Decision Science News

LOSS AVERSION AND RISK AVERSION ARE CORRELATED (AT LEAST WITH OUR TECHNIQUE)

A student recently emailed us asking for some data from our 2008 paper:

Goldstein, Daniel G., Johnson, Eric J. & Sharpe, William F. (2008). Choosing outcomes versus choosing products: Consumer-focused retirement investment advice. Journal of Consumer Research, 35(3), 440-456.

In particular, they were interested in the correlation between estimates of risk aversion and loss aversion within a person, which can be seen in the above plot (Figure 4 in the article).

We thought, why not make the data open to the whole world. Here they are:

Goldstein, Johnson, Sharpe (2008) Loss Aversion and Risk Aversion Data

subject: An anonymous identifier indexing the unique human who submitted the data
dist: participants submitted two distributions (one right after another) in Year 1. They were invited back in Year 2 to submit distributions again. There was some dropout. This column tells you which distribution you are looking at.
riskAversion: this is the coefficient of relative risk aversion, commonly referred to as alpha
lossAversion: this is the coefficient of loss aversion, commonly referred to as lambda

Note: There are some extreme values in the data that will throw off your correlations. In the figure above, as noted in the paper, we just plot cases in which lambda is < 25. Here we’ll show how this affects the correlations:

PEARSON CORRELATIONS BETWEEN ALPHA AND LAMBDA: CASES WHERE LAMBDA IS < 25
Distribution Correlation
Year 1 Dist 1 0.65
Year 1 Dist 2 0.60
Year 2 Dist 1 0.52
Year 2 Dist 2 0.48

PEARSON CORRELATIONS BETWEEN ALPHA AND LAMBDA: ALL CASES
Distribution Correlation
Year 1 Dist 1 0.65
Year 1 Dist 2 0.88
Year 2 Dist 1 0.65
Year 2 Dist 2 0.56

Clearly, using the method of Goldstein, Johnson and Sharpe (2008), estimates between risk aversion and loss aversion are quite correlated. The method forces the correlation somewhat: the most loss averse data one can submit will necessarily be quite risk averse.

Feel free to download the data and draw your own conclusions, but please cite the above paper as the source if you do.

SCP WINTER SUBMISSION DEADLINE AUGUST 12, 2016

Society for Consumer Psychology
Annual 2017 Winter Conference
Palace Hotel, San Francisco, California
February 16 – 18, 2017

The Society for Consumer Psychology (SCP) will be holding its Annual Winter Conference from February 16-18, 2017 at the Palace Hotel in San Francisco, California. The Society for Consumer Psychology conference provides opportunities for a high level of interaction among participants interested in consumer research and in advancing the discipline of consumer psychology in a global society.

We are seeking proposals for symposia, original competitive papers, and working papers for presentation at the conference. We encourage a diverse set of ideas and approaches to consumer psychology. We also welcome diverse methodologies, including experimental research, survey research, conceptual and/or theoretical developments, or other methods relevant to the study of consumer psychology.

GENERAL SUBMISSION GUIDELINES:

Submission Deadline

All symposium, competitive paper, and working paper submissions are due by Friday, August 12, 2016. We will send notification of acceptances in November 2016.

The conference website will be available for submissions between Monday, June 6, 2016, and midnight PST of the deadline, Friday, August 12, 2016.

SYMPOSIA

Symposium sessions focus on a specific area of research. Submissions may share similar theoretical or methodological bases, or they may approach the same research question from different perspectives. Each session is 75 minutes and should include either three or four papers. The symposium chair is expected to lead the discussion—there will be no space in the program for discussants. Symposium chairs are responsible for submitting all materials by the deadline and ensuring that all session participants receive copies of each paper or presentation prior to the conference.

Symposium proposals should include the following:

• The title of the symposium
• A brief proposal describing the symposium’s objective, topics to be covered, likely audience, stage of completion of each paper, and how the session contributes to the field of consumer psychology.
• The name, contact information, and affiliation of the symposium chair
• The title of each presentation, with a listing of the authors and their affiliations and contact information. For multi-author papers, please underline the presenter.
• A 75-100 word short abstract of each presentation (for publication in the conference program)
• A 750-1000 word extended abstract of each presentation (for evaluation by the Program Committee)

COMPETITIVE PAPERS and WORKING PAPERS

Competitive Papers. Competitive papers present completed work and address substantive, methodological, or theoretical topics in consumer psychology. We will be grouping four competitive papers into a single 75 minute session. Authors will have 15 minutes to present their work, followed by approximately five minutes for questions.

