The Chicago Board Options Exchange Volatility Index, or VIX, is often cited as a market-timing indicator. Traders have created a variety of strategies to exploit supposed correlations between this “fear index” and future market returns. Using Palantir Finance, we can effortlessly examine the exact nature of these correlations to make more informed trading decisions. The inspiration for this study came from a report from the Credit Suisse Quantitative Trading and Derivatives Strategy group. The original report can be found here.
The VIX measures the implied volatility of index options based on the S&P 500. Plotting both the VIX and the S&P 500 on a chart with a split axis, a pattern immediately emerges.
The VIX and the S&P 500 appear inversely correlated, especially during two notable periods – the tech bubble and the real estate bubble. This is an interesting observation, but it isn’t especially helpful. We want to know whether the VIX can act as a leading indicator for market returns. To achieve this end, we look at the VIX’s current standing relative to its recent trend, or “relative level”, defined as the ratio of its current value to its N-day moving average, minus one. This isn’t a commonly used financial metric, but we create it easily enough in the Custom Metric tool.
The Custom Metric tool offers a great deal of power and flexibility. With it, we can create powerful functions and seamlessly integrate them across all of Palantir’s applications. Here, we use the Chart tool to plot the relative level of the VIX to future returns on the S&P 500. The following chart shows the VIX relative level using a 90-day moving average, plotted against 3-month future returns of the S&P 500.
Although it appears that these two plots may be positively correlated, it’s hard to infer anything with too much confidence. When money is on the line, it’s important to be confident in our analysis. Ideally, we’d have some quantitative measure of their correlations. There are two major parameters we can vary to adjust the results: the number of days to include in the VIX moving average, and the window in which we examine returns on the S&P 500. Using the scatterplot’s color coding functionality, we can create a heatmap to visualize exactly how the correlations vary as we change these parameters.
The x-axis varies the number of days we use to calculate the VIX relative level, and the y-axis varies the length of S&P 500 forward returns. The color of each point represents the correlation between these two series. For example, the color of the point at coordinates x = 100 and y = 100 represents the correlation of the VIX relative level using a 100-day moving average with the 100 day forward returns of the S&P 500. As shown by the scale on the right, red color means higher correlation, while blue means lower. Take a closer look at the color scale. The values range from 0.1262 to 0.3036. This means there is a positive correlation over the entire domain of both parameters (50 – 250 days for both). We can adjust the color scale to isolate points with very high correlation. In the scatterplot below, the color scale minimum has been increased to 0.2750. All points with correlation below 0.2750 are colored blue.
A horizontal band centered around the y-value of 150 emerges as having the highest correlation, suggesting that the VIX predicts the market best in the medium term, about 150 trading days into the future. The pattern of horizontal streaks indicates that the results are more sensitive to the forward return window than the VIX moving average window. Now we can open up the Table data view and sort the values by correlation to find the point that gives the absolute highest correlation.
The highest correlation between the VIX relative level and the S&P 500 forward returns occurs when using a VIX moving average of 148 days and looking at 146-day forward market returns. Combining the power of custom metrics with the utility of the scatterplot tool, we were able to generate a visualization of the sensitivity of a well known market indicator against several parameters rather easily. Using this analysis as a starting point, a good follow up exercise is to create a strategy to test the historical profitability of a trading scheme based on using the VIX as a leading indicator for medium term market performance.