Finding Portable Alpha in Fixed Income

My colleagues and I have been doing some interesting work in developing non-credit based fixed income overlays.

By way of introduction, portable alpha received lots of negative press in 2008. The primary reason for this was the after -the -fact discovery that many of the alphas being ported were, in fact, betas. These betas were wrongly characterized alphas as a result of measurement against inappropriate benchmarks. And although many of these betas were uncorrelated to the portfolios they were being ported on to, the liquidity meltdown of 2008 drove correlations of seemingly independent betas to 1. This conditional correlation had the end result of leaving allocated portfolios with leveraged beta exposure, probably the last thing managers needed to tell their investment committees at that time.

But don't shoot the delivery mechanism. Done properly, portable alpha as a useful tool in both risk mitigation and return enhancement, as we will demonstrate. 

 

We set out generate a strategy that reduced the Barclay's Intermediate Term Bond Index's overall volatility while increasing total return. Additionally, our goal was to  do this without adding corporate credit risk.  


To achieve a meaningful result, strategy generation must be robust with respect to  survivor bias, rule generation, and implementation. All elements of any strategy must be formulated out of sample and all rules must be achievable in practice. Additionally, we propose a different way of looking at the portable alpha problem. Our main concern is the overlay's utility as a condition risk mitigation tool. Rather than simply juicing the strategy with normally uncorrelated returns, we seek to mitigate benchmark loss and improve the overall risk profile of the investment. We believe that this results in a better definition of alpha for our purposes.

The resultant strategy (abbreviated AO), partitions the yield curve into multiple maturity sectors and runs multiple behavioral models in each. The models have zero correlation between them and thus extract quite different patterns from the market. Holding periods range from about 5 weeks down to  well under 1 week. With no overlap, the models find opportunities in all market conditions.

For purposes of risk analysis, we have highlighted two  exogenous event types that  both asset and liability  driven investors should find of particular interest. Months in which significant Capital Markets dislocations occurred are highlighted in yellow. Months in which significant Insurance Loss events occurred are highlighted in red. We labeled 9/11  orange as it represents both. Event details and statistics are noted below. 

The following  table illustrates AO’s monthly returns to Capital Market dislocations. The conditional  returns are strongly positive and statistically significant, whilst remaining unaffected by large Insurance Loss events.

	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec	TOT
1997		0.03%	1.18%	0.21%	0.44%	0.25%	0.97%	0.44%	0.38%	0.02%	0.46%	-0.01%	4.38%
1998	0.32%	0.62%	-0.07%	0.02%	0.62%	0.76%	0.43%	0.71%	1.46%	1.36%	0.32%	0.34%	6.90%
1999	0.78%	0.56%	-0.53%	0.48%	1.24%	0.49%	0.05%	0.27%	0.24%	-0.18%	-0.19%	0.66%	3.86%
2000	0.91%	-0.21%	1.16%	-0.20%	-0.20%	0.38%	0.40%	0.66%	0.54%	0.21%	0.53%	1.29%	5.45%
2001	1.49%	0.95%	1.01%	0.43%	0.51%	0.32%	0.14%	1.13%	1.36%	1.20%	-1.92%	1.10%	7.73%
2002	-0.28%	0.66%	0.87%	-0.36%	-0.17%	-0.04%	2.25%	0.43%	0.49%	-0.13%	0.05%	-0.13%	3.65%
2003	-0.24%	0.52%	-0.74%	-0.18%	0.21%	0.84%	2.89%	0.55%	0.32%	0.19%	-0.54%	0.46%	4.28%
2004	-0.18%	0.37%	-0.73%	1.23%	0.23%	0.95%	0.73%	0.37%	-0.31%	-0.60%	0.19%	0.06%	2.31%
2005	0.40%	0.83%	0.20%	1.02%	0.26%	0.24%	0.37%	0.08%	-0.11%	0.96%	-0.24%	0.77%	4.76%
2006	-0.06%	0.22%	-0.22%	0.96%	0.32%	-0.23%	0.41%	0.58%	0.64%	0.26%	0.75%	0.64%	4.28%
2007	0.26%	0.93%	0.58%	0.05%	0.33%	0.18%	0.54%	0.38%	0.33%	-0.03%	1.78%	0.59%	5.92%
2008	1.15%	0.24%	0.92%	0.03%	1.20%	1.19%	-0.29%	0.05%	1.25%	0.33%	2.88%	1.86%	10.82%
2009	1.25%	0.63%	0.69%	0.07%	0.24%	0.12%	-0.01%	-0.46%	0.04%	-0.09%	-0.52%	-0.15%	1.81%
*All returns are Net of Fees

*All returns are Net of Managment and Performance Fees. All returns net of financing and transaction fees

The following tables illustrate AO's non-correlated  returns to the Barclays Aggregate and Government  bond benchmarks while significantly outperforming during Stressful Market Events.

