Sunday, March 24, 2013

Who will benefit from AR-TPD "cost-savings"?

Reflecting on the second NAC in-person meeting and a few brief discussions about the use of Administrative Records and Third Party Data (AR-TPD), I was reminded of an old saying, which I’ll paraphrase:

“There’s never been a time-saving device that’s created a minute of leisure.”**

My interpretation of that saying is this: Lots of technological advancements are advertised as doing menial work, so that we have more time for relaxing or doing more meaningful work.  But that’s rarely what actually happens.  For example, at my job, I have a computer and it’s frustratingly slow sometimes.  In those moments, I think, “Gah!  If only I had a faster computer, I could be done with this work faster!”  And that’s true.  But if I had a faster computer and finished my work faster, what would happen?  Would my boss say, “You finished that right quick, guess you’re done for the day!”  

Probably not.  

A faster computer wouldn’t create more leisure time for me -- it would just create more time for me to do what I already do: work.  Because that’s what “time-saving devices” are really for: getting us to do more work, not allowing us to do the same amount of work and then having some time off.  

And when I think about the use of Administrative Records and Third-Party Data for the Census 2020, I have similar concerns.  Bear with me here.

Congress has mandated that the Census Bureau produce Census 2020 for approximately the same cost as Census 2000.  That’s a major budget cut.  

So, the Census Bureau is looking for ways to cut costs and use what we might consider time-/labor-/money-saving devices.  But, as with all devices, there are pros and cons to using Administrative Records and Third-Party Data.  On the plus side, they almost certainly will be cheaper than the current ways that the Census Bureau tries to gather data from “hard-to-count” populations (e.g., people who the Census has to chase around because they don’t send in their Census forms right away).  But, on the major minus side, many AR-TPD strategies are also not so good at counting “hard-to-count” populations.  For example, most AR-TPD strategies produce more accurate data about White people than about People of Color -- and the Census is already not as good at counting some of those People of Color.  And that’s a problem.  

A few of us raised this point at the March 14-15 NAC Meeting.  A Census Bureau administrator told us that, while they know that AR-TPD has some problems, it could “save” them money and they could then use some of that savings to spend on strategies that are better at counting “hard-to-count” populations.  To try to do this argument justice, I’ll present it as a personal analogy:
I need to cut my expenses by $1,000 this year and that means I can’t do my job the way I have been.  Using Strategy X, I can reduce my costs by $800 -- so Strategy X could be a valuable part of my overall cost-reduction plan.  I know that Strategy X isn’t so good with some of my clients, but that’s okay, because with the money I save using Strategy X, I can pay to use a little bit of Strategy Y, which is more expensive but also more effective with those clients.  So, overall, using a lot of Strategy X could help me reduce costs and maybe allow me to pay for a little bit of a more expensive strategy, too.  Then the work still gets done AND I meet my new reduced budget.

Now, let me be clear, I believe that that administrator really does believe that AR-TPD will reduce costs and that they really do want to take some of that “savings” and spend it on more expensive and effective methods.  But when I heard the Bureau administrator offer that argument, I started thinking about that “time-saving devices” saying.  Because, yes, ideally, using AR-TPD will allow the Bureau to save money and then spend some of that savings on  more expensive & effective Non-Response Follow-Up (NRFU) methods.  But then, ideally, a faster computer would allow me to save time and then spend that time going for a walk or talking with a friend.  But that’s not what really happens.  Really, with a faster computer, I finish my work faster and my boss adjusts her expectations: she gives me more work and expects me to do it at the faster pace.  And that’s what I’m worried will happen with AR-TPD: it may allow the Bureau to do its work for less money, but that “savings” might not actually get spent on the NRFU methods that’d be needed to fix, say, racial disparities in the accuracy of the counts.  Because when you show that you can do your work faster or cheaper, you don’t get time off or the same budget: you get more work and/or a smaller budget next time.  And Congress has already cut the Census Bureau’s budget -- and I don’t imagine that trend is going to change much in the next decade or two.

So, one important question that remains to be answered and one important issue for folks to keep an eye on: 

If the Census Bureau uses AR-TPD, a strategy that produces some inequitable results, will enough of the cost “savings” be redirected to other NRFU strategies for “hard-to-count” populations, so that the Census can address the inequities in its counts?

The answer, I think, will not be determined by technology -- it will be determined by people’s ability to organize, mobilize, and ensure that those “savings” really do get reallocated, rather than simply being erased from a budget spreadsheet.

**For a bit of historical commentary on the idea, check this out.