Saturday, May 16, 2015

problems with measuring personality

At lunch we were discussing Big Five traits (a common 5-dimensional categorization of personalities, measuring openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism) and this reminded me of the strange phenomenon in my Myers-Briggs categorization that I've noticed over time. In 8th grade I first scored INTP - entirely Introverted vs Extroverted,  strongly iNtuitive vs Sensing,  entirely Thinking vs  Feeling, and borderline Perceiving vs Judging. Nowdays I'm only mostly I and slightly J. I've only taken the Big 5 test recently so I unfortunately don't know the trend, but I assume something similar would have happened.

I don't feel like my personality has changed so this has puzzled me. But now I think I understand the source, and it is concerning with regard to the use of Big 5 in research (mostly in psychology but more frequently as control variables in economics.)

If people are free to maximize their utility of time use,  the marginal utility of any activity should be equal at the bliss point. Obviously we work with many constraints, so I don't expect this to be exactly true*, but I expect it to be closer to true the more freedom of choice we have.

In 8th grade I had no choice but to spend 8 hours a day around people and to follow the strict schedule of activities in general. This was far too much human interaction for me and I would nearly always opt for alone time when given the chance. Nowdays I've carefully engineered my life to have as little compelled socializing as possible, and will usually opt to go to any party friends of mine might have. Similarly,  in San Francisco Matt used to not believe me when I said I was more introverted than him, because I was usually the one wanting to go out and do stuff after work and on weekends. But I mostly worked at home alone,  while he spent 10 hours a day in an open plan office. I would (/did) have a nervous breakdown in that environment.

Along with having more freedom to be alone, I also have more control over my schedule and environment. So while previously the rigid structure and organized environment imposed on me was more than enough, I now realize that I do generally like having a plan and an organized approach to things, hence P became J.

I can't think of what might have suddenly given me more freedom to trade off S/N or T/F, so I'm not surprised those have remained steady.

What does this mean for using these scores in regressions? My first thought is that asking about marginal preferences to measure averages will make people look less variable than they are and will understate the importance of personality. But that's on average. I bet there are plenty of circumstances in which the measure is actually biased.

*but maybe surprisingly close since we should really be equating the marginal present value of activities,  not immediate happiness, which makes work and sleep seem a lot more attractive.

Tuesday, May 12, 2015

awesome broken windows theory tests

This conference on "social norms and institutions", as you might expect, is quite cross-disciplinary. I'm not sure what the breakdown is, but there are plenty of behavioral economists, sociologists, psychologists, and even law professors and political scientists. So, I'm collecting quite a list of interesting new sources from lectures and conversations with these non-economists.

The most fun, and incredible, set of experiments I've learned about are the ones run by Siegwart Lindenberg and Linda Steg and colleagues at Groningen University. Despite having an unusually high prior*, I'm honestly stunned by their results, both in terms of the sheer size of the effects they're finding, and the stability of those large effects across a great number of slightly different situations.

They're studying, essentially, the broken windows theory. This theory hypothesizes that if there is evidence of some law/norm breaking behavior, that will cause people to break other laws/norms at a higher rate.** This is not due to inferences about levels of enforcement, because the findings are just as strong in purely normative (prescriptive, in addition to proscriptive) settings. In a world of marginally statistically significant results, it's a thrill to see such a long list of replications with huge and extremely statistically significant findings.
  1. Bikes are parked in a row next to a fence with a conspicuous "no graffiti" sign, and flyers are attached to each bicycle such that they must be removed to use the bike. If no graffiti is on the fence, 33% of subjects will litter their flyers. If graffiti is on the fence, 69% will. This was so surprising that a news station paid the researchers to replicate the study while they watched from rented rooms looking down on the area. The finding replicated very closely, and now the Netherlands requires immediate removal of graffiti.
  2. A temporary fence is set up around a parking lot. Two police ordinance signs on the gate state "no trespassing" and "no locking bikes to fence". People were directed 200 meters around to an alternative entrance, but the gate was left about 50cm ajar. If four bikes nearby were not locked to the fence, 27% of people squeezed through the gap in the main gate. If the four bikes were locked to the fence, 82% did.
  3. In a grocery store parking lot, a sign states "please return your shopping carts", and flyers are left under the driver-side windshield wipers on cars in the lot such that people had to remove them to safely drive away. When no stray carts were left in the lot, 30% of drivers littered their flyers. When several carts were improperly left in the lot, 58% of drivers littered.
  4. Bikes parked in a shed have flyers attached to them as in example 1. In the control condition, 52% of subjects littered their flyer. If fireworks were set off, within earshot but out of sight, which is a well-known offense prior to December 31 in the Netherlands, 80% littered.
  5. An envelope is left partially hanging out of a mailbox and visibly has a 5 Euro note inside, showing through a window in the envelope. In the control condition, 13% of people who passed the mailbox on foot stole the envelope. If the mailbox was covered with graffiti, 27% stole the envelope. If the mailbox was not covered in graffiti but litter was on the ground next to the mailbox, 25% stole the envelope.
  6. A stamped addressed envelope is on the ground next to a mailbox. In the control condition, 24% of passersby put the envelope in the mailbox. If bags of garbage are on the ground nearby (illegal in the Netherlands), 10% did.
  7. A bike is on the ground in an alley, having apparently accidentally fallen off its stand. If passersby have just entered the alley from an empty, clean street, 20% of individuals and 27% of groups right the bicycle. If garbage bags had been left on the street, then 6% of individuals and 5% of groups did so. If prior to entering the alley, passersby passed by someone who dropped an aluminum can and then picked it back up, 34% of individuals and 35% of groups picked up the bike.
  8. A person on the sidewalk accidentally drops some oranges just before meeting another pedestrian. Normally, 40% of passersby help the stranger pick up their oranges. If approximately 20 yards earlier, the passersby had witnessed someone drop an aluminum can and pick it up back up, 64% will help the stranger. If 20 yards earlier, the passerby had witnessed someone (a private citizen) sweeping the sidewalk, 82% helped the stranger.
  9. I can't find the numbers for this one, but in a mall where you have to walk down a hallway to get to a smoking zone, a certain percentage of people will start smoking before they reach the zone. If a group of typical looking Germans are standing around smoking prior to the zone, though, more passersby will light up early. If a group of goth teens are smoking prior to the zone, though, a higher than normal percentage of passersby will wait until they get to the smoking zone to light up.
Amazing. And while each individual study has many issues with interpretation and is targetting a different nuance of the phenomenon, the evidence from all 9 considered together is undeniably striking.

References: 1-8 come from here and here.

~~~

* Perceived unusuality is based on my reading of the literature (see next footnote). I've always been sure (without solid evidence) that the mechanism must be quite important.
** This motivated a famous crackdown on petty crime in NYC, which did undergo a drastic reduction in crime, but causality has not been convincingly established and the theory remains controversial.