I spent the first 10 years of my working life as an analyst and fund manager in the City, but subsequently worked overseas for many years as a project economist. When I returned to the UK investment scene, in the early 1980s, I soon realised that many private investors were operating at a disadvantage as they did not have access to the vast amount of company analysis and information that is available to the professionals. For many, the options were either to pay outside advisers and managers a substantial sum to do what is often a poor-quality job, or to spend a huge amount of time trying to track down information on quoted companies from a number of disparate sources. Admittedly, things have improved markedly since then, as there are now several free online information sources that make such information available. But nevertheless, the catalyst for embarking on my own share-screening process was my dissatisfaction with the options on offer 20 years ago.
The STAR Process
The basic alternative routes were either fundamental or technical. So because my background training as an economist and fund manager was essentially dealing with fundamentals, this was the route I chose. I reckoned that the first stage in developing a share-selection and equity-portfolio-management platform was to see if there was any correlation between the expected growth in a company's earnings and its subsequent share price. As there was no way that I had the time or ability to do the research on over 200 companies, I enrolled the help of former colleagues to assemble the detailed earnings projections for leading companies from a number of the leading brokers.
The estimated earnings data for several hundred larger-cap companies as produced by a number of leading analysts was then tabulated on a monthly basis over a 10-year period. All the shares covered were then ranked using nine different equations, labelled as Versions A through to H. The equations covered key variables such as expected future earnings growth rates, current share valuations, dividend yields, sector weightings and, not least, step changes to expected profits and earnings growth. The best results were from Versions A, B and H.
Of the nine permutations tested, versions A and B involved ranking by share price multiples derived from dividends weighted according to forecast future growth in earnings. Version H was somewhat different in that future estimates of earnings growth were related to future price earnings and dividend ratios - variations on price-earnings growth (PEG) factors. They were intended as the first stage in a two-stage process aimed primarily at capital growth share selection, whereas the objective of the other versions combined both income generation and capital growth.
The best results were achieved from a broadly based selection of shares that combined above-average projected earnings with below-average current valuations. These really did produce above-average share price gains over the longer term. This was the case in some seven years out of 10. More importantly, the gains substantially outweighed the losses. The results from the three best-performing sets of equations are shown in the table below.
It was from this basic analysis that the Share Tracking and Ranking (STAR) share selection and management methods evolved.
The Method Selected for Smaller Portfolios
Although two versions, not shown here, appeared to give the closest overall correlation between ranking list order and annual capital appreciation, the actual 10 share portfolio selections derived from the ranking lists of versions A, B and H, in the table, tended to provide the better final result - albeit with a possible increase in inter-year volatility. Version B, which was initially used for selecting high-income portfolios, had the added advantage of attaching greater weight to existing income and therefore seemed more appropriate for the selection of income portfolios. In fact, subsequent work showed that the best way of selecting income portfolios was to employ the basic growth criteria and then overlay a dividend yield filter below which shares would not be selected.
As the capital performance figures do not include income, the results from the income-oriented ranking versions tend to understate their total outperformance. The average yearly difference in gross running yield between the income and growth share selections (versions B and H) over the nine-year period was approximately 250 basis points - a not insignificant additional benefit over a 10-year period.
The reason that the Version H 10 Share portfolio results are much better than would be expected from the first-stage ranking data is that the second-stage analysis adds an extremely effective filtering process, which eliminates a large number of shares from the initial ranking.
The Method Tested Over 20 Years
One of the problems with investment is that fashion, or perhaps more accurately emotion, plays such a big part in shaping the hopes and fears - and therefore the actions - of most investors. There are therefore bound to be periods when an objective share selection and management approach is going to be wrong-footed. It is partly for this reason that I believe the long-term record of the STAR approach is so vital. As my share-screening process is essentially based on value investing, it obviously becomes less effective in periods when this approach is out of favour. I believe that the 20-year period over which I have been analysing the STAR growth selections gives a fair impression of the usefulness of this basic methodology. The table, STAR 10 Growth Share Results 1985-2006 records the yearly gains and losses achieved by the nominal investment of £1,000 in each of the 10 shares selected for growth in early January each year.
These selections were based on the middle-market prices ruling as at the close of business on the relevant day, with a limited number of switches signalled during most years. The aim of the exercise was to show how the method compared with that of the wider market as represented by the All-Share index. Although dealing costs were excluded, so were dividends. The final column in the tables therefore indicates the stock-picking ability of the STAR process rather than a total compounded return over the period.
The first table shows the absolute annual gains or losses...
...while the second shows the cumulative effective since the start of the period.
Variations on the Basic Theme
I have used the basic STAR methods to select portfolios designed to produce income as well as growth and, more recently, I have been testing their effectiveness in highlighting speculative investments that are likely to outperform the market within the following few weeks and months.
Using STAR for income portfolios is a straightforward process. An additional filter that excludes all shares whose dividend yield falls below the median of the whole group is applied to the growth share selection procedure. In effect, this means that the income portfolios tend to incorporate a number of shares with lower expected earnings growth rates than the basic growth selections. In some years, this results in the income portfolios achieving lower annual capital growth than the growth selections, but there is usually the compensation of a dividend yield that is between 100 and 200 basis points (1 per cent and 2 per cent) higher for the whole year.
The table below shows that, between early January 1999 and January 2006, the index had risen by 4 per cent, whereas the STAR 10 share growth selections had risen by more than 220 per cent, excluding costs.
In the same period, the larger 20 share growth selections rose by more than 260 per cent and the 10 share income selections had advanced by more than 150 per cent, excluding any benefit from above-average dividend income.
The idea behind using STAR for selecting speculative investments is that the change in consensus estimates over the previous month is used to rank shares according to the alteration in the expected rate of growth in earnings over the next year or two. In theory, those companies that meet the criteria of above-average earnings growth rate plus a below-average valuation should get a boost from investors when their forecast growth rates are increased by a significant margin. So far, the jury is out on the success of this as a reliable tool for picking shorter-term investments. It seems to be most successful when the method highlights shares that are within a sector that is being re-rated. For instance a number of oil and mining shares were highlighted early in 2006 as they appeared fundamentally cheap and were also being subjected to quite substantial earnings upgrades at the time.
The current STAR share selections
I run the STAR ranking methods at the beginning of every month and currently cover 307 of the larger companies. The 10 shares ranked for growth as at 4 October are set out in the table below. However, it should be noted that prices have moved a lot and the STAR method might well produce a different selection at current prices. More background on the STAR method may be found at www.companynews.co.uk/star/index.htm
