www.CartoonStock.com
Up to now I have never been tracking the Permanent Open Market Operations (or "POMO") activity of the FED (it wasn't in my arsenal of seasonalities like the release dates of the Job Report and the Consumer Price Index, or the FED announcement sessions), but a recent comment to one of my postings inspired me to dig in and check if - and to what extend - a tradable edge might be provided on (or before / thereafter) days where the FED conducts Permanent Open Market Operations (means the FED is 'reinvesting principal payments from agency debt and agency mortgage-backed securities (MBS) in longer-term Treasury securities', see Permanent Open Market Operations - FEDERAL RESERVE BANK of NEW YORK).
What caught my eye at first glance was the fact that on September 24, 2010 (last Friday), the FED conducted it's 9th coupon purchase of the current month, and there will be 2 more (September 28 and 30) this month for a total of 11 operations. Unfortunately historical market data for Permanent Open Market Operations is available from 8/25/2005 to present only (there were 196 operations since 8/24/2005).
I found a couple of related articles on the web / blogospere (e.g. an excellent one from Bob English at The Precision Report, as of August 2, 2009), almost all of them assuming / stating that on POMO days (liquidity intervention by the Fed) money is being shuffled into the stock market, with a statistically significant ramp up in major market indexes, especially after the end of the auction (operations are regularly scheduled to begin around 10:15 AM and close at 11:00 AM ET) into the close. But instead of checking what historically happened intraday during the second half of the session on those days where the FED's Permanent Open Market Operations occurred, I thought it would be more interesting (you'll know why at the end of the posting) to check for the market's
- end-of-day performance, and
- successive short- and intermediate-term performance where those POMO days accumulated.
(If you're interested in a short summary only, please click here)
First of all Table I shows the SPY's historical end-of-day performance (since 8/24/2005) on those days where the FED's Permanent Open Market Operations occurred (assuming one went long on close of a session immediately preceding a POMO day).
At first glance (Compound Return, Geometric Growth per Trade, Profit Factor) it seems that a tradable edge might be provided, but the above-average profitability on POMO days is mainly due to the significantly below-average percentage of violent losses on those days (e.g. max. loss -3.46%). In fact the probability for a higher close on a POMO day (54.59%) more ore less equals the benchmark probability (random or at-any-time chance) for a higher close, and furthermore the median trade on a POMO day (+0.05%) as well as the Distribution of Returns (the setup's median return in comparison to the benchmarks's ranking of daily returns) slightly undercuts the benchmark's median trade (+0.06%) and Distribution of Returns respectively.
A conclusion might be that even if the FED's liquidity injection is flowing into the stock market on POMO days, this may be limited to those occurrences only where opportunity is provided (lower prices at the end of the auction) - at least mitigating otherwise potential worse losses - , but not chasing the market and entering into highly priced positions after they already have (sometimes rapidly as if on last Friday) increased during the first part of the session.
Especially with respect to the fact that Friday, September 24 marked the 9th occurrence of a POMO day in the current month, the next step was to check for the markets's successive short- and intermediate-term performance where those POMO days accumulated. Table II shows the SPY's historical performance (since 8/24/2005) assumed one went long on close of any session where at least the following number of FED's Permanent Open Market Operations occurred during the recent rolling 20-day time frame (the respective last 20 sessions). Up to now the maximum number of Permanent Open Market Operations during any 20 consecutive business days has been 15:
- Strat. #1 'POMO == 0': none,
- Strat. #2 'POMO >= 3': 3(or more),
- Strat. #3 'POMO >= 6': 6(or more),
- Strat. #4 'POMO >= 9': 9(or more) .
The FED's Permanent Open Market Operations seem to have a significant short- and intermediate term impact on the stock market's performance during those time frames where POMO days accumulated.
Going long on close of any session where there hasn't been even a single Permanent Open Market Operation during the last 20 sessions would've significantly under-performed the benchmark with respect to Compound Returns (negative), (percentage of) Winning Trades, the Median Trade (especially considering a significantly above-average median losing trade), the Profit Factor and the Distribution of Returns., while going long on any session where there've been 3, 6 or 9 (the more the better) Permanent Open Market Operation during the last 20 sessions would've provided a significant edge.
