Sell-side analysts are often highly-trained professionals who exploit data, sector expertise, complicated financial models and enviable company access to produce research suggesting how much a company should earn and what its shares should be worth.
Investors, recipients of said research, are also often highly-trained professionals who exploit data, sector expertise, complicated financial models, enviable company access and, yes, sellside research to decide when a company’s shares are worth having.
This is a notionally elaborate relationship — at the quantitative meeting point between conception and action — but numbers are scary for some people, so sellside analysts also stick labels on their notes that say things like “BUY”, “HOLD” or “SELL”. Company executives love them as well, for example.
Citi:
In this piece we study a simple question for Asia markets: are investors able to consistently generate investment returns by following the most basic of broker recommendation — the stock rating. That is, should investors buy what most analysts rate as “buy” and avoid what they rate as “sell”? An analyst’s stock rating is the one-word summary of their view on a stock’s share price, encapsulating analysis from their company’s financial and operating model, industry model, sector dynamics, macro impact, management, competitive landscape, etc.
Finally: systematic investment for dimwits. We know that people are instinctively (and, to be fair, not incorrectly) sceptical of sellside ratings, but let’s shelve our cynicism for a few minutes and take this work at face value.
Chris Ma and his team at Citi have explored the power of these written words in a variety of ways — starting, obviously, with testing a portfolio hinging entirely on ratings.
Ignoring hold/neutral ratings, they ran through companies in the MSCI All-Country excluding Japan index and made an alternative basket thusly:
We apply a weighting based on the numbers of ratings. For example, if stock A is covered by 10 brokers, with 7 buys and 3 sells, then we buy $7 of it and sell $3 of the same stock. This way, the portfolio weighting reflects how consensus an idea is. The portfolio is rebalanced monthly, and ratings are from monthly IBES consensus data.
The output is an index that would have cumulatively beaten its benchmark over most of the past two decades — until recently.
There is a consistently higher number of “buy” than “sell” ratings across the index. In part, this is what you’d expect from sellside analysts who need to keep management teams sweet. It presumably could also show an optimistic bias: if, during ‘normal times’, stocks tend to go up, it makes some sense that many analysts typically believe most of the stocks they cover should also tend to go up.
This has worked well in practice. Most of the outperformance appears to have come from the inherent value of running with the bulls during a bull run — and analysts’ apparent unwillingness to switch ratings in response to “short-term volatility”.
But potential problems are apparent early in the series — look at the inverse performance of the Buy-Sell index versus the benchmark six months after the financial crisis. The problem, writes Ma, is:
. . . the GFC came as a surprise to most observers. Most analysts had a hard time adjusting their ratings in such a short, volatile period, as the market continued its dramatic selloff.
It’s obviously hard to rebase by eye, but to us it looks like this approach is being quite heavily flattered by the first half of the past decade — an X-axis starting around 2015 (anyone remember any big Asian macro stuff that year?) might have produced quite a different outcome.
An apparent collective inability to deal with large-scale drama has been very apparent since the pandemic:
We have seen a shift in recent years; fast forward to 2021, when markets were hit by the slow recovery from COVID and geopolitical conflict on multiple fronts happening in a very short period. Unsurprisingly, the strategy suffered along with the overall market. However, it has continued to underperform up until now, even when markets have been leveling out. Most analysts have been making the wrong calls in the current macro-driven environment, which is a challenging environment for stock calls. While we do see signs that the macro environment is marginally improving, the macro influence on equity markets is still elevated above any other period during the last 20 years, including the GFC.
So far, perhaps so obvious: those highly-trained professionals who exploit data, sector expertise, complicated financial models and enviable company access to produce research suggesting how much a company should earn and what its shares should be worth are much effective when the factors driving those earnings are no longer idiosyncratic to the company.
In this context, the fairly large number of “holds” — sellsidespeak for “sell”, while actual sell ratings should be read as “GTFO” — troubles us. It seems large enough that it should somehow enter the equation.
Whatever. How else can simplistic ratings be used?
Citi tried a different approach — this time focusing on the change in ratings, not their absolute level — with seemingly superior results:
We extend the strategy to buying/selling stocks not by the absolute level of ratings, but by the change in ratings. For this strategy, take the following scenario as an example: a stock has 10 broker ratings, with 5 of them revised up from the previous month, 3 of them revised down, and 2 remaining unchanged. In this case, the long portfolio buys $5 of the stock and the short side sells $3 of it. Note that a rating being upgraded does not necessarily mean it is buy-rated, but could be upgraded to neutral from sell rated, and vice versa…
Not only is the overall return positive, but it also shows significantly less volatility than the market and even stayed positive during some of the worst market crashes. This suggests that there is strong empirical rationale to pay special attention to analysts immediately when they adjust their ratings, with stock prices reflecting the ratings changes over the subsequent month.
Well, yes . . . but it has taken until the current moment in time for the outperformance to be sizeable, carrying a whiff more of lucky macro break than reliable strategy.
Bored of words, Ma et al. then tried numbers — roughly recreating the ratio of earnings revisions deployed above, but utilising forecasts of “earnings and other metrics” instead of just rating labels:
We calculate the number of ratings upgrades less downgrades as a percentage of total number of ratings (and only include companies with at least three ratings). The ratio is then ranked by quintiles across the universe, and we measure the long-short return by buying the top quintile and shorting the worst quintile.
This approach also appears to work in both bull and bear markets…
. . . but:
Interestingly, although EPS is often the most examined line item, the EPS revisions strategy had the lowest return and Information Ratio amongst all the net revisions strategies. Ratings revision had the highest IR and lowest volatility. This once again shows that rating revisions is a useful signal, and more directly tied to share price performance than estimates of other financial line items.
¯\_(ツ)_/¯
We’d be tempted to say that somewhere here, a cart is being put before a horse: after all, the raw ratings should theoretically reflect the underlying estimates. So if they’re more stable, that probably just reflects that buy/hold/sell are too broad to capture the subtleties of changes in those estimates. Which is . . . obvious, no?
Citi go on to break their analysis down by country and sector…
. . . and by intensity of coverage, where fewer analysts (intuitively) = more influence per analyst:
All right, what’s been learned? Citi see a working framework (for Asian stock traders at least):
Investors who are inundated with incoming broker reports may sometimes be tempted to simply ignore them and avoid the noise. One simple rule-of-thumb to filter through the more meaningful reports is to pay special attention when the report is a ratings upgrade or downgrade. Our study empirically confirms this intuition, as we show that ratings changes actually do generate consistent returns, in our methodology.
The flip side is however that the absolute levels of ratings matter little to share prices within our methodology — thus, the corollary may be to remain cautious on trading based upon unchanged ratings. The exception for this is on low-coverage stocks — the actual ratings levels do help drive returns on average, in our methodology.
Much to consider. But this is, of course, all theoretical: if you’re an investor who has absolutely meatloafed through your career by unflinchingly following analyst ratings, please send your life story to the usual address .