While the sector population index is useful for intra-regional analysis - comparing sectors within the region - an aggregate measure is needed for comparisons between regions. This is the ``phase population index'', and it consists simply of the sum of the sector population index values for all of the sectors within a region pertaining to a particular chronological phase. As a measure derived from the sector population index, the phase population index inherits all of the problems and ambiguities of its parent. In addition, it suffers from the problem of sector contemporaneity. Sanders puts it thus:
A major methodological problem is chronology. Most researchers deal in blocks of time involving hundreds of years. Conceivably in the history of an area, there could have been a period in which settlements were less sedentary and village sites were frequently shifted, followed by a period in which people lived in more stable communities. In this case a simple calculation of the total amount of habitation area [as I use here] or number of ruined houses of the various sites from each period would not give an accurate picture even of the relative population size of the two periods. ([Sanders 1972]: 102)The percentage of sectors within a single given phase which are occupied simultaneously at any point will vary from phase to phase, and also between different types of sites within a single phase. In the present case it seems likely that a large percentage of the sectors were occupied simultaneously and continuously throughout the Formative and Tiwanaku periods on the Taraco Peninsula, since most of these sites were large villages, and they are few in number. In the Pacajes phases, however, most sites represent individual household compounds, probably occupied for no more than a generation or two, notwithstanding a small number of Pacajes villages, which were probably continuously occupied.
The phase population index as defined here - as a simple sum of sector population index values per phase - will therefore tend to exaggerate the contribution of small and ephemerally-occupied components of a total settlement system relative to sedentary and continuously-occupied components. In other words, farmsteads will count for relatively more than will villages, towns and cities. It will also tend to exaggerate the population estimates for phases characterized by predominantly transitory settlement vis a vis more settled or urban phases. In the present case, this means that the phase population index values of the post-Tiwanaku phases will be inflated relative to those of the Formative and Tiwanaku phases. These are caveats which must be remembered when using the phase population index for analytical purposes. These distortions may of course be corrected given accurate information on average occupation spans of different types of sites for a given phase. At the moment, however, we lack the fine-grained excavation data necessary for such calibration, and I will be forced to do without.