Mapping Equality: Locating Racial Bias

| |

Mapping demographics reveals how important location is to a number of things. Where a person lives determines certain benefits and disadvantages. These include opportunities for education, careers, housing options, and sometimes even delivery services.

Amazon, the online retail giant, has gradually been rolling out same day delivery services in cities across the country. The determination of where to offer such services is predicated on a number of business factors Amazon keeps closely held: the number of subscribers to its Prime service in a certain zip code, the proximity of distribution centers, and the availability of drivers or courier services.

So when Bloomberg broke the article titled “Amazon Doesn’t Consider the Race of Its Customers. Should it?” I looked closely at the maps they had created, along with the facts about where service is and is not offered. After some study, I formulated some questions and came up with some answers.

Which Came First?

Amazon created its own maps as it does when researching expanding any service. Director of PR Scott Stanzel said in response to Bloomberg:

“There are a number of factors that go into determining where we can deliver same-day. Those include distance to the nearest fulfillment center, local demand in an area, numbers of Prime members in an area, as well as the ability of our various carrier partners to deliver up to 9:00 pm every single day, even Sunday. We will continue expanding our delivery capabilities and are adding more zip codes rapidly.”

In short, there is no evidence that Amazon used race at all in making their decision about where to offer same day delivery. In fact, Craig Berman, vice President of Global Communication said as much: “Demographics play no role in [our decisions]. Zero.”

Map comparing delivery area against demographics in Boston. Map Credit: Bloomberg 2016
Map comparing delivery area against demographics in Boston. Map Credit: Bloomberg, 2016

Basically, Amazon’s customer map came first, and then the racial data was added later by a third party. Rather than revealing a bias on the part of Amazon, instead it reveals facts:

  • The demographic of the Amazon Prime Customer Base
  • Delivery services in certain areas
  • The variety of local demand.

These are the same reasons Prime Same Day Delivery and even two hour delivery is not available in all areas.

Where does the data come from?

There are two sources of data used to develop the Bloomberg maps: only one of them is public. What is the relationship between these sets of data, and should Amazon look at it?

Customer Information and Big Data

The Amazon data comes primarily from their customer database, which of course they do not share. This is information voluntarily offered by customers, even though they are sometimes unaware of how it is being used. While some things about the demographic of Amazon customers can be assumed, at least in those cities where same day delivery is offered, it cannot be known for sure.

The data Amazon and other retailers are gathering is not just being used for geographic data, but also other customer spending habits. Almost all of them try to use predictive and prescriptive analytics as well as standard data analysis to not only anticipate what customers will buy, but determine how to influence that behavior.

Population Data

Bloomberg uses population data compiled from block group figures from the 2014 American Community Survey 5-Year estimates tables. They describe their methods as:

Table B03002—Hispanic or Latino Origin by Race—provides population figures by racial category, including the following subsets: white alone, black or African-American alone, Hispanic or Latino, Asian alone, and other races. The data were released on Dec. 3, 2015 and are the most recent local population data available from the ACS. All ACS figures are estimates with a 90% confidence interval and are subject to a margin of error. City-level figures presented in the graphics and charts are compilations of individual block group estimates, and share the same 90% confidence level.

This data is freely available for anyone with the capability to search it out. It is the most accurate population data currently available for these areas. So why didn’t Amazon use this data in their analysis of where to offer Prime Now services?

What Data is Relevant?

What role should racial demographics play in this type of business decision? The answer is none.Utilizing such data introduces a racial bias in one way or another. There are things that the Bloomberg article does not tell us about Amazon’s customer base and the availability of Prime delivery, primarily because we do not have all of the data Amazon used:

  • Isolated population data does not reveal the true demographic of Amazon Prime subscribers. Unless Amazon volunteers this information coupled with location data, we can’t accurately map it.
  • Population data does not reveal logistical or sales data. The decision to offer same day delivery likely hinged on the average dollar amount of customer purchases, and the cost and availability of delivery services to specific zip codes.
  • Data about current delivery areas does not account for expansion of services based on changing demand.

Amazon claims race was not a factor in their decision, and I am inclined to believe them. Retail is certainly not the only place where demographic data reveals racial discrepancies that appear to be tied to location.

Nursing Student Demographics. Credit: National League of Nursing
Nursing Student Demographics. Credit: National League of Nursing

In the nursing profession, less than 10% of Registered Nurses are men, and 75% are white, according to an article outlining demographics in nursing by Maryville University. This disparity persists even in areas where minority populations are the majority. There are still more disparities within nursing specialties.

Location and demographic data tell a part of this story as well, but leave out much information about educational opportunities, localized job markets, and changing trends.

Minority home ownership is on the rise, but again this data varies by location. While overall minority home ownership is on the rise and white home ownership has experienced a slight decline, the numbers fail to reveal a more localized picture about equality. The change in percentages correlate to population demographic shifts, but still not perfectly in all locations.

This is particularly true for African American buyers, While in part this has to do with an aging population, the declining African American homeownership rate also reflects a failure of policy and market trends to address the African American homeownership gap.

What does this mean for GIS technicians? We are often called upon to create mapping products of various types for different clients. When deciding what data to use, there are several things to consider.

  • The purpose of the map. Who is the map targeted to? What information do decision makers need? Information included in the map will vary depending on whether these are customers ready to buy or marketing departments studying demographic data to formulate strategy.
  • Bias that may be introduced. While it may be useful to the marketing department to have as much consumer data as possible to tailor advertising efforts to specific audiences, managers dealing with issues such as the Amazon example above may not want to be biased by information about race, sex, or age of those in specific geographic locations.
  • Research Budgets. Gathering data takes time, which costs money in the long run. The ideal situation or contract allows adequate time for thorough research, but in some cases there is only enough money and time to complete essential research.

Mapping demographics is important for a number of reasons. However, including it in some maps may introduce racial bias instead of removing it. Determining the source of data and the relevance, and what effect it will have on a project overall is vital to determining what to include.


Share this article

Enter your email to receive the weekly GIS Lounge newsletter: