January 2022 - Marketing Systems Group

Is it possible to take traditional RDD cellular telephone sampling and make it better? The answer is yes, thanks to one of our newer products, Advanced Cellular Frame (ACF).

Advanced Cellular Frame is built upon the traditional RDD frame. It takes all possible telephone numbers in the RDD frame and adds something more. This makes for a much more versatile sampling tool, both for doing RDD and a targeted sample.

How is ACF Compiled?

Think of it as one database with two components inside. First, there’s the matched component. In the old days, this was known as “Listed.” Second, there’s “everything else”: the unmatched, unlisted, unassigned telephone numbers.

We take the original RDD frame, which includes every single thousand block that was dedicated to cellular servers. Next, we identify all the telephone numbers from a set of half a billion or so. We advance the frame by attaching as much ancillary data to the numbers as we can: names, addresses, individual demographics, household demographics, and geography.

Let’s say you’d like to do a targeted sample. For that, we would go into the database and fish out those telephone numbers matching the specific geographic and demographic criteria that you are targeting.

If you want to do an RDD, we go in and include everything within the geography you are sampling. All numbers have an equal probability of selection both from the listed (matched) component and the unlisted component.

KEEPING DATA FRESH

The database is updated quarterly. It’s true that with any database on the marketplace, there’s always going to be aging. There will be lag time between the vendor compiled data and loading it into a production environment. We compensate for the lag by sending selected telephone numbers out for a real-time name and an address append.

Because people tend to move over time for one reason or another, this method ensures that we are appending the most current information available in terms of names and addresses, for the sample we provide.

ADDRESSING THE MIGRATION PROBLEM

So, what happens when for example, a person in the listed portion (name and address) was geocoded, but that person actually moved? Many of these people will have carried their existing telephone number from one geography to the new one. Will Advanced Cellular Frame RDD move that number to the new frame?

Yes, the person will be identified based off the new address. That’s the beauty of the ACF frame. It does an excellent job at addressing migration. You can include for your target geography (like a state) all phone numbers from all area codes across the country of people who it so turns out have actually moved into that geography.

The converse is true as well. Anyone who has moved out of state will be excluded from the frame.

This improves your coverage and the quality of data collection and cuts down on collection costs.

HOW ACCURATE IS ACF?

Advanced Cellular Frame pulls on new technology to try to accurately link a telephone number to a name and address. It utilizes ID authentication, the technology used to validate transactions online. That information is used to help clean up and tighten up the ACF frame, which significantly boosts the matching accuracy.

WHAT ABOUT WORKING RATE?

You would expect a 75% to 80% overall working rate in ACF. That working rate jumps up to about 95% within the listed portion because there’s so much information known about those telephone numbers that they’re actually known to be working. The result is a much higher working rate than a traditional RDD frame.

LEARN MORE

To learn more about ACF, click here and check out the first video episode of our YouTube series “Coffee Quip” an informal series of information talks with a panel of MSG experts. In this episode, Hillary McDonough, Raj Bhai, and Greg Pizzola chat about Advanced Cellular Frame with subject matter expert David Malarek, Senior Vice President of sampling and database services.

In an ideal survey research world, it is preferable to work with a single probability-based sample as it provides the best representation of the target population. In the real world, however, cost and feasibility often prohibit the luxury of using purely probability-based samples. This is where different sampling methods come into play to reduce cost and improve feasibility, especially those that rely on online panels. All in all, online sampling isn’t ideal, since such samples are void of “organic” representation. If you can’t get generalizable results from your surveys, then what’s the point?

A blended (hybrid) sampling approach can offer an effective and practical alternative, through which multiple frames are used for sample selection—oftentimes a combination of probability-based and convenience samples from online (opt-in) panels. Further, we might start with a fully probability-based sample from a telephone or address frame, but then tap into online panels to supplement what we get from the main probability sample.

Taking a hybrid sampling approach sounds well and good, but just because you’ve gone hybrid doesn’t necessarily equate to unbiased survey results. Sampling from online panels is always a little tricky because if you don’t know what you’re doing, you can end up taking a seemingly inexpensive sample component, mix it with your precious probability-based sample and end up with a poor combination.

Sure, theoretically it’s preferable to have all or most of the samples be probability-based, but they are expensive. At the same time, you don’t want samples from opt-in panels dwarfing your precious probability-based sample. As a general rule of thumb, something on the order of no more than 50% of your sample should be coming from opt-in panels. Keep in mind that budget and other factors may dictate a higher or lower contribution.

The selection of samples from opt-in panels needs to be carried out sensibly. Equally important is the way you blend the probability and nonprobability-based sample components to produce a single database capable of producing reliable conclusions. It’s a little bit like chemistry when different materials are tossed into the mix to produce an alloy with higher-level properties; you have to be measured about it and get the ratios down just right using correct weighting and calibration adjustments.

As response rates continue to decline into single digit territory, even with fully probability-based samples, geodemographic weighting of survey data becomes essential. This is proven true since nonresponses are always different in nature. However, this issue will magnify with hybrid sampling when part of the sample may come from opt-in panels. Hence, in addition to basic weighting, additional calibration adjustments become necessary as well. This means going beyond geodemographics and applying corrections based on attitudinal and behavioral characteristics to ensure respondent representation for their population.

If you are looking to enhance your phone or address-based surveys and supplement them with samples from online panels, survey research scientists at MSG have decades of knowhow and hands-on experience to support your hybrid sampling methods. Our experts can assist you with sample selection, survey administration and questionnaire design, as well as state-of-the-art weighting and calibration procedures. Additionally, we can support you with reporting and analysis of data from complex surveys.

To learn more about our hybrid sampling products and services, click here, or contact one of our specialists.

For a deeper dive, watch Episode 06 of our Coffee Quip YouTube series, wherein the panelists discuss the intricacies and benefits of Hybrid Sampling!