Mar

30

The hypothesis is that at the end of a quarter in which bonds are up while stocks are down, institutions need to rebalance their asset allocations by selling bonds and buying stocks.

I found 14 such quarters since 2002, not including the current quarter. In the last 5 trading days of those 14 quarters, SPY was up 8 times and down 6 times, with an average net change of 0.9% with a t score of 0.76–statistically insignificant.

My Python code that I used to obtain the above results.

Big Al responds:

That's an event I hadn't thought about in a long time. It's hard to imagine a lot of big institutions running a simple strategy like that these days, which doubt your study would appear to support. But it does make me wonder if there are other, more complex balances or relationships that big players do manage on a calendar basis.

Alex Castaldo comments:

The general idea of trying to take advantage of "fixed behavior" by others is a good one IMO.

Paolo Pezzutti agrees:

It's like finding regularities end of month or Holiday's behavior or several others. I think there may be many still uncovered. Steve on Github has made public a number of Python notebooks. Very nice work to stimulate curiosity in searching patterns. It's not rocket science based on Artificial Intelligence, but I think this methodology has still value.

Asindu Drileba writes:

The rebalancing edge is real. In BTC for example, I realized that the most consistently active, "high activity" period is the time around 0:00 UTC (Server time). Something interesting is always happening during that period.

It turns out alot of people trade BTC daily and it just makes sense to rebalance the position size at midnight. I too even choose it sub-consciously. I don't think many people are choosing 03:00 UTC , 17: 43 UTC etc. Unfortunately, you need second by second, price quotes over many days, weeks, months and years to investigate this activity further. So I put it on pause. But the "activity" still exists.

M. Humbert adds:

Window dressing at quarters end is probably still occurring as well.

William Huggins writes:

several years ago i followed in Markman's steps of investigating the S&P500 drops and additions for irregularities (they did exist but have since been arb'd out). the driving mechanism was that index fund managers were paid to minimize tracking errors, not maximize performance so they would all trade at the same time, causing a secondary effect on the day the change actually took place (there was a preliminary change the day of announcement). it was a pretty basic academic event study but the most valuable part was uncovering "why" big money was doing a thing that created opportunities for fast moving traders (email me if you want it, but the trade doesn't work anymore)


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search