Imagination and Analysis: Refining a research question (part 2)

In the previous post, I was trying to illustrate how I approached a question, trying to detail my ideas and the process through which my thinking developed.  The purpose was to highlight the importance of using one’s own imagination and judgement in developing a research question. Many graduate students with whom I have worked, have struggled because they didn’t develop their own ideas, but rather tried to follow the ideas of other people. To contribute to research in a field, however, requires the confidence to challenge old ideas and develop or refine theories. As I have argued elsewhere, it doesn’t take vast brilliance, just willingness to use natural basic abilities with care and attention to detail.  The example of reasoning presented in these two posts is offered as an illustration of the way someone (me) who is naive in a subject (politics), can develop a complex, detailed range of issues that are potential subjects of research as I try to answer a fairly simple question (simple in the sense that it came from my initial, offhand intuition/curiosity, not any detailed analysis), and complexity arose as I imagined possibilities.

Wild imagining

One intellectual exercise for a scholar is to try to imagine what other possibilities could be added to those lists of possible issues affecting voting results. In this area, imagination is the key factor. This imagination of possibilities could also be called “generation of hypotheses,” if I wanted to frame it in more formal (and perhaps intimidating) terms.

It can be useful to suspend reason and logic to help imagination flow: for example, I can imagine that maybe ballots are being improperly counted because extra-terrestrials are tampering with the ballots.  That’s kind of ridiculous, but it is an explanation, and maybe there are even a few people who would believe it. Maybe there are some Satan-worshipping pedophiles who are working to actively miscount. That also seems ridiculous, though in this case, it seems likely that many people would believe it. Or maybe the election system has been tampered with by the Russians.  This seems less ridiculous, given what is known about Russian cyberactivity, or, for that matter, hostile cyberactivity from a range of sources.  To research the possibility of cyberattacks affecting the vote counts, we would want to understand Georgia’s election protocols: at what point could hostile actors affect the vote counts? Does Georgia use electronic voting machines where individual votes can be changed (e.g., I click on “Ossoff,” but the machine registers “Loeffler”)? Does Georgia have a centralized computer tally that could be hacked and altered? Diving deeper, we could ask about technologies used by different groups of potential attackers—maybe Russian cyberattacks use different techniques than Chinese cyberattacks. (And maybe there’s a nefarious actor who pretends to be the friend of the US but also carries out cyberattacks?) Not only could we dive down those technical holes of whether and how the voting system could be compromised, we also are led to ask why: why would someone want to help Perdue but not Loeffler, or help Warnock but not Ossoff?

Anyway, my general point is that as I ask questions—sometimes ridiculous questions—it’s possible for new ideas to arise that might be worth some investigation. Generating ideas for what could be researched—research hypotheses—is an act of the imagination. Therefore it’s useful for imagination to operate freely, to  be able to propose the absurd as well as the “reasonable,” in order to generate hypotheses for investigation.

Yet more possibilities

So far, all the questions I have been asking were generated from looking for an explanation that allowed me to retain the assumption that people would not split-tickets in this election.  Once I start to entertain the notion that people might split tickets, a variety of new questions arise: who would do so, and why? (Also note that these considerations do not rule out any of the previous considerations—in addition to ballot errors and tampering, people could also split their ticket—the observed vote totals could be influenced by all of these factors.)

And just in asking this, I realize that there are multiple ways to “split” a ticket: you could vote for one R and one D, or you could vote for one R (or one D) but not vote at all for the other, or you could vote for one R (or one D) and a third-party candidate (well, actually not in this election because it was a run-off with only two candidates, but if that weren’t true, you could get people saying, e.g., “I’m not voting for Ossoff because he’s not progressive enough, so I voted Green party”).

Now, again, it’s necessary to start to use imagination: why would people split the ticket and cross party lines? Maybe:

  • 1. Democratic women (or feminists) crossed party lines to vote for Loeffler (a woman).
  • 2. Black Republicans (or anti-racist Republicans) crossed party lines to vote for Warnock.
  • 3. Democrats who voted for Warnock didn’t vote for Ossoff because
    • he’s not progressive enough
    • he’s a white man
    • he’s too young
    • he holds a specific position on a specific issue to which they object (I don’t know a ton about his campaign, so…)
    • he was the kid they hated most in elementary school (I’m reaching for the absurd here—we wouldn’t expect this kind of explanation to affect large numbers of people, but it is a possible, if silly, reason that someone might choose not to vote for Ossoff—again, I’m exercising my imagination)
  • 4. Republicans who voted for Perdue didn’t vote for Loeffler because
    • she’s too much/too little like Trump
    • they didn’t like her for some position on some specific issue.

Again, these lists are probably not exhaustive—there are probably many other reasons that Dems might vote for Warnock but not Ossoff (and vice versa) and that Reps might vote for Perdue but not Loefller (and vice versa).

