Sunday, November 29, 2015

Another Leftist attempt to deny the obvious

They do a lot of that.  They need to

That liberals are the ones who find society all wrong and want to change it is definitional of liberalism.  So that suggests that liberals are the unhappy people.  Would you want to change the world about you if you were happy with it?  And that little bit of logic gets repeatedly confirmed by the survey data.  Republican voters (for instance) always report greater levels of happiness than do Democrat voters

But that has begun to get under the skin of some Leftist psychologists.  In their wisdom they think conservatives are the misfits and liberals are the regular guys.  So how can misfits be happier?  There are several possible answers to that and some Leftist pychologists have tested some of the answers.  Their preferred hypothesis is that conservatives are not telling the truth about how happy they are.  And they have research evidence to prove it.  But do they?

There is a much reprinted article by Tom Jacobs which summarizes some of that research.  It's longish so I am not going to reproduce it but I do want to look at the detail behind it.  Have they in fact proved anything?  We will see.

I will take the "evidence" for Tom's claims one by one.

The first study quoted by Tom was by Cara MacInnis and Michael Busseri of Brock University.  It actually concludes that "Extreme Right Wingers" were happier than others.  So how it supports Tom's  claims is puzzling.  Whether it does or not, however, hardly matters.  It was based on two totally invalid questionnaires.  The RWA scale gets roughly the same level of endorsement among voters of the Left and Right. Right-wing Leftists?  If that were not odd enough, the high scorers in Russia tend to be Communists.  Right wing Communists??  So scores on that set of questions tell you nothing certain that I can think of. It is just a bit of academic nonsense.

They also used another invalid scale called the SDO, but I have said enough about that piece of junk elsewhere. In short, the research was so ill-conceived that it proved nothing

The second study by Sean P. Wojcik et al. seems to be the main one for Tom.

Their Study 1:  Even Wojcik et al agree that the results of that study are ambiguous but that is the least of their problems.  The main problem is the tiny size of the effects observed.  A correlation of .1 explains .01% percent of the common variance between the two variables.  Combine that with the fact that the sample was of visitors to an internet site and you have a big problem indeed.  You have to have really strong correlations among such an unrepresentative sample for it to be of interest.  So again, the study proved nothing. The tiny correlation was statistically significant but that just reflects the large sample size -- N = 1433.

Their Study II was of greater interest. They did a content analysis of speeches by Congresscritters.  But again they found little.  I quote:

"Greater conservatism was associated with a small but significant decrease in positive affect word use (b = –0.16, P <0.001). Conservatism was not significantly associated with the use of negative affect words, joviality-related words, or sadness-related words".

So of the 4 relationships they examined only one was significant and it was again very low.  But again, that may not be the big problem of the study.  Content analysis can very easily be biased and strong precautions have to be taken against that. Wojcik et al list no such precautions.  So again no firm conclusions can be drawn from the work.

But they include another highly inferential piece of research in their Study II.  They analysed the smiles of Congresscritters!  Again, however, the correlations were tiny.  I quote:

"We observed only marginally significant differences in the intensity of smiling behavior in the muscles lifting the corners of the mouth (AU12: b= –0.10, P=0.096), but conservatism predicted significantly less intense facial action in the muscles around the eyes that indicate genuine happiness"

So again, their findings were negligible.  And, in the circumstances, we have to ask whether inferences drawn from eye muscle movement tell us much anyway. Eye muscle movements might tell us something in a gross sort of way but where the differences are very slight, do they tell us anything at all? Thinking in terms of Venn diagrams, the tiny overlap indicated by a .10 correlation could be entirely outside the overlap between eye-movements and happiness -- and thus tell you nothing about happiness.

And I liked this bit of modesty about their results:

"Of course, elected political leaders are not representative of liberal and conservative individuals more generally, and it is unclear how well speech and facial expressions occurring within the confines of Capitol Hill reflect similar happiness-related  behaviors  in  less  overtly  political contexts"


Their Study III also raises questions. I quote:

"We analyzed the statuses of individuals who subscribed to (“followed”) the official Twitter pages of either the Democratic or Republican Party, excluding those following both, under the assumption that users who followed one party exclusively were likely to share that party’s political views"

I suppose I can pass that as a reasonable assumption but it again raises sampling problems.  I don't think we have yet got to the point where the man in the street uses Twitter -- so the representativeness of the sample would appear to be deficient, thus limiting or even vitiating generalizations from it.

But leave that aside.  They  found:

"Republican Party subscribers’ updates were significantly less likely to contain positive emotion words, joviality words, and happy emoticons, and significantly more likely to contain negative emotion words"

This time they reported their statistics in terms of odds ratios rather than correlations.  But again the findings were utterly trivial. An odds-ratio has to be of 2.0 or above to be taken seriously and none of theirs were. Most were in fact below .1. In other words, their findings basically indicated "No effect".

Is my view of what is required of odds ratios just my opinion?  Not at all.  The Federal Reference Manual on Scientific Evidence, Second Edition says (p. 384): "the threshold for concluding that an agent was more likely than not the cause of an individual's disease is a relative risk greater than 2.0."  Odds ratios and relative risk are not exactly the same but with weak effects such as we have here they are much the same.

Their Study IV was another study of photographs. They found that:

"smiles were marginally more intense among employees at ideologically liberal organizations"

And "marginally" was the word again.

And that's it!  There's your proof that liberals really are happier. Generalizations based on extremely weak effects and highly indirect measures of happiness.

And none of the studies examined general population samples. There was no sampling at all, in fact. There was no attempt at representative sampling of conservatives or liberals at any point.  And without representative sampling of a group, you cannot make generalizations about that group.  So the study proves nothing.  Its reliance on crinkles in the corner of people's eyes is rather hilarious in fact.  You couldn't make it up

Tom Jacobs does quote one extra study but gives no name for it, no authors for it nor any link or journal citation for it.  My Google searches for it were in vain.  Did he just make it up?  Who knows? We have seen that Liberals do get desperate for confirmation of their beliefs.


No comments: