Valérie Golaz

INED senior researcher Valérie Golaz studies interactions between population and environment from a spatiotemporal and demographic perspective.

(Interview conducted in April 2024)

Relations between population and environment are complex: they are both direct and indirect, and they unfold at different levels over varying timespans. In fact, much of current debate focuses on those different levels of observation, analysis, and action, which are sometimes conflated with each other. For example, global environmental changes such as climate change are tracked using international indicators. Yet those changes manifest locally, by way of isolated or occasional events affecting on-site groups, and they are managed by public policies defined at the regional or national level. Another example is air pollution, which can be a local phenomenon (due, for example, to leakage from an industrial site) or a phenomenon linked to activities practiced across the world but occurring in particular conditions and spaces (on road, highways, freeways, for example, or around certain crops). Yet pollutants also move through the atmosphere or into the ocean, thereby affecting regions or countries that did not produce them and may themselves be minimal polluters. In this way, pollution escapes national and international regulations. 

All of this means that it is not enough to study planetary issues by way of likewise planetary, or global, data. Yet determining connections between local or even national data is not satisfactory either if we want to analyze these phenomenon comprehensively. We therefore have to localize and contextualize the information, whether it concerns a population or the environment, as well as to account for larger spaces. It is here that the discipline of demography can be extremely useful, with its individual data on populations, such as those collected by censuses, surveys (ranging from harmonized international ones to strictly local ones) demographic tracking data; and administrative data. 

That’s a considerable issue. Yes, findings may contradict each other depending on the spatial or temporal scale chosen to measure interactions between population and environment. And that’s normal. Typically, long-term trends include short periods where the opposite trend is observed. Likewise, an indicator showing an average value for an entire country may show extreme values at a much smaller, local level. An example here is demographic growth: growth at the scale of a country at large is not the same as growth experienced throughout the local area an individual lives in. Locally, the population of some areas grows faster than the population of others—for example, urban growth is very strong in many Global South megacities­—while actually decreasing in specific contexts. The larger the unit studied, the less sharply we can see environmental and demographic heterogeneity.

But overly localized data are not very instructive either. The air an individual breathes is not only that of their principal residence but also of their place of work and any other places they spend time in. 

So how do you strike the right balance for the purpose of analysis? The answer clearly depends on the phenomenon studied, the research question, the data available, of course, and the scale at which they were defined and for which they are accessible. This applies to information on both population and environment. And it is here that compromises must be made sometimes.

For demographic data, the fact is that at precisely the moment that the technological possibilities of GPS tracking are enabling researchers to do extremely fine-grained spatial analyses, data producers have to restrict diffusion of that information to researchers in order to protect respondent anonymity. In many Global South countries, for example, national census data on individuals are only accessible at level 3 of the IPUMS-International administrative unit breakdown (see the IPUMS-International model); that is, the equivalent of the département level in France, or via geolocation of demographic survey sites that have been repositioned randomly within a 2 to 10 kilometer radius from the center of the surveyed areas (see the Demographic and Health Surveys model). 

Concerning the environment, many international databases exist today, e.g., local measurements, observation station data, satellite data. Since 2015, when France’s National Space Studies Center first developed Sentinel satellites, we have rapid-pace access to certain finely detailed data, e.g. data on land use and air quality. And with material from the European Space Agency and NASA we can go back further in time using smaller, more precise spatiotemporal breakdowns. For other environmental indicators—such as more specific analyses of the physical chemistry of soil, water, and air—we still need local measurements. Numerous international databases today are producing environmental indicators on areas that have been subdivided into squares as small as 1 x 1 km. These data can then be aggregated to calculate indicators corresponding to the desired administrative unit, or for any surface area of interest around inhabited areas, for example.

One of the main difficulties of quantitative analysis here is linking environmental and population data in relevant ways. In any case, the matching of physical and demographic data is governed by demographic data, the use of which is in turn governed by comparatively strict rules of ethics, thereby protecting respondents but limiting the scope of the analysis. The spatial scale at which an environmental measurement has meaning is often smaller than the one for which demographic data are accessible. More than ever, then, we need to be cautious in interpreting results. 

This assessment of the situation leads me to three conclusions. First, not all demographic data lends itself to linkage with environmental data. Second, we need to start designing demographic data collection projects with an eye to possible secondary analyses of how those data relate the environment. Third, fieldwork remains crucial as a complement to the increasingly abundant secondary data.