Friday, April 4, 2008

Will The Employment Report Cause Large Range Expansion?

Compared to the recent volatility the last two days have been very tame. One line I’m hearing is that traders didn’t want to take big bets before Friday’s employment report. The expectation seems to be that the employment report will spark a big move Friday one way or the other.

To test this I compared the 2-day average true range with the 20-day average true range. The ratio as of Thursday’s close in the SPY was about 0.53. I ran a test to see what happened when this ratio had dropped below 0.55 or lower going into a report. The basic expectation was that the contraction in volatility would reverse after the news was released and lead to an explosion in volatility.

That did not hold true. The table below shows the results.

The third column shows the true range on the day of the release vs. yesterday’s 20-day average true range. The average for the 9 instances was 0.94 – meaning the true range after the report failed to reach even average size (1). In the last column I showed the percentage gap that SPY opened the next morning.

You’ll probably hear a lot of hype about the importance of the number Friday morning. Don’t be too surprised if it turns out to be just that – hype. Historically after such contractions it hasn’t led to the volatility explosion that you might expect.

If you would like to explore the action leading up to and on employment days in more detail and you use Tradestation, you may purchase the study here.


thelonelytrader said...

This stuff is very difficult to quantify for me. Once I come up with a result, I seem to come up with more questions.

I've run my own tests on the majors in the spot currency markets for employment, inflation and capital flows data -- and the hype, as it turns out, is largely true for the very short term (15 mins to 4 hours). For the longer term, not so much, but you can still build a model with a decent chance of success -- just not very spectacular in terms of gains. (I can't, because my connectivity and platform aren't fast enough.) Like anything else, the results need a lot of qualification. According to what I learned, the most important factors going into post-release price action were:

(1) Market expectations going into the move;

(2) Price action in the two sessions (Sydney/Tokyo and London opens) prior to the move, and;

(3) How much the actual result differed from expectations.

So I needed to put a little more nuance into the testing than I initially thought...for example, I've been keeping a tally for the last two years of inflation-related releases and there are positive correlations in the following areas:

(1) If the market expects a "big number", there will likely be frontrunning of these releases -- but this frontrunning has different time horizons for each release;

(2) If the results are in line with expectations, there of course is no "explosion" in volatility *unless* there is an underlying statistic that casts a nuanced view of the headline figure;

(3) If the results beat expectations by a significant margin, there is a short term explosion in volatility...which can turn into a longer term move depending on a number of factors.

More qualifications come into play:

i - Was the result in line with the prevailing fundmental and technical themes that were already driving price action for that particular currency pair?

ii - Did the release occur near a key technical level (e.g. round figures, stop clusters, etc.), where the mechanics of order flows (microstructural factors) were conducive to a large expansion of volatility? Were there large barriers preventing such expansion?

iii - Were underlying data in-line with or out-of-line with the headline figure? And how does underlying data, or for that matter the headline figure, combine with other macro- and micro-factors driving price action for a given currency?

There are a number of other questions, but I suspect you've already looked at these issues. While it is possible to quantify these events, the results can change dramatically depending on what factors are introduced and how the data is parsed.

Have you looked at any of these questions for SPY?

Anonymous said...


you nailed it, good call and research! (one thought, what would be the result if it was tested based on a big big move during the week?)


Rob Hanna said...


Thanks for sharing your knowledge in this area. You've put a lot more effort into assesing these reports than me. Studies I've done with regards to employment or other econnomic statistics are fairly simple.

The reason I haven't explored them so in depth is that - for my trading - the edge often appears after the reaction (overreaction) to the report. I tend to look for edges based on price, volume, breadth, and/or sentiment following economic reports and less often try and read into those reports.

Thanks again for your thoughts,