Bruce Chapman and I have released a paper looking at the effect of the HECS repayment threshold. In 2003-04 (the most recent year we looked at), earning $1 over the threshold reduced your post-tax income by $760, making it – on one view – a 76000% tax rate. The abstract:
Do Very High Tax Rates Induce Bunching? Implications for the Design of Income-Contingent Loan Schemes
Bruce Chapman & Andrew Leigh
We test whether very high marginal tax rates affect taxpayer behaviour, using a unique policy. Under the Higher Education Contribution Scheme â€“ an income-related university loans scheme in Australia â€“ former students with a debt face a sharp discontinuity. At the first repayment threshold they are required to repay a percentage of their entire income, resulting in an effective marginal tax rate that could be regarded as being as high as 76,000 percent. We formally model the taxpayer decision, and then use a sample of taxpayer returns provided to us by the tax office to investigate whether taxpayers bunch below the repayment threshold. We find a statistically significant degree of bunching below the threshold, but the effect is economically small. On net, we estimate that both the deadweight cost and the budgetary loss are less than A$1 million per year, a small fraction of the amount annually repaid through the Higher Education Contribution Scheme. The result has an important implication for the design of income contingent loans for higher education, such as those being introduced in the UK for tuition in September 2006. This is that it is possible to design arrangements in which the first income threshold of repayment is apparently high, but which are still able to deliver relatively high revenue streams in the early stages of income contingent policy reform without important tax payment avoidance consequences. Our findings also reinforce earlier research suggesting only minimal bunching around kink points in taxation schedules.
Apologies for the gratuitous British references – we’re taking a shot at a UK journal first.