Update: I had a look at the ATO Taxation Statistics, specifically tables 8 and 9 of the corporate tax statistics. I calculate two measures of effective tax rate. First just the Net Tax to Net Income and second Net Tax plus Royalty expense to Net Income. Results are shown below in the table. Note that Mining has the highest effective tax rate after accounting for Royalty payments.
Mr Swan used independent analysis included in the Henry tax review to argue “ordinary workers” were getting ripped off by the likes of BHP Billiton and Rio Tinto.
“In Australia, wholly-domestic mining companies paid an effective tax rate of only 17 per cent and multinational mining companies paid an effective tax rate of only 13 per cent,” Mr Swan said in his latest economic note.
“Both (are) dramatically below the headline company tax rate of 30 per cent.”
By contrast, domestic manufacturers paid 25 cents of company tax for each dollar earned, while local retailers paid 27 cents in the dollar, during the period 2003 to 2007.
An average worker pays 30 per cent tax on every dollar earned above $35,000. Wealthier Australians pay income tax of up to 45 per cent.
“An ordinary worker who earns an extra dollar through their hard work pays higher tax, but a mining company that earns massive amounts pays the same flat, low rate of company tax,” Mr Swan said.
“That’s simply not fair.”
Either Swan doesn’t understand what it is Treasury are telling him, or Treasury (and the Henry Review) don’t understand the underlying paper that they are quoting (subscription required).
That paper is an unpublished NBER Working Paper Do Multinationals or Domestic Firms Face Higher Effective Tax Rates? by Kevin Markle and Douglas Shackelford. In their paper Markle and Shackelford compare the effective tax rates paid by domestic and multinational firms across 10,642 firms from 85 countries over the period 1988 to 2007. So far so good. The work horse equation in the paper is show here.
The explanation of all of that is found on page 7 of the paper.
To estimate the corporate income taxes paid by multinationals and domestics around the globe, we regress firm-level ETRs on categorical variables for the domicile of the parent and whether the company is a multinational. The regression coefficients on the categorical variables provide estimates of country-level ETRs for both domestic firms, i.e., those operating in the home country only, and multinationals, i.e., those domiciled in the home country but operating in at least one other country. These ETR estimates enable us to compare domestics with multinationals, both within countries and across countries, industries, and years.
I also want to pick out two of their footnotes where they provide valuable additional information.
11 To estimate equation (1), one industry and one year have to be excluded from the regression. To determine which industry to leave out, we calculate the mean ETR in each industry (two-digit NAICS) and then determine the median of those means. The industry with the median mean (code 31) is the one left out. We implement a similar procedure on the years, resulting in 2005 being the excluded year. To improve comparability across estimations, we exclude the same industry and year from each regression.
NAICS industry 31 is (some sort of) manufacturing. They do need to leave out one industry and one year so that the model doesn’t have what is known as a singularity problem, but that normally means that the coefficients must be interpreted relative to to omitted industry and year and that information is captured by the intercept terms. In this case there is no intercept and I expect that its all captured in the beta (zero) coefficient. That coefficient captures the average effective tax rate for each country given the country, industry and firm characteristics on average over the period being analysed. As they admit in footnote 12 (emphasis added).
12 Note that the magnitude of the domestic and multinational ETRs cannot be directly compared with the actual ETRs from the financial statements, which serve as the dependent variable. The domestic and multinational ETRs are the tax rates, conditional on industry, year, and size. That said, our empirical analysis shows that the estimated ETRs are very similar to the actual ETRs from the financial statements.
They then estimate their equation (1) and find the average effective marginal tax rate for various countries and differentiate between domestic and multinational firms. Overall they find the model fits well and the estimated effective tax rates and fairly close to the actuals. Then they estimate industry specific equations – that is what the Henry Review and Gillard and Swan are referring to.
This is how Markle and Shackelford describe what they do when coming to the industry analysis.
To assess whether ETRs vary across industries, we estimate equation (1) using industry groupings. We form the industries using two-digit NAICS codes and the 2003-2007 sample with total income tax expense in the numerator. We group two-digit codes to ensure that each reported industry has at least 900 firm-years. All observations are included in the regressions, but only cells with twenty or more observations are reported. Manufacturers comprise 49% of the firm-years.
Note what they are doing. A comparison between domestic and multinational corporations across countries within the same industry. For mining, to take one example, Australian domestic and multinationals don’t have a statistically significant difference, but UK multinationals do. Or we can compare across countries. Given firm characteristics Australian mining firms have the same effective tax rates as do Canadian firms. And so on. It seems that the table is best interpreted up and down the columns and not across the rows. Across the rows are results from different regressions and the results should not really be compared to each other. Markle and Shackelford do not test to see if the estimates are statistically different across the rows so we can’t really know if they are or are not. When looking across the rows we are not comparing apples with apples. Given the difference in firm characteristics across industries do we expect all firms to have the same effective tax rate? I would think not. (It is true that Markle and Shackelford do comment briefly on the variatuon across industries within the US, but they don’t make much of that issue). In order to test whether the different industries have significantly different effective marginal tax rates Markle and Shackelford would have to re-estimate their equation and interact the country and industry variables. I don’t know if they have done that exercise but they certainly do not report the results of an exercise like that in their paper.
The Henry Report shows this table based on the Markle and Shackelford industry analysis.
To make the argument that the Henry Review does, and Gillard and Swan have repeated, we would need to see tests of statistical significance across the industries – something that Markle and Shackelford do not do. Rather they present tests of statistical significance across domestic and multinational firms. The evidence does not support the argument the Henry Report, Gillard and Swan have made. It is unclear what the Henry Report means when it says, ‘Their country-specific estimates show significant variation in effective tax rates across sectors’ (emphasis added). There are no tests of statistical significance in the Markle and Shackelford paper to support that statement.