Annoying Surveys as a Useful Government Platform

What do Deloitte consultants, presidential campaign interns, federal government economic policy analysts, and your grandmother all have in common (other than probably [hopefully?] voting for Hillary Clinton this November 8th)? Most of them, including grandma, have directly interacted with the American Communities Survey (ACS) since its inception in 2010 or in its prior form, the long form Census. Hundreds of millions more citizens across the country have been indirectly impacted by others use of the ACS.

So what exactly is the ACS? Taken literally, it’s a survey that looks like this that is mailed to randomly selection sample 3.5 million households in the U.S. each year, rolled out on a monthly basis. Recipients are legally required to complete the form accurately and to return it to the government. Once there the information is collated and made public through the Census bureau’s slightly clunky, but altogether not that bad, data portal.

The ACS provides access to all kinds of random information ranging from the obviously useful (median wages and education levels across geographies and demographic groups) to the obscure (did you know 0.2% of households in Middlesex County, Mass work in “Agriculture, forestry, fishing and hunting, and mining”?)

I’d argue that while the ACS at face value may appear to be a service carried out by the government, its better conceptualized as a platform that meets a variety of needs of a variety of customers far beyond its narrow scope in government and has spawned a small but growing ecosystem of applications built on top of it. Thinking about the ACS as a platform rather than as a service helps us better appreciate the value that is provides as well as justification to keep it running despite some political backlash.

Professor Eaves described a “platform” in class as “a group of technologies that are used as a base upon which other applications, processes, or technologies are developed.” Since its inception, various external parties have started to build applications on top of ACS that allow customization and further leveraging of the data.

One of the best examples I have seen is a tool built by a team from Deloitte, MIT, and Datawheel that provides searchable, well visualized ACS data products alone and in combination with other open data sources. I’m told by consulting classmates that it has become a first stop for market research on communities in a way that it never would have been locked in a government database. Other examples exist of applications built to leverage ACS and combine it with other data, such as ESRI’s ACS mapping products which are also targeted at private sector clients.

The nice thing about ACS being an open platform is that it allows any member of the public to access the data and built applications on top of it. This summer I did just that, working in politics conducting outreach to specific ethnic voting blocks. The ACS allowed me to build an application that combined county level demographic data with election data (from other government sources) to figure out where campaigns should be targeting resources. For example, which counties in Pennsylvania have the most Polish-American voters who may be receptive to certain foreign policy and economic messaging simply based on ACS data. My guess is that this was not the intended consequence of the ACS when it was first dreamed up, but has been a natural outcropping of the data being opened up on a platform.

Less surprisingly, the ACS also serves a lot of purposes in the day to day functioning of government — tracking local economic conditions, monitoring health insurance rates, and tracking education enrollment profiles, all of which matter in the management of large federal and state programs.

What’s interesting about the ACS, and a lot of other platforms in government, is that only the government is positioned to create such a platform. The private sector would have a very hard time replicating the ACS, but not because it doesn’t have survey capacity. Rather because it can’t legally oblige households to answer a survey — making the task of creating a near comprehensive dataset that much more expensive to collect.

Fitting well to the description of a platform, the ACS also embodies the tradeoff inherent in most platforms between control and efficiency / cost / customizability. By not collecting their own data, private firms may not get the exact demographic data or economic data that they would prefer to build into their applications and models. Researchers who use the demographic data may wish that more ethnic groups were default options (why is Samoan an explicit option on the ACS but not Bangladeshi?). Which questions and how soon they get published are key powers retained by the government that are undoubtedly shaped by strong political and administrative priorities.

However, these downsides are met by extreme savings in cost per use of the data and a standard framework that allows for applications to be built to merge the ACS with other data sources and applications. The ACS in this way is part of a broader open data system that is itself a platform. I’m looking forward to the class discussion on how open data platforms differ from other types of government platforms that we have discussed so far.

In the meantime, here’s to hoping that opponents in Congress don’t kill the ACS as it has tried to in the past in the name of liberty.