miscellaneous factZ

Courtesy or a kind invitation from Richard Cave, tomorrow I’ll be heading over to Warwick University to present in their IO Seminar. The talk will be focused on my main ‘IP papers’: Cumulative Innovation, Sampling and the Hold-Up Problem and Imitation and Innovation with and without IP, but if there’s time I might also get the chance to discuss another paper of mine on the Control of Porting in Two-Sided Markets.

Yesterday I attended the first of a series of “Workshops on Well-being” at the LSE organized by Paul Dolan, Richard Layard and Andrew Oswald. Below can be found some (very) impressionistic notes.

Talk by Andrew Oswald: Does Higher Job-Status Make a Person Healthier? A Longitudinal Test of the Whitehall Hypothesis

Basic (well-known) facts:

• Strong inverted u-shape in depression/anxiety over life-cycle peaking in mid-40s to mid-50s
• Need to move away from GDP to well-being in the next century
• More collaboration across discipline
• Across countries wealth correlated with happiness
• Within country across time (i.e. repeated cross-sections) no real growth in happiness (though growth in money)
• Now have data for Britain, Belgium and Netherlands and we reject null of no change in GHQ over time (so there is a decline in mental health over time).
• Can repeat across EU countries.

New social welfare functions:

• blood pressure, obesity, height.
• Life Satisfaction (LS) from NCDS: LS=f(high blood pressure, personal controls). high blood pressure enters negative.
• Well-being and height (guided by John Komlos)
  • danes and netherlands have been getting taller faster than anyone else (in US it may even be going down recently i.e. since ~1990)
  • this is interesting because danes and nthlands are happiest (and US is pretty unhappy)
  • height and happiness are correlated in rich EU countries (this still holds with deltas of height and happiness)
• Weight and well-being
  • BMI enters negatively in regressions for LS, Happiness, well-being (GHQ)
  • Christakis and Fowler The spread of obesity in a large social network over 32 years. NEJM 2007
  • Summarizes his paper on BMI with Powdthavee (where relative effects are significant)
  • u is concave in relative position then upward spirals, convexity then deviate from herd.
  • Relative BMI enters negatively (along with absolute BMI)
  • Carol Graham on US and Russians

Status and well-being.

• Look at the ‘Whitehall effect’ using longitudinal data. Contrary to existing cross-sectional results finding a robust effect they do not find such a result. This suggests that cross-section results may be picking up causality the other way and the resulting selection effect (people who are healthier get promoted).
• Redelmeier and Singh, Annals of Internal Medicine (Oscar winners live longer). But lot of issues statistically (not enough controls)
• Rablen and Oswald (2006), look at Nobel prize winners (vs. nominees). Get 1.6 extra years.
• Final aside: did experiment looking at reporting function on height. Found perceived height was linearly related to actual height.

Discussion

Paul Dolan

• When do we compare?
• Inequalities (higher moments …)
• Mental health and well-being (v. blurred distinction)
• International comparisons (are they useful)
• What measures if we move away from GDP?

Misc

• Range frequency theory
  • Parducci etc
• How happy they are with weight
  • Actual weight vs. actual weight
• Gilevich study
  • won a medal. Silver medallists less happy than bronze medallists.
• Increase in reports of morbidity but less actual illness.

Own questions

QU: if we’re hoping for a reorientation of public policy in relation to happiness and GDP) one would want to ask why hasn’t there already been a reorientation in relation to other areas (e.g. environment and GDP) — or at least why has it been so slow.

QU: Is relative effects in obesity coming from status stuff or from signalling (i.e. I use other people’s weight to compute what the optimal weight is). Has importance to determining policy impact.

• 2008-06: There is now a pdf version of these comments available. This may be a better way to read this as some people have reported difficulties with viewing the maths in the page (requires mathml support — via firefox + mathml fonts or IE).

Back last October I was a participant in a “Networks” reading group that worked through “An Economic Model of Friendship: Homophily, Minorities and Segregation” by Currarini, Jackson and Pin (2007). The discussion at the time encouraged me to expend some effort in order to try and derive some of the paper’s results in what I believe is a simpler, and more intuitive, manner.

Having finally got round to writing/marking up my rough notes into computer friendly format (markdown + latexmathml) I’ve now posted them online at:

http://www.rufuspollock.org/economics/papers/comments-on-currarini-jackson-pin-2007

Herewith are further (partial, impressionistic) notes from the second day of the two-day workshop (programme) on Rationality and Emotions organized by Miriam Teschl at Robinson College here in Cambridge.

S-Shaped Probability Weighting and Hyperbolic Preference Reversal - An Intimate Relationship by Herbert Walther

Walther has published these results as a 2003 paper in Journal of Economic Behaviour and Organization.

http://www.robinson.cam.ac.uk/academic/robinson_rationality_walther.pdf

Overview

• Empirical regularities:
  • hyperbolic discounting
  • sign effect: loss discounted less than gains
  • preference reversal:
  • magnitude effect: preference for losses before gains
  • s-shaped prob weighting (Gonzales and Wu 1999, Fehr-Duda, 2006 et al.)
    • prob weighting can explain Allais paradox
• How to resolve?
• Ans: EU maximizer considers anticipated emotions reactions to resoluton of uncertainty
  • prob weight derived via intertemporal state dependent EU max
  • using this can explain most empirical effects

Model and Results

Part 1: Generating the S-shape prob distbn

• EU of some binary prospect L(p, w1, w2), w1
• EU of wealth
• EU of elation (if you win)
• EU of disappointment (if you lose)
• last two both fade away over time

Part 2: Empirical regularities

Having generated the S-shaped result Walther goes on to show how this can generate most of the empirical regularities we are interested in.

• Look at some payment/contract whose probability of payment fulfilment is falling over time (this way we get probability in which we need)
• Now have some S-shaped setup and probability that goes into this S-shape is dropping over time (contract is less likely to be fulfilled).
• Hyperbolic discounting: can also now generate hyperbolic discounting within this same framework (other explanations e.g. Souzou 1998, Dasgupta and Maskin (2005) only do it on its own).
  • Logic underlying hyperbolicity: at start contract is very likely to be fulfilled so if it does not lots of disappointment — so (exp) discount rate is very high. Over time prob falls and S-shape prob distbn kicks in (so elation outweighing disappointment) and discount rate falls.
  • prediction: poor will show more hyperbolicity than rich
• The sign effect: gains discounted more than losses
  • Logic: again simple. If loss is very likely little disappointement but a very certain gain has lots of potential for disappointment.
  • prediction: again the sign effect is more pronounced for poor than the rich.
  • magnitude effect for losses: higher losses have higher impact that lower losses (because straightforward wealth utility becomes more important than disappointment/elation effect).
• preference reversal
  • poor will prefer losses before rich subject but gains after rich subject
  • preferences are same but marginal utility of wealth is different

Summary

• Simple model that is a small extension of basic EU maximization most of the empirical regularities.
• If diminishing marginal utility of wealth poor people will behave ‘less rationally’ than rich people despite having same preferences
• For the future: Why is prob weighting evolutionary sustainable?
  • potential answer: in hunter gather society there are externalities in that (large) gains and losses are shared (this would => S-shaped prob distbn).

“It’s a boy! Behavioural and Neural Evidence on Self Delusion” by Danica Mijovic-Prelec

• Deficits (due to lesion) on right side of brain lead to deficits in left hemisphere
• Furthermore these patients are not aware of the deficit and deny its existence (to the extent of confabulating experiences)
• Sackheim-Gur 1979: self-deception in social psychology
  • played people mixture of their own phone and others
  • averse to your own voice
  • people would not hear their own voice and furthermore physiological measure of stress indicated it went up when ‘not hearing their voice (when it was there) — i.e. when people were self-deceiving
• Sackheim-Gur criteria:
  • individual holds 2 contradictory beliefs
  • beliefs held simultaneously
  • individual is not aware of of holding one of the two beliefs
  • nonawareness of this belief is motivated

Experiment

• shown korean figures and asked to classify as male/female
• first stage: get figures and must classify (5c for each correct prediction — correct measured against classification by some control group)
• second stage: must also predict gender of next figure (and then classify)
  • paid like before for classification but bonus for being in top x% of predictors

