Archives for posts with tag: intra-national disparity

(note: I haven’t written a blog in weeks because I’ve moved recently and have been working a great deal since then.)

One thing that is too often absent in the debate about immigration in developed countries has been the so-called “Brain Drain.” This term is used to describe the trend of mass emigration of skilled/highly educated workers (also called “human capital”) to a few high interest places. This trend has been documented both between countries and inside of countries. In the US, this has been described as “rural flight” while the international phenomenon continues to be called either “human capital flight” or simply the “brain drain.” Most developing countries see some form human capital flight, be it Ethiopia’s loss of doctors to Chicago or the high percentage of Chinese who study abroad and choose to stay abroad after graduating. While these countries lose human capital, many developed countries gain human capital as a result of this.

America and other developed countries benefit immensely from large amounts of highly educated individuals migrating to their shores:

(source)

The trend isn’t new, either. From the Chinese source we can see that even in the 1970s Chinese students were moving to the US with only a minority choosing to return. Similarly, the trend of educated individuals fleeing the countryside in America isn’t new; this map shows US migration from 1970 to 2000:

Something that’s always interested me when it comes to maps/graphs about the brain drain is they often break down either by country (which country is gaining and which is losing) or they only focus on recent statistics (who is moving where right now). Instead I would like to look at which parts of the US have the highest concentration of educated/skilled individuals. This is not intended to rank how intelligent parts of America are, or to malign the parts of the US where educational attainment is low. Instead I want to look at where the highest concentration of people with at least post-secondary education reside in the US. I decided to use the American Human Development Project in part because it allows you to measure these indicators by state but also by Congressional District (CDs). I think this is the best metric because US congressional districts have to be at least similar in population, only small states like South Dakota have wildly divergent congressional district populations. Of course there are other problems that can threaten congressional districts as units of measure: they can be gerrymandered to include disproportionate percentages of groups for political reasons. But while this could threaten the reliability of statistics on educational attainment, I don’t think it threatens the reliability of broader observations about educational attainment in the US.

Alright, now for my map and my methodology: I used a ranking of the top congressional districts in the US by percent of population with at least a Bachelor’s Degree. I made the threshold 32%, which produced 121 congressional districts. (You can find a table of this data here)

Many things can be noted from this so I will just list them first. The the states with the most CDs with at least 32% populations with a BA are: California with 18, New York with 13, and then Massachusetts with 8, several states then tie with 7. States not on the list include: Alaska, Arkansas, Delaware, Hawaii, Idaho, Iowa, Kentucky, Louisiana, Maine, Mississippi, Montana, N. Dakota, Nevada, New Mexico, Oklahoma, Rhode Island, S. Carolina, S. Dakota, West Virginia, and Wyoming.

What I found interesting was the heavy concentration in particular areas. Despite its smaller population, the Bay Area in California has nearly as many CDs (8) as the rest of the state (10). Massachusetts is similar in this regard: only two CDs in the state aren’t counted, the entire Boston Metropolitan Area is included. Massachusetts actually has more CDs on this list than Texas (with 3.8x as many people). Boston’s high concentration of respected universities is likely a factor in this, take a glance at this map I found on Radical Cartography of Boston:

Similarly, California’s Bay Area includes Standford University and UC Berkeley, which probably contributes to its high percentage of college graduates. But does my list adequately grasp which parts of the US have the highest concentration of skilled/educated workers? I decided that pruning this list down by making the criteria higher: what congressional districts have a 45% or higher population with at least a Bachelor’s Degree? lets see what we found (my updated list can be found here):

This pruning yields some interesting results. While many metropolitan areas were represented on the former map, this updated map excludes many large cities. only 4 CDs from southern states are listed, which makes them notable. The Research Triangle in North Carolina (a region where UNC, NC State, and Duke intersect) is represented, 2 CDs from the Atlanta metro area and one from Houston are included. The Boston-Washington DC corridor is well represented with 16 CDs, while California has 7. What’s notable about California is that when you raise the threshold to 45% it actually puts the Bay Area on top with 4 CDs, compared to the rest of the state’s 3.

Finally, I wanted to corroborate these findings with a final metric. We know now where the highest concentration of people with at least 4-year degrees are, but where do people who go beyond that tend to live? I took the 30 congressional districts with at least 18% or higher having a Master’s or Professional Degree (think medicine/law/engineering degrees). Here we find similar results with the previous graph:

Here we see that like in the other graph, the Boston-Washington corridor and California are leading the nation with 16 CDs and 7 CDs respectively. In addition to these dominant regions, Seattle, Houston, Chicago, Atlanta, and the NC Research Triangle appear to have to highest concentration of human capital.

