Monday, March 15, 2010

A look at bond market seasonality

Currently I have a neutral tactical call with a bullish tilt (see Temporarily back to neutral dated March 9) and a neutral strategic duration call. However, my view is that Q2 will be difficult for international bond markets amid a temporary re-acceleration in growth following a severe winter. Furthermore, bond market seasonality has its strongest negative effects in April, a topic I will explore on below.

When looking at bond market seasonality, several issues arise:
- seasonality studies need a long time history in order to make the results somewhat statistically meaningful (even if one uses 20 years of data, that still gives only 20 data points per month)
- benchmark bond yield histories face the drawback that the underlying benchmarks change over time which distorts the yield history
- bond yields can trend for decades but in a seasonality study one does not want such trends to influence the outcome (for example bond yields have been in a bear-trend over the past 25 years).

One time series which offers itself for a seasonality study are the constant maturity treasury yields (CMT) as provided by the St.Louis Federal Reserve. They go back to 1962. Firstly, this provides a longer than usual time series (49 years of data which, however, is still not that much). Secondly, given that they are CMT yields, they do not suffer from benchmark changes. Thirdly, in 1962 10y CMT were at approx. 4%, i.e. not far away from the current level of around 3.70%.

The chart below shows the average change in basispoints during the various months. In addition, it shows the net number of months during which yields rose in between 1962-2009. As can be seen, during this time period yields rose on average from January to May and fell thereafter until year end. Furthermore, this was not achieved by some outlier months with huge yield changes but rather by a significant larger number of months seeing yield rises than yield falls (from Jan-May) and vice versa especially in June and from September-December.
Strong seasonality in the US 10y CMT rate from 1962-2009
Source: St. Louis Federal Reserve, Research Ahead

To better highlight this seasonality, the chart below shows the p&l from being short bonds from Jan-May and from being long bonds from June-December. As can be expected from the average yield changes, both, the short and the long trade produce significant profits over this time period. What is more, while the long trade only really starts to perform from the early 1980s onwards (i.e. once the structural bond bull-market sets in), the short trade performed during both, the bear-market of the 1970s and more importantly the bull-market of the 80s/90s and 00s.
A bond-bearish move until May seems to be a feature of a bull as well as a bear market
Source: Research Ahead

In order to test for a different behaviour during the Great bull and bear markets, I divided the history in two sub-segments from 1962-1981 (when the yield high was reached) and from 1982 to 2009. The table below shows the average yield changes during both sub-periods. It is interesting to note, that during the bear market, yields on average rose almost throughout the year (except during November) with the largest yield rises in January, February and July. During the bull market, however, the dispersion of average yield changes was much wider (26bp vs. 13bp) and yields fell on average only during 6 months (June and August-December).

Average monthly change during the Great bear market and the Great bull market

Month

Jan62-Dec81

Jan82-Dec09

January

8

2

February

9

1

March

1

8

April

3

13

May

2

0

June

2

-5

July

9

0

August

5

-9

September

3

-11

October

6

-13

November

-4

-11

December

2

-4

Source: Research Ahead

Furthermore, while during the bear market, the average rise in yields in January and February was larger than during the bear market, the average rise in yields for March and April was much higher in the bull market than previously.
Now, if we do not know whether we are in a structural bull or bear market and the above pattern still holds, this would suggest that shorts are most preferable in March and especially April while longs seem to be advisable especially in November. But it also suggests that even if one assumes that the bond bull market has not ended, being invested early in the year does not pay on average and one should be more active in H2 of any given year.
While I do not advise to invest purely on historical seasonal patterns, it supports my stance that Q2 might be a difficult quarter for international bond markets.

3 comments:

  1. Really, this is great job!Also,this is one of the memorable things ever.Your graphical presentations about market seasonlity is awesome....Very thanks....


    markets

    ReplyDelete
  2. Great article! Thanks for sharing borsa.tv @Freda

    ReplyDelete
  3. I noticed this seasonality effect today and searched for this topic. Possible reasons: 1) Home buying season in spring tends to drive up rates, 2) "Sell in May" - some of that stock money goes into bonds, while the demand for home loans dries up and supply/demand tilts in favor of lower rates. Any other possible explanation?

    ReplyDelete