Results

• Focus on items that were ‘well-classified’ by control
• First classification: 65% accurate
• Anticipation: 50% accurate (as expected since randomized)
• Second classification: <65%
  • anticipation effects classification
• stronger for males: anticipated as male results in classification as male 72% (for females like first time round)
• calculate self-delusion index for each subject
  • four options for response pattern (starting with female) of form 1st classif, anticipation, 2nd classif:
    • FMF: honest
    • FMM: self-deluding
    • FFM: inconsistent
    • FFF: consistent
  • need to subtract inconsistency from self-delusion percentage to get ‘true’ self-delusion
  • index = % self-delusion / % inconsistency (could use difference)
• fMRI
  • expect that self-deluding subjects behave differently from inconsistent (and honest and consistent)
    • notably don’t show this activation on consistent trial (when they also confirm their prediction)
  • this is what they find (v. significantly)
  • in attentional and cog. control regions
    • self-deluding and inconsistent is similar
    • however big difference in parahippocampal gyrus (associated with memory)
• [rp] question: could some of this come from a priming effect combined with better recall. I anticipate X, which primes me. Then suppose I see the figure and have a vague recall from before. Suppose that people experience different priming effects — then those with a strong priming effect feel conflicted and have more stress (i remember Y sort of but do I really or I just doubtful because of having seen X) which means more fMRI anomalies and and means they are more likely to ’self-delude’ while those with weak priming simply aren’t sure what they think (not really excited/conflicted) and just go randomly with M/F (so ’self-delusion’ or ‘inconsistency’ are equally likely).

Herding and Social Influence in Economic Decision Making by Michelle Baddeley

• Solomon Asch
  • Length of line experiments (everyone says line is B when actually A)
• Task design: stock-picking
  • two charts for past prices of a stock
  • shown faces along with their associated choice (controlled by experimenter)
• Results:
  • strong effect of other decisions on own decision (on average 72% vs. 50% choose the one chosen by herd)
  • perhaps not very surprising here given the lack of info about stocks (and their underlying equivalence) — a small piece of information should have a dramatic effect

Today I attended the first day of a two-day workshop (programme) on Rationality and Emotions organized by Miriam Teschl at Robinson College here at Cambridge. The mix of economics, psychology and neuroscience has so far been fascinating and below I include some general ‘impressionistic’ notes from some of the sessions so far.

Stress and Euphoria on a Trading Floor by John Coates

Overview

• Better title might have been ‘Fear and Greed’.
• Managed a trading desk on Wall Street.
• Are fear and greed exaggerated by a steroid induced shift in risk preferences?
• Cortisol rises in a market crash, raises risk aversion and accentuates crash.
• Little literature on hormones
  • one paper on oxycytocin and trust
• Steroids have very widespread effects in body (everything from body shape to cognitive function).
• Cortisol follows adrenalin as a stress response.
  • Acute exposure: euphorogenic, increases motivation (+ve basically)
  • Chronic exposure: bad. Selective attention to negative precedents (Erikson 2003) etc.
    • Decrease risk preferences
• Traders
  • 17 males, 19-38 from City Trading Floor. Healthy and no outside source of stress.
  • No overnight positions, no salary, no bonus. Just given capital and trade (each with own deal: basically a percentage or profits)
  • Annual P&L range: -10k to +5m
  • Experiment was live because you could not replicate high stakes trading in the lab.
  • Trading US Bond Futures adn Bund Future, Dax, and Euro/$
  • German market main component of P+L so that was the focus
  • Use calendar of econ stat announcements as ’stress’ events (some traders make almost all their money just when US employment info comes out)
• Data:
  • Sampled 2x a day (11am,, 4pm) to bracket NY 0830 news releases
  • P+L at 11am and 4pm
  • Long term P+L
  • Plus other info

Results

• Massive volatility in cortisol: cortisol should drop by 30% in a normal subjects but in 40% of subjects upwards slopes were increasing by 500%
• No relation of cortisol to P+L at all (surprising: but this was not a a very volatile period).
• But a strong relation of cortisol and std dev of P+L
• Next look at effect of expected volatility (measured by implied volatility measures used to price derivatives) to see whether this affects cortisol release
  • incredibly good fit (people were excited because volatility means there is money to be made)
• Conclusion: only saw acute exposure effects (so mainly excitement)
  • prolonged exposure could be very different: ‘irrational pessimism’ (but no evidence in this paper)

