GLD Return Distribution & How It Changed Since Summer 2011

Yesterday GLD dropped 5.3% to close at 164.29. Most of this decline happened intraday (from open to close: -5.1%). In fact, GLD lost some 7-8 dollars in mere 60-75 minutes between 10:00 and 11:15 ET.

Although in the recent years we got somehow used to big daily percentage changes in major macro markets including gold, yesterday’s fall was the 6th largest close-to-close fall since the launch of GLD in November 2004 and the second largest open-to-close fall (the one larger was 10 October 2008 with -7.1% open-to-close and -7.4% close-to-close on GLD).

GLD Chart

GLD chart, daily, since summer 2011

Distribution of GLD Returns

The following is an overview of the distribution of GLD daily returns: close-to-close, close-to-open (gap), and open-to-close (intraday). The underlying data is in percentage points, always as % of previous day close.

A few key points that the data reveals:

  • Although gold was clearly trending up all the time since the launch of GLD, there were slightly more big down days than big up days (compare the 1st percentile to the 99th or the 2.5th to the 97.5th and also see negative skewness). Median close-to-close performance was +0.11%, arithmetic mean +0.08%.
  • On average, vast majority of the overall appreciation of GLD can be attributed to overnight moves (gaps). Arithmetic mean is 0.075% for gaps, 0.006% intraday, and 0.081% total.
  • On the other hand, many of the extremely big up moves and big down moves happened intraday rather than overnight. Compare the 99th, 1st, 2.5th, 5th, and 10th percentiles for close-to-open and open-to-close. Also note the positive skewness and very high kurtosis for intraday moves.
Distribution of GLD daily returns, all time, close to close Distribution of GLD daily returns, all time, close to next open (gap) Distribution of GLD daily returns, all time, open to close (intraday)

GLD Return Distribution since Summer 2011

The picture changes markedly if you only look at the period since August 2011 (shortly before GLD made its all-time high). You can notice a change in GLD price behaviour even by looking at a chart only, but the numbers speak more clearly.

  • For most of the time since summer 2011 gold has been a consolidating, sideways, but quite volatile market. Of course the close-to-close arithmetic mean was lower (+0.038%) compared to the previous years, but the average overnight performance was still much better than intraday, which was negative (-0.044%). In fact, the average overnight move was even a bit higher than for the all time stats (+0.083% vs. +0.075%).
  • GLD realized volatility (standard deviation) has been up since summer 2011 for all three periods (close-to-close, close-to-open, open-to-close). There was still very little difference between standard deviation of overnight and intraday price changes.
  • There have been much fewer days with very small price changes (positive or negative – compare the 25th and 75th percentiles all-time and since summer 2011) and more days with big changes. For example, daily declines greater than 2% were about two times more frequent since summer 2011 (see 5th and 10th percentiles for all time and since summer 2011) and it was similar for big daily increases (90th and 95th percentiles).
Distribution of GLD daily returns, since summer 2011, close to close Distribution of GLD daily returns, since summer 2011, close to next open (gap) Distribution of GLD daily returns, since summer 2011, open to close (intraday)

What It Means

As usual, I won’t speculate regarding the future price direction of gold and GLD. The statistics above means just that gold/GLD in the last 6-7 months was quite different from the gold/GLD before. Differences in return distributions can have huge consequences for profitability and risk/return profiles of various option trading strategies.