Capital Markets Events   All Returns are Gross for Purposes of  Comparison
	Date	AO  Return	2YR  Gov Index*	Intermediate  Agg. **
Russian Financial Crisis / LTCM	Sep-98	1.46%	0.38%	0.41%
Dot Com Bubble Burst	Jan-01	1.49%	1.18%	1.61%
9/11 Terrorist Attacks	Sep-01	1.36%	2.07%	1.50%
Enron Bankruptcy	Nov-01	1.10%	-0.48%	-0.48%
US ISM Survey Tightening Fears	Mar-02	0.87%	-1.53%	-1.33%
WorldCom  Bankruptcy	Jul-02	2.25%	2.00%	1.20%
Rate Hike Fears	Jul-03	2.89%	-2.19%	-2.41%
Fed Rate Tightening Fears	Apr-04	1.23%	-2.26%	-2.15%
Bear Sterns Collapse	Mar-08	0.92%	0.65%	0.39%
Lehman Bankruptcy	Sep-08	1.25%	0.67%	-0.97%
Bank Liquidity Fears	Oct-08	0.33%	0.67%	-1.79%

Event Risk Summary All Returns are Gross for Purposes of  Comparison
Average Returns for	AO	2YR Gov Index*	Intermediate Agg**
Capital Markets Events	1.37%	0.10%	-0.37%
All Other  Months	0.35%	0.49%	0.56%

The AO active management of Government Securities creates a risk and return benefit that cannot be reproduced by adding a passive Treasury Benchmark, as evidenced  by the correlation comparison below.

Correlations Comparison
AO			2YR Gov Bond Index*
	To 2YR Gov Index*	To Barclays Intermediate**	To Barclays Intermediate**
For Capital Markets Events	-12.63%	5.18%	86.68%
For Non Event Months	32.95%	30.40%	88.36%
* 2 Year Barclays Govt Bond Index **Benchmark is the Barclays U.S. Aggregate Bond Total Return Index (BBG Code LC08TRUU) Note: For analysis Sep-01 data removed from insurance correlation as it is in Capital Markets Correlation Data. Francis and Charlie data used once in correlation

The following tables highlight  AO's utility.

An overlay would have added value to the Barclays Intermediate Bond Index bench mark in every year since 1997, regardless of Interest Rate environment, or market conditions. 
On a monthly basis, the AO overlay outperformed the benchmark only return 92% of the time, added no benefit 6% of time, and underperformed 2% of the time. There was no serial auto-correlation to the under performance.

