A decision as personal as moving house or emigrating from one city to another also depends on how many people did the same last year, according to a new study.
Demographic changes in large cities depend on millions of individual decisions, but the population evolves depending on two factors: what 'reminds' them of their recent past and the existence of other urban areas around them.
This is the proposal supported by a group of researchers through algorithms, which show how American cities have a 25-year-old memory and interact with others 200 km away while in the case of the Spain these values are 15 years and 80 km.
Knowing what each inhabitant in a city will do in the coming years, with their own motivations and feelings, is a practically impossible task; but if all the individual decisions are analysed together as a set some demographic patterns appear or a 'collective coherence' which can be predicted.
The team has developed some algorithms which reveal that what happens in a given moment in a city on a demographic level depends on what happened in previous years, as well as the presence of other large cities nearby.
"We can say that the urban systems have an inertia or memory of their past," said lead author, Alberto Hernando, from Lausanne Federal Institute of Technology, Switzerland.
"It may sound obvious, but this implies that a decision as personal as whether to move house or emigrate also depends on how many people did the same the year before independently, people that in reality you have never even met!" said Hernando.
The researchers have applied their algorithms to cities in Spain and the US.
In the first case they used demographic data from the Spanish National Institute for Statistics (INE) for the period 1900-2011, and in the latter case, registers from the US Census Office between 1830 and the year 2000.
The results show that Spanish cities with more than 10,000 inhabitants have a medium-term memory of 15 years.
The amount of people that in one particular year moved to another city is highly related to the figure that did so the previous year, but that correlation tails off as time goes by and after 15 years the correlation has dropped to half.
The finding was published in the Journal of The Royal Society Interface.
