature, alternative data – information about firms in the moment – is useful mostly for short-term forecasts. Longer-term analysis – from one to five years into the future – is a much more important judgment.
Previous papers have proved the common-sense proposition that analysts have a limited amount of attention. If analysts have a large portfolio of firms to cover, for example, their scattered concentration begins to yield diminishing returns.
We wanted to know whether the increased accuracy of short-term forecasts and declining accuracy of long-term predictions – which we had observed in our analysis of the I/B/E/S data – was due to a concomitant proliferation of alternative sources for financial information.
To investigate this proposition, we analyzed all discussions of stocks on Stocktwits that took place between 2009 and 2017. As might be expected, certain stocks like Apple, Google or Walmart generated much more discussion than those of small companies that aren't even listed on the Nasdaq.
We conjectured that analysts who followed stocks that were heavily discussed on the platform – and so, who were exposed to a lot of alternative data – would experience a larger decline in the quality of their long-term forecasts than analysts who followed stocks that were little discussed. And after controlling for factors such as firms' size, years in business and sales growth, that's exactly what we found.
We inferred that because analysts had easy access to information for short-term analysis, they directed their energy there, which meant they had less attention for long-term forecasting.
The consequences of this inundation of alternative data may be profound. When assessing a stock's value, investors must take into account both short- and long-term forecasts. If the quality of long-term forecasts deteriorates, there is a good chance that stock prices will not accurately reflect a firm's value.
Moreover, a firm would like to see the value of its decisions reflected in the price of its stock. But if a firm's long-term decisions are incorrectly taken into account by analysts, it might be less willing to make investments that will only pay off years away.
In the mining industry, for instance, it takes time to build a new mine. It's going to take maybe nine, 10 years for an investment to start producing cash flows. Companies might be less willing to make such investments if, say, their stocks may be undervalued because market participants have less accurate forecasts of these investments' impacts on firms' cash flows – the subject of another paper we are working on.
The example of investment in carbon reduction is even more alarming. That kind of investment also tends to pay off in the long run, when global warming will be an even bigger issue. Firms may have less incentive to make the investment if the worth of that investment is not quickly reflected in their valuation.
The results of our research suggest that it might be wise for financial firms to separate teams that research short-term results and those that make long-term forecasts. This would alleviate the problem of one person or team being flooded with data relevant to short-term forecasting and then also expected to research long-term results. Our findings are also noteworthy for investors looking for bargains: though there are downsides to poor long-term forecasting, it could present an opportunity for those able to identify undervalued firms.
Thierry Foucault is a professor of finance at the HEC Paris Business School .
This article is republished from The Conversation under a Creative Commons license. Read the original article.