In censuses, population counts are based on either the place where a person spent the previous night during the time of census enumeration (de facto) or the place where the person usually resides (de jure). The daytime population meanwhile refers to where people spend their time during regular working hours. Similarly, nighttime population refers to where people are located during night hours. By design, the census does not accurately capture or estimate the daytime population even though estimating it is highly important, particularly in urban areas. Daytime population estimates have several important uses. For instance, it can be crucial for efficient urban and transportation planning purposes. Additionally, it can be used to optimally plan for evacuation during disaster scenarios, as it is highly useful to have accurate data about the density of people in an urban area at various times of day.
There is little literature on methods for estimating the daytime population. This is partly due to the fact that for specific planning purposes such as those in the transport sector, the daytime population is in one way or another inferred from origin-destination (OD) matrices. Recent work by Boeing (2018) shows how to estimate the daytime population for the San Francisco Bay Area at the census tract level. He uses census commuting and population data to achieve this. Since Boeing's work is not comprehensive and does not evaluate the results of this approach, it is valuable to turn to older work by Schmitt (1956) on methods for estimating the daytime population. In the paper, Schmitt provides four statistical techniques to estimate the daytime population at the census tract level, including the proration (or prorata), censal ratio, correlation, and component methods.
When MPD is used as a non-demographic data source, it has been seen to have tremendous potential to provide useful data that matches with the correlation method and the component method. This is particularly the case in lower-income countries, where valuation data can be scarce and in some cases impossible to obtain. In such situations, MPD perhaps becomes the only viable source of non-demographic data to estimate the daytime population. In fact, the correlation method has been used in several studies either to estimate the population in order to provide a comprehensive overview of the use of mobile phones for estimating population density (e.g. Deville et al. (2014) or to evaluate how well MPD estimates populations.
In conclusion, in order to estimate the daytime population in low- and middle-income regions where valuation data based on taxes and housing may be scarce, MPD has huge potential. It can be used on its own to provide more frequent updates to the daytime population based on prior surveys. It can also potentially be used with remote sensing-based approaches. Furthermore, MPD can be used to generate OD-matrices which can provide estimates on the daytime population.