How does imagination match up with the real world?

As I start to lay out these different possibilities, it raises questions of how these hypotheses might be reflected in the data. 

If, for example, the differences are caused by damaged or incomplete ballots, what kind of data patterns would we see? To answer this question, we can look for old empirical data: what does previous election data show, with respect to damaged/incomplete ballots? Given the standards set by the historical data, we could compare to see if the damaged/incomplete data would predict the data that we’re seeing—would we see the kinds of discrepancies we see, on the basis of that kind of problem? Have past elections had enough damaged ballots that we could see the differences that we see in this election?  Alternatively, we can use imagination—what would we expect if there were a lot of damaged ballots? Would we expect them the ballot errors to be distributed evenly across all candidates (i.e., the number of votes lost by Warnock due to damaged ballots would be equal to the number of ballots lost by Ossoff)? What if damaged ballots were coming from one specific location, because of a damaged machine, perhaps, and perhaps that machine was damaged so it failed to read both elections and only read one of the two?  (We would expect an error like this to be quickly discovered in checking ballots—someone at the precinct would notice that they weren’t getting votes for a single candidate.)

Even simple questions get complicated very quickly

I’ve gone through all this detail to show 1) how imagination plays a key role in finding hypotheses for research, and 2) how quickly a question can branch out into many questions—even this brief, informal analysis identified a number of different concerns that could lead to further research. I didn’t even begin to ask questions about how I might gather any supplemental data that could support inferences about the vote totals.

My final steps with the voting question

I’m not researching the question of why Warnock got more votes in any formal way. It was mostly a passing curiosity, but I wouldn’t be able to the put an answer to use in any way. So I didn’t go far, but I’m going to briefly mention my final steps in my “research,” just to give an angle on yet more details that crop up in research.

Before I decided to start working on this blog post, here’s what I did: I compared the number of votes received by Warnock and Ossoff—Warnock received about 19,000 more votes when I looked. And I compared the difference between the votes received by Perdue and Loeffler—Perdue had about 19,000 more.  This similarity of numbers was highly suggestive of people splitting their ticket because each person who splits their ticket, voting for Warnock and Perdue, adds one to both their totals and takes one away from Ossoff and Loeffler—a mirroring.  The similarity in numbers could be coincidence, of course (it would require further analysis to study), but it is suggestive of a group of about 19,000 people who split their ticket, voting D for Warnock and R for Perdue.  If the difference was caused by errors in reading or filling ballots, or by people voting for Warnock while leaving the other vote blank, we wouldn’t expect that mirroring. Again, my interpretation of these basic numbers requires imagining how different voting patterns would be reflected in numbers. <y imagination may be wrong, but having written out my premises, I can begin to test them, and other people can check me and, if necessary, correct me.

Why did Warnock get more votes that Ossoff?

Here’s my guess at a simple explanation for those numbers: there is one group that seems most likely to explain people splitting a ticket between Warnock and Perdue: Black Republicans. It seems plausible that some Black Republicans would cross party lines to vote for a fellow Black person. Doing some rough numbers just as estimates: about 5,000,000 votes cast in the GA election; GA is roughly one-third Black–estimate that as roughly 1,500,000 Black voters—roughly 12% of Black voters in GA are Republican (according to Pew Research Center), so that’s roughly 180,000 Black Republicans in GA—far more than the 19,000 in the Warnock/Ossofff difference. If one in ten Black Republicans decided to cross party lines for Warnock, that would explain the observed difference. It’s also worth noting that because the Democrats needed to win both seats to win control of the Senate, it’s possible that a Republican voter might think that voting for Perdue would be their step to preserve control of the senate (“As long as Perdue wins, we keep control, so I can vote for Warnock”). Let me reiterate that this is a simplistic conclusion that probably misses real world truth, but at least offers an easily understood explanation. (Real world explanations might include differences in specific policy positions held by Warnock and Ossoff, but I have not studied them closely enough to do any analysis based on their policy recommendations.)

Another group might also explain the same pattern of data: misogynist Republicans, who might vote for Perdue but against Loeffler because she is a woman. This seems less likely, just on the basis of how many women have previously been elected by GOP voters. (Continuing down the path of imagination, we can conjure up a group of racist Democrats who vote for Ossoff but not Warnock, or feminist Democrats who vote for Loeffler over Ossoff because she is a woman. But these groups would give more votes to Ossoff than Warnock, so don’t help explain the observed data.)

Conclusion

On many levels, what I have offered above is simplistic analysis. Despite my performing a quick analysis, the various considerations and possible questions proliferated. I didn’t do any research beyond looking up the numbers of votes cast.  I could have looked more deeply. I could have looked at different details (what if I had looked at county-by-county breakdowns? Those might provide some counter to the ideas I used above).