miscellaneous factZ

Last week I attended FLOSS 2008, the second international workshop/network meeting on FLOSS (Free/Libre/Open Source software) in Rennes, France. I was presenting my paper Innovation and Imitation with and without Intellectual Property Rights (and would have offered discussant comments but the author of the paper I was scheduled to discuss had to pull out at the last minute). In addition to this I got to hear a variety of interesting talks. On some of these I was able to take notes which I have included below for the ‘delectation’ of anyone else who is interested.

Mikko Valimaki: IPR and Open Source Software

• Goodman and Myers (2005) — the 3G standard.
• Leveque and Meniere 2007: what does RAND mean
  • reasonable royalty is R = c (v1-v2)p where c is incremental costs of licensing, v1-v2 is gain from using this pattern over second-best.
• Other questions for royalty-setting
  • quality of volume of patents
  • early or late innovators
  • cumulative royalties or one-time fees
• But all models he knows of have non-zero royalty fees
  • [ed]: not surprising given that you will always get interior solutions
• Windows/Samba discussion
  • specific sets of terms
  • provide RF for the open source community
• Commission Decision para 783
  • “On balance, the possible negative impact of an order to supply on Microsoft’s incentives to innovate is outweighed by its positive impact on the level of innovation of the whole industry.”
• Nokia to acquire Symbian:
  • “a full platform will be available … under a royalty-free license … from the Foundation’s first day of operations … the Foundation will make selected components available as open source at launch.”
  • [ed]: Motivation here is clear: Nokia care about the hardware and for them software is a complementary good — which they therefore wish to be as cheap as possible. But this raises question as to what is being made open: is hardware patents or pure software patents (and if so how big a deal is this)

Stefan Koch: Efficiency of FLOSS Production

• Question of efficiency of open source development
• How much software did we get for our effort
  • Is OS a waste of resources?
• Discussion without much empirical basis
  • Claim: fast and cheap, high quality, finding bugs late is inefficient (actually large effort) — see IEEE Software 1999
• Completely unknown as no-one keeps time-sheets. So
  • Effort based on participation data
  • Effort based on product — look at software and ask how much effort would be needed in commercial environment
• Empirical research in open source
  • Mainly case studies
  • Helpful but need proper large-scale analysis
• Mined software repositories [ed: cf. today FLOSSMatrix, FLOSSMore]
  • 8,261 projects
  • 7,734,082 commits
  • 663M LOCs
  • resources and output is skewed: top decile of programmers: 79% of code base, second decile: 11%
• Effort estimation based on actual participation
  • active programmer months (define active as committing in a given month)
  • high correlation with LOC added in month
• Cumulate this number for each project
  • But not equal to a commercial person-month
  • How do we scale: use 18.4 h/w taken from stats for committers on Linux kernel
  • [ed:] this is the key assumption. The whole point is that FLOSS effort is not observed and they are using a measure of output (committing) and trying to infer actually activity
• Manpower function modelling:
  • Norden-Rayleigh model (1960)
  • Some set of problems N (unknown but finite)
  • Probs are solved independently and randomly (following Poisson)
  • This fits ok but has eventual decline in participation which does not occur
  • Modify this: in particular to allow introduction of new problems
    • Introduce in prop to original no. problems, in prop to current set of problems etc
    • Also have different learning rates
    • [ed: but isn’t the setup a little different. Really it is a question of success vs. non-success in terms of acquiring users + some kind of bound on amount of participation due either to fission or complexity]
• Product-based estimation
  • COCOMO 81 and COCOMO 2
• Results:
  • Comparison COCOMO - Norden-Rayleigh
  • For COCOMO 81 cannot find parameters favourable enough to explain Norden-Rayleigh curve
  • For COCOMO 2 can find parameters but very favourable
  • Suggest (roughly) that FLOSS very efficient (but not very rigorous)
• More formal estimation using all models etc
  • Norden-Rayleigh significantly below prodcut-based estimates (factor of 8 in mean)
• Interpretation
  • FLOSS v. efficient (self-selection for tasks etc)
  • Extremely high amount of non-programmer participation (1:7 relation …)
• [ed]: not sure about this generous view. Other explanations
  • No quality measurement (also mentioned by Koch)
    • OK: lot of code but low quality
  • (Related) Many sourceforge projects are incomplete, easy bit at the start
    • Later comes a lot of refactoring/writing documentation. This may display significant diminishing returns
  • Many FLOSS projects come from what were originally commercial projects. In that case:
    • code may have already been written
    • conceptual components have been done already
  • Trade-off of time vs. productivity
    • May be more productive to only work 10h a week but then product might not be ready for 10 years
• Form discussion
  • interesting point: Nokia thinking of moving to more FLOSS in-house because they can’t manage their 5-10k programmers centrally any more

Mickael Vicente: Shift to Competences Model: A Social Network Analysis of Open Source Professional Developers

• Robles 20007
  • Statistics on Debian showing increasing corporate involvement
• Social network extraction
  • Get repo logs
  • Create link between 2 developers if they have committed on the same file (non-directed graph)
    • Simplification: the best collaboration of each developer (directed graph) — pick other developer with whom they have committed most files in common
  • Longitudinal analysis
    • extract clusters
• Correlation with professional career
  • CV collected on Internet, personal web page etc (96% collected)
• Interesting data

Nicholas Radtke: What Makes FLOSS Projects Successful: An Agent-Based Model of FLOSS Projects

• Positive Characteristics of FLOSS
  • High quality (Low defect count: Chelf 2006)
  • Rapid development
  • Violates Brooks law (Rossi 2004)
  • Risky Business
• for every successful FLOSS project there are dozens of unsuccessful projects
• Corporate IT manager survey (2002)
  • 41% mention inability to hold someone responsible for software
• Attempts at Simulating FLOSS
  • SimCode (Dalle and David 2004)
  • OSsim (Waggstrom et al 2005)
  • …
  • K-Means stuff
• Simulate across landscape
  • Not social network
  • Focus on developer decision to join/contribute to projects (Agent-Based Modelling)
• Defining Success and Failure
  • Traditional metrics do not work well (on budget?)
  • Completion (Crowston et al. 2003)
  • Progression through maturity stages (Crowston and Scozzi 2002)
  • Number of developers
  • Mailing list activity
  • Project outdegree, Active developer count (Wang 2007)
• The Model Universe
  • Agents and projects
  • Agents:
    • Consumption: 0-1
    • Producer: 0-1
    • Resource: 0-1.5 (1=40h)
    • Memory: agents only aware of some subset of projects
    • Needs vector (preferences)
    • utility: linear sum of: similarity match + current popularity (current resources) + cumulative resources + download + f(maturity)
  • Projects:
    • resources needed
    • current resources
    • cumulative resources
    • download count
    • preferences: same as agent but converges towards those had by agents working on it
• Agents choose between projects each time period
  • have some randomness in that use multinomial logit: prob choose project i ~ exp(mu * Utility of project i)
• Results
  • Simulate over 250 time steps ~ 4 years
  • calibrate [ed: in a way I was not quite clear about]
  • compare simulation with empirical data from sourceforge
    • developers per project
    • projects per developer
  • Find that (from simulation data) downloads and cumulative resources are not important

Fabio Manenti: Dual Licensing in Open Source Software Markets

• Benefits of Going Open Source
  • feedback from community
  • network effects (usage)
  • competitive pressures (e.g. Netscape) [ed: not sure this is a benefit]
• Dual-licensing
  • Kosky (2007): 6% of representative sampl of European OSS business firms employ DL strategies

