In March, there was extensive media coverage (e.g. in the Guardian Australia) of shadow assistant treasurer Andrew Leigh’s company tax research which found companies paying lower effective company tax rates (i.e. actual tax paid/profits) did not create more jobs than those paying higher effective tax rates, and may actually create fewer jobs. The research paper was published in Economic Analysis and Policy (EAP), the online journal of the Economic Society of Australia (QLD) of which I’m Vice President, although this is a personal comment on my part, and I am not on the editorial committee of EAP.
I am very pleased that RMIT’s Sinclair Davidson has converted his critique of Andrew Leigh’s paper, originally published at Catallaxy Files, into a submission to EAP. Davidson’s paper has been published along with a reply by Andrew Leigh, which in my view doesn’t effectively counter the major criticism Davidson makes, that Leigh has used a mis-specified econometric model. The abstract of Sinclair Davidson’s comment is:
The paper presents a critique of the recent paper published in this Journal by Dr Andrew Leigh (Do firms that pay less company tax create more jobs? Volume 59, September 2018, Pages 25–28). Besides the model misspecification, omitted results and data availability bias of the regression used to inform the overall results, there are no controls for confounding factors that influence the effective marginal tax rate. These omissions are detrimental to the validity of the paper. This paper seeks to illuminate the sources of those errors and argues that the paper is sufficiently flawed that no weight should be given to its analysis when considering the potential consequences of a company tax rate cut.
I agree with Davidson’s conclusion. Indeed, as Judith Sloan pointed out in the Australian, it does not make sense to relate jobs created to the effective tax rate paid by companies in the way Andrew Leigh did, because the effective tax rate can differ from the statutory rate (i.e. 30% for large companies) for a variety of reasons, some of which would affect employment growth. So it would not be possible to say there was a one-way causal relationship going from the effective tax rate to jobs growth. A firm could have a low effective tax rate and low or negative employment growth for the same underlying reason (e.g. a previously poor financial position which means it can now take advantage of previous tax losses, but which would still make it reluctant to hire workers). In her Australian article, Judith Sloan was alluding to the fact the effective company tax rate is endogenous, which makes it perilous to estimate a relationship between the effective tax rate and employment growth. Judith Sloan noted in her Australian article:
What Leigh doesn’t seem to understand is that there is a variety of reasons why companies pay an effective tax rate below the statutory one.
Take mining companies that have made large investments. These companies will bring the cost of these investments to book over a long period of time, bringing down the rate of company tax they pay. You would also expect the number of jobs in those companies to be lower after the investment has been completed.
Andrew Leigh attempted to respond to one of Sinclair Davidson’s criticisms in his reply by including average profits in the regression equation. But econometric theory tells us this alone would not deal with the fundamental endogeneity problem that has probably given Andrew Leigh a biased and meaningless equation.
Andrew Leigh knows well the challenge of estimating causal relationships based on historical data from uncontrolled experiments, and it’s why he rightly advocates the analysis of natural experiments (e.g. in his work on the impact of the baby bonus on the timing of births) or the use of randomised controlled trials for policy evaluation. His EAP article should have been clearer about its limitations. In summary, Andrew Leigh’s findings are of questionable value to the policy debate over cutting the company tax rate.
Originally posted at Queensland Economy Watch.