Environmental Justice and Your Block

We want­ed to cre­ate a visu­al­iza­tion that shows how income varies by block to facil­i­tate com­mu­ni­ty envi­ron­men­tal jus­tice orga­niz­ing. In some neigh­bor­hoods there are major dif­fer­ences between blocks, and these dif­fer­ences are used by cor­po­ra­tions to decide what blocks to tar­get with tox­ic facilities.

So we recent­ly fin­ished a high­ly exper­i­men­tal lay­er that esti­mates medi­an house­hold income by block. This could be used as an Envi­ron­men­tal Jus­tice Index. It is com­pa­ra­ble to the EPA’s EJScreen’s EJ index. Both are esti­mates, but this block income index has a high­er geo­graph­i­cal res­o­lu­tion, and may also be more accu­rate (as the block group data that they use has very large con­fi­dence inter­vals). In the US there are 6 mil­lion pop­u­lat­ed blocks and 200,000 block groups.

The Cen­sus Bureau does an annu­al sur­vey, the Amer­i­can Com­mu­ni­ty Sur­vey, which is less accu­rate than the decen­ni­al cen­sus as it only sam­ples 1% of the pop­u­la­tion. So for income data the best res­o­lu­tion you can get is the cen­sus tract — or typ­i­cal­ly 4000 peo­ple. And for that res­o­lu­tion, due to the small sam­ple, the 90% con­fi­dence inter­val is often very sub­stan­tial — often plus or minus 20%. So most high res­o­lu­tion maps of medi­an house­hold income will use cen­sus tracts.

How­ev­er, the Cen­sus col­lects race data from every­one and that is avail­able at the cen­sus block res­o­lu­tion! Cen­sus blocks typ­i­cal­ly have 1 to 200 people.

So we cre­at­ed a visu­al­iza­tion that com­bines the cen­sus tract income with the racial demo­graph­ics of each block. We then apply a smooth­ing algo­rithm to it. So each block is influ­enced by the sur­round­ing blocks.

The result is high­ly exper­i­men­tal. It should be seen as more of a visu­al tool than an esti­ma­tion. It has error from the cen­sus tract income con­fi­dence inter­val, using the lat­est cen­sus block race data from 2010 (US pop­u­la­tion is up 5% since 2010), assump­tions regard­ing the medi­an house­hold income by race, and the smooth­ing algo­rithm. How­ev­er, I’m esti­mat­ing that this is more accu­rate than oth­er mea­sures and that peo­ple will often want to visu­al­ize block by block differences.

How accu­rate does this look for your neighborhood?