Alexia Gaudeul: Blogs and the Economics of Reciprocal (In-)Attention

• What blogs are
• Reasons for blogging
• Question: do you befriend (link) because of content produced or do you produce content because of friends
• General points
  • Market interactions only part of wider class of reciprocal relations
  • Time vs. money economics
  • Unique dataset, very detailed and complete, to test networked relations
• Model — but left out due to time
• Dataset: livejournal 2006
  • Sociology: teenagers to young adults (15 to 23), female (67\%), Americans (70\%)
  • Fast growth: created in 1999, 8M accounts, 1.3M active
  • FLOSS but for-profit (SaaS)
  • Great part from self-referential
  • Lively: 4 comments per post on average
  • Federated by communities: no. of communities per person 15
  • Journals updated for more than 2 years on avg
  • 70\% have posted in last 2 months
  • No. of entries: 1 every 2 days
  • No. of friends: 50 avg
  • Balance between friends and friends of
  • Balance between comments received / made
• Friendship patterns
  • May be balance but does not explain no. of friends of diff. individuals
  • Need to distinguish
    • Norm of reciprocity: more promiscuous bloggers accumulate friends
    • Content attractiveness
      1. Quality/freq. of posts
      2. Interactivity (comments per post)
• Regressions
  • Reciprocity: No. blogs read (friend) = b * number of readers (friend of) + error
  • Activity: No. readers = cX + error — X = matrix of ind. variables
  • Endogeneity issues [ed: all over the place)
  • Regress: ln(Friends) = ln(Friend of) + … (with instrumenting Friends Of on Activity so solve endogeneity issues)
    • Saturation around 400 friends seemingly (few with more)
  • Max no. of friendship when your no. friends = no. friends of (maybe)
    • A norm of reciprocity
  • Issues with endogeneity of activity (which was used to instrument friends of)

Sylvain Dejean

• Does ICT lead to the Internet lead to a global village or a cyber-balkan
• What leads to emergence of virtual commmunities
• Is the heterogeneity of contributions an impediment to self-organize
• How to manage virtual communities
• Agent-based model:
  • Individuals defined by some characteristics
  • Herfindahl index measures degree of self-organization [ed: why self-organization]
  • Communities change via selection and variation

Yesterday I attended the sixth and final of the series of “Workshops on Well-being” taking place at the LSE (I missed the fifth workshop as I was away and so the last one I attended was the fourth workshop back in April). This time the presentations were given by David Clark of KCL and Martin Knapp of LSE and KCL. Below are some heavily impressionistic notes.

Presentation by David Clark (KCL): Developing Effective Psychological Treatments for Common Mental Health Problems

1. Anxiety disorders

   • ~ 1/2 of mental health problems
   • overly pessimistic view on outcomes etc
   • can become obsessional (+ fear that thoughts are self-realizing)
   • If beliefs are inconsistent why do they persist
   • panic attacks (~ 30% have them once/v. occasionally but realize that they are not dying). But in the disorder people might have had them 5000 times — how can they still think they are dying when it happens again?

2. Research Strategy:

   • identify core cog. abnormality
   • …

3. Example: social phobia

   • most common anxiety disorder (lifetime prevalence: 12%)
   • persistent: natural recovery rate is 37% over 12 years
   • marked underachievement
   • persists because:
   • shift to internal focus (which means ignore external)
   • use of internal information to infer how one appears to others (and as they are anxious this unreliable)
   • safety behaviour
   • test some of this
   • Do high socially anxious individuals have an internal attentional bias (Mansell, Clark, Ehlers 2003)
   • Evidence that socially anxious individuals have a distorted external perspective (Hackmann, Surawy and Clark 1999)
   • Evidence that onset of phobia correlated some stressful (bad) social event
   • Does negative self-image affect relation with others. Yes, to some extent (another Clark paper)
   • treatment (Cognitive Therapy)
   • attention training
   • drop safety behaviours (to test no adverse consequences)
   • video feedback
   • rescripting early memories
   • does CT pass the randomized controlled trial: YES
   • compare against no treatment
   • placebo
   • at least as effective as medication

4. Common disorders where CBT is effective as a sole treatment (recovery rate, controlled effect size):

   • Major depressive disorder: 50%, -
   • Panic disorder 75%, 2.8
   • PTSD: 80%, 2.3/1.2
   • Social phobia: 75% 2.6
   • Generalized anxiety disorder: 50% (77%)
   • OCD: 45%, 1.5
   • Also show that effects of CBT persist for anxiety (unlike psychotropic interventions where there is a high relapse rate)
   • depression slightly different as naturally recurrent — though CBT still effective (and complementary to medication). Hollon et al (2005) (Arch Gen Psychiat) compare medication vs CBT over long-term and shows CBT better.

