From Headlines to Alpha: News Sentiment in Credit

研报英文原文证据摘录

From Headlines to Alpha: News Sentiment in Credit

Barclays | QPS FICC

the rapid incorporation of information into stock prices. As a result, extracting incremental

predictive signals from news remains challenging in equity markets.1

Credit markets may offer a more favourable environment for sentiment-based strategies.

Corporate bonds typically trade less frequently than equities, depend more on dealer

intermediation, and often exhibit slower price discovery. Consequently, sentiment shocks may

be incorporated into bond prices more gradually, creating opportunities for return

predictability.

In this paper, we examine the role of news sentiment in corporate bond markets and its

applications in systematic credit investing. Using sentiment measures derived from news, we

evaluate their relationship with subsequent bond performance and assess whether sentiment

provides incremental information about future credit market outcomes.

News Sentiment Data

Our analysis uses news sentiment data from RavenPack, which transforms unstructured textual

information into a structured, machine-readable dataset. Generating sentiment signals from

news requires several processing stages. First, each article needs to be mapped to the entities it

references, such as companies, individuals, countries, or products. Because a single article may

mention multiple entities, RavenPack identifies the primary subjects of the story and links

corporate entities to unique identifiers. It also assigns an entity relevance score to distinguish

entities central to the article from those mentioned only incidentally.

Second, articles must be classified according to the events they describe. A news story may

cover one or more events, such as earnings announcements, analyst rating changes, mergers