Where Z scores first

The inevitable questions also arise whenever a former FTSE 100 company, such as Carillion, goes belly up: could it have been foreseen? And that quickly morphs into ‘how could that have been missed?’ Hindsight is wonderful. Additionally, to the extent that mounting debts did for Carillion – at its collapse its bank debt stood at £925m and its liabilities totalled almost £7bn – a more specific question arises: would the use of Altman Z-Scores have revealed the coming collapse? Would those Carillion shareholders who were smart or lucky enough to apply Z-Scores to the group’s financial data have spotted the impending demise and have escaped with some of their invested capital intact?

To assess that, we must explain Z-Scores and – more important for equity investors – we must quantify their usefulness not so much as a tool for predicting bankruptcy than as a means of separating out those companies whose shares are likely to produce above-average performance from the also-rans and the failures. As it happens, Z-Scores seem to be a helpful screening tool for predicting above-average investment performance, but more of that in a moment.

First, what are Altman Z-Scores? They are named after Professor Edward Altman, a US academic who has spent most of his career at New York University’s Stern School of Business. That Z-Scores focus on predicting corporate bankruptcy is almost incidental. The purpose of Professor Altman’s 1968 paper introducing them was to assess – even to revive – the use of simple ratios in rigorous economic analysis. Professor Altman did this using ‘multiple discriminant analysis’ where several ratios are combined to assess a probability. In his case, he used five ratios appropriately weighted to generate an overall index – the Z-Scores – that would measure a company’s vulnerability to bankruptcy.

Details of the eight financial variables that are the basic raw material, the five simple ratios into which the variables are combined, the weightings given to each and the resultant Z-Score are explained in the box 'Build your own Z-Score' below. However, what was by today’s standards a small-scale project has had a very big impact. The original data for Professor Altman’s 1968 paper was drawn from just 66 US quoted companies, half of which filed for bankruptcy between 1946 and 1965. The test accurately predicted failure two years ahead of the event seven times out of 10.

In many subsequent – and much bigger – tests in the following decades the accuracy of Z-Scores in predicting failure steadily ran at over 80 per cent, while generating false negatives (ie, predicting bankruptcies that did not materialise) of less than 20 per cent. Meanwhile, the methodology of Z-Scores was tweaked to include private companies, where balance sheet data was substituted for the stock-market data of listed companies, companies operating in emerging markets and those engaged in operations other than manufacturing; that said, Z-Scores don’t work for finance-industry companies whose balance sheets are structured differently.

Much of the merit of Z-Scores lies in their simplicity. As the box copy explains, they are simple to calculate and easy to understand – literally, the score indicates the likelihood that a company will slide towards financial immolation. In his original paper, Professor Altman divided companies into three categories, according to their Z-Score. Those that got a score above 2.99 were deemed ‘safe’. Those whose score was between 1.81 and 2.99 were in a grey zone and those that scored 1.8 or below were predicted to go bankrupt.

However, time and the ability of companies to run on much leaner balance sheets has changed all that. Nowadays Professor Altman says that the zones published in 1968 are no longer valid. After all, today the average Z-Score for a company whose corporate bonds are rated ‘B’ by the ratings agencies is 1.65, yet not even half of those bonds, which are at the speculative end of the ratings spectrum, go into default.

Instead, Professor Altman reckons that today’s cut-off point between companies in the grey zone and those that are distressed is a Z-Score of zero. Taking a cue from that, in constructing Z-Scores for components of the FTSE 350 index, we have added an extra category so that those companies scoring between 1.8 and zero, which would previously have been ‘distressed’, are now in the grey zone and only those scoring below zero are distressed.

The question we have addressed is not whether Z-Scores are likely to predict bankruptcy, but – far more relevant for equity investors – whether they give any guidance about the likely performance of company shares. Table 1 'Z-Scores and share price performance', summarises our findings.

Essentially, we take the current components of the FTSE 350 index and, where our database has the relevant figures, calculate Z-Scores for each of the years shown in the table. Having entered each company into one of the categories – safe, okay, etc – according to its score on that year’s financial data, we then quantified the performance of its shares for the following year, thus making the loose link between Z-Score and subsequent share price performance.

Take the column for 2016’s data in Table 1. The sample size is 286 companies, mostly because all investment trusts and most banks were removed. Of this, 69 companies (24 per cent of the sample) scored above 2.99, putting them in the ‘safe’ category. In the following year – 2017 – the unweighted average share price performance of this 69 was a 20.2 per cent gain relative to the change in the FTSE 350 index. Over the same period, the small number of companies that were classified as distressed – just 5 per cent of the sample – generated a 2.6 per cent return relative to the FTSE 350.

Arguably what’s most interesting is that those companies in the safe category consistently perform well. For three of the years – 2012, 2014 and 2016 – they turned out to be the best-performing of the four categories. In two of the other three – 2008 and 2010 – they were the second best.

Simultaneously, the comparatively small number in the distressed category proved to be the worst performers, being bottom of the pile in four of the six years – 2010, 2012, 2014 and 2016. In only one year – the 2008 batch – were they the top performer. That was logical to the extent that 2008’s group was assessed by 2009’s price movements when markets rebounded from the panic caused by the collapse of Lehman Brothers the year before. Thus shares in weak companies that had been pretty well written off during the mayhem bounced back when it became clear they weren’t about to vanish.

Clearly enough, the relative movements of the four categories of companies suggests an investment plan – go long those that are safe, maybe short-sell those that are distressed and ignore the bulk – about two-thirds of the sample – that are in-between. Obviously this is easier said than done, especially the short-selling. Besides, it is debatable whether that part would be a winning tactic since, for the most part, punters would be selling into a rising market – during the six periods under review, the FTSE 350 only fell twice. And, despite their iffy status, the distressed stocks on average produced gains relative to the 350 in four of the six years.

But, to quantify this tactic in very crude terms, first imagine that the years under review in the table were successive and, second, that the performance data was absolute and not relative. In those contrived circumstances – and making no allowance for dealing spreads or costs – then short-selling the distressed group would produce a small loss – £100 would become £97.20. Applying the same terms to going long the safe category would turn £100 into £213.

Meanwhile, there are clear limits to the predictive value of Z-Scores on the performance of individual stocks. This is shown in the table by the data for ‘slope of line' and ‘R-squared’. These deal with the line of best fit from a scattergram, which, in this case, shows the extent to which Z-Scores drive relative share price performance.

The values for slope of the line are positive and, happily, that means the line slopes the ‘right’ way – share price performance improves with rising Z-Scores. That said, the line is pretty flat. For every unit that Z-Scores rise, relative share price performance never rises by more than 0.2. Besides, the R-squared values in the table show that the data points are all over the place. With such big samples that’s probably to be expected. Yet even the highest R-squared value – 5.8 in 2014 – indicated that Z-Scores could account for just 5.8 per cent of share price movements, while other factors – whatever they were – accounted for the little matter of the remaining 94 per cent.

No matter. The loud message coming from this exercise is that Z-Scores have something useful to say about share price performance; in particular, that the winners are more likely to come from the top-scoring group of ‘safe’ stocks. Perhaps portfolios should wholly comprise shares from the safe zone. At the very least, investors should add a Z-Score calculator to the tools they use to screen for would-be investments.