Capital Crime

True, spending on fixed assets must be vital if a company is even to hold its own, let alone to progress. Existing plant and equipment has to be upgraded to the latest technology and sparkling new kit must be added if customers are to be grabbed from competitors and new markets explored. And finance directors never knowingly confess that they have little idea which chunks of capital spending are worthwhile and which are money down the drain. To hear the average finance chief in PR mode you would think that every pound laid out fulfilled its intended return and that claim would be backed by data on the pay-back period, the internal rate of return, the net present value and all the other metrics they teach in business school.

However, when we crunched the data relating to FTSE 350 companies’ capital spending during the five years 2012 to 2016 we found that the link between capex and ‘success’ – however you want to measure that notion – was rarely strong, tentative only some of the time and downright illusory much of the while. But before we discuss the data in detail, let’s sketch in the background.

It’s hardly an exaggeration to say the great and the good of British business are obsessed with the UK’s productivity – or, rather, the lack of it. It used to be axiomatic that productivity – usually measured by output per hour worked – rose inexorably, even if the UK’s rate of improvement often lagged its competitors. But after the 2008-09 financial crises, that stopped. The UK’s productivity dipped then barely recovered and the widening performance gap between the UK and its competitors prompted economists to talk about ‘the productivity puzzle’.

While it remains unclear exactly why the UK has suffered especially poor productivity since 2009, one factor that economists latch onto is the country’s comparatively low rate of capital spending. That’s logical since lower rates of capex imply workers having to cope with sub-standard infrastructure and using less efficient equipment than their tooled-up overseas rivals. Result – they produce less in a given time.

To the extent that is true, it tells us that capital spending must be good. From that, it should follow that those companies that do the most capex, related in some way to their size, will have the most productive workers. And if their workers are particularly productive, it also follows that those companies should be the most profitable ones since the fruits of the extra productivity will go to shareholders as well as – and often more than – employees. And the final link in this chain connects productivity-driven profitability to excess share price performance.

But if this process starts with capital spending – if cap-ex is the driver of improving productivity – then there should also be a discernible connection between capital spending and share-price performance. That is the purpose of this charts-based exercise – to spot and to quantify this connection; to examine the proposition that those companies that do the most capital spending are also the best ones to invest in.

The feature is best viewed in the above Excel workbook that takes readers through the charts – 24 of them in all – which provide a regression analysis of the link between capital spending and share price performance. Under scrutiny are the component companies of the FTSE 350 index excluding investment trusts, which leaves 298. In addition to the charts, the workbook provides additional commentary on each chart plus the full raw data, which readers can also use for their own analysis. Charts 1 and 2 sketch out the macroeconomic background just mentioned. Chart 1 clearly shows that UK productivity – though less than its main rivals – kept pace with their improvement until 2007. Thereafter, a sharp divergence began. While productivity in the rest of the developed world recovered rapidly, the UK’s hardly progressed at all. Perhaps this was linked to capital spending. It is easy – and tempting – to think so because, as Chart 2 shows, the UK’s rate of capex always lags its rivals.

Charts 3 to 12 get down to the nitty-gritty, looking at companies’ capital spending in the aggregate, examining various aspects of the relationship between their capex, their productivity and their share price performance (always relative to the FTSE All-Share index).

A summary of the charts’ findings is shown in Table 1. The important columns are for ‘Slope of the line’ and ‘Closeness of fit’. In a basic linear regression model, the slope of the line defines the extent to which the so-called ‘independent variable’ (ie, the driving factor) influences the dependent variable. By convention, the data for the independent variable are plotted along the horizontal (or ‘x’) axis, while the data for the dependent (or influenced) variable are plotted along the vertical (‘y’) axis. A positive value for the line’s slope means that the independent variable influences the dependent one in the same direction of travel. So, for example, a value of 1.0 would mean that for every unit that the independent variable rises, the dependent variable would rise by the same unit. A negative value for the slope implies that the dependent variable moves in the opposite direction. Usually, that’s not what you want. It suggests that causality runs in the opposite direction to what intuition suggests; it’s even a perverse outcome. Yet, as Table 1 shows, there are several perverse outcomes in the relationship between capital spending and the influence it is supposed to have on the performance of companies and their share prices.

