Frontier Economics director Jason Hall is presenting today at the Australasian Finance and Banking Conference in Sydney. Jason will be explaining some recent work estimating the market value of imputation credits, outlined below.
Imputation tax credits are the credit an investor gets with a dividend for corporate tax already paid in Australia. The imputation system has been around for 30 years and there is still no consensus about how much the credits are worth. This is an issue because credits are not separately traded. You either get a credit with a dividend, or you don't, and only Australian resident investors get a cash flow benefit from the credits. For example, they are not useful for U.S. investors.
This creates a problem. As an Australian you get a large cash flow benefit from receiving a credit and if it was only Australians buying Australian-listed shares, you would pay more for a share that gives you a credit than a share which doesn't. But the international investors don't care about the credits. So when you have a market with Australian and international investors, it is an open question about how much higher the share price is if the company pays dividends that are accompanied by a tax credit.
It is considered important amongst regulators and regulated entities because the regulator makes an estimate about the value of the credits. The higher the assumed value of the credits by the regulator, the lower the revenue stream the regulated entity is allowed to earn. So the estimation of this single parameter affects the aggregate revenue stream of regulated entities by hundreds of millions of dollars.
The paper, co-authored by Stephen Gray, allows us to estimate the market value of imputation credits. We developed a novel technique that allows us to jointly estimate the value of cash dividends and credits (recall that it is hard to separately value them because credits are not traded). Our technique involves making an assumption about the distribution of error terms in regression analysis and then generating thousands of simulated samples in order to estimate confidence intervals.
Further, our research method has applications amongst any empirical analysis in which two of the explanatory variables are correlated. For example, suppose you were measuring the relationship between running times, testosterone and gender. There is correlation between testosterone and gender (if you code males = 1 and females = 0 there will be positive correlation between testosterone and gender); or suppose you were measuring the relationship between consumers' willingness to buy a train ticket as a function of (1) income and (2) distance to the CBD - people living close to the CBD have, on average, higher income so there will be negative correlation between income and distance. Our method has applications to those cases.
Click here to access the PowerPoint presentation (17 mins duration) and a transcript. (Please open the audio PowerPoint presentation as a slide show: the audio should automatically play).
The full research paper is available here.