Working Papers. In contrast, working papers typically report the results of research in its early stages. Authors of accepted working papers will present their work during a Focused Reports Session during the conference. This is a new format for the working papers that will replace the poster session. Authors of accepted working papers should plan to present a 5-minute summary of their work to an audience in a focused session. Authors will be allowed to have a power point deck to aid their presentation, but presentations must be kept to 5 minutes. This will be followed by a reception, in which interested parties can ask questions about the research presented.

Detailed guidelines about the focused presentations will be sent with the acceptance notices.

Competitive Paper and Working Paper submissions should include the following:

• The title of the paper
• Nature of submission: Competitive or Working Paper
• The name, contact information, and affiliation of the author(s). For multi-author papers, please underline the presenter.
• A 75-100 word short abstract (for publication in the conference program)
• A 750-1000 word extended abstract that summarizes the motivation, conceptualization, methodology, and major findings (for evaluation by reviewers)

Note: Please indicate if the first author is a PhD student. (If so, the paper will be considered for the Best Student Paper Award.)

GENERAL GUIDELINES

Submissions will be judged on the following criteria:

• Quality of the research
• Contribution to the field of consumer psychology Interest of the topic to SCP members.

Each SCP participant may present in no more than two sessions. When submitting a symposium or paper to this conference, you must agree to be available at any time on both days of the conference (Friday 2/17 and Saturday 2/18) to give your presentation. If you will not be available on one of the days, please arrange for a co-author to give the presentation. We will not consider date/time change requests for presentations unless a presenter has been inadvertently scheduled to give two presentations in the same time slot.

SUBMISSION INFORMATION

All submissions should be single-spaced Microsoft Word documents.

Submissions should be made electronically through the conference website at http://www.chilleesys.com/scp/. The website will provide additional information about the conference and serve as an interface for authors and reviewers.

To submit your proposal, please follow these steps:

1. Sign up for the submission system: When you first enter the conference website, you will be required to sign up to use the website submission system. Here you will provide your name and contact information and be provided with a login name and password. You will use this login whenever you navigate the submission system. Please keep track of this information.

Some e-mail addresses are already signed up in our database. Please use the website password reminder function if you see the following message: “The E-mail address you entered has been already registered with our database. Please proceed to Log In page. If you forgot your password, please click here.”

[Note: When you complete this step, you will have only signed up with the conference website. This is NOT the registration for the conference.]

2. Enter the submission information: Once in the submission system, you will be asked to submit the information requested above for the symposium, competitive, or working paper submission. Please note that in order to facilitate reviewer assignment, you will also be asked to provide content and methodological area codes.

DOCTORAL SYMPOSIUM

As in recent years, there will be a day-long doctoral symposium immediately before the main conference, that is, on Thursday, February 16. Relevant details will be announced separately by the symposium co-chairs Kelly Goldsmith (Northwestern University) and Cassie Mogilner (UCLA).

HOTEL INFORMATION

The Palace Hotel is located at 2 New Montgomery Street, San Francisco, CA 94105. The telephone number is: 1 (415) 512-1111.

When making reservations you must mention that you are with the Society for Consumer Psychology to obtain the $229.00/night rate.

Visit the hotel website at: http://www.sfpalace.com

If you have questions, please email the conference co-chairs at: scp2017@sauder.ubc.ca

SOCIAL EVENT

As in recent years, there will be a social event on the evening of the last day of the conference (Saturday, Feb 18). Relevant details will be coming soon and it is guaranteed to be an epic evening!