Finding out where the largest concentration of human capital is a much easier question to answer than why that concentration exists in the first place. Here are some of my less informed ideas about these maps and why we find concentrations of human capital in a few places. One immediate assumption I had was that these elite regions must have the highest wages in the country and thus attract the most human capital as a result. But surprisingly, the top 30 districts didn’t directly correlate with higher wages. Comparing Income Index with % Master’s/professional degrees produced almost as many mismatches as it did hits. None of the Boston metro districts made it to the top 30 in Income Index and Seattle and 8 others were also not in the top 30 Income Index districts. I am by no means asserting that wages and education don’t correlate here, just that they don’t perfectly correlate this instance.

But some broad observations do appear to hold true: places with more than one elite university in near proximity tend to have more human capital (UC Berkeley+Stanford in the Bay Area, Boston’s plethora of top schools, The Research Triangle’s UNC+Duke, etc.). There also appears to be a based on type of industry for at least a few districts. Microsoft and Amazon.com are headquartered near Seattle and Boeing was headquartered there and still operates a large plant in the region. The importance of Silicon Valley cannot be understated: Facebook, Apple, Google, and Intel are all based there. Washington DC’s role as the epicenter of national political life and most federal agencies makes the region’s concentration of human capital almost inevitable (Imagine the human capital the CIA, FBI, NSA, DoD, DoS attract?). Finally, places of commerce like Houston and Chicago probably attract a great deal of human capital with a combination of incentives.

I think the question of critical importance for regions that lack these advantages in attracting human capital (be it a lack of quality universities or lucrative industries) is to attract human capital using alternative methods. According to one source, making housing attainable in the urban core where they say amenities/diversity/jobs/social life are usually concentrated. A compelling blog was written about attracting smart people to cities that rates cities by their potential and actual college graduates by sq/mile. Personally, I just hope that my city doesn’t get left behind in the zero-sum competition over human capital.

I found these cartograms from an article in the Telegraph and was immediately impressed. The cartograms originated here and use data from the Global Rural-Urban Mapping Project as to create the intriguing images. You can use the map in the previous link to look at any country’s population cartogram, here are a few that stood out to me:

First I would like to show three countries that had their national capitals moved from a heavily populated coastal city, to an inland location.

Istanbul’s historical significance cannot be understated. As Constantinople it was the seat of the Byzantine empire before becoming the capital of the Ottoman empire for centuries. But in 1923, after allies had occupied Istanbul at the close of WWI, newly independent Turkey moved its capital to Ankara.

Pakistan’s capital used to be in Karachi but was moved in 1960 to Islamabad. Perhaps this was to disperse the population of Pakistan more evenly, or to protect the government’s critical infrastructure from a naval attack.

For multiple centuries Rio de Janeiro was the capital of colonial and independent Brazil, until in 1960 when the capital was moved to the planned city of Brasilia.

Now I would like to look at examples of countries where populations tend to be focused in one place and are not evenly distributed.

Russia represents a interesting example of a nation having an East-West divide. Geographers often divide Russia along the Ural mountains, with the west often called “European Russia” and the east called “Asian Russia.” 78% of Russians live in the western part of Russia, in about a quarter of the country’s landmass.

Argentina’s population centers around its capital Buenos Aires and only small nothern cities like Mendoza and Cordoba figure.

This shows a population trend that’s been observed in the US (to a lesser degree): densely populated coastal cities and an “empty quarter” in the center. I say that this occurs to a lesser degree because the US has many large cities in the midwest and other inland locations:

Another great model for US population density comes from the Times article “Where we live.” The image was created by Joe Lertola:

The US has many big patches of population such as the Acela megaregion (AKA Bos-Wash) which includes everything between Boston and Washington DC. This appears to be the biggest patch, possibly followed by Southern California. Demographers and the like have tried to anticipate growth in the US by grouping large populations into megaregions like Bos-Wash and focusing on infrastructure and land reforms that accommodate these growth trends. One of the most obvious ones is the America 2050 initiative. This group puts out a map of what they consider to be the most important US megaregions in the next 50 years:

