By Jonathan McGeachen
I got onto a very wide tangent from the post on health care I was working on, and started wondering about framing government as “us.” Why do people do it?
“We should raise/cut taxes.”
“We shouldn’t have invaded Iraq.”
“We need to throw that person in prison.”
“We” aren’t really doing any of these things. They are the job of a very specific subset of “us.” To call the state “us” is only slightly more accurate than calling a favorite sports team “us.” I do this sort of thing sometimes too, of course. Only the most dedicated Ron Swansons would not.
I began wondering if this phenomenon factored at all into consent for government action. Can people be substantially manipulated by in-group framing? I started writing “if I could perform a study on this…” but then I realized I kinda could, thanks to Al Gore’s Information Superhighway.
If I frame a government policy proposal along the lines of “we must do xyz,” then it will have higher approval than the exact same proposal framed along the lines of “the government must do xyz.” This may vary by respondents’ ideology.
I constructed an 8-question Google Forms survey. The first question asked if the respondent identified as liberal, conservative, or moderate/in between/neither. The second question asked the first letter of their last name, the answer to which directed the participant to one of two versions of the next 6 questions. Both versions asked, in one way or another, whether government should intervene to address a given problem, and presented two answer options, constructed from common talking points. One answer advocated government action, the other government inaction.
Version 1, given to last names A-M, in-grouped the state for purposes of action. I asked whether “we” should carry out the proposed policy, and retained the “we” phrasing on the action response, while framing the inaction response as “they.”
Version 2, given to last names N-Z, out-grouped the state for purposes of action. I kept the exact same wording, except replacing “we” with “the government,” and “they.” I moved the “we” wording to the government inaction response. So survey was as follows:
Version 1 Q1: Should we raise taxes to pay for healthcare for everyone who needs it?
A1: We have a responsibility to provide for those in need.
A2: They need to let people be responsible for themselves
Version 2 Q1: Should the government raise taxes to pay for healthcare for everyone who needs it?
A1: They have a responsibility to provide for those in need.
A2: We need to let people be responsible for themselves.
V1 Q2: Should we expand electronic surveillance to fight terrorism and crime?
A1: They need to respect people’s privacy.
A2: We need to do what it takes to keep people safe
V2 Q2: Should the government expand electronic surveillance to fight terrorism and crime?
A1: We need to respect people’s privacy.
A2: They need to do what it takes to keep people safe.
V1 Q3: Should we intervene militarily to defend innocent civilians around the world (example: Syria)?
A1: We need to use America’s power for good.
A2: They need to stop wasting tax dollars fighting wars that don’t concern us.
V2 Q3: Should the government intervene militarily to defend innocent civilians around the world (example: Syria)?
A1: They need to use America’s power for good.
A2: We need to stop wasting tax dollars fighting wars that don’t concern us.
V1 Q4: Should we keep marijuana illegal at the national level?
A1: We need to protect people from this dangerous drug.
A2: They need to stop putting people in jail for something so harmless.
V2 Q4: Should the government keep marijuana illegal at the national level?
A1: They need to protect people from this dangerous drug.
A2: We need to stop putting people in jail for something so harmless.
V1 Q5: Should we punish businesses for discriminating against customers?
A1: We need to make this a country where people feel included and have access to goods and services.
A2: They have no right to order businesses to serve anyone.
V2 Q5: Should the government punish businesses for discriminating against customers?
A1: They need to make this a country where people feel included have access to goods and services.
A2: We have no right to order businesses to serve anyone.
V1 Q6: Should we expand gun control laws?
A1: They need to respect people’s right to defend themselves as they see fit.
A2: We need to keep people safe from dangerous weapons.
V2 Q6: Should the government expand gun control laws?
A1: We need to respect people’s right to defend themselves as they see fit.
A2: They need to keep people safe from dangerous weapons.
I presented this survey to my immediate social network on Facebook and Twitter. I also placed it in a few places on Reddit in order to draw as large a sample size as I could. I didn’t expect ideological sample bias to matter, because I wasn’t looking at the actual results, but the differences in results between the two surveys.
There were two “rounds” of gathering participants from these sites. For the second round, it occurred to me to include a “where did you hear about this survey?” question, so I did. Glancing at the results, I also realized that many weren’t answering all questions, so I marked them “required” for the second round (give me a break, I’m not getting published in a journal, I’m just some dude with a computer).
I will discuss limitations of this method later.
Results and Analysis
Data is here. Additional sheets along the bottom are separated into categories for convenience.
There were 278 responses. 150 respondents (52.4%) identified as Liberal, 95 (34.3%) as Moderate/In Between/Neither, and 32 (11.6%) as Conservative. 1 respondent failed to identify, prior to my making it a required question.
There were 174 (62.6%) A-M names sent to Version 1, and 104 (37.4%) N-Z names sent to Version 2.
I’m not going to parse out every last detail, but I’ll point out what I found most interesting, both for and against my hypothesis.
The first thing I should say is that overall, there was a slight increase in approval for state action among Version 2 respondents:
The Conservative sample was especially small, so it should be taken with a larger grain of salt. I won’t dwell on it too long for this reason, but there were substantial differences between how people reacted to re-framing each question.
Here were the results for moderates:
And here were liberals:
There is something of a pattern here: approval increases for public healthcare, surveillance, and military intervention, when it’s “the government” doing it vs. “us.” It decreases or stays the same for marijuana prohibition (can’t decrease below 0: see below), and consistently decreases for gun control. Discrimination is the only area where there was a qualitative inconsistency, and this may have been an artifact of the small conservative sample size.
Not related to anything I was looking for, but only 10 total respondents across all versions and ideologies, or 3.6%, opposed marijuana legalization: 3 Conservatives, 7 Moderates and 0 liberals.
I threw the survey together in half an hour. I tried to keep it simple to make it easy to answer. What could go wrong?
1.) It would have been good to include “where did you find this survey?” from the very beginning. I would have been able to divide up responses by audience type.
2.) Sample size (and sample biases?). I have no means to randomize participants, so I can only try to artificially “balance” demographics. Even though apparently I live in a social media environment where conservative voices disproportionately dominate, the actual population I have access to is mostly liberal.
3.) Perhaps some question about how informed they consider themselves or how much political news they consume. I’d expect low-information participants to be more easily swayed by question framing than participants with clear and consistently-reinforced ideas.
4.) It wouldn’t have hurt to randomize question order, but I didn’t know how that worked.
My hypothesis, at least as it was stated, is clearly not supported. Perhaps there is no relationship between in-group framing and approval, perhaps it is more nuanced. The consistency by question across demographics is interesting. That big change for moderates on the intervention question is also interesting. I don’t see any solid conclusions I can draw from it, though.
But if I hadn’t done this, I probably would have just assumed most people are tribal atavists at heart. Having done this gives me pause. Thanks, Science!