Featured

Defining Small Area Geographies with Radius Sampling

In this blog we look at different ways to define small area geographies. How can we define a trade area or location using radius sampling (defining a geography around a particular point at different size ranges)? It sounds simple enough. You pick a point and draw a circle around the location. Done, right? Not exactly. The issue is how you identify geography within the circle. How precisely can we get at those addresses?

You can look at census blocks or block groups, or you can look at zip codes. It all depends on how large the radius is. Typically, we use census blocks for 5, 8, and 10 mile diameters, and because the blocks tend to fit well around the edges, we only need to do some minor carving.

With smaller radii, it becomes more of a challenge, though. Census blocks or block groups tend to fall way out of that range, or we get under coverage because not enough of the blocks or block groups fit inside the circle.

In episode 7 of our Coffee Quip video series, subject matter experts David Malarek (Senior Vice President, Sampling & Database Services) and Dennis Dalbey (Manager, Geodemographic Services) demonstrate the two basic ways to define radius geography using census blocks within a 10-mile radius—area inclusion using polygons and block centroids.

For example, we can start with a 10-mile radius around a point. Then we overlay census blocks (or block groups). On the overlay all block groups that intersect or have some relationship to the 10-mile radius can be seen. Where the color overlay falls outside the radius, we can decide which block groups to keep in or out of the sample frame.

The Polygonal Approach

By using polygon geometry, we can apply area inclusion. We can determine or make a cut on the percentage of the polygon area within the radii. For instance, any block group where the geographic area is at least 50% or more within the radius can be selected. When this is mapped, you will find holes along the edges of the circle where the geography once existed but was cut out because the inclusion was under 50%. Note that no matter how you carve the fringe, there will always be some overstatement of the sample frame or some under coverage depending on how well the block group geography fits the radius.

The Block Centroid Approach

Another way to do this is to use the center point of a polygon to assign geography to a radius. It’s a different type of geometry being put to use. Instead of using polygons where we’ll make an inclusion, we are actually using what’s considered the center of the polygon—the census block centroid. This method will allow us to select blocks that are theoretically 50 percent or more within the radii without having to apply an inclusion like we did with the polygon earlier. It is a more efficient way of doing it, but keep in mind that we are making a 50 percent cut across the board. You will encounter tradeoffs such as introducing under coverage or over coverage here, too.

One of the benefits of using the block centroid is that we can vary the distance. Let’s say we are not really sure whether the 10-mile radius is going to meet your quota. We may want to overstate the geography at 20 miles. With block centroid we can apply the distance and go from 10 to 15 to 20 miles until we meet the population or household quota. Note that this can be applied to polygons as well, but it is much easier to do with the block centroid.

The Address Level Approach

Yet another way to do this is to plot all the known addresses that fall within the blocks or block groups touching the circle and exclude any address that falls outside the circle. This is the most accurate way of getting accurate household counts and eliminates under and over coverage. It’s a two-step process and a more involved methodology but it is also the most accurate. One downside to this approach is demographic data is only available at some larger aggregation of geography such as block group and not by individual addresses.

Dave and Dennis compare the polygon and block centroid methods, and why they sometimes yield different results. Sometimes we actually need to plot ABS address locations in the blocks within the radius and remove the ones outside the radius, so we get a better fit for the overall target population. This is really the best methodology in terms of making sure that households are completely within a radius without having to worry so much about the regular geography, where part of it is in and part out.

Find more details and visualizations of these methods, in our Coffee Quips #7 video. It is the first in a series of geodemographic coffee quips. For additional information on our geodemographic services click here.

Featured

Key Advantages of Advanced Cellular Frame (ACF) Over Traditional RDD

In a previous blog we introduced you to one of Marketing Systems Group’s newer products, Advanced Cellular Frame (ACF). This time, we will discuss how the frame is compiled and the key advantages it brings you compared to traditional RDD by itself.

Let’s start with a few key facts. When we do a traditional RDD sample of a county or group of counties, the telephone numbers within that frame are defined using rate center geography. The RDD frame has zero bearing on your ability to place numbers geographically outside of a rate center geography. It’s just not built for that capability.

With ACF, we have a better ability to actually place numbers within the correct geography, because we have more known information about more of the numbers (approximately 45% of them).  

The Power of Split Frames

ACF can split an RDD frame into two pieces:

  • The listed / matched phone numbers we have information about for a particular geography
  • Everybody else, including unlisted / unmatched

The listed component we can put precisely within whatever target geography you’re doing, if it’s a ZIP Code, county or a CBSA.

The unlisted or unmatched numbers get pulled in using the old traditional rate center methodology, but we pull only the unlisted ones. All listed and matched numbers are excluded from this component.  

A County-level Example

To illustrate how this all works, check out our Coffee Quip videos, specifically episode #3 featuring Subject Matter Expert David Malarek. There you will see him explain a case involving Multnomah County Oregon, home to the city of Portland.

The ACF frame does a better job targeting an RDD sample within Multnomah county. Dave shows how you can take the listed and unlisted portions and create a split frame, in which you can sample the listed’s independent of the unlisted’s or unmatched.

The ACF RDD universe is about 1.5 million numbers. About one-third of them are listed. We know they are in Multnomah county. The balance of the records (993,000) are coming from the rate centers that best fit Multnomah county geography.

But remember, rate center geography does NOT conform to any census geography we are accustomed to. It doesn’t conform to counties or city boundaries. Rate center geography is really just based on where the telephone companies run their wire lines.

This means you will have some over coverage and some under coverage, because you could miss a spot within the county or flow into adjacent counties outside the county.

With a split frame, you get the two components of ACF RDD that best fit Multnomah county.

How We Address the Migration Problem with ACF

Now we hit the biggest advantage of ACF: how it addresses migration — people who have a cell phone in one part of the country then move to another part of the country but keep the old cell number tied to the old rate center geography.

For the “known” listed cellular telephone numbers, ACF allows us to identify people who have migrated from other parts of the country into Multnomah County, Oregon. Conversely, we can exclude from the frame everybody who moved out of Multnomah County.

If you had defined the frame by rate center and just did a traditional cellular RDD sample, a huge chunk of those listed numbers would actually be outside the county. And you would be excluding 27% of the numbers inside Multnomah county from your sample frame. Not good.

Simply put, ACF does a better job in terms of getting the people who really live in your target geography into the sample frame and leaving out the people who’ve left.  

What About Lower-level Geographies?

ACF is great for smaller geographies, especially city limits not easily defined by rate center, which tend to be fairly large. Rate centers might serve four or five dozen different communities; whereas with ACF (at least for the listed portion) you can pinpoint a town or city you want to sample and exclude the other towns around it.

You can see why we are so excited about Advanced Cellular Frame; it enriches the sample and empowers you to target smaller geographies with more precision.

For more details, check out Coffee Quip episode #3, with Subject Matter Expert David Malarek or click here.