Here you can see that the US population isn’t quite as diffuse as the earlier cartogram would indicate. The same group estimates the population of Bos-Wash to be 49 million, or nearly a 5th of the total US population. Overall, this group says a majority of Americans live inside one these megaregions. By 2025 they predict that 75% of Americans will live in these megaregions. the regions themselves are loosely defined; for instance the “Texas Triangle” includes Oklahoma City despite it being nearly 200 miles away from Dallas or any other city in the megaregion. Similarly the “Front Range” region connects Albuquerque to Denver, a distance of 330 miles separates them. If the distance between large populations inside of these megaregions seems daunting, the distances between megaregions is an entirely different beast. Especially in the western US, megaregions are spread very far apart. Separating the “Front Range” from “Cascadia” (my megaregion : ) ) is nearly a thousand miles of mountainous frontier. Maybe the lesson from this map is that the US should focus more on connecting megaregions within themselves instead of paying to connect them with one another. This idea was actually adopted by Obama when he announced plans to build several high speed rail lines in the US. One of the proposed lines would connect Vancouver, BC to Eugene, OR with a high speed rail.

I think its important to note that models showing population trends can help convey the trends that might not be apparent to the casual observer. Once one sees how the population of a country is placed, one can start to ask how it affects policy regarding those trends.

When I was getting vaccinated for a trip to the Middle East, I was surprised to find the doctor asking me which part of Turkey I’d be visiting. She pulled out a map that looked something like this:

I was shocked because Turkey had to be the most developed country we would visit on that trip; how could they the only one where Malaria shots are necessary in certain parts. I felt even more shocked after I visited Istanbul, where its level of development seemed otherworldly compared to Damascus, Amman, Beirut, or even Jerusalem. They have a transit system that is efficient, people sort of obeyed traffic rules, and everything looked much better maintained. It felt like a European city, while the others felt like something else. I marveled at Istanbul’s unique mix of secularism  and development. Yet apparently for someone living in Diyarbakir, Malaria is a part of life.

What’s more startling is that Turkey’s HDI for 2010 is 0.679, behind Jordan and Tunisia and not far ahead of Algeria [source]. How could a city that seemed so modern be in a country less developed than resource-starved Jordan, who has some 13 miles of coastline and a mostly desertous landscape. Jordan’s GDP (PPP) per head in 2010 was $5,400 while Turkey’s is more than twice that at $12,300 [source]. There has been research from multiple good sources on the matter:

A great research publication called “Regional Disparities and Territorial Indicators in Turkey: Socio-Economic  Development Index (SEDI)” written by Metin ÖZASLAN, Bülent DINCER, and Hüseyin ÖZGÜR (found here) delves into this question with depth and authority I can’t match, so I’m going to just post some of their findings and briefly summarize them. They use 58 different indicators from myriad sources to measure provincial development and collate them into one index called the SEDI. Unfortunately this means that, like the HDI value from measureofamerica.org we cannot compare these values directly to other countries. Fortunately they do go into great detail in the article on their methodology and it appears to check out. Time for some cool maps thanks to this article.

Many things stand out. Most of Turkey’s most developed regions are in the western part of Turkey, with the lowest SEDI scored provinces all being in the east.Four Cities+suburbs stand out as the most developed provinces in Turkey: Istanbul, Izmir, Ankara, and Bursa. These provinces (and a province that includes suburbs of from Istanbul) have a combined population of 26.14 million, according to Turkstat. This means that of Turkey’s 73 million people, just over a 3rd live in these five most developed provinces [source]. This shouldn’t seem so troubling, when you look at the US (as we did in the previous blog) you can see its not so uncommon for states (especially ones with big cities) to score higher on development indicators. The problem with Turkey is it’s development curve among provinces is much steeper than the disparity among US states. A graph from the same source illustrates this effectively:

Istanbul’s score dwarfs the others, its more than four times that of 6th ranked Eskişehir province. The top five provinces themselves dwarf the remainder of the provinces.

Here is a map of the geographical regions of Turkey using the same source. I edited it to show which regions have above average and below average scores (note: mediterranean is almost perfectly at the average):

Now here is an unedited graph showing the regional SEDI scores from the same source:

This paints a picture of Turkey having three regions that this source claims, drive most economic growth in Turkey, with Marmara far outpacing the rest of the country. Meanwhile the two easternmost regions of Turkey experience the least amount of growth or development. This doesn’t perfectly coincide with the Malaria map I showed earlier, but I suspect that map was geared towards ease of use and probably wanted to include the entire southern border region to aid tourists traveling by land.

