To see the problem, let's start with the standard way of measuring defensiveness. All we do is plot the asset's return against the market return and then estimate the line of best fit. My chart does this for a classic defensive sector, utilities. Each point represents a monthly return in the past five years.
You can see several things here. One is that the trendline has a slope of less than 45 degrees. This tells us that each 1 per cent return on the All-Share is associated with a less than 1 per cent return on utilities - actually 0.63 per cent. This is utilities' beta, its response to market returns. This is relatively low, which is the classic sign of a defensive asset.
Now look at the line of best fit at the point at which All-Share returns are zero. At this point, the line is slightly above zero on the vertical axis. This tells us that, on average, zero All-Share returns are associated with a positive return on utilities. In the jargon, utilities have a positive alpha. They do slightly better than you'd expect from their beta. This is typical of defensives. It's why I like them.
But here's the problem. The points don't lie neatly along that line of best fit. Instead, they are scattered around it. This tells us two things. One is that our alpha is not assured. Look at the points along the horizontal line around zero - representing the months when All-Share returns were zero. Some of these are to the left of the vertical axis, showing that utilities fell then. This tells us that a positive alpha doesn't ensure a positive return when the All-Share does nothing. It's just that a positive return is slightly more likely than not.
Secondly, a beta of 0.63 doesn't mean that utilities return 0.63 per cent for every one percentage point return on the All-Share. Instead, it tells us that 0.63 per cent is the average response, and there's a lot of variation around this average.
One way to see this is to look at the four lowest points in my chart, representing the four months when the All-Share index fell the most. In these four months, utilities' returns were -9.3 per cent, -2.8 per cent, -0.2 per cent and -5.4 per cent. In two of those four months utilities actually fell by more than the market - although in the two months when it outperformed, it did so more significantly. This tells us that even the most defensive sector is only probably defensive, not certainly so.
What causes this variation? One thing is dumb luck. If a share happens to get bad news at a time when the market rises, it will move in the opposite direction to the market and this will tend to depress its beta. This is what happened to utilities in, for example, February 2015.
Utilities' beta since 2012
Some companies, and especially smaller volatile ones, can have enough idiosyncratic bad news on good days for the market (or vice versa) to produce negative betas. Such coincidences, however, would be accidents which you can't rely on to persist into the future. Personally, I usually regard negative betas on shares as things that are too good to be true.
You might think there's an obvious solution to this danger of measured betas being distorted by accidents: we should take a bigger sample so that our measure isn't distorted by outliers.
Generally speaking, this is good advice. But it runs into other problems. One is that for newly floated stocks we just don't have a big sample.
Another is that betas can change over time, a fact that a big sample can hide. For example, since 1990 the telecoms sector has had a beta of 0.9, implying it has been slightly defensive. But this single number hides the fact that telecoms had a highish beta during the tech boom and bust, a low beta in the early 2010s, and the sector has recently seen its beta rise. The sector has gone from being defensive to speculative and back to defensive since the early 1990s.
This is a widespread problem. For example, during the tech bubble IT stocks had high betas and miners had low ones. But the opposite has been the case recently. This is because investors' sentiment changes. During the tech boom and bust, IT stocks were most sensitive to this sentiment, but during the mining bubble and burst, miners were most sensitive.
There's another issue. Shares' co-movements with the All-Share index can vary depending on what causes the market to fall. For example, if trouble in the Middle East were to cause a spike in oil prices, oil stocks might well be defensive, as they'd do well as the market fell. But if the market falls because of worries about weak global growth, oil stocks could do badly as demand for their products weakens.
We should ask of any stock: to what is it defensive and to what is it not?
So, what should we do about these problems?
Here, I'll argue against my usual position. Ordinarily, I prefer to use statistics rather than judgment. And this method has worked: my no-thought low-risk portfolio is based on shares' simple historic betas and it has beaten the market for years. There might, however, be a case for assessing defensiveness by using judgment as well. We should ask of any stock: does it have resilience against possible shocks?
One source of resilience is market power, or what Warren Buffett calls "economic moats" - things that protect a company from competition, such as high capital requirements or brand names.
Another source of defensiveness is a lack of cyclicality in demand. Food producers can be defensive because people keep eating even in recessions. Yet another source is simple familiarity: in uncertain times, people hold on to stocks that seem reassuringly familiar.
Such questions direct us towards bigger stocks such as Unilever, Diageo, Reckitt Benckiser and GlaxoSmithKline. These might not be at the top of lists of stocks with the lowest betas. But I suspect they are nevertheless defensive - or at least that reasonable baskets of such stocks are. And, in fact, very many income funds do have big weightings in such shares.
This is not to say we should assess defensiveness purely by judgment. We know that judgments can be warped by all sorts of biases. Two especially dangerous ones here are the endowment effect and wishful thinking. The moment we own a stock we become attached to it, which means there's a danger of us overestimating its resilience. A mix of judgment and statistics might be better.
Another thing: All this shows the virtue of "eyeball econometrics". A scatterplot can tell us more about a share's qualities than one or two numbers such as alpha or beta.
