A 2003 paper by Pan and Poteshman entitled The Information in Option Volume for Stock Prices (found here), and highlighted by this 2006 New York Times article suggests that events in equity options markets can serve as predictors for trends in the stock market. Their fundamental claims are two-fold: first, that options traders are better informed than the majority of the market, so optimism in the equity options market will tend to lead good performance in corresponding stocks, and vice versa; and second, that this phenomenon is more observable in firms where information flow is less efficient. The academic study uses information from a proprietary database offered by the Chicago Board Options Exchange (more on that later). Here, we will replicate the study using less descriptive data, but the basic workflow and thought process remains identical.
Pan and Poteshman measure the mood of the options market for each firm using the ratio of put volume to call volume. More puts relative to calls generally indicates pessimism while more calls to puts indicates optimism. Efficiency of information flow is more difficult to measure, and we will make the same assumption as Pan and Poteshman that information flow is more efficient in larger firms. To test these hypotheses, we construct the following strategy: divide a market into large, mid, and small cap stocks; for each of these groups, construct a portfolio that buys the bottom quintile of stocks by put/call ratio and short sells the top quintile; and rebalance the portfolio every few days. Using the Explorer Group tool, we can quickly parcel out the groups of instruments we are interested. The screenshot below shows stocks in the top third of the NYSE by market capitalization, and bottom quintile by put/call ratio.
In our strategy, this group of stocks comprises our long positions, while the top quintile comprises our short positions. We can quickly adjust the filters to grab the groups we are interested in, e.g., the top and bottom quintiles of large, mid and small cap stocks when ranked by put/call ratio.
The Strategy tool allows us to test the historical performance of this idea. We construct a custom backtest to build a portfolio of long and short positions on these quintiles, and rebalance as the quintiles change over time. The screenshot below shows the results of our strategy for large cap stocks starting in 2008, rebalancing once a week.
Our strategy has two major parameters that could affect its performance: the size of the firms considered, and the portfolio rebalance interval. Instead of manually altering the parameters in the strategy setup — a tedious task if the parameters exist over even a small domain — we can use the Twiddle feature to examine how the performance changes as we adjust these variables. Below, we plot the annualized return of our strategy, incrementing the holding period from one to seven days along the X-axis and market cap from small to large along the Y-axis. If we believe this is primarily a short term strategy and more effective with smaller stocks, then we should observe higher returns nearer to the bottom left corner of the table.
The success of this strategy does appear to increase with lower holding periods and smaller firms. We observe dramatic gains using small cap stocks and rebalancing our portfolio everyday. With a rebalance period of greater than three days, the annualized return on the strategy is relatively weak, indicating that any excess information in the equity options market has likely been transmitted to the stock market, and accounted for in stock prices.
As mentioned earlier, the paper by Pan and Poteshman uses a custom database that distinguishes option orders according to whether it opens a new position or closes an existing position, and by the type of investor initiating the trade (e.g., proprietary trader, public customers of discount brokers, etc.). Presumably, an order that opens a new position is more likely to occur as a result of new information, and thus should be weighted more heavily. An interesting followup would be to integrate this additional data into the study and observe the results.