How Understanding Yard Signs Helps Us Understand Text Analysis

Filed Under: Best Practices, Market Research, Reporting, Tools & Techniques, Online Qualitative Research


Walt Dickie

The town I live in is about to have elections, and while walking my dog this weekend I noticed that yard signs were blooming with the early crocuses. Here is what you’d see if you traveled around my neighborhood:

  • Blue signs with the names of four people running for the village board on them, with the website of a local voters’ organization printed below. There are a LOT of these signs.
  • Orange signs with the name of one person who is running for a position on the high school board. There are a lot of these, too, though maybe not quite as many as the blue signs with the four village board candidates.
  • Red signs with the name of one person who is running for village treasurer. Again, there are a lot of these; I can’t tell if there are more or less than the number of orange signs.
  • Blue-and-red signs with the name of someone running for school trustee. Not as many of these compared to the orange or red signs.
  • White-and-blue signs with the name of someone running for village president.
  • White signs with green lettering bearing the name of a voters’ organization different from the one on the blue signs.

How should we analyze these yard signs? We could come up with a block-level sampling plan and count the frequencies of the various sign types. That would suggest certain kinds of findings – and some Presidential “polls” were conducted this way. We could also code what little text copy that appears on the signs, but that approach wouldn’t discriminate well, since what text is there is duplicated across many signs (for instance, the title or office the candidate is running for, or claims about “experience.”)

Those approaches wouldn’t tell us much about how the various signs relate to each other, and we need that answer to understand the political meanings reflected in the signs. And, that’s the key; meaning is found in the relationships among objects. We want to know how the signs go together with each other, not just how many of each kind there are. To find that, we need to know some correlations. (Because I didn’t have the time or resources to do an extensive count and statistical analysis, I took notes and made some qualitative observations instead.)

  • Almost any sign can be on the same lawn as any other sign, but the blue signs never appear on the same lawn as the white signs or the white-and-blue signs.
  • Orange signs and red signs often occur by themselves, without any other signs on the lawn.
  • Blue-and-red signs are occasionally found alone, but most of them are with other signs, often orange ones. Frequently, there aren’t any signs for village board positions on these lawns.
  • The village board signs (blue and white-and-blue) are often found alone, although the white-and-blue and white signs are together on many lawns that have no other signs.
  • There generally seem to be more yard signs on lawns that had Obama signs in the previous election (2008), but there were so many of those in my immediate neighborhood that it’s hard to tell.

Here’s what I think the signs are telling me about politics in my village:

  • The blue candidate for village president is part of a slate of candidates, which may represent a local political “party.” There seems to be no strong relationship between this party and the national Democratic and Republican parties.
  • The white-and-blue candidate running for village president is running against the presidential candidate of the blue slate for village board; that’s why the blue and white-and-blue signs are never on the same lawn.
  • The blue-slate party is identified with a local voters’ group, according to its sign. It’s likely that the white-sign voters’ group opposes the blue-slate group and backs the white-and-blue candidate, although no sign actually states this relationship.
  • There are at least two political contests going on in the Village. One is over seats on the Village Board and the other is over school issues. Some households are involved in both, others in one or the other, but there is no apparent party alignment between school candidates and village board candidates.
  • The orange and red candidates aren’t firmly allied with either of the local parties. They’re apparently ambitious individual campaigners.
  • The blue-and-red candidate and orange candidates for school positions may be allies.

How This Relates To Text Analysis:

Text analysis tools analyze texts the same way this example analyzes yard signs.

  • A text is made up of words, which are like the names and voters’ organizations on the yard signs.
  • Text analysis looks at how words co-occur the same way the example looked at what signs appeared together on lawns. Most tools represent this graphically as a network of associations of varying strength.
  • From these relationships, the tool makes inferences about the meaning of these clusters, which we can think of as themes; in this case, the political parties, issues, and alignments. The local “blue party” is a theme, as are the ideas of “school issues” and “village board issues.” We can easily imagine these themes being represented as circles encompassing the varying concepts and demonstrating their logical relationships across the network of co-occurring yard signs.
  • This is the same kind of inferences that humans make about relationships that are implied but never explicit. This sort of analysis is absolutely critical to understanding texts because normal language relies so heavily on the human ability to infer meaning.
  • This approach is diametrically opposed to traditional marketing research coding. Coding imposes an external set of analytic categories on data – it would code all the signs claiming that a candidate was “experienced” together. Text analysis can produce this result, but it does so by finding that theme within the data. In this example, text analysis would discover the “experienced” theme, but it would not be useful for sorting the signs.

We could push this approach further by wandering into other neighborhoods to see if we could link the themes that emerge in my village with the themes in others, or maybe we could use the linkages between these village election signs to the broader themes that were expressed on yard signs for the recent national elections. We could probably take in other signage (to stay in the visual realm) as expressed on bumper stickers and lapel pins: correlations forming clusters, which themselves correlate with other clusters, revealing higher-level clusters.

explore featured
Case studies

Hey, get our newsletter

join 5,000+ market research professionals
who “emerge smarter” with our insights