While carbon pricing is considered the most effective policy to cut emissions, offshoring of emissions, defined
by the carbon leakage rate, is potentially a major drawback. Recent literature on carbon leakage from the EU
Emission Trading Scheme (EU ETS) exhibits a major discrepancy between analytical forecasts and empirical
observations of carbon leakage. The former predicts a leakage rate of 2 - 50%, and the latter is unable to show
any significant leakage. Aside from synchronising the theoretical and empirical literature, this paper aims to
tackle this discrepancy by investigating trade data from the EU Aluminium sector. The Aluminium sector is in
a unique position because most of its emissions are indirect which means it does not receive free allowance
allocations. The proposed regression model uses EU ETS monthly allowance price data from all four phases to
predict monthly net aluminium imports while controlling for domestic industrial demand. A number of com-
monly used autoregressive tests and corrections are applied to reduce the serial correlation in the error term.
This paper additionally proposes a new autoregressive correction, which minimises the Lagrange Multiplier.
For phase IV of the EU ETS, a percent-increase in allowance price is associated with an increase in Aluminium
imports of =C1.12 million, which empirically demonstrates the existence of significant carbon leakage from the
EU ETS. The serially correlated error terms call for further investigation of EU Aluminium trade to gain a more
robust understanding of the magnitude of carbon leakage.