AO 130/30 (AP)   vs.  Barclays Intermediate 1997 – 2009 (all Returns in percent)															Added		AP  vs. Index
	AP	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec	Annual Net	Value In %	Count	Monthly Win Loss
1997	AP	0.00	0.25	-0.44	1.39	1.01	1.08	2.25	-0.28	1.31	1.11	0.39	0.84	8.82	1.15	9	Outperforms
	Barclays	0.00	0.24	-0.78	1.33	0.88	1.01	1.97	-0.40	1.20	1.11	0.26	0.84	7.67		0	Underperforms
1998	AP	1.28	0.21	0.34	0.53	0.89	0.80	0.54	1.52	2.41	0.27	0.29	0.51	9.40	1.79	10	Outperforms
	Barclays	1.19	0.03	0.36	0.53	0.71	0.58	0.41	1.31	1.99	-0.12	0.20	0.41	7.60		1	Underperforms
1999	AP	0.84	-0.88	0.56	0.51	-0.33	0.04	-0.34	0.12	1.31	0.34	0.05	-0.11	2.01	1.01	9	Outperforms
	Barclays	0.62	-1.04	0.72	0.37	-0.68	-0.10	-0.36	0.04	1.24	0.39	0.10	-0.30	1.00		3	Underperforms
2000	AP	-0.32	0.90	1.40	-0.15	0.04	2.05	0.83	1.52	1.13	0.64	1.58	2.13	11.60	1.42	9	Outperforms
	Barclays	-0.58	0.96	1.07	-0.10	0.10	1.94	0.71	1.33	0.97	0.58	1.43	1.76	10.18		3	Underperforms
2001	AP	2.04	1.06	0.97	0.04	0.75	0.39	2.00	1.29	1.89	1.89	-1.54	-0.17	10.40	2.01	11	Outperforms
	Barclays	1.61	0.79	0.67	-0.08	0.60	0.30	1.96	0.97	1.50	1.54	-0.99	-0.48	8.39		1	Underperforms
2002	AP	0.63	1.15	-1.08	1.69	0.83	0.86	1.84	1.32	1.48	-0.10	-0.10	1.68	10.10	0.95	6	Outperforms
	Barclays	0.71	0.96	-1.33	1.79	0.88	0.88	1.20	1.20	1.34	-0.06	-0.12	1.71	9.14		6	Underperforms
2003	AP	0.03	1.26	-0.15	0.57	1.30	0.26	-1.58	0.59	2.30	-0.64	0.01	1.09	4.94	1.12	8	Outperforms
	Barclays	0.10	1.11	0.06	0.62	1.24	0.02	-2.41	0.43	2.21	-0.70	0.17	0.96	3.82		4	Underperforms
2004	AP	0.60	1.06	0.44	-1.79	-0.28	0.80	1.08	1.74	0.08	0.56	-0.59	0.71	4.34	0.61	8	Outperforms
	Barclays	0.65	0.95	0.65	-2.15	-0.35	0.53	0.87	1.64	0.17	0.73	-0.64	0.69	3.73		4	Underperforms
2005	AP	0.46	-0.28	-0.33	1.43	0.93	0.47	-0.61	1.09	-0.77	-0.34	0.32	1.00	3.26	1.24	10	Outperforms
	Barclays	0.34	-0.52	-0.38	1.14	0.86	0.40	-0.71	1.07	-0.74	-0.61	0.39	0.78	2.02		2	Underperforms
2006	AP	0.11	0.30	-0.67	0.31	0.01	0.10	1.38	1.51	0.95	0.69	1.24	-0.18	5.62	1.12	9	Outperforms
	Barclays	0.12	0.23	-0.61	0.04	-0.08	0.17	1.26	1.35	0.76	0.61	1.02	-0.37	4.51		3	Underperforms
2007	AP	0.10	1.60	0.37	0.49	-0.54	-0.14	0.95	1.29	0.86	0.80	2.25	0.49	8.37	1.54	11	Outperforms
	Barclays	0.03	1.33	0.21	0.48	-0.63	-0.19	0.79	1.18	0.76	0.81	1.74	0.32	6.83		1	Underperforms
2008	AP	2.09	0.26	0.65	-0.22	-0.24	0.19	-0.08	0.86	-0.61	-1.69	3.38	3.38	7.66	2.80	11	Outperforms
	Barclays	1.76	0.19	0.39	-0.23	-0.58	-0.15	0.00	0.85	-0.97	-1.79	2.55	2.84	4.86		1	Underperforms
2009	AP	0.14	-0.03	1.54	0.67	0.76	0.36	1.37	0.79	0.90	0.61	1.13	-1.41	6.64	0.32	7	Outperforms
	Barclays	-0.22	-0.21	1.35	0.65	0.69	0.33	1.37	0.92	0.89	0.63	1.28	-1.37	6.31		4	Underperforms
* All AP  Returns are Net of Fees.

 

             AO130/30 Overlay vs. Barclays Intermediate Aggregate Bond Index 1997-2009

KEY STATISTICS	Barclays Inter. Agg*	AO Net of Repo**	130/30 Overlay	AO Real Money***
Annualized Avg. Return	5.88%	5.12%	7.41%	7.25%
Annualized Std. Dev.	2.98%	2.18%	2.97%	2.90%
Information Ratio	1.97	2.35	2.39	2.50%
*Barclays Intermediate Aggregate Index includes Repo Return **AO Real Money defined as  AO with no Net Borrowing

The key driver to the attractiveness of AO as an overlay candidate is the strategy's low conditional correlation to exogenous events that tend to drive lowly correlated betas to high correlation. From an allocators prospective, these periods are precisely the times when low correlations need to occur.