5. Evidence that benefits of CBT extend outside of targeted syndrome. Beneficial effects for:

   • other mental health problems
   • work, family, social adjustment
   • employment (less sick days, moving to work)
   • but these effect sizes are lower bound (overall want SWB scores …)

6. Developing more effective (shorter) treatments

   • Traditional approach is 1h/w for 3-4 months
   • but 1-2h of ‘homework’ per day between visits
   • Now trying intensive 1w course (~ as effective at least for PTSD)
   • Also treatments with extra-focus (e.g. social phobia + work: found big impact on time to get back to work)
   • CBT with well-being emphasis. Fava et al (2005) (Psychotherapy and Psychomatics). Find CBT-WB > CBT but tiny sample, no blind assessment etc.

7. Major policy changes underway to increase access to CBT

Martin Knapp (LSE + KCL): Economics of Mental Health: Some Open Research Questions

1. Why mental health is different

   • breadth/multiplicity of need
   • association with crime + violence
   • associated with suicide
   • compulsion, stigma, complex links with ethnicity

2. Leading policy/practice themes

   1. stigma/rights/social exclusion
   2. funding
   3. Balance of Care
   4. Treatments
   5. Prevention

3. Social exclusion, stigma, etc

   • Participation-based approach
   • opportunities, socio-economics roles
   • Rights-based approach
   • stigma, discrimination, compulsory treatment
   • If i were suffering from mental health problems I don’t want anyone to know (Scotland): 50% in 2002 to 41% in 2006 (following a big campaign)
   • evidence in UK actually may be getting worse (16% 2000 to 22% in 2007 on similar question)
   • Equity: great variations (inequality greater for mental health than for income), esp by ethnicity.
   • Costs:
   • total cost of depression £9 billion (Thomas and Morris, Brit J Psychiatry 2003)
     • mostly productivity effects (not service or morbidity)
     • prob. underestimate as also have staff turnover, presenteeism
   • major impact of psychosis on life-time development [ed: not exactly surprising …]
   • homicide: Taylor and Gunn (Brit J Psychiatry) show that across various European countries between 5 and 20% or homicides committed by those who are mentally ill

4. Funding

   • Mental health spend as %tage of total public spend: England is highest in EU [ed: is this good or just that England has a lot of mental health issues]
   • Good efficiency arguments for intervening (cost-effective)
   • Schizophrenia: total cost ~ £54k per person per year (only a 1/3 hits the health system)

5. Balance of Care

   • Massive reduction in number psychiatric beds (personal preferences, social preferences etc)

6. Treatments

   • Does it work?
   • Is it cost-effective? etc
   • In 2000 only 53% of people with depression received treatment compatible with NICE guidelines
   • More attention to non-health interventions
   • particularly risk factors such as bullying, family violence, uncontrolled debt

7. Prevention

   • Inner London Longitudinal Study (ILLS)
   • Study of all 10y old in part of London in 1970
   • Categorise into groups from: “no problems at school” to “conduct disorder”
   • Estimate costs to society per child from 10 to 28 (education, criminal justice, social services etc)
     • no problems: ~ 7k, conduct: ~ 24k, conduct disorder: ~70k (mostly criminal justice)
   • 1970 British Cohort Study
   • earnings at age 30 by childhood need at age 10
   • no probs: ~24k, behavioural (lowest quartile): same, Cognitive (lowest quartile): 15k, emotional (not a great effect but interacts in a minor way with cognitive). Another study finds same effects for behavioural at age 32 but extended to 48 finds same -ve effect of cognitive issues.

