Parts of the modern investment industry appear to view reading through company accounts as like hand-washing dishes or waiting in the rain to hail a taxi — a loathsome task that will eventually be eliminated by technology.
Leading this trend are so-called smart beta funds — a booming area of investment management that seeks out magic formulas to screen for stocks with characteristics proven to “beat the market”, and distils them into simple, low-cost index-like products.
The strategy, which has enjoyed tens of billions of dollars of inflows, has come under criticism of late for its vulnerability to suffocating itself. If too much money chases the same winning “factors”, such as cheapness or price momentum, then the advantage gained from doing so will be arbitraged away.
Less frequently discussed is the mechanistic approach these smart funds employ to screen for supposedly “fundamental” features of the companies in which they invest. Stock screens were in the past intended to be exclusionary rather than inclusionary tools, meaning they filter out investments that do not match certain criteria and allow for closer examination of those that do. By inverting this process — seeking to invest solely on the basis of what the screen says — certain smart beta funds appear at risk of making some dumb mistakes.
One popular type of smart beta approach is to screen for “quality income stocks”, or shares that offer a solid and dependable dividend. Billions of dollars have in recent years poured into exchange traded funds that invest on this basis, with investors relying on faith that the quantitative models will keep their cash safe. Société Générale, for example, publishes a screen seeking out “attractive and sustainable dividends” using a methodology very similar to some of these smart beta ETFs. Its model right now indicates that AstraZeneca, the UK-listed pharma company, is a strong pick.
SocGen’s quant screen that tells us AstraZeneca’s high dividend is solid is based on two neat formulas. The “quality” of a dividend payer’s earnings is calculated using what is known as a Piotroski score, which uses various ratios to give a company a score out of nine. The “sustainability” of the dividend is then calculated using the so-called Merton model — sometimes referred to as “distance-to-default” — which measures the probability that a company will default on its debts.
A look at AstraZeneca’s accounts, however, raises several questions about the stability of its dividend that this “quant” analysis appears to miss. AstraZeneca has been reporting its own version of earnings, “core earnings per share”, that has started to increasingly diverge from earnings based on generally accepted accounting principles (GAAP). This “core EPS” includes gains from one-off events but strips out restructuring and amortisation charges, among other things. The difference between AstraZeneca’s GAAP earnings and “core earnings” has become noticeably large over the past two years. In 2014 “core” EPS was $4.28 while GAAP EPS was just $0.98, and in 2015 “core” EPS was $4.26 while GAAP EPS came in at $2.23.
Once you look past the apparently strong earnings and instead to AstraZeneca’s cash flow generation, the solidity of its high dividend appears less certain. The past two years there has seen a sharp deterioration in AstraZeneca’s free cash flow while it has sharply increased its net debt. In 2015 it would not have been able to pay a $3.4bn dividend out of free cash flow, instead borrowing $4.3bn to plug the gap. The quantitative model appears to be painting a highly divergent picture from the company’s own accounts.
Whatever the marketing tells you, we are still far away from a time when computer can do all the boring investment chores for us
There are similar problems with using the Merton model to monitor the sustainability of a dividend. Certain dividend ETF providers use it as a forward-looking indicator as to the stability of payouts. Morningstar, whose Merton-style “distance-to-default” rating powers BlackRock’s $6bn iShares Core High Dividend ETF (which does not hold AstraZeneca shares), for example, argues that a deteriorating rating has historically proven a strong indication that a dividend is about to get axed.
The Merton model uses the Black-Scholes options pricing model to assess the likelihood that a company’s assets will fall below the value of its liabilities, and is mainly used by credit analysts as a quick and dirty way of measuring the likelihood a company will default. Yet unlike a debt coupon, the decision to pay a dividend is an arbitrary one made by company management — it depends on capital allocation decisions and use of cash flow that are unable to be represented by inputs into the model.
The second issue is that by being based on Black-Scholes, the Merton model is heavily skewed to volatility as the main measure of risk. Shares in large-cap stocks such as AstraZeneca tend to be less volatile, meaning the model is biased to making them appear less risky. By the time the volatility of the shares starts to spike in anticipation of a cut, it is likely to be already too late to avoid large losses.
The weakness of using these models as a “smart” way to invest is that they can only be as good as the assumptions and inputs that go into them. Whatever the marketing tells you, we are still far away from a time when a computer can do all the boring investment chores for us.
Sample the FT’s top stories for a week
You select the topic, we deliver the news.