The column for closeness of fit quantifies how well the trend line in a basic linear analysis fits the individual data points. In effect, it is the average distance between the line itself and all of the data points; the higher the value, the better the fit. A value of 1.0 (or 100 per cent in the table) would mean that every data point lies exactly on the trend line. That’s almost never going to happen, least of all when there are many data points, as with the charts here.

Indeed, for Charts 3 to 12, the values for closeness of fit (or R-squared, in statistical talk) indicate little linkage between the independent and dependent variables. Take the fit for Chart 10, which connects share price performance to profit per employee. The value says that almost 8 per cent of the changes in the share price are driven by changes in profit per employee and it is the highest R-squared value of all the charts. Yet put the proposition the other way around and the regression says that 92 per cent of the price changes are down to factors other than profit per employee; in other words, that particular factor accounts for very little. And when closeness-of-fit values drop to below 1 per cent, as with four of the 12 charts, then, in effect, the regression line is simply joining up a series of random data points, none of which has anything to do with the others. Quite likely we would get a superior fit if we correlated share price performance with, say, changes in the price of potatoes.

Granted, the findings of the charts – or, rather, the lack of findings – are open to question – most obviously that we did not churn enough data. After all, we confined the analysis to the five years 2012-16, which may have given a distorted picture. A bigger sample, taking different stages in the business cycle – 2012-16 were basically years of hesitant recovery – may have produced different results.

Then there were problems of definition – exactly how do you measure capital spending in order to make comparisons across many companies of differing sizes and capital intensity? In relation to the size of the company, yes, but that still leaves options open. To make the comparison, we quantified capex both in relation to a company’s gross profits (ie, after the cost of sales but before other input costs) – see Chart 3 – and in relation to its capital employed (Charts 4 and 5).

Next, we fine-tuned what is meant by ‘capital spending’. Perhaps capital spending in excess of the depreciation charge is the stuff that really adds value (Chart 6). Alternatively, maybe it’s another form of capital spending altogether that does the trick – so we plotted spending on acquisitions against share price performance (Chart 7).

One danger was that we were dragging ourselves into a data-mining exercise. It’s a play on the joke about monkeys and typewriters – churn enough data and you’re bound to come up with what looks like a significant finding. Even so, we ploughed on and compared growth in capital spending with share price performance (Chart 9).

We also looked at the link between capital spending and productivity (defining ‘productivity’ as gross profit per employee). Here, there was a tantalising connection. At least, Chart 12 implies that growth in capex influences profit per employee in a fairly clear way, even though Chart 11, which deals with something similar, indicates that changes in capital spending between 2012 and 2014 had next to no effect on raising employee productivity between 2014 and 2016. Related to that, Chart 12 seems to complement what we had seen in Chart 10, where improving productivity (measured by the change in profit per employee between 2012 and 2016) seemed to feed strong price-relative performance over the same period.

We concluded with another thought – that measures of capital spending and productivity may provide a better predictor of share price performance when applied to individual companies. This presents another problem with the data – too little of the stuff, rather than too much. The difficulty is that, with companies, one can really only collect data once a year and few data banks offer full accounts details for companies stretching back, say, 20 years. Even where such data is available, its usefulness may be limited because at the end of the period the company may be quite different from at the start; in other words, the data points won’t compare like with like.

With that in mind, we chose six long-established FTSE 100 companies whose character has remained stable but which have experienced differing levels of success. Tables 2 and 3 summarise the key findings of the regressions we ran where we juxtaposed, first, profit per employee and, second, capital spending per employee against the share price at the end of the 15 years 2002 to 2016.

For some companies the exercise works better than for others. At Unilever (ULVR), the fit between both capex per employee and profit per employee and the share price is uncanny. It is tempting to think there is predictive value in both these metrics; that the regression line on the charts denotes the point where the shares are cheap or dear for a given amount of capital spending or profit per head (below the line indicates a cheap price; above it, the shares are dear).

At Sage (SGE) and the soon-to-be-swallowed GKN (GKN), the closeness of fit indicates that capex and productivity are saying something useful. At Marks and Spencer (MKS) and GlaxoSmithKline (GKN), the charts are as unhelpful as most of those that use aggregated FTSE 350 data. The two companies are doing capital spending – at Glaxo, much more per head than at the other five – but it seems to have little impact. Neither profit per employee nor the share price has shown anything more than stuttering progress over the 15 years. Surely each of these is a case where the finance director should acknowledge upfront that not even 20 per cent of the company’s capital spending is likely to be worthwhile. Well, yes, but don’t hold your breath.