Conference Co-chairs

Kate White – University of British Columbia
On Amir –  University of California San Diego

CENSUS DATA ON EDUCATIONAL ATTAINMENT BY COUNTY

People ask us “what are the most educated counties in the USA”? It turns out the census keeps track of this sort of thing. We found a table called ACS_14_1YR_S1501.csv in the American Community Survey; look for stuff on educational attainment by county. And it was an easy bit of R to get the answers. We computed two things:

• The percentage of the 25 and older population with a graduate or professional degree
• The percentage of the population (of any age) with a bachelor’s degree or higher

Here’s how it came out:

Top US counties by percentage of people 25 and up with graduate or professional degrees in 2014:

Top US counties by percentage of people with Bachelors degrees or higher in 2014:

R Code to follow along at home

Photo credit: https://flic.kr/p/aQM3Z

CALL FOR PAPERS: DEADLINE JUNE 1, 2016

The Tilburg Institute for Behavioral Economics Research is happy to announce the 15th TIBER Symposium on Psychology and Economics, to be held on August 26, 2016 at Tilburg University.

The Symposium
The goal of this series of symposia is to establish contact and discussion between Economists, Psychologists, Marketing researchers and others who work on Behavioral Decision Making, either in individual or interdependent settings. We look for empirical contributions from diverse fields, such as Individual Decision Making, Consumer Behavior, Bargaining, Social Dilemmas, Experimental Games, Emotions, Fairness and Justice, Rational Choice, and related subjects.

The symposium consists of two keynotes, a number of parallel sessions with presentations of 20 minutes, and a poster session. We are proud to have Dan Goldstein from Microsoft Research and Lise Vesterlund of the University of Pittsburgh as this year’s keynote speakers.

Call for Abstracts
If you would like to contribute to TIBER by presenting your research, we invite you to submit an abstract of max. 250 words.

Dates
June 1 Deadline for submission of abstracts
June 15 Selection of speakers
August 26 Symposium at Tilburg University

You can submit your abstract and find more information about the the symposium on our website:
http://www.tilburguniversity.edu/tiber15.

If you have any questions regarding the symposium, feel free to contact Arnoud Plantinga : a.plantinga at tilburguniversity.edu

Kind Regards,
Arnoud Plantinga, Ilja van Beest, Rik Pieters, Jan Potters, and Marcel Zeelenberg,

MAKING SURE THE REVIEWERS REFLECT THE MEMBERSHIP

Click to enlarge

Perhaps because it is an election year in the US, we’ve been thinking a lot about proportional representation.

The Society for Judgment and Decision Making (SJDM) is a diverse academic society, with people coming from about seven academic fields (or groups of related fields). See above.

Nina Mazar (SJDM Program Committee Chair) and Dan Goldstein (SJDM President) have been thinking about the following question: What can we do to help assure the papers accepted to the SJDM conference reflect the interests of the membership?

This lead to the “representative reviewer project”. Based on the membership chart above and a simple R script, we tested candidate sets of reviewers to see how well they matched the interests of the society. We tweaked the set of reviewers until we got as close as we could to proportional representation of fields.

We’ve sent out invitations to a representative set of reviewers. If you’ve been invited, please say “yes”. You’re representing a whole category of researcher.

We keep saying “representative”, but one might ask “representative of what?” The survey of the membership just includes faculty (as opposed to students members) who have paid dues in the last three years. So it’s a bit backward looking, which is okay. If the society is going to change focus, it should do so slowly. We at the Decision Science News feel that the Society for Judgment and Decision Making should be about judgment and decision making. If one is not careful, the field can run away from it’s name, like how the field of social psychology is no longer about social psychology.

ANALYSIS OF THE SJDM (SOCIETY FOR JUDGMENT AND DECISION MAKING) MEMBERSHIP

In order to make sure that decisions made for the Society for Judgment and Decision Making (SJDM) reflect the membership of the Society, we wanted to see some stats about the membership. Unfortunately, good stats on the members don’t exist. Sure, we did some crude analyses a while back, mostly about geography. However, this time we we wanted more precise information about gender balance and academic disciplines. So, we undertook an analysis. This was a lot of work. Please see the end for the methods. But first, graphs!

Gender overall

Count in all subfields

Click to enlarge

Count in major subfields. In this figure:

• Psych comprises Cognitive Psych, Social Psych and Psych (Other)
• Business comprises Marketing, Org Behavior, and Econ

Click to enlarge

Fields within gender

Click to enlarge

Methods

• Started with the member database
• Excluded people who haven’t paid dues since 2012
• Excluded student members
• Excluded members who didn’t specify an institution in their profiles

This left 695 names. We then:

• Went person by person through the list and looked them up online
• Coded each person’s gender. We couldn’t determine it 5 times and coded it as NA
• Coded each person’s academic field. If we couldn’t determine the field, usually because a lack of a CV, we coded it as NA. If the person worked in industry, we coded it as NA. There were 67 NAs for field overall.