The Psychology of Gambling Behaviour by Dr Luke Clark

Overview

• Lots of Gambling and it has been increasing (9.6bn a year)
• Structural characteristics lead to overestimation chances of winning
  • Erroneous verbalizations using ‘think aloud’ (Gaboury and Ladouceur 1988) — up to 80% of their thoughts are irrational (problem gamblers show more)
  • Gamblers develop ‘illusion of control’ (Langer 1975)
  • Failure appreciate independence of turns (Wagenaar)
• Investigate particular items
  • Effect of ‘near-misses’: people play more when more near-misses (Kassinove and Scharre 2001 find maximum at 30% of near-misses)
  • Effect of personal control:
    • craps: people bet more on their own throw than on others’ throws
    • roulette: higher bets when palyer versus croupier throws ball
    • lottery: players demand more to exchange self-selected tickets ($9) than lucky dip tickets ($2) (Langer 1975)
• Experimental design like a slot/fruit machine (but with participant having chance to select value on left ‘wheel’)
  • Thus can investigate both control and near-misses at same time
  • Also elicit feeling at particular points of time

Results

• Behaviour
  • control results in significant increase in ‘happiness with situation’
  • happier when they win if ‘in control’
  • interaction of control and near-miss: when in control near-miss increases desire to continue (but this is not so when computer ‘in control’)
• Arousal:
  • near-miss impact on arousal depends on control (w/o control no response)
• fMRI
  • participants who show more erroneous beliefs in questionnaire show more activity in response to near-misses

Summary

• Gambling-related cognitive distortions can be elicited on a laboratory task
• Personal control increased perceived chance of winning and pleasure at winning
• Near-misses are aversive but encourage continued play (*when player ‘in control’)

Time and Emotions by Stephane Luchini

Overview

• People prefer to ‘consume’ something unpleasant before something ‘pleasant’ (if they have a choice)
  • seems problematic as with discounting would expect the opposite
• One explanation: people ‘consume’ an event both at the time it happens and before (thus want the good thing second so you ‘anticipate’ it for longer).
• But what about the effect of the anticipated outcome on the subjective experience of time
  • Significant amount of evidence that emotions affect time perception
• Emotions are generated by differential/passage between two situations (not time dependent)
• Anticipated duration is function of actual clock time and the basic emotion which depends on difference between current state and future state.

Results

• To obtain time reversal (negative time preference) emotions must have strong impact on time discounting
• Only can occur if the the future date is not too remote.

Over Christmas I’ve had some spare time. This has permitted me to get the Atlas of Economics Models off the ground. This is a project I’ve been thinking about for some years, first motivated really by the experience of trying to discover what variations had been done on the basic Hotelling-line model of ’spatial’ product differentiation and competition (previous allusion earlier in the Autumn here).

So what is the Atlas supposed to be? From the front page:

The Atlas of Economic Models is a comprehensive list of the basic ‘building-block’ models used by economists. It also includes additional information, for example worked out analytical solutions to special cases and details as to how models inter-relate (hence the ‘Atlas’ in the title). More about the atlas can be found on the about page about page.

Other important features of the Atlas are that it is:

• Community Editable: the Atlas is a community-based project with most content editable by anyone who wishes to contribute. Specifically we’re managing the content in a wiki and to edit any given page all you need to do is click on the edit button at the bottom of that page.
• Openly Licensed: all content is openly licensed. That is all material is made available under a license that permits it to be freely used, reused, shared and redistributed by others. Further details on the license page.

As yet, it obviously does not have much content but that should be gradually remedied over the coming months. And if you’re economically inclined why not head over there and help out …

A bit of playing around with the excellent bibtex2html and my bibtex economics bibliography is now up online in html format:

http://www.rufuspollock.org/economics/biblio/

It has a fairly good coverage of published material in the areas of IP and innovation, though by no means complete, particularly on the working papers front as I don’t tend to enter these unless I’m specifically citing them in a paper.

I’ve put a basic primer on demand theory up online in the economics section. I wrote this several years ago and haven’t had much chance to update it and there are still several sections marked with TODOs but it does provide a rapid and concise overview of some of the main results. Originally I hoped it (and its companions) might expand into a rudimentary open textbook but at present I see them more as a basis for a putative atlas of economics models — an alternative and perhaps grander project about which I plan to post at some later point.