The next article I am going to use comes from the World Bank, titled “Turkey: Country Economic Memorandum Volume I – Main Report” it can be found here and the section I will cite begins on chapter 6, page 29 (41 in adobe). This article compares regional GDP per head variation among European countries and shows its findings in this graph:

Turkey ties with Belgium, a country known for its Flemish/Wallonian divide (a north/south divide in this case). I am surprised by the other results in this publication, as I had assumed Italy’s regional GDP variation would exceed the UK’s.

This article points out that from 1980 to 2000 Turkey’s regional disparity has either increased (gotten worse) or stayed the same. It points out that while Industrial activity has expanded in the Western half of Turkey, the East remains primarily employed by agriculture, and that hours/employee are considerably lower in the East. Finally, the article gives some explanations for why Turkey’s institutions might inhibit growth in the Eastern regions; it states that Turkey’s centralized planning and allocation of resources for things like infrastructure and public works projects gives local officials few options to raise their provinces from poverty and underemployment.

A few quick statistics can be found in this report from the European Commission titled “Second report on economic and social cohesion: Regional Features in Turkey” found here. It states:

“between east and west: two-thirds of the population were concentrated in the west of the country in half the land area, accounting for 82% of national GDP, and with GDP per head 23% above the national average (41% of the EU average). In the east, GDP per head was 53% of the national average, much the same as 10 years earlier”

One word that is missing from all of these articles is “Kurd” which is surprising because Kurds make up the largest ethnic minority in Turkey with 15 million living there, most of them are located in the Eastern part of Turkey. Here is a map I found from the University of Texas here that shows where most Kurds live:,

I found another, more recent map here that looks at recent elections results in Turkey in 2011:


It’s worth noting that Turkey has a unique 10% electoral threshold that prevents most Kurdish interest parties from electing members into parliament; the easiest way to circumvent this rule is to run candidates as independents.

There seems to be a strong correlation (using these two maps and the first map) between Kurds and low development. I am not trying to imply that Kurds don’t work as hard, but simply pointing out that like Appalachia and the Mississippi Embayment, the Kurdish region of Turkey appears to lag behind the rest of Turkey. Something I would be very interested in seeing is how a partition of Turkey that removed part Eastern Turkey from the rest would effect the HDI value Turkey currently enjoys. Using the 3 sources from the beginning of this article, it seems clear that Turkey’s Western half would benefit (at least in its HDI score) if its indicators were measured separately from the Eastern/Kurdish part. Of course the political ramifications of such an outcome would be significant. I’ll leave that debate for the citizens of Turkey, be it Kurds or Turks.

Unfortunately this isn’t a subject where I can compare countries and their regions to other countries. Instead, here is a selection of articles and related maps that deal with the problem of uneven growth across various counties. I want to stress that these graphs use different measurements from different time periods and thus cannot be compared with each other. I will provide links for the maps I use and give a brief summary of the research that corresponds with them.

The first example I would like to present is from a familiar source, the USA. here we can use a brilliant website designed by social/political scientists to display a variety of statistics relating to US development. using measureofamerica.org you can access an HDI map of the US that is divisible to the congressional district level. here is a state level map of the US using their latest dataset.

I got this map here it covers most US cultural regions that I accept with its omission of Appalachia NOTWITHSTANDING.

look at measure of america. wow here is the HDI of the US

this website is really interesting.

Look at what happens when you gauge obesity and diabetes in the USA:

This shows a tendency of Appalachia and southern US states to be have comparatively bad health. The West and New England do well here. How about Diabetes?

between these two maps, the unhealthiness of Appalachia stays strong. What are they doing wrong?

Diane Sawyer has some ideas about this

here are some other maps from Appalachian Regional Commission 

This shows college completion rates in the region and compares it to the US average.

Another troubled area represented on the HDI map is the US South. Breaking the region down by Congressional District allows us to look closer at regional disparity by showing disparity inside states.

Using the HDI data from our earlier source, lets see what the bottom 100 US Congressional districts look like:

It’s important to note that the bottom 100 districts is an arbitrary measure and many districts with similar HDI values were excluded. Nonetheless it includes the important bottom quintile with about 10 districts from the next lowest quintile. It also provides us with nearly a quarter of the 437 US congressional districts so I went with it. This map took a long time to make so please feel free to verify my findings here

In many states there are examples of urban poverty as well as rural/agrarian poverty being represented. In my region (Pacific Northwest) the eastern districts in Oregon and Washington are examples of rural poverty. NYC provides an intriguing example of urban poverty. The district NY-16 is one of the lowest HDI scores in the US, it sits nearly adjacent to NY-14, the district with the highest score in the US. The difference in the scores (8.79 vs 3.20) shows how geography can mean little when defining a region’s development.

