Before delving into each source of change, it is important to discuss the limitations of our ability to understand what causes populations to grow. There are many factors that may influence each of these drivers of population growth. In studies of population growth, separating causes from effects can be challenging. When we observe that X and Y are correlated, it is possible that X causes Y, but also that Y causes X. Another possibility is that some third factor causes both X and Y. Finally, X and Y could just be coincidentally related in the data, having no causal relationship.[*] In short, for population growth, many factors are intertwined, and causation is difficult to isolate.
Consider places that have both well-funded government programs and high population growth rates. It may be the case that their robust government spending promotes population growth, but it also could be the case that the area’s healthy economy is the source of both the population growth and government revenues. There could also be causation running in both directions, where these factors mutually reinforce each other. Sometimes, factors can move in conflicting directions. For example, higher housing prices may deter in-migration, but they can also be evidence of high demand stemming from in-migration to an attractive location.
Economist Ardeshir Anjomani describes the difficulty in decomposing these “complex interrelationships.”[4] While careful researchers can leverage statistical and econometric techniques to peel back these layers, the multitude of interrelated factors and the constantly changing economic realities and individual preferences make causal claims tenuous.
Further, past trends may not continue. A 2014 paper by Glaeser, Ponzetto and Tobio notes, “[F]ew, if any, growth relationships remain constant” when looking at regional change in the United States.[5] As technology, preferences and tradeoffs change over time, so do people’s choices of where to live and how many children to have. Thus, relationships that have held in the past will not necessarily continue to do so.
[*] For more on these and similar issues, see: Will Koehrsen, “Lessons on How to Lie with Statistics” (Towards Data Science, July 28, 2019), href=https://perma.cc/ZVJ8-8HN4.