Using Bond Market Data To Trade Stock Index Futures and ETFs

One of the most useful concepts when constructing trading strategies is the idea that key markets are related and interconnected.  This concept is particularly applicable to the relationship between stocks and bonds.

The stock and bond relationship – Basic concept review:

On one side, asset managers see stocks and bonds as their primary asset allocation options.  Indeed, many investment programs (Pension funds, mutual funds, and structured ETFs) are based upon the idea that the investor should maintain a constant percent allocation to each of these asset classes.

For example:  One of the most common asset allocation schemes is the, “60% stocks, 40% bonds” allocation.  Some “lifecycle” funds attempt to provide a “one stop choice” for retirement planning.  They operate by gradually increasing the allocation to bonds as the “retirement date” approaches.  Such funds might start at 80% stocks, 20% equities, and end up with 20% equities, 80% bonds at the retirement date.  On the other side, bond prices (and hence interest rates) directly impact the rate at which corporations are able to borrow.  All else equal, lower interest rates decrease borrowing costs, which increases profits and corporate valuations.

Large moves in the relative price of stocks and bonds trigger a whole host of necessary re-balancing activity at asset management firms.  This is due to the need to maintain a desired asset allocation, and also because the relative desirability of each investment class changes as prices change.   Large, persistent changes in the relationship also alter the fundamental conditions of corporate valuation.  This is true both at the macro level, and also at the individual firm level (For example, the acquiring firm with the lowest cost of capital can afford to pay the highest rational price).

Trading application:

With these fundamental concepts covered,  let’s think about how this relationship might be helpful in simple trading terms.   In order to have a predictive relationship, we know we can only use data that is available at the time the trading decision is made (we can’t use today’s change in bond prices to predict today’s change in stock prices).   Lets start by simply evaluating how yesterday’s (open to close) relationship in bonds might effect today’s return in the S&P 500 stock index futures.   For those of you who are unfamiliar with S&P 500 (Emini) futures, 1 point is equal to $50 per contract.

First, lets find the open-to-close return for all days in our study period.  This can be used as a benchmark when evaluating our bond ideas.  Here it is:

sp500.all.days

Next, lets look at the S&P 500 futures return after bonds were up yesterday vs. when they were down yesterday:

sp500.if.bonds.upsp500.if.bonds.down

What we find is that over our evaluation period, the current day’s S&P500 return increases (relative to all days)  if bonds went up yesterday, and decreases (and in fact turns negative) if bonds were down yesterday.

Our understanding is that a larger move in bonds might require a larger need for re-balancing.  We also know that a larger move up in bond prices means lower interest rates, which in turn impacts corporate values.  So lets add magnitude to our study.  How do the results change if we only look at the S&P500 return when bonds were up or down greater than one full point?

sp500.up.plus.1.yesterday.second.trysp500.if.bonds.down.over.1.point

What we see is that as the magnitude of yesterday’s bond price change has increased, the expected impact on today’s S&P500 change has also increased.  The average day-session gain after bond prices were up 1+ points yesterday is now greater than 6  times larger than the benchmark return.  Lets add a factor from the price behavior of the S&P 500 futures contract itself.  What if the S&P 500 was down yesterday, and bond prices were up by greater than one point?  here are the results:

sp500.bonds.up.plus.one.stock.down.2

If you have read my other S&P 500 articles (or have studied the market yourself), you know that on average the expected return for the current day has historically gone up after a down day, and down after an up day.  As a result, the results of adding a “down day” rule to this study are not completely surprising (During our test period, buying after a down day boosted today’s return from .16 to about 1 point).   With the two ideas working in conjunction, the total profit is now  372 points, or $18,600 per contract and $137 per contract per trade

Lets do one more evaluation.  We know that after a lower close in the S&P500 futures yesterday, the average return is on average higher today.  We also know that when stocks close lower and bonds are up by over one point, the average return today has increased higher still.  What has happened when stocks closed down yesterday, and bonds closed down by greater than one point?   Lets benchmark this idea against the average return after a down day, rather than against all days.

stocks.down.bonds.down.2

I find these results to be extremely interesting.  Out of a sample with an above average return (buying after a down day) the “bonds down +1, stocks down” idea has pulled out a significantly negative return.   For the statistically minded, the T-Test statistic for the above was .03.  At the very least, the above suggests that tracking bond price moves might be very handy for those who are looking to buy dips in stock index futures and ETFs.

The stock-bond relationship is a fundamental economic concept that can yield many  potentially profitable insights.  These tests are  just the tip of the iceberg in terms of potential patterns.  Keep in mind, though, that it is always important to update studies and observations regularly.  Basic conditions (and patterns) change over time.   If you have read my other trading strategy articles, you should have more than a few ideas about how these results can be improved upon and extended.

If you have not already done so, join my mailing list here.

Good trading.

 

 

 

 

 

Leave A Reply (9 comments so far)

  • http://www.pickmonitor.com/z01vc evo34

    What index do you use to measure bond performance? Thanks.

  • http://twitter.com/NASTrading NAS Trading

    I am using 30 year bond futures (day session) data. Contract symbol is ZB. Thanks for the question.

  • Yaba Qi

    Could you define your evaluation period? Thank you.

  • http://twitter.com/NASTrading NAS Trading

    Start of 2008 to present. I will add that into article text, thanks

  • mhp

    I think the caption in your table with the 2.76 average should read “bonds up > 1 point yesterday and stocks down”, correct? I wonder how much of this effect is really due to bonds, vs. simply what the stock futures themselves did yesterday. For instance, when bonds are up > 1, it generally means stocks were down substantially, so mean-reversion the next day is more likely. Could a similar result be seen by simply requiring stocks down more than a certain threshold?

  • http://twitter.com/NASTrading NAS Trading

    First, thank you for the correction, I think i have the table labeled appropriately now.

    Second, great question, I was hoping someone would ask this. What I have found is that when you overlap different types of signals (stock and bond signals) you find some very counter-intuitive results if bonds were simply reflecting the inverse of the prior day stock price or had no predictive value – yet logically consistent with the notion it is a partially-independent predictive factor. It is not fully independent because the markets are related and intertwined, as described in the article. The last study (bonds down, stocks down) started to show this, I think. It is partly a matter of not wanting to give away too much in one article, and partly that it is a judgement call.

  • Luis

    Thanks for the analysis, it’s truly interesting. My question: are you always talking of open to close returns in all cases?…then you are not taking into account the yesterday’s close to today’s open gaps, am I right?…

  • nastrading

    I look at both things, there are not fixed rules.

  • Sam

    Hi, are you de-trending data?