But while NY-14 sticks out, it pales in comparison to the overwhelming poverty of the US South+Appalachia. To corroborate my view of the US South look to this wikipedia page

Of the 100 lowest HDI scores, this combined region contributes 59 districts (59%). When you count the bottom 50 this region contributes 30 (60%).

But when you count only the bottom 25 you get a staggering 20 Southern+Appalachian districts or 80% of the bottom 25. This map illustrates the disparity:


  Here’s a chart of the 100 lowest HDI scores and my Southern+Appalachian selections in red:

California

20

2.60

Kentucky   

5

2.82

West Virgini

3

3.16

New York

16

3.20

Texas

29

3.23

Missouri   

8

3.24

Oklahoma   

2

3.33

Mississippi

2

3.34

Alabama    

4

3.37

Arkansas   

1

3.39

Alabama    

7

3.46

Kentucky   

1

3.50

Tennessee  

4

3.50

Virginia   

9

3.50

Arkansas   

4

3.50

South Caroli

6

3.52

Louisiana  

5

3.52

North Caroli

1

3.53

Georgia    

2

3.55

Alabama    

3

3.61

Georgia    

12

3.66

Louisiana  

2

3.68

Tennessee  

8

3.69

California

34

3.69

Arizona

4

3.70

California

18

3.73

Texas

15

3.74

California

31

3.78

Texas

28

3.78

California

43

3.80

Illinois

4

3.80

Tennessee  

1

3.81

Pennsylvania

1

3.86

Florida    

3

3.86

Louisiana  

7

3.87

Texas

27

3.88

Texas      

1

3.89

Texas

30

3.90

Texas

13

3.92

Texas

20

3.92

Georgia    

1

3.93

Louisiana  

3

3.94

Michigan

13

3.95

Alabama    

2

3.95

Oklahoma

3

3.95

New Mexico

2

3.95

Ohio       

18

3.98

Louisiana  

4

3.99

Texas

9

3.99

Mississippi

1

3.99

Texas

19

4.01

Ohio       

6

4.04

Mississippi

4

4.04

Georgia    

8

4.06

Arkansas   

3

4.06

South Caroli

5

4.07

Alabama    

1

4.07

Tennessee  

9

4.08

Missouri   

4

4.09

North Caroli

7

4.09

Texas

18

4.10

Texas

11

4.10

California

47

4.11

California

2

4.11

Michigan

14

4.13

California

21

4.13

North Caroli

10

4.13

Tennessee  

3

4.13

North Caroli

2

4.14

Michigan

1

4.15

West Virgini

1

4.15

West Virgini

2

4.16

Texas

17

4.17

South Caroli

3

4.19

North Caroli

3

4.20

Texas      

5

4.20

Pennsylvania

12

4.22

Indiana

7

4.22

Missouri   

7

4.22

Mississippi

3

4.23

Washington

4

4.24

Virginia   

3

4.24

Oregon

2

4.26

Arizona

1

4.26

Nevada

1

4.26

Florida    

1

4.27

Kentucky   

2

4.27

Virginia   

5

4.27

Ohio

17

4.27

Oklahoma

4

4.28

New York

23

4.29

Pennsylvania

9

4.29

Indiana

8

4.30

Tennessee  

6

4.30

Texas

16

4.32

Illinois

17

4.32

Georgia    

9

4.32

Florida

23

4.32

North Caroli

8

4.34

North Caroli

11

4.34

Feel free to disagree with my assessment of the South or of Appalachia. This was excluding a large number of districts that seem to have an inconclusive regional definition. For example, Florida-23 straddles the Miami metro region and wasn’t included on the list. I only included two districts in Texas (CD1 and CD5) because the rest had mixed definitions for culture; Oklahoma-2 was the only district included. The northern district of OH-17 was excluded because only part of it is included in the ARC regional map. Missouri’s regional definition produced conflicting results but the southern districts of MO-8, MO-7, and MO-4 appear to reliably count as “southern.” The rest are 100% Southern and/or Appalachian. I wouldn’t have included Northern Virginia but none of their districts had a low HDI score so it didn’t matter.