Childfind: Data and the 4YO children in Nashville (Part 1)

Well, this is going to be complicated.  

As I wrote at the outset of this project, one of the policy/planning questions that needs to be addressed is how many four-year-old children live in Davidson County? Generating a smart estimate has turned out to be more complex than I anticipated initially.

In short, no wonder nobody has tackled this process before!

Turns out, isn't really the responsibility of any one particular government entity to track children in this age cohort. Or, rather, lots of entities track some of these kiddos, but nobody tracks all of them. So I need to bring the relevant data together and build the model myself.

Initially, my plan was to (a) fit kindergarten enrollment numbers, by year and by school cluster, and then (b) use that model as a proxy for the number of five-year-old children who would have been eligible for four-year-old (universal, voluntary) pre-K had it been in place in the previous year. From this figure I could determine trend lines for PK need, and predict/forecast  But as I discovered, there are some characteristics of the data itself that will necessitate adjustments to the plan.

School Geography is a Cluster...

Before you scold me for the pun implied above, you need to know that MNPS is organized into twelve geographically delineated units called "clusters" -- a zoned high school, and its feeder elementary and middle schools. You can access the (updated as needed) cluster maps on the district website, but here is how the clusters were organized in AY1314(1):

Each of the clusters can include zoned and charter schools alike. If you open it in another window, you can zoom in, etc. The colors alone should give you a sense of how these clusters are jumbled across the county.


It is quite possible to generate reasonable estimates of the county-level(2) trends for numbers of young children (zero to five). But that doesn't help us get a sense of which clusters have the most un-met need for PK4. So I need to find a way to aggregate or dis-aggregate figures in such a way that we can look at cluster-level assets and un-met need.

So, here are the resources I am thinking about for this problem (i.e., datasets that I am thinking about using for this model).

Five Year Planning Data from the District's School Assignment Services office

Fortunately for me, MNPS has staff members who keep track of zoning and enrollment issues. In October 2012, their office published a five year projection for each school's enrollment trends. Definitely my first exposure to using these particular indicators -- such as new residential units, potential new residential lots, and undeveloped agricultural land. You can see the whole 200+ page report here on the district website.

So, this will be a very useful dataset... if I can obtain it in some raw form, from the folks in Student Assignment! I will ask nicely, since they seem very open to helping me.

Live births in Davidson County 

For those children born in the district(2), the county vital records division could theoretically tell me the number of live births for various time periods across the entire county. So, I could incorporate past historical information into the model, as compared to, say, enrollment at age six. But that's not everyone - families move all the time, within the county and in/out of the county. Plus, I need to be able (ultimately) to organize my data at a smaller unit of geography.

ACS Five-Year Dataset(s)

Call me boring, but my first instinct for population estimates is the good ol' Census Bureau. The American FactFinder search engine tells me that I need Table B14001 -- with the very shouty title "SCHOOL ENROLLMENT BY LEVEL OF SCHOOL FOR THE POPULATION 3 YEARS AND OVER". I can drill down by geographic level to "All Census Tracts in Davidson County TN" -- I leave it to you, Dear Reader, to navigate their query interface as you see fit! -- for the five-year American Community Survey. Ultimately I pulled estimates for five tables. Here's the (lazy) checklist of what was available for download (dl) from the five-year estimates:

ACS Five Year: Table Label & Title	ACS 2012	ACS 2011	ACS 2010	ACS 2009
B14001: School Enrollment By Level Of School For The Population 3 Years And Over	DL	DL	DL	DL
B14006: Poverty Status In The Past 12 Months By School Enrollment By Level Of School	DL	DL	DL	n/a
B14007: School Enrollment By Detailed Level Of School For The Population 3 Years And Up	DL	n/a	n/a	n/a
S1401: School Enrollment	DL	DL	DL	DL
DP02: Selected Social Characteristics In The United States	DL	DL	DL	n/a

I downloaded each of these tables as a zip file, and provide them here (zipped into one file) for your enjoyment. You're welcome, by the way.

MNPS Clusters and SABINS

So thus far I've set myself up pretty well to estimate county level estimates. But I still have not figured out how to align the Census tracts with the MNPS cluster geography. In short, I need to overlay these two non-aligned frames on top of each other, and see how they overlap (or don't).

Enter the miracle that is SABINS, the School Attendance Boundary Information System brought to you by the NSF, the College of William and Mary, and the Population Center (MPC) at the University of Minnesota. From their website:

The School Attendance Boundary Information System (SABINS) provides, free of charge, aggregate census data and GIS-compatible boundary files for school attendance areas, or school catchment areas, for selected areas in the United States for the 2009-10, 2010-11 and 2011-12 school years.

I honestly don't know how well-known this resource is , but damn if it isn't gonna be helpful in this process! More on the SABINS resource and the "layering" process in future posts.


Alright, nerds, what a project I've got in front of me! I'm off to clean and merge some files.



*** Notes ***
(1) That's Academic Year 2013-2014. I don't know where I learned that abbreviation but it makes a lot of sense.
(2) Metro Nashville Public Schools (MNPS) district is contiguous with Davidson County, the governance for which in turn is administered jointly (is that the technically correct way to describe it?) with the city governance of Nashville. Confusing, I agree.