The Unusual Correlation Between StockTwits & Market Volatility

Two years of StockTwits Volatility

Example 1. shows 2014-2016 correlation between StockTwits message volume and VIX futures price. Charts and data processing provided by @MKTSTK.

Volatility in markets is inevitable.

In a world of incomplete information, dislocations between expectations and reality can become quite severe. In the options and futures markets, contracts with similar expiries may have a much larger discrepancy in price than similar measures in equity markets. Dislocations often remain until there’s a rebalancing of expectations, causing rapid snapbacks in price, and some of the storied 7,000% trades.

For the purposes of this paper, rather than focus on the specific discrepancies in options contracts, or the discrepancy between front month and second month, we thought to look at something a little different: Pure volatility, and how it directly relates to the StockTwits ecosystem.

Behavioral Risk Modeling.

To take a page from the behavioral finance lexicon; this is a story of risk, and the avoidance of that risk, in realtime. During periods of “risk-off”, and depending on what side of the trade you’re on, there’s an opportunity to pile into a trade, or disassociate yourself with that exposure. On StockTwits, what we see is a conversation flow. There’s a “flight to safety”, and an avoidance of risk, which is measurable, and repeatable. This translates to not just trading volume, but a shift in asset classes as well: conversations flow from equities, into broader indexes, and different trading instruments altogether. During heightened volatility, conversations around ETFs and Index BOOMS, and rightfully so; nobody want’s to be caught red-handed holding high-beta solar names or microcap biotechs when markets are volatile.

Q3 Volatility Indicator

Example 2. shows how the ETF message volume percentage (red) moves lock-step with VIX futures (blue). 

Methodology behind the construction.

How does it work? StockTwits users can tag zero or more $TICKER symbols in each message they produce. To measure the saturation of ETF message volume we treat each of these symbol mentions as an unique ‘event’. This allowed us to measure ETF’s symbol flow over any given rolling look-back window (we chose about 3.5 days for this research). To arrive at the Rolling ETF Message Percentage (denoted in red), we take the total number of ETF symbols mentioned and divide it by the total number of symbols mentioned over the look-back window. Updating this percentage every very 5 minutes, we get a consistent view of ETF message saturation across the StockTwits ecosystem. 

Q3 2015 Volatility Indicator

Example 3. Shows the discrepancy between acceleration and deceleration of message volume and VIX price. 

VIX futures construction.

The futures prices in the above and below charts are a systematic blend of the front and second month VIX futures contracts. The weights (and price) are proportionate to time left to expiry in the first month futures contract. Duration targeted is one month.

Q1 2015

Example 4. pictures Q2 of 2015, where volatility trended downward, and StockTwits traders were stock pickers. By @MKTSTK.

Thoughts on correlations.

Without comment from our in-house expert, “As the charts above lay bare, there is a persistent relationship between the current level of ETF symbol saturation and the level of volatility in the stock market. Over the sample period, 2014 to the start of 2016, the correlation between near month VIX futures prices and ~3.5-day rolling ETF saturation is 74%. Using a 30 minute look-back window drops the correlation to around 51% as the ETF series becomes much more erratic and possibly subject to intraday seasonalities. Nevertheless the results are encouraging from a volatility modeling perspective.” – @MKTSTK.

Q2 2015

Example 5. looks at the dramatic uptick in Q3 of 2015, where we see a relatively flat volatility market, followed by a huge spike.  

Applications & future research.

Our studies have been eye opening, but are just the tip of the iceberg. These findings, as preliminary as they are, have far reaching implications for any market participant sensitive to US equity volatility. And further research on the data bodes well for volatility predictiveness, as well as providing early warning signs for risk managers and regulators who monitor market participants. More research is needed to determine the causal structure at work. It is also quite possible that a subset of ETF symbols make up for the bulk of the correlation with volatility. As we look to further refine our research, we could see the potential for the creation of a volatility index or indicator based solely off social media data.

We hope our partners will take an active role in pursuing such initiatives. If interested in learning more, or to gain access to the underlying data, reach out to us (data@stocktwits.com). Also, be sure to sign up for our monthly data-focused newsletter.

This paper was produced in partnership with our in-house expert, @MKTSTK, whose work is quickly gaining recognition in academic and financial circles around the globe. Download the full piece here.


Tickers: , , , , , , , ,

blog comments powered by Disqus
StockTwits Blog