Stock Market Visualization: the S&P 500 since July 2nd 2015

The last two weeks have been anything but boring. For a brief moment, it looked like a combination of China and Greece might very well pull us over the edge. During this period the VIX swung from 16 to 20 then down to 12. At one point last week the VVIX, or volatility of volatility, briefly touched 110 which marks its highest level since January of this year. As we have harped upon in the past, spikes in VVIX represent periods where traders are revising their estimates at a rapid rate and thus can’t even trust their own perception of the market. Our trader’s intuition says we shouldn’t be surprised to see wild swings from lows to highs in this kind of environment, but this is a hypothesis we should test in short order.

In the meantime, we wanted to present a more detailed picture of correlation during this turbulent time. Thus we have produced our first Minimum Spanning Tree created from intraday data: 1-minute log-changes over the two weeks from 7/2/2015 to 7/16/2015. You can take an in-depth exploration of this tree for yourself by downloading SliceMatrix.

Screenshot from SliceMatrix of the S&P 500 from July 2nd to today

Screenshot from SliceMatrix of the S&P 500 from July 2nd through the 16th

Each graph plots components of the S&P 500. The bigger a node, the more connected it is to the rest of the network. The thicker an edge between two nodes, the greater the correlation between the stocks at each node. The color of each node is colored using a red to blue to green colormap where bright red means the stock’s return was negative, blue is unchanged, and green is up. The edge color is a blend of node colors.

The most important stock by some margin is Marsh & McLennan Companies (MMC) which shares 55 connections to other stocks in the S&P 500 index. It is important to remember that this chart was generated from a correlation filtering algorithm, not subjective tags like industry classifications. Thus it is important we understand MMC’s business model so as to properly interpret the fact that so many other stocks fall within its sphere of influence. Here’s MMC’s business summary (emphasis ours):

Marsh & McLennan Companies, Inc., a professional services firm, provides advice and solutions primarily in the areas of risk, strategy, and people worldwide. It operates in two segments, Risk and Insurance Services; and Consulting. The Risk and Insurance Services segment offers risk management services, such as risk advice, risk transfer, risk control, and mitigation solutions, as well as insurance, reinsurance broking, catastrophe and financial modeling services, and related advisory services… The Consulting segment offers health, retirement, talent, and investments consulting services and products; and specialized management, and economic and brand consulting services…

MMC is strongly connected to insurance company MetLife (MET) with an intraday 1-minute correlation of 80%, which in turn is connected via correlation to Nucor (NUE), a producer of steel. At first the connection between a financial services firm and a steelmaker might seem odd, or at least not very intuitive. Investigation of the two time series, however, shows a similar factor could be driving the action in both stocks (subject to tradable idiosyncratic shocks):

met_nueThe intraday correlation between NUE and MET is close to 70%. What could be causing such a high correlation? Perhaps insurance and steel-making are more deeply linked than appear at first glance. Further investigation is warranted, but when you think about it, the two products are highly cyclical and levered to the health of the overall economy. Both the creation of new businesses and the expansion of existing ones leads to an increase in marginal demand for both insurance and steel products.

Following along the chain, NUE is connected to ConocoPhilips (COP) an Oil & Gas exploration company. This connection makes more intuitive sense as both business model’s rely on natural resources, which may themselves be highly correlated. COP, in turn, serves as a local hub within the graph. It is connected to many other Oil & Gas companies such as Schlumberger (SLB) and Noble Energy (NBL), to name a few:

Close-up view of Oil & Gas stocks

Close-up view of Oil & Gas stocks

zoom_in2

Traversing a different branch of the network emanating from MMC takes us to other local hubs like cruise-line operator Carnival Corp (CCL) and 3M Corp (MMM).

For anyone new to SliceMatrix, please check out our video tutorial, SliceMatrix 101 on YouTube.


Check out the Beta release of SliceMatrix: a unique tool for visualizing the stock market, including views of filtered correlation networks and minimum spanning trees


Want to learn how to mine social data sources like Google Trends, StockTwits, Twitter, and Estimize? Make sure to download our book Intro to Social Data for Traders


Enter your email address to follow this blog and receive notifications of new posts by email.


follow_st


3 replies »

  1. Interesting weird correlations you’ve outlined here!

    Could the correlation between such seemingly unrelated stocks be something external and/or artificial? Like perhaps the same fund or group of funds using the same trading techniques on those stocks? If for example you’re a HFT fund, you might pick stocks that had low correlation in the long term, but use the same trading system on them, causing the intraday data to have a high correlation?

    Just a thought.

    Liked by 1 person

  2. yes yes and yes. theres a lot to reply to here, but we’ll take an initial try…

    a lot of what we see is caused by algo’s for sure. these days a lot is due to a proliferation of quant trading. its alot easier to open a medium term stat arb shop than invest the millions of dollars in tech that a true HFT firm takes to open. HFT shops burn money on computers, programmers, and colo space…

    its all kind of fractal in nature. consider 3 different people trading the same group of stocks: a high frequency trader, medium term trader (MTT) and long term trader (LTT)…

    HFT’s are trading all the time, so when a MTT or LTT trader makes a bet on the same spread as an HFT, they are by construction taking a trade away from HFT’s. In concert, however, they are all keeping the correlation in tact.

    In a way the existence of the three groups is self-stabilizing. HFT’s trade spreads very tightly, so news (or just the natural relative movement of stocks) can push spreads to the point that HFT’s “puke” their positions. They find liquidity in MTT and LTT who are keen to profit off a divergence, but are willing to take more “heat” due to their enlarged timeframes.

    This kind of process would generate ebbs and flows in ST and LT correlation. periods when ST correlation is high would attract more and more HFT’s until the market becomes destabilized: so many people are on the same side of the trade, their collective presences can push a spread to the limit. We would see breaks in ST correlation which would be stabilized by longer term traders

    Like

  3. Thanks for the thoughtful response to my comment! Lots to think about. 🙂 Some sort of realtime correlation indicator comes to mind as being useful somehow. Probably a bit like watching ocean waves ebb and flow.

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s