But going long on any session where there've been at least 9 (or more) Permanent Open Market Operations during the last 20 sessions would've resulted in a significantly above-average profitability, with a median winning trade of +0.94% (benchmark: +0.57%) almost doubling the median losing trade of -0.49% (benchmark: -0.65%), and a Distribution of Returns significantly above the benchmark's Distribution of Returns of almost 50%.
The benchmark's Distribution of Returns of almost 50% indicates that daily returns are relatively evenly distributed on both sides of the mean, typically - but not necessarily - implying a symmetric (normal) distribution, while the setup's Distribution of Returns of 56.99% indicates that the bulk of the values (including the median) lie to the right of the (benchmark's) mean. For demonstration purposes figure I below shows the respective 'profitability density function' for being long on any session where there've been at least 9 (or more) Permanent Open Market Operation during the last 20 sessions, while figure II shows the benchmark's (SPY) 'profitability density function' for the time frame from 8/24/2005 to present.
Figure I
Figure II
It seems as if 'someone' always absorbs (into the close) otherwise potential severe losses during those time frames where Permanent Open Market Operations heavily accumulated.
Table III below now shows the SPY's respective periodic returns (broken down into weekly, monthly, quarterly, semiannual and annual time frames) and a couple of performance metrics (since 8/24/2005):
Even more amazing are the respective periodic returns. Going long on close of any session where there hasn't been even a single Permanent Open Market Operation during the last 20 sessions would've resulted in a below-average probability of a positive periodic return and a negative median return during any of those periods (from weekly to annual), while going long on any session where there've been at least 9 (or more) Permanent Open Market Operations during the last 20 sessions would not only have resulted in a significantly above-average probability of a positive periodic return, but in a significantly above-average profitability as well (e.g. a median weekly return of +1.42%, and a median quarterly return of +14.08%).
And last but not least Table IV below shows the SPY's probability for at least one higher | lower close over the course of the then following x sessions (with x going from 1 - the next session's close- up to 63, approximately 3 month later), and the SPY's probability for trading higher (posting a higher close) exactly x sessions later (since 8/24/2005) as well:
Whenever there hasn't been even a single Permanent Open Market Operation during the last 20 sessions in the past, this has resulted in an at least slightly below-average probability for at least one higher close over the course of the then following x sessions and a slightly above-average probability for at least one lower close over the course of the then following x sessions, but the market was trading at a higher level 2 and 3 month later on only 1 out of every 3 occurrences (or 36.50% of the time), significantly below the benchmark's probability (58.67% of the time) and random chance for trading at a higher level 2 and 3 month later.
But whenever there've been at least 9 (or more) Permanent Open Market Operations during the last 20 sessions, chances for at least one higher close over the course of the then following x sessions are (partly significantly) exceeding the respective benchmark's probabilities (and chances for at least one lower close over the course of the then following x sessions are (partly significantly) undershooting the respective benchmark's probabilities), and up to now the SPY posted at least one higher close during the then following 10 sessions on every out of those 144 potential occurrences (trades).
And even more amazing: Up to now (144 potential occurrences/trades), the SPY was always trading at a higher level 3 month later (latest), but on 9 out of every 10 occurrences (or 89.63% of the time) already 1 month later, significantly better than the respective benchmark's probability (or random chance) for trading at a higher level 1 (60.13% of the time) and 3 (58.67% of the time) month later.
Conclusions:
Although with respect to those session where the FED's Permanent Open Market Operation occurred no significant end-of-day edge will be provided, FED's Permanent Open Market Operations accumulated ( e.g. at least 9 during the last 20 sessions) historically had remarkable and statistically significant positive implications with respect to the market's short- (e.g. at least one higher close over the course of the then following 10 sessions) and intermediate term performance looking 1, 2 and 3 month ahead (trading higher 3 month later on all of those 144 potential occurrences/trades).
This could very well be a reason for the market's above-average month-to-date performance despite a couple of (negative) seasonalities (e.g. the historical negative week immediately following September's triple witching) and setups (e.g. the down-day immediately following the Labor Day exchange holiday).
This signal had been triggered on close of Friday, September 14 and will be triggered during any of the following sessions until the running count of the FED's Permanent Open Market Operations accumulated during the then recent last 20 sessions closes below 9. With the FED's Tentative Outright Treasury Operations Schedule showing 2 additional POMO days at the end of September and at the beginning of October, the running count will not fall below 9 for the foreseeable future.
Successful trading,
