A high sensitivity to the allocation method has been reported for LCA results when evaluating carbon intensity and fossil energy consumption for bioethanol pathways (Beer and Grant, 2007; Kim and Dale, 2002; Malca and Freire, 2006). Allocation refers to the distribu­tion of environmental burdens between coproducts in the LCA of a multifunctional system. The issue of allocation is one of the weaknesses of biofuels LCA. The ISO 14040-series (ISO, 2006a, b) recommends avoiding allocation whenever possible either through division of the whole pro­cess into subprocesses related to coproducts or by expanding the system limits to include the additional functions related to them (often referred to as system expansion or substitution and treated in the literature as an allocation method of its own). A complete subdivision is not possible for joint production processes due to the dependence between coproducts’ flows. In fact, subdivision is only feasible when unit subprocesses are physically separate in space or time (combined production), so it is only on exceptional occasions that the allocation problem can be completely eliminated. Ekvall and Finnveden (2001), after screening a large sample of LCA case studies, did not find a case study where this was the case. In the present study, sub­division is applied to the stages downstream of the distillation process. According to Kim and Dale (2002), system expansion is based on the assumption that function-equivalent production systems have equal environmental impacts; that is rarely the case. Furthermore, this approach requires highly accurate data and can be subject to a high degree of uncertainty and/or inaccu­racy; its implementation is difficult as the result depends significantly on the substitute that is chosen in the reference system. Finally, estimating the impact of this substitute may lead to another allocation problem. If "avoiding" allocation is not possible, then the ISO series recommends using a method that reflects the physical relationship between the environmental burdens and the coproducts. In that sense, allocation can be carried out by mass (wet or dry), carbon content, energy content or volume. Allocation on a weight basis relates products and coproducts using a physical property that is available and easy to interpret. But some researchers claim that it cannot be a good measure of energy functions (Malca and Freire, 2006; Shapouri et al., 2002). Energy allocation is mostly used in US biofuels studies by the US Department of Agriculture (Shapouri et al., 2002) and the Argonne National Laboratory (Wang, 2005). It is also the methodology chosen in the European Union (EC, 2009) and conse­quently applied by ADEME (2010). However, an objection can be made against this approach in the case where the coproducts are not meant for energy purposes. When physical properties are not appropriate, ISO recommends the use of other basis for allocation such as the economic value of the products. The rationale for economic allocation is that environmental burdens of a multifunctional process could be allocated according to the share on sales value, because demand is the main driving force of the production system. Price variation, subsidies, and mar­ket interferences could however imply difficulties in its implementation (Bergsma et al., 2006; Shapouri et al., 2002; Wang, 2005). In an LCA carried out in order to determine the net energy value (NEV) of bioethanol production, Shapouri et al. (2002) do not recommend this method because prices are determined for a number of market factors that are not related to the energy content. Gurnee et al (2004) state that in spite of considerable price fluctuations, the shares on the total sales value remain quite constant, particularly in the longer term. According to some researchers (Weidema, 2003), attributional LCA requires market allocation, while consequential applications require system expansion. Allocation by mass is applied in ADEME-DIREM-PWC

(2002) . System expansion (or substitution) is used in CONCAWE-EUCAR-JRC (2008), GM-LBST (2002), and VIEWLS (2005). The latter is also tested in ADEME (2010) by means of a sensitivity analysis. Kim and Dale (2002) investigated an expanded system including ethanol production from dry and wet milling, agricultural corn production, soybean oil and soybean meal from soybean milling as well as the urea production system for animal feed. Economic allocation was used in EMPA (2007a) and Gnansounou and Dauriat (2004). Elsayed et al.

(2003) used alternatively economic allocation and substitution, depending on the biofuel pathway considered and the availability of data.