Analyzing intraday time series in various markets is a familiar strategy, and has usually been applied to markets which have liquidity 24 hours of the day, such as FX. A good recent example was Deutsche's report on "How To Make Money Trading FX? Just Wake Up At 3AM." And while such regional time-series comps have been mostly conducted with currencies, over the weekend Morgan Stanley's Matthew Hornbach did a similar analysis with rates.
What he found was startling.
As Hornbach explains, a key feature of his team's work over the years has been to what extent and when have investors in Japan placed downward pressure on Treasury yields. This led him to consider whether or not the effect occurred during the Tokyo trading day or outside of the Tokyo trading day. Morgan Stanley then created 4 indexes to track the changes in 10y rates during 4 sessions of the global trading day: (1) the Tokyo session, (2) the London session, (3) the NY morning session, and (4) the NY afternoon session.
The exchibit below shows the cumulative change in 10y swap rates during each session since November 2012, while the exhibit on the right combines the morning and afternoon NY sessions. To avoid calendar effects on yields, MS excluded any day that includes a holiday in one of the three regions. The bank also used swap rates due to the availability of historical intraday data on Bloomberg, however Hornbach urges those with access to a fuller set of data to use Treasury futures prices: the outcome should not be different.
And now his finding: what both charts suggest is that the entirety of the decline in 10y rates in 2015 and 2016 occurred during the Tokyo trading session. In fact, since October 8, 2014, 10y rates have fallen by 141bp cumulatively during the Tokyo trading session (Tokyo open until 2:00 AM ET). During the same period, 10y rates rose by 109bp cumulatively during the London session (2:00 AM ET to 8:00 AM ET), and rose by 37bp cumulatively during the NY session (8:00 AM to 5:00 PM ET).
Narrowing down the time series to just since the US presidential election, reveals that 10y rates have risen by 35bp during the London session, risen by 37bp during the NY afternoon session (1:00 PM ET to 5:00 PM ET), fallen by 18bp during the Tokyo session, and fallen by 1bp during the NY morning session (8:00 AM ET to 1:00 PM ET).
The implication of these findings, simplistic as they may be, is that the buying pressure on Treasurys is entirely the result of Japanese action, while the "rest of the world" has been a net seller. While the MS analysis does not definitively confirm that the buyer is an actual Japanese organization, it suggests that the buying entity does operate mostly during Japanese trading hours, for whatever reason.
But while it remains to be proven that the buyer is indeed Japanese, or operating out of Japan, it suggests that anyone wishing to trade bonds - at least as long as this pattern holds - should buy Treasuries during Japanese trading, and sell them mostly during London trading hours. Of course, now that the pattern has been exposed, it is unlikely that it will persist but it never hurts to try.