This took about 10 hours, mostly done after dinner, in front of the television.

We coded things in the following way:

• People working in Marketing departments were classified as Marketing
• People working in Management or Organizational Behavior departments were classified as Org Behavior
• People who had mostly Cognitive Psych or Cognitive Science publications were classified as Cognitive Psych
• People who had mostly Social Psych publications were classified as Social Psych
• People working in Neuroscience, Developmental Psych, and other kinds of psych were classified as Psych (Other)
• People with a wide variety of psych publications were classified as Psych (Other)
• People in Econ, Accounting, Finance, and OR were classified as Econ, Acct, Fin, OR
• People in Policy, Law, and Medicine were classified as Policy, Law, Med

ADDENDUM

Mark Horowitz made a “Tree Map” from the data and sent it in. Thanks!

NOT WHAT WE WERE EXPECTING

Over here at Decision Science News, we like Engineering All The Things.  Accordingly, in an obsessive coffee phase, we acquired:

• A Coffee Maker that lets us set the exact brewing temperature
• A Digital Thermometer that lets us measure drinking temperature
• A Digital Scale that lets us weigh the grams
• A Coffee Grinder that lets us set the coarseness of the grind

If you want to experiment, knock yourself out. What the figure above shows is that something around 55 grams of coffee per liter of water is considered ideal, but you can play around in the 50 to 65 grams / liter range. It’s harder to set where you want to be on the red lines without expensive equipment, but generally as you grind the coffee finer and brew longer you move up and to the right.

We have:

• Experimented with the brew temperature from 195 – 205 F (91 – 96 C)
• Experimented with the drinking temperature
• Experimented with 50 to 65 grams of coffee per liter water
• Experimented with the coarseness of the grind, burr, and blade grinders
• Experimented with methods like Aeropress, French press, pourover methods
• Even experimented with fancy coffee, filtered water, etc.

The result of all this informal experimentation on ourselves was not what we were expecting.

The result of all this experimentation was that, for overall taste, the thing that matters most is the temperature at which you drink the coffee. Given that you are making coffee with standard parameter ranges, it all basically tastes the same holding drinking temperature constant.

The effect of cooling on the taste of coffee is substantial. I am not the first one to have noticed this, but it seems to be a rather unappreciated point.

Takeaways:

• Drink it too hot, it tastes like garbage
• Drink it too cold, it tastes like garbage
• Drink it between 130 and 140 F (54 – 60 C) and it tastes amazing
• Within normal ranges, nothing else matters much

Given our admittedly American tastes, this is how we make our coffee (because you’ve got to choose something):

• Coffee maker
• 195 degrees (low brew temps do taste a bit better, see below re: Aeropress)
• 55 grams of coffee / liter water
• Ground so it looks like what you see when opening a can of supermarket coffee
• No milk, no sugar
• Consume at 135 degrees (57 C)

Exception: In all our playing around, we did find that an Americano made using the Aeropress is an amazingly smooth cup of joe. They recommend brewing at about 170 degrees F. We also found that a lower brew temp, even with a coffee maker, does taste smoother.

But drinking temperature matters most.

Figure credit: http://www.engineerjobs.com/magazine/2013/engineering-perfect-cup-coffee.htm

SOCIETY FOR JUDGMENT AND DECISION MAKING NEWSLETTER

The quarterly Society for Judgment and Decision Making newsletter can be downloaded from the SJDM site:

http://sjdm.org/newsletters/

best,
Dan Goldstein
SJDM President & Newsletter Editor

OOPS

So, we’re on the Twitter and we see this animated GIF from the Wall Street Journal. It says

• $53,647 is the 24th percentile of Americans in 2014
• $53,647 is the 15th percentile among blacks
• $53,647 is the 13th percentile among black women
• $53,647 is the 29th percentile among black women with college degrees

Wait. What?

It says $54k is the 24th percentile of Americans. That means 24% of Americans earn less than $54k. The rest (76%) earn more. Americans are wealthier than I thought.

