There is theoretical work on the matter in general. In fact, many standard models of optimal consumption taxation will typically yield a distortion that is increasing and convex in the rate of taxation. This means that increasing a tax on a good from 5% to 6% results in higher distortion than increasing it from 2% to 3%. Following this logic, it is better to tax many goods at a low rate than to tax few goods at a high rate, as the distortion would be smaller. This means that removing the exemptions (taxing goods with a currently lower rate of taxation than the rest) is better than increasing the tax rate on previously taxed goods (since they have a currently higher rate than the exempt goods). To conclude, it means that the distortion from taxing exempt goods, which would mean going from 0% to e.g. 20% (assumed current VAT in Australia) is lower than going from 20% (current tax rate on non-exempt goods) to 40%, or so.
For a reference, see for example slide 30 (and a few before) in this slide set I found online. Its from Mankiw's Microeconomic Principles.
This however does not take into account any "fairness" considerations or welfare implications for the case of households not being able to afford enough food as a result. So the standard theory may not apply in cases where taxing food at all may lead to starvation, as starvation is not considered. However this is likely not the case anyway for such a rich economy as Australia that has other means of providing welfare (incl. food) to the needy.