Lazard Asset Management has developed a Covid-19 data model to more accurately reflect the health of corporates throughout the virus outbreak.

Over the course of the pandemic, updated company information has been difficult to access, and firms like Lazard AM have found it difficult to establish views of pricing and performance that accurately reflected reality, Paul Moghtader, a portfolio manager at the firm, tells WatersTechnology.

At the beginning of March, Lazard AM set out to resolve these data gaps and inconsistencies. To help with the unprecedented unavailability of data, the firm is now leveraging fundamental analysis, including updated public filings, company announcements, pattern valuations, and corporate information.

Whether it’s balance sheet data or it’s other proprietary data that we have that doesn’t get updated as frequently, those are the things that are really concerning us in this environment. We don’t want to be misled by rapid changes in price, with other factors trailing, and so that’s where the proprietary work around Covid really came into play.

Paul Moghtader

Within the first two weeks of March, the firm had over 100 of its fundamental analysts measure and rank 1,500 company business models based on how negatively or positively they would be impacted by the pandemic. The analysts used a ranking system based on a five-point scale, ranging from one point for “very negatively impacted,” three for “neutral,” and five for “positively impacted.”

Using data science techniques and granular, industry-level data, Lazard mapped those 1,500 business models to the 4,000-plus companies used by the asset manager to develop investment portfolios. The strategy involved collaboration across its global fundamental, quantitative, and data science teams. 

“We’ve combined fundamental insights and data science techniques to come up with a Covid-19 score for the companies in our universe, to try to determine how reliable the data that we have is going to be. Because, ultimately, if you have companies that are severely impacted by Covid-19, it’s quite possible that some of this data we’re looking at may be either unreliable or outdated,” Moghtader says.

The firm also uses natural language processing (NLP) and sentiment analysis from alternative data sources such as online content, news, and transcripts, to complement its Covid-19 investment strategy. In March, WatersTechnology spoke with Jai Jacob, managing director, and portfolio manager at Lazard, on how the firm has developed themed asset categories, and how it uses NLP and network theory to identify new associations between companies, looking at areas such as their supply chains, location, or mentions in the same press articles. 

In some cases, mapping the business models to individual companies was straightforward, Moghtader says. However, in cases such as conglomerates, which have multiple peripheral businesses, the firm had to come up with a revenue-weighted measure of each subsidiary to see how it was impacted by the pandemic. This measure was then used to adjust the firm’s Covid-19 score.

Lazard uses FactSet’s Entity Data Management tool to map the firms to their parent companies. The FactSet product works by using a single entity-level identifier to expose links and hierarchies between entities, securities, people, and funds.

Moghtader says problems emerge when fast-moving signals, such as price movements, fail to reflect or align with slow-moving signals, such company filings. However, the new Covid-19 model is designed to reduce the firm’s investment portfolio’s exposure to historical data that may be unreliable, and avoid potential “value traps.” Now, Lazard’s investment managers can compare existing online textual data with the newly developed Covid-19 data model to see if it matches their view of the virus’ impact.

“Whether it’s balance sheet data or it’s other proprietary data that we have that doesn’t get updated as frequently, those are the things that are really concerning us in this environment. We don’t want to be misled by rapid changes in price, with other factors trailing, and so that’s where the proprietary work around Covid really came into play,” Moghtader says.

Implementing the new Covid-19 data model was the easy part, says Moghtader. As the economic situation post-coronavirus outbreak continues to evolve, the model will have to be frequently adjusted in parallel.

“The next step is making sure that as new information comes into the market, the scores are updated by the fundamental side of the business,” he says.

And as corporates start to update their own information online, it will begin to match with the up-to-date pricing and market data, and Lazard will have to remove some of the controls on the model.

“And that’s going to depend on the speed at which [companies update information] in different markets, and the frequency of reporting on some of this data. All of that is going to play a role in how the adjustment is modified over time,” Moghtader says.

The Covid-19 model is expected to remain in use until fundamental data catches up with market movements and the reality of what is happening all over the world, which is important as countries and cities come out of lockdown in different stages, or reimpose lockdowns. 

“We’re really trying to avoid the asymmetry between data that is not fully reflective of the Covid environment and data that is. So as the models continue to update, and as new information comes in, that’s really where the judgment comes into play … and when these types of models can be unwound,” Moghtader says.