The two ongoing political scandals of corporate behaviour currently featuring in public debate in many countries, concern company tax and executive pay.
There is growing awareness and widespread criticism of large companies for not paying enough tax, with many paying no tax. This is paralleled by other criticism of large companies regarding excessive payments given to chief executive officers, other senior executives and board members.
The lobbyists for business defend the low (or zero) tax some companies pay by arguing that they have made losses, either now or in the recent past, and so it is entirely fitting that they pay no tax.
By that logic, the firms that pay little or no company tax would also be paying their CEOs poorly, because their financial performance had been so poor. This is the “corporate model” of CEO tax and pay.
But what if the reverse was the case? What if, instead, “paying no tax” was rewarded by higher CEO pay, rather than being a sign of poor performance? It would suggest that the corporate lobbyists’ arguments were misplaced, and a different explanation for both those things was necessary.
To test this, I compared information from two recent datasets. One was 2013-14 taxation data at the company level, that covered 1700 large companies, from the Australian Taxation Office (ATO). Almost a third of firms in that group paid no tax. The other source was the 2014 Annual CEO Salary Survey data collected and published by the Australian Financial Review on CEO pay in the top 300 firms (that is, the 300 firms that pay their CEOs the most). I linked those datasets using company names and found 221 matches. That overlap wasn’t surprising, as the biggest single predictor of CEO pay in a number of studies is company size, so both datasets ended up having a lot of large companies in them.
You can’t just do a simple correlation, because company resources or size is the biggest predictor of CEO pay. So you have to control for company size. The simplest way of doing this is with what’s called “ordinary least squared regression”. In effect, this statistical technique enables you to ask “what is the relationship between tax paid and CEO pay, if you can hold company size constant?”.
In some studies, size is measured by revenue, assets, market capitalisation or number of employees. In this dataset, we can use revenue — in particular, “total income” (excluding deductions) or “net income” (“total income” minus deductions and minus tax paid).
The results are in the table below. In short, the results suggest that a company that pays no tax gives its CEO around 20% per annum more than an equivalent company that pays tax — but there are a number of important caveats to bear in mind.
These caveats are as follows. First, there are four different equations below. Two use total income (to measure size), and two use net income. In each pair, one equation uses industry (there are 13 of them) as control variables, to account for average industry differences in legitimate tax deductions, and one does not. As it turned out, industry did not have much effect.
Size is probably better measured by total income, and after controlling for that paying no company tax is worth roughly an extra 23% in CEO pay.
Net income is a weaker measure of size. It also reflects performance and the ability of firms to claim deductions. By the “corporate model”, this would be a better predictor of CEO pay, but it is not. Without including “paying no tax”, equations using net income would be barely significant. When we control for “net income”, then “paying no tax” adds two-thirds or more to CEO pay, but that probably reflects the weakness of net income as a measure of a firm’s resources.
Still, it’s feasible that if more variables were available, we might reduce or remove the impact on CEO pay of not paying tax.
Similarly, if sample size was higher we might find more significant industry effects. By concentrating on the highest-paying firms, we might also be oversampling those with dodgy practices (explained a bit more below) and overstating the average effect of “not paying tax”.
Second, the data only relate to Australia, to a particular period, and to a particular group of companies. I have compared CEO pay in one year with predictors from the previous year. This use of “lags” is common in studies but does not affect the results much. (If the explanatory variables were measured in 2015 rather than 2014, the net effect on CEO pay of “paying no tax” changes from 23% to 26%.)
*Read the rest at John Menadue’s Pearls and Irritations