Then says $54k is the 13th percentile among black women. That means 13% of black women earn less than $54k. The rest (87%) earn more. However, according to the graphic, only 71% of black women with college degrees make more than $54k. So the subset with college degrees earns less than those with or without college degrees?

None of this makes any sense, unless of course you are using the word “percentile” incorrectly. The Xth percentile of a distribution is a value such at X% of the population is at or below that value. If your test score is at the 99th percentile, you’re one percentage point from the top and 99 percentage points from the bottom. If you score is at the first percentile, you’re one percentage point from the bottom and 99 percentage points from the top. Here’s a refresher on percentiles.

The WSJ used the term percentile backwards (100 minus percentile instead of percentile) in their Twitter infographic and they made the same mistake on their website in their Real Time Economics piece What Percent Are You?

Have a look below. With the slider moved to $100k income, the graphic reads “8th percentile”. Nope. That’s the 92nd percentile. 8th percentile: relatively poor. 92nd percentile: relatively wealthy.

The trouble comes, we think from them wanting to say “1 percenter” for the richest one percent. But that doesn’t make it the 1st percentile, which is the lowest income hundredth of the population.

Maybe you don’t believe us because we’re just a crummy little blog and they’re the Wall Street Journal. So we’ll turn it over to the dictionaries:

Merriam Webster’s Dictionary:

a value on a scale of 100 that indicates the percent of a distribution that is equal to or below it

.

American Heritage Dictionary of Student Science:

The percentile of a given value is determined by the percentage of the values that are smaller than that value. For example, a test score that is higher than 95 percent of the other scores is in the 95th percentile.

.

Random House Kernerman Webster’s College Dictionary:

Ninety percent of the values lie at or below the ninetieth percentile, ten percent above it.

.

And just to make sure it’s not an Americanism, here are two English dictionaries from England where they invented English:

Collins English Dictionary:

The 90th percentile is the value of a variable such that 90% of the relevant population is below that value.

.

Oxford Learner’s Dictionary:

Overall these students rank in the 21st percentile on the tests—that is, they did worse than 79 per cent of all children taking the test.

ANALYSIS OF A LOW-SCORING SPORT

Recently, our friend and co-author Sid Suri predicted that a soccer match would end 2-1. This got us thinking. Suppose you didn’t say which team would score 2 and which would score 1. And suppose people considered you “close” if you got within one goal of 2-1 (so 1-0, 1-1, 2-1, 3-1 or 2-2 would be close). How often would you be close by guessing 2-1?

We’ll save the answer for the end of this post.

To figure this out, we went to http://www.football-data.co.uk and pulled every English soccer match from 1993 to 2016 (so far), for the following leagues: Premier, Championship, League 1, League 2, Conference, Division 1, Division 2, and Division 3. In all: 52,017 matches.

First, let’s establish that soccer is a low-scoring sport. We see below that a two-goal match is the most likely outcome in soccer. The average number of goals per match is 2.6 (median 2).

This leads to certain low scores (where a score is a high-low pair) being quite common. The graph at top of this post shows that 1-0 (or 0-1, which is the same thing by our definition) is the most common outcome of a football match there is. Nearly 20% of matches end 1-0.

Now to our key question. What score is just one goal away from the most soccer matches’ scores? It’s close, but the winner is 2-1. 52.9% of matches end within one point of that score. So if you can be vague about who you predict will win, just confidently proclaim that it will end 2-1 and you’ll be right most of the time. Same deal for saying 2-0 or 1-0. All three of these scores are within one point of most soccer matches’ outcomes.

One cute math-y thing about this is that some scores are one point away from more possible scores than others. See below. 2-1 is within one point of five scores (including itself), but 2-0 is only within one point of four possible scores. Same with 1-0. To consider why, realize that scores can’t be negative. Despite being neighbors with fewer scores, 2-0 and 1-0 are within one goal of roughly the same number of matches as 2-1 is. Even 1-1 does quite well with just two neighbors beyond itself.

The figure above shows the counts of each of the neighbors of (from top to bottom) 2-1, 2-0, 1-0, and 1-1.

Want to play around more? Here’s R code to play around. Thanks to https://twitter.com/hadleywickham for ggplot, dplyr, httr