Back in the dog days of last summer, the Government announced a package of reforms to corporate governance. Among those reforms was the welcome requirement for “listed companies… to publish pay ratios between chief executives and their average UK worker”. On the back of gender pay gap reporting and the commitment to transparency expressed in the Government’s response to the Taylor Review, this is a good step forward.
But as has been highlighted with regard to the gender pay gap, the devil is in the detail if we want to strive for a fair picture of what’s going on in the UK’s biggest firms. First, there’s the issue of whether it’s the total compensation received or just hourly or annual pay. For many chief execs, their remuneration package will include a cash salary, shares and a pension. So it’s welcome that the government has said that the calculation should be based on the CEO’s total annual remuneration. Doing this accurately for a typical employee may be slightly more challenging. But given the policy is restricted to roughly 900 large, publicly-listed companies who will use payroll software, calculating the salary and pension received by its workers should be achievable.
The second big question relates to what exactly the ‘average’ worker means. So far, the government has stated that it will be the “average total remuneration of the company’s UK workforce”. However, how this is to be calculated has yet to be decided. With the publication of the secondary legislation that will make this commitment a reality imminent, the decision needs to be made now.
The two most common ways to define the average worker are the mean (add up all the salaries and divide them by the number of employees) and the median (the person in a company who is higher paid than half of employees and lower paid than the other half). The mean is easier to calculate, but the median is a more meaningful metric.
The challenge is greater for firms that are split into subsidiaries and would need to bring together pieces of information from different parts of the organisation. This could make identifying the median employee more difficult, especially if, say, different payroll software is used by different subsidiaries which may not sync seamlessly. But given large firms comprised of a number of entities such as, for example, Shell have been able to combine this information in order to report their median gender pay gap, this should be a surmountable hurdle.
To get a sense of why choosing the mean or median matters, consider this illustrative example. Take a company with 100 employees. In this company, 15 workers are lower-paid support staff each earning £15,000 per year. In addition to them there are 80 employees on £25,000 per annum and an executive team made up of five members. Four members of the executive team are on £1 million per annum, while the CEO earns £2.5 million a year. The mean pay in this firm is £87,000, whereas the median is £25,000. The ratio of the CEO’s pay to the mean is 29, while to the median it is 100.
Now suppose that this company has a good year and increases pay for all staff. Pay for the lowest-paid employees rises to £17,500, and pay for the majority of employees rises from £25,000 to £30,000. Perhaps unsurprisingly (given recent trends in executive remuneration), pay for the executive team (in absolute and percentage terms) rises most, from £1 million to £1.5 million for four of them and from £2.5 million to £3.5 million for the CEO. This change increases mean pay to £122,000 and median pay to £30,000. It has no impact, however, on the ratio of the CEO’s pay to the mean. The ratio of the CEO’s pay to the median on the other hand rises to 117 (approximately the ratio reported by the High Pay Centre in its most recent report on executive pay).
Finally, suppose that this company decides to outsource its lower-paid support staff. Moving those 15 roles outside of the company would have no effect on median pay or the ratio of the CEO’s pay to the median. However, it would reduce the CEO to mean ratio from 29 to 25, a fall of 14 per cent. This example shows why relying on the CEO to mean ratio would be a poor choice.
The argument against using the mean is also backed up by analysis of actual wage data. As the chart below shows, there’s a large and consistent gap between the mean and the median wage across all firms. In 2017, the mean annual pay (£28,300) was 24 per cent higher than the median (£22,900). So choosing mean over median makes the gap between the highest earners and more typical employees appear less wide.
Source: RF analysis of ONS, ASHE 1999-2017
The government’s proposal is that only quoted companies will be required to report this information. To that end, we can limit our analysis to large firms and compare the wages of high earners (those at the 90th percentile or ‘p90’) to the average or typical employee.
To calculate the figures in the chart, we calculate the ratio of p90 to the median wage for all firms of that size, and the same in relation to the mean. For the largest firms (with at least 10,000 employees), p90 is 2.6 times higher than the median while p90 is roughly 1.9 times higher than the mean. The percentages presented in the chart below illustrate the differences between these two metrics, and show that the difference is consistently bigger for larger firms.
Source: RF analysis of ONS, ASHE 2000-2017
We can also track the gap between mean and median ratios within firms. In order to match the firms most likely to be affected, our analysis selects those that have at least 10,000 employees and with a sufficient number of employees covered in the data. The metric used in the chart below is the same as above, except it uses the earnings of the worker at the 99th percentile (p99) within the firm, versus mean and median.
For some firms (on the far left of the chart) the median ratio is actually lower than the mean. However for the vast majority, the median ratio is higher. These figures don’t take into account the mix of part-time and full-time working, so a firm with a particularly high or low share of people working fewer hours could be affecting these gaps. To create a more standardised metric, firms could be allowed to published a pro-rata version alongside the ‘raw’ gap. That said, repeating this analysis using hourly wages, thereby overcoming the relevance of the number of hours worked, reveals a similar pattern in terms of the size of the difference.
Source: RF analysis of ONS, ASHE 2015
In summary, it’s welcome that the government is taking action to shine a light on the issue of pay inequality. But as always, policy-making is about the detail and the best way to make a good idea reality. The median is required if the ratio is to be meaningful; choosing the mean would risk achieving nothing. It is technically more difficult for firms to calculate the median. Firms should therefore be provided with clear guidance on what is expected from them – as has been the case in the US where detailed advice has been published by the Securities and Exchange Commission. There will be a burden for some companies but given that the government has taken a bold decision, it’s worth ensuring that it has the maximum impact.
This work contains statistical data from ONS which is Crown Copyright. The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research datasets which may not exactly reproduce National Statistics aggregates.