Matthijs van Veelen and Roy van der Weide (2008) (henceforth VW) compare the axiomatic and economic approaches to choosing multilateral index numbers for international comparisons of real income. They are right to stress the potential importance of accounting for preference heterogeneity in general, and to warn against misinterpretations of the approach advocated by Neary (2004), in particular: the preferences of the reference consumer used in the comparison may not resemble those of any actual consumer in the countries compared. VW go on to claim, however, that the axiomatic approach can make meaningful comparisons without the assumption of homogeneous tastes. They may be correct, but they offer little in the way of concrete suggestions to back up their claim. By contrast, the economic approach offers a clear path to dealing with preference heterogeneity. First, at a conceptual level, it avoids the pitfalls associated with interpersonal comparisons of utility by proposing that a particular (actual or hypothetical) consumer be selected as reference, and asking how well off that consumer would be when faced with the prices in each of the countries to be compared. Second, at a practical level, the existence of a common utility function is testable (as discussed in Neary 2004, footnote 20). If a common utility function is acceptable, then searching for an appropriate parametric system of preferences (as done in Neary 2004) is valid. If a single utility function is rejected, then we can ask how many different utility functions are needed to rationalize the data, and partition the data into corresponding subgroups. Perfectly valid international comparisons can then be made within groups. Equally, if the data reject common preferences, it might still be reasonable to use the approach in Neary (2004) to proceed on an “as if” basis and to impose, as a reference consumer, those preferences that best fit the observed demand patterns. The economic approach therefore offers very promising scope for an intellectually coherent “Plan B” (VW, p.1722) as well as a “Plan A.” To illustrate this procedure, we reexamined the data in Neary (2004) using algorithms from Ian Crawford and Krishna Pendakur (2008). We found that the data on 11 commodity groups in 60 countries are rationalizable by a single non-homothetic utility function, but not by any homothetic utility function. This is good news for the approach followed in Neary (2004) in two ways: (a) it justifies estimating a single set of preferences using the pooled data; and (b) it shows that the usual justification for the EKS method (the main alternative to the Geary method, which underlies the Penn World Table) is shaky. Finally, if we take homotheticity seriously and ask how many different sets of homothetic preferences would be needed to rationalize the data for these 60 countries, we found that the answer is 18, although only 4 were required to explain 75 percent of the data; the rest are singletons.