The retail stock boom in Russia is gathering momentum and, like elsewhere, is increasingly being fueled by social media, according to a survey by VTB Capital (VTBC) titled “To the moon: Scoring social stock sentiment “, published on September 22.
After six years of almost continuous rate cuts by the Central Bank of Russia (CBR), retail investors have abandoned traditional high-interest bank deposits and turned to stocks in search of an investment with a better return. .
The CBR has long complained about the low level of financial literacy of the population, but thanks to the deep penetration of the internet – there are more people online in Russia than in all of Germany, the most populous country in the world. ‘Europe – and the sophisticated fintech applications that the big Russian banks are developing, this ignorance is disappearing.
The coronavirus crisis has only accelerated the process, as Russians forced to stay at home spent more time playing with their phones and computers when they searched for a place to park their savings and protect them from the ravages. high inflation and instability of the ruble.
As recently presented by bne IntelliNews, the St. Petersburg-based company SPB exchange was the big winner in the switch to actions by ordinary Russians. Retail investors used its platform to buy shares listed overseas, but the Moscow Stock Exchange (MOEX) has also seen the number of individual accounts soar into millions over the past few years and also started offering access to overseas-listed stocks last summer.
“Investors in the Russian retail stock market have become a mature and material investor base whose investment models and preferences have a huge influence on market performance,” said Alexander Isakov, chief economist of VTBC in a note.
“We are trying to better understand the sentiment of the collective retail investor base on a granular level by studying unstructured text data from Russia’s most active stock market-focused social media and applying modern language processing tools. natural machine learning, ”he added.
Retail investors now account for around 40% of the turnover of the Russian stock market and increasingly affect its performance. For example, in September, international investors sold some of this year’s gains, but Russian retail investors were still buying aggressively, keeping the market buoyant, according to BCS GM. International investors selling should have reduced exposure to Russia, but over the course of this year it has increased and they are currently overweight Russian stocks, mainly because they have bought a slew of IPOs over the course of the year. of the last 12 months. Retail investors, on the other hand, rushed into Russian blue chips, with banking stocks being by far the preferred investment.
VTBC results show that for a subset of publicly traded stocks, the data suggests a potential causal link between social media sentiment and price changes, and therefore information about social sentiment can help inform asset allocation decisions of professional investors.
Below we reproduce the set of questions and answers from the VTBC survey that were published in a report on the impact of social media on retail investment decisions in Russia:
Q: What is the size of the retail investor base in Russia? A: It is large and growing: the CBR reports total retail holdings of 6 trillion rubles ($ 82.5 billion) in 1Q21 (up from 2.6 trillion rubles in the first quarter of 19); MOEX reports some 14 million unique retail customers, while MOEX officials report that more than 15% of recent IPO books have been filled with retail orders.
Q: You want to capture the sentiment of retailers, but in what way?
A: Yes, we are trying to capture social sentiment towards individual actions.
More concretely, we collect large-scale publicly available textual data, identify discussions of publicly traded companies and map them on a five-point scale ranging from “strong sell” equivalent to “strong buy” equivalent messages. , and all the rest.
A: In short, public social media posts + modern word processing machine learning methods.
Most notably, we are following text updates posted by financial market focused groups with significant audience (i.e. 1k +) on one of the most popular social media platforms in Russia, Telegram .
We then fit a set of text classification ML models to a manually evaluated seed message sample, our models ranging from standard complexity (i.e. regularized logit) to state of the art ( i.e. a classification neural network head above the Russian language deep learning model SBERT). The resulting sentiment score is a set, i.e. consensus, of our models.
Q: What is the scale of the exercise? Is it representative?
A: The total number of audiences covered is approximately 1.5 minutes, which represents approximately 11% of unique accounts and 75% of active retail investor accounts on MOEX.
The channel audience (channels are essentially author-centric communities where messages are translated to the audience) certainly has a substantial intersection and the unique audience is considerably smaller. Nevertheless, we believe that the universe of opinion is sufficiently covered.
We are deliberately focusing on a stock picking segment of social media and are not general “news aggregators” in order to reduce signal-to-noise ratios.
Nonetheless, even in the stock picking community, “non-price cross-reading fact statement” messages dominate the flow.
Q: What’s the most important thing to remember from an equity investor’s perspective?
A: It depends on the investor – and the sector covered – because stock prices, according to our figures, are more or less sensitive to public opinion.
In order to estimate this sensitivity, we perform a Granger causality test for 15 social sentiment lags influencing changes in stock prices and calculate the corresponding pvalues. The null hypothesis here is that there is no causality, so the lower the p-value of a stock, the stronger the evidence that the influence of public sentiment on that stock’s prices should not be excluded.
The bar chart below left shows some of the stocks with the lowest average p-values in our sample spanning the period January 2019 through August 2021. For their holders (actual and potential), social sentiment tracking could make sense and improve asset allocation.
Q: What is the current social sentiment per share?
A: In the graph below, we exclude neutral sentiments (i.e. messages regarding the statement of facts) and only show those who have a strong current sentiment. We also censor tickers with the most negative social sentiment.
Q: Have you tested the performance of a hypothetical portfolio allocation strategy based on social sentiment?
A: Yes. The charts below show a single equity strategy performance based on equity sentiment versus “buy and hold” from January 2019 to August 2021 based on daily returns of US dollar denominated stocks.
For illustrative purposes, we have estimated a strategy in which short stocks are “hated” or “hated” and long stocks are “liked” and “liked”. The charts below show that for some stocks, social sentiment outweighs simple buy and hold, and for others, it lags.