>

Migration in the United States

Where We Came From and Where We Went

"Where We Came From and Where We Went" and "Mapping Migration in the United States" are two linked visualization showing the migration of population in the United States between 1900 and 2012. Those visualization were realized in 2014 by Gregor Aisch, Robert Gebeloff and Kevin Quealy for the New York Times.

They present the data about the migration flow of population in the US in two different visualization styles: on the left, for each state, they represent where the people living in the given state come from; on the rigth, the same information is presented with a alluvial diagram moreover the user can switch between the incoming and outcoming migration flow.

We will see which kind of information can be extracted from both the graphs and which one conveys the infomation at the best.

How to use

What can you do?

In this section the features of the visualization are presented, highlighting the main controls.

The Mapping Migration in the United States shows a map of the US states, in each states there are the percentages of population who lived in the state at a given year. This is a particular useful visualization for people working in the demographic area and in general can be exploited by professionals who want to understand the flux of people among the US states. For instance, if a country as much more foreign citizen compared to the average then some political decision may take place to re-organize the state accordingly to its population.

Legend:

The map is colored accordingly to the legend shown on the left, in this way the states are grouped by their geographical position.

Zoom:

The map comes with a handy feature that allows the user to zoom to select a particular state (in this case Illinois), in so doing the names of states belonging to each category can be shown, moreover if the user mouse hover a given state a pop up gives additional information about the exact percentage of population selected and which his the year we're refering to.

Year selection:

The data can be shown in three different years and the one selected is in bold. What is immediately obvius here is that the gap between years is not constant, but we will analyze better the pros and cons of the visualization in the later sections.

The visualization with the alluvial diagram uses the same patterns. States are divided in categories accordingly to their position, and a pop up shows additional information. The x-axis represent the timeline and the y-axis the percentages, this configuration is particular useful to the user who want to extract the patterns abount the migration flows from one US state to another or more generally e.g. from the south states to the mid-west area.

The user can switch between:
  • the flow graph about the people which moved TO a state (e.g.California in this case)
  • the flow graph FROM that state to any of the other US states.

Finally one can explore the states either by clicking it on the little reference map or by selecting the state from the drop menu.

Data

The authors state very clearly where the data come from and which type of assumptions they made in order to aggregate the data, in particular they grouped foreign citizen in an overall 'Other' category and they also combined residents of states born in Puerto Rico, Guam, American Samoa, the U.S. Virgin Islands who in early Census years were identified as being born in Native American tribal territory into 'US others'.

The charts were compiled using Census microdata obtained from ipums.org at the University of Minnesota Population Center. The microdata are records containing the characteristics of individuals compiled from a representative sample of Census forms.

Critique

Things I like

Missing values

For Alaska and Hawaii the data relative to years before 1960 was missing. I really liked the way the authors dealt with this problem and I believe the resulting visualization was well studied.

In the map chart they put a sentence saying that data was missing for Alaska and Hawaii when the user selected 1900 or 1950; in the aluvial graph they simply started from 1960, what is noticeable is that they did not rescale the graph to make use of the all space, this lets the user compare even Alaska with other countries because they keep the same scale.

Although direct comparison is not really possible with the application.

Legend

The colors of the legend are meaningful, light, and divergent; the number of categories is 5 so the user has no problem identifying the clusters. This is exactly how the choice of categories should be done.

Alluvial chart

The aluvial chart to represent incoming and outcoming flow from a state is just RIGHT. It gives the user the possibility to extract trends over times, see patterns between different states and studying the variations in the composition of the population.

Things I do NOT like

Misleading dimensions

The biggest limitation of this visualization is the impossibility to compare states using the map view. The problem is that if we look at the states whose popularation is mostly made by native born[errata was:foreigns](grey) it seems like the 68% of texas is the highest percentage because it has the biggest circle, anyway Lousiana which is just on its side and has 79% of natives [errata was:foreign population] has actually a smaller circle.

This problem arises because the shapes are constrained to the state shape. This can greatly mislead the user the compares states. Given this, the map visualization is only useful if one wants to dig deeper in the composition of a certain state and,in case this is the ultimate goal, there is no reason to use a map as underling representation when it could be used a small map as geographical reference as they correctly did in the alluvial diagram.

Multiple graphs unnecessary

The authors decided to have an alluvial graph for each state. I think this was a really poor choice and it would have been much more usuful to be able to select a state and see the alluvial graph updating. Also I felt that being able to put side by side to graphs would have been a real useful feature to compare states but this was totally missing.

Uneven gap of years

Why one would ever split year like this? Moreover why did they limit the year selection to a set of three when they had the data for each year? The map graph was really a bad visualization overall and little useful.

Findings

FLORIDA

The florida chart for incoming citizens presents a couple of interesting trends. First of all the Florida population composed by migrants has been growing year after year. The graph should not mislead the user to think that the local population is less present than before but the real insight is that the percentage of population coming from out of Florida has grown a lot. An other trend can be seen between the composition before 1960 and after: before 1960 most of the population of Florida was made by Georgia migrants and in general the South states whilst after 1960 most of the migrants came from New York.

GEORGIA

Georgia shows an increasing percentage of migrants, mostly from the surrounding states. Alabama was the first source of migrants in Georgia around 1960 whilst know Florida is one of the greatest source, given the previous finding it seems like people are switching between Florida and Georgia.

WASHINGTON

This chart clearly shows a pattern: 83% of the people born in DC do migrate to other states. In particular Maryland seems to be one the favorite destination (36%), followed by Virginia (13%)

WAShington 2

If we have a look at the opposite chart (incoming) of Washington DC we can have a more complete understanding of the situation. The Washington graph is very interesting because the user can exploit the power of colors to get a sense of what is going on in that state. We can see that the district has always had a great influx of people, in particular in the last decade 63% of the population was not born in D.C. Moreover we can see that Virginia and Maryland used to be the greatest source of migrants before 1960 whilst now their contribute is no longer stronger with respect to other states. So, taken together the incoming and outcoming flux of migrants in DC it seems like DC is an extension of Maryland and Virgina where people go to live for a while and then come back to their original state. Anyway this is just an hypothesis which can be starting point to discover some interesting trend between those 3 US states.
It is worthed to notice that 1960 seems to be a recurrent meaningful year in all the graphs, infact it usually correspond to a switch in the trend of migration in the state.

Umberto Di Fabrizio

Master Student in Computer Science at Univerisity of Illinois at Chicago.
B.S. in Computer Engineering at Politecnico di Milano, Italy.