Research house of the year: UBS

Risk Awards 2020: Swiss bank is pushing its analysts to be more like scientists

Daniel Dowd

In forcing banks to unbundle their research offering from their trading business, Mifid II has sparked still-unanswered questions on both the sell side and the buy side, about what consumers want from standalone research and how much they are willing to pay.

The UBS thesis – an appropriate word – is that sell-side research departments will become more like scientific laboratories, testing propositions with the help of proprietary data. A second bet is that some investors will pay for access to that data.

This year, the bank rolled out a research protocol that draws on practices in social science to guide its increasingly empirical studies into what’s driving stocks. It separated Evidence Lab, the bank’s data-gathering unit set up in 2014, from the research division to create an independent entity able to sell data directly to clients. And it delivered reports markedly different from those of competitors: from the narrow – on cobalt stockpiles in Africa, as an example – to the broad – on the future of commercial space travel.

Daniel Dowd, UBS’s global head of investment bank research, describes the division’s approach as “the scientific method applied in business”.

It’s an approach other banks have shied away from, because of the degree of change and investment it implies.

So, how is the thesis holding up so far?

It’s early days, and Dowd is a little cagey, but in a tough environment for sell-side research generally, the bank has dropped none of its coverage nor reduced or juniorised research staff. Revenues have been “in line” with what the bank expected, he adds.

“We have published research informed by primary data on a scale that others have not. We navigated the separation of Evidence Lab from research so it could sell data directly to clients. We have made aggressive efforts to help our analysts deepen their knowledge of research design and the kinds of data sources Evidence Lab collects, and we continue to train our analysts on how to use new types of data,” says Dowd.

Around a thousand active users have trialled the Lab’s subscription service, which allows access to raw datasets ranging from results of proprietary surveys to satellite images. “Very few” have dropped out of the pilot programme and some clients are now paying for the service, Dowd says.

For some investors, at least, the approach makes sense.

“Most investors or sell-side analysts focus too much on quarter-in, quarter-out noise rather than really thinking about what’s happening in an industry, so you might get the price action for three days right, but you get the price action for 12 months totally wrong,” says Ken Xu, founder and chief investment officer of BosValen Asset Management in Hong Kong, and a former portfolio manager with Point 72.

Thesis maps

Dowd took on his role in January from Juan-Luis Perez, who was elevated to group head of research and data analytics as part of a wider push to maximise the use of data across the bank.

“We are committed to hypothesis-driven empirical research,” Dowd says. “I mean that in a technical sense: formulating falsifiable hypotheses. We want to go out and collect data in the real world to adjudicate those hypotheses rather than passively interpret information that comes to us.”

Much of his work has gone towards instilling a scientific culture in the division. The bank wants as far as possible to eliminate the possibility that industry experts “pontificate” in a way that is not “rigorously evidence-based”, Dowd says.

Across all longer-form equity reports, and increasingly for other asset classes too, UBS has implemented a research protocol that requires analysts to set out a “thesis map” detailing the “pivotal questions” that could move the price of the stock concerned. The aim is to focus research on “unknowns” and systematically to gather evidence to address them.

We are committed to hypothesis-driven empirical research

Daniel Dowd, UBS

In April, the bank hired Amar Hamoudi from JP Morgan Chase, where he worked in consumer research, to advise further on how best to frame these empirical studies. Previously a professor at Duke University, Hamoudi is a leading expert in research design.

UBS has set up half-day workshops in which small groups of analysts work with Hamoudi to turn their questions into hypotheses and to develop a data collection and analysis plan to test them. The programme launched in the autumn with 40 analysts starting so far. The bank is airing a series of videos internally to make Hamoudi’s training more widely available.

The bank has also assembled a 13-person product management team headed by Jose Saiz, global head of product management, to work with analysts on the more important pieces they are writing, Dowd says, particularly on how to make questions more testable and to ensure the highest priority data requests are funnelled to Evidence Lab first.

This year, the bank expects to publish up to 4,000 reports backed by Evidence Lab data. In the past 12 months, those reports have attracted about a third of the total readership for research reports.

Making sense of data

The bank’s approach reflects how client needs have changed, Dowd says. “Research used to be about evaluating and interpreting public information. That’s still important. Clients care about what our analysts think of quarterly releases and how the year is evolving. Now they also expect us to bring them data – data that is not easily available elsewhere, data that we’ve collected ourselves.”

But specialist data collectors are often unsure what data to collect, and research analysts at banks are sometimes uncertain how to get hold of it, Dowd explains. UBS figured that a team like Evidence Lab’s combining both sets of skills would be valuable, not least because the data gathered would become more valuable as the new unit gathered more of it.

The abundance of data means clients need help to make sense of it, even in cases where the findings are not new. Proving that a consensus is correct can save time for buy-siders for whom time is precious, Dowd points out. “None of our clients wants to be fire-hosed with even more data today than yesterday,” he says.

The firm put its data capabilities to work last year in a series of reports on how cobalt supplies might affect battery production for electric cars – finding that concerns about possible shortages are unfounded.

UBS used satellite imagery to track activity at mines in the Democratic Republic of Congo, the world’s biggest producer of cobalt. The team also identified privately owned discarded cobalt from historical copper mining in the region that adds up to a supply surplus that will last six years at least.

Now clients also expect us to bring them data – data that is not easily available elsewhere, data that we’ve collected ourselves

Daniel Dowd, UBS

Even the realms of science fiction can be subjected to fact-finding, UBS has shown. In another report, the bank drew on a survey of business travellers in addition to other data to assess the growth potential of the space travel industry.

“Our client base is heterogeneous. Some want a short-term perspective. Many want a long-term view. Most want both. These kinds of pieces provide clients with a roadmap for how change could play out and what to look for,” Dowd explains.

Xu points to UBS research on the solar power industry as another highlight. That research found the sector was at a “global cost inflection point”, with solar electricity generation becoming cheaper than fossil fuel energy, he says.

The Evidence Lab separation has also been helpful in driving demand for published research, Dowd says: “When clients start using data from Evidence Lab, they typically want to talk more to our publishing analysts.” The publishing analysts often have years of experience working with the data, he points out. “One of the most valuable things a buy-side client can do is phone an analyst and get their perspective. It leads to more conversations, not fewer.”

As for the future, Dowd sees the balance of skills required of analysts changing in favour of those most comfortable handling data. Research at UBS today is a “relentlessly evidence-based” process, he says.

Going forward, expect UBS to hire more technicians than talking heads.