These are a first set of impressionistic notes taken from various sessions yesterday.

Wednesday AM: Janos Kornai: Life and Works

No real notes from this `session’ but some comments from the Q&D caught my attention:

Janos Kornai: Let me paint a dark picture: today we have become a machine for manufacturing economists. All papers are very uniform, with a very uniform style on subjects that are all in acceptable areas that are already established.

If you have a young PhD student today, very original say, with a very original paper but that isn’t really that polished. Well they could never go out into the `job market’ with that paper, they would take their standard paper in a well-establisehd area, which all the reviewers will be happy with.

…

To get money from Brussels, you have to be very specific about what you are going to deliver, what you are going to provide in one year time etc. But real, new, original work is not like that — you can’t say what you will exactly be producing in one or two years time.

Assar Linbeck: what has happened over the last few decades there has been a vast improvement in the level of technical training of PhDs. That is wonderful but you then you spend your whole time on technicalities, searching under the lamp-post as it were. Nowadays we have a lot of government funding and that leads to very specific funding which are all aimed at solving some big public problem (the environment etc etc). We need to be more free in our funding approaches.

Wednesday PM: Development Panel

Daron Acemoglu

Should do more detailed work, especially on the empirical side, to get better models of growth (which then might help us do policy stuff).

Esther Duflo: Aid Optimists and Aid Pessimists

• Deplores focus on either pure optimism (Aid can solve the development `problem’) or pure pessimism (Aid is useless)
  • This is not very surprising, as public discourse has a tendency to get polarized. The attention span in the public sphere (and rock concert) rarely extends beyond one sentence.
• …
• On the optimists side we have lots of low-hanging fruit
  • But lots of rubbish in there too (look at Sachs or the World Banks list: some items with lots evidence of their effectiveness but lots with zero evidence).
• On the aid pessimist side these are OK but lack glamour
  • These are modest interventions. No government or donor wants to back these.
• Let’s be modest: we are plumbers not nuclear physicists. The devil is going to be in the details. We can’t just work from models have to get our hands really dirty. Need plenty of evaluation — and this has payoffs for economic research — especially in estimating from micro-data key parameters such as returns to physical capital (De Mel et al.) returns to human capital (Thomas et al.).

William Easterly

Engaging, witty presentation attacking 3 main fallacies:

• We know what causes growth
• We know how Aid interacts with these causes
• We know who is the ‘we’ in the statement:we must achieve development/growth'' (orwe must solve poverty” etc)

All kinds of data and info whose main message is we don’t have a lot of evidence that large scale Aid disbursements work.

[ed:] Best summary of Easterly’s arguments in recent JEL review of Sach’s latest screed (End of Poverty) which made almost exactly the same points as Duflo:

• Focus on discrete small/medium size projects for which we have good evidence they work and do small experiments to develop new ideas.
• Stay away from large projects, do not just throw Aid money at countries and avoid promising the earth (as e.g. Bono and Sachs do regarding `Ending World Poverty’).

Conclusions from talk:

• Really need to keep the Aid and Development separate. Focus on what Aid can do to help people in direct ways now and leave development as something else.
• Aid: do only what can be evaluated, and evaluate only what can be done.
• Development: don’t trust experts, focus on home-grown solutions.

Soros

Talks about some of his work particularly in relation to resource curse and accountability for revenues from natural resources.

Paul Collier

Argues strongly that economists need to engage with policy-making more. Complains that too often now economists only address their peers (and only care about their opinions). What we write for our peers is completely incomprehensible to other to the general public. Says that nowadays regularly writes 2 versions of his research: one for the academe and one for the public.

Re-mentions the wonderful story told by George Soros about accountancy exercise there which led to publication of data on the level of oil revenue going to each subregion. This resulted in death threats for the data providers and the biggest ever daily sales of newspapers in Nigeria.

This week I’m in Budapest for the EEA/ESEM annual congress. I’ll be presenting Cumulative Innovation, Sampling and the Hold-Up Problem in the ESEM stream on Friday afternoon in the 1500-1700 slot. I’ll gradually post further notes as I write them over the week.