I would insist that social stratification systems are far more complex than mere income categorizations. Not only financial capital but also social capital (positioning within a social network) and human capital (education, health) influence an individual's abilitiy to access resources and opportunities. Taking income alone could be misleading since a country's income distribution might be dependent on very different social structures thant those of another country. Altough outcomes are certainly comparable (earnings per day for example), such comparisons may hide important structural differences between societies. A $3 earner in Malawi may have very different capabilities than a $3 dollar earner in Syria or Cuba. Hence, I feel the criteria to categorize the middle-classes should be tailored to the type of research that is to be pursued (provided that the limitations of such categorizations are understood and duly noted). Cross-country comparisons would of course demand a much simpler method (perhaps related to the aforementioned three), whereas case studies may be able to afford the luxury of a more holistic method of a higher level complexity (where capabilities of different sort, not only income, can be taken into account).