| January 20, 2015
This study is one of the first looking into trading sovereign debt instruments with sentiment analytics from news and social media.
This study has been conducted by Saeed Amen, Quant Strategist at The Thalesians and former Executive Director at Nomura.
The author uses the RavenPack Analytics Global Macro data to create news-based economic indices (NBESI) for the U.S., E.Z, U.K. and Japan which he then tests against sovereign
prices and spreads. He also tests the indices against common FX strategies.
Specifically, Saeed finds that, since 2001:
Figure 1 and Figure 2 show the return profiles for NBESI-based models on G4
and US Treasury spreads, respectively, since 2001.
Saeed also used the indices on a combined filtered G10 FX carry and G10 FX NBESI basket which had risk adjusted returns of 1.11 and a max drawdown of 6.7%.
Before creating any sort of trading rule based on news data, we need to understand the relationship between markets and economic sentiment. It seems relatively intuitive that there should be a relationship between economic data and the price action in bonds. As economic data improves, we would expect central banks to adopt a more hawkish tone to keep inflation in check, which would be accompanied by rising yields as the market prices this in.
By contrast as economic data gets worse, we might expect central banks to become more dovish, which would be reflected in lower bond yields. There is the obvious caveat, that there can be periods where high levels of inflation can occur during periods of poor growth, which is called stagflation.
Can data confirm our hypothesis? We can take a look at economic surprise indices to help answer this question. Economic surprise indices are popular in the market. Many banks produce their own versions including Nomura, where I created their growth surprise indices. These measure the difference between actual data and economist expectations. Hence, we can use them as indicators of economic sentiment. Creating such indicators can be non-trivial from a data collection perspective.
In Figure 3, we plot Citi’s US economic surprise index, which is the most well-known of the various surprise indices, against 3 month changes in UST 10Y yields. We find, at least on a stylized basis, there is a strong positive correlation between changes in bond yields and changes in economic surprises. We note that broadly speaking economic sentiment data has mean-reverting properties. This seems quite intuitive, if we consider how the marke interprets economic data.
As data improves, the market updates its expectations higher. Eventually, the expectations become so elevated that data starts to miss expectations. We then see a peak in market sentiment with respect to economic data, which coincides with the medium term peak in yields. At this point economic sentiment begins to mean-revert, as do yields. We see a similar process in reverse, when economic sentiment keeps worsening and it creates a trough, which coincidences with the local low in yields.
In Figure 4, we look at the relationship in a more systematic manner, conducting a regression between daily changes in UST futures and Citi’s US economic surprise index. We report T stats, which are statistically significant. We note obviously, that the sign is negative, because bond futures have an inverse relationship with bond yields. As we might expect, S&P500 has a positive correlation with US economic surprises, whilst EUR/USD has a negative correlation (the rationale is that worse data results in lower UST yields which tends to be bearish USD, thus pushing EUR/USD higher).
The idea behind creating news based economic sentiment indices, is that they will have a much richer dataset than economic data surprise indicators. Later, we shall discuss how we can use the relationship between economic sentiment and yields to enable us to create trading strategies to trade bond futures, when using news based economic sentiment indices.
On a broad basis, there are two ways we can trade economic sentiment indices, one using a momentum based approach, which takes advantage of the fact that assets are correlated with economic sentiment. We can also take a longer term approach, fading economic sentiment, given that over the longer term, sentiment is mean-reverting and it tends to be bounded.
So far we have only looked at UST futures. However, what is the relationship between USTs and other G4 bonds? In Figure 5, we plot the returns for UST 10Y, Bunds, long Gilts and JPN 10Y bond futures. We have adjusted for the differences in volatility. We see that there does appear to be a strong relationship between the various bond futures. In Figure 6, we compute weekly correlations between these various bond futures from 2001-present. We find that there are generally quite high correlations. We shall later use the highly correlated nature of G4 sovereign bond markets to enable us to use both US based and local news indicators. The rationale behind using US based news indicators, is that the US is likely to be a major driver for other bond markets.
Please use your business email. If you don't have one, please email us at email@example.com.
We will process your personal data with the purpose of managing your personal account on
RavenPack and offering our services. You can exercise your rights of access, rectification,
erasure, restriction of processing, data portability and objection by emailing us at firstname.lastname@example.org. For more information, you can
Your request has been recorded and a team member will be in touch soon.
High inflation has returned in developed markets after decades of lying low. In our latest paper, we show how to build an inflation-based asset allocation strategy using sentiment data and we illustrate that sentiment-based strategies outperform models that depend merely on past observed inflation values.
This year's RavenPack Research Symposium brought two intense days of knowledge sharing in London and New York, from 25 top experts in natural language processing, quantitative investing and machine learning. Together, we explored how firms can leverage new language models to generate alpha, better manage risk and respond to calls for more sustainable investment practices.
Human capital is at the heart of value creation. Our latest research demonstrates how unprecedented workforce insights, sourced from over 200 million job postings, can generate more alpha.