One the major things I’ve been working on since last summer (other than the work on Trading Funds) is a paper on search engines such as those provided by firms like Google, Yahoo! etc. The first complete version of this is now ready for public consumption. Entitled Is Google the next Microsoft? Competition, Welfare and Regulation in Internet Search I’ve posted it online at:

http://rufuspollock.org/economics/papers/search_engines.pdf

Abstract

Internet search (or perhaps more accurately ‘web-search’) has grown exponentially over the last decade at an even more rapid rate than the Internet itself. Starting from nothing in the 1990s, today search is a multi-billion dollar business. Search engine providers such as Google and Yahoo! have become household names, and the use of a search engine, like use of the Web, is now a part of everyday life. The rapid growth of online search and its growing centrality to the ecology of the Internet raise a variety of questions for economists to answer. Why is the search engine market so concentrated and will it evolve towards monopoly? What are the implications of this concentration for different `participants’ (consumers, search engines, advertisers)? Does the fact that search engines act as ‘information gatekeepers’, determining, in effect, what can be found on the web, mean that search deserves particularly close attention from policy-makers? This paper supplies empirical and theoretical material with which to examine many of these questions. In particular, we (a) show that the already large levels of concentration are likely to continue (b) identify the consequences, negative and positive, of this outcome (c) discuss the possible regulatory interventions that policy-makers could utilize to address these.

I’ve added a reasonably detailed treatment of Stackelberg Competition to the Atlas (of Economic Models).

After attending the IIOC conference last year I was back this weekend at the 2008 IIOC event which took place at Marymount University in Virginia. I presented the latest version of two of my papers: The Control of Porting in Two-Sided Markets and Forever Minus a Day? Theory and Empirics of Optimal Copyright Term.

I also provided discussant comments on Christopher Ellis’s and Wesley Wilson’s paper entitled Cartels, Price-Fixing, and Corporate Leniency Policy:What Doesn’t Kill Us Makes Us Stronger. In addition I include below some very partial notes on some of the sessions I attended — though activity in this regard was rather limited by the fact that, though there were more papers overall than last year (388 in total), sessions were organized into more breadth and less length.

Transaction Costs and Trolls: the Behaviour of Individual Inventors, Small Firms and Entrepreneurs in Patent Litigation (Gwendolyn Ball and Jay Kesan)

• Explore settlements in relation to patents. Questions:
  • How often do settlements happen relative to litigation
  • Are small firm and entrepreneurs at a major disadvantage in defending their patents
  • Or do patent trolls' use the threaof litigation toextort’ payments
    • NTP vs. RIM ($612M)
    • Saffron vs. Boston Scientific ($412M to individual doctor who had an infringed heart stent patent)
  • Does nature of defendant/plaintiff (L/M/S) affect likelihood of settlement
• Existing databases not so great
  • Only list trial outcomes not pre-trial outcomes
  • Often only list primary plaintiffs
  • Fix this and link patent litigation to companies
• Results
  • Claimed usually that 95% cases settle
    • In fact 8% are resolved at pre-trial (still expensive)
    • 4% settled at trial
    • so ~ 88% settle
  • Troll stuff:
    • 97 licensing firms as plaintiffs (none as defendants). These may be classic trolls but they are a small part of overall litigation.
    • Evidence shows that entrepreneurs and small inventors are very active (so do not seem particularly disadvantaged) and often sue each other rather than larger firms
    • Crudely: small inventors more likely to pursue a case to the end than large litigators
• Discussant comments:
  • Bessen and Meurer find $28M hit on firms facing litigation
  • Issues of correlated errors across cases
• My comments:
  • probably need to disaggregate across areas — after all no-one has suggested ‘trolling’ is an issue in traditional pharma
  • (for me) it would be useful to have an idea how many cases ’settle’ at the ‘letter stage’, that is, before anything even turns up in the court system. After all you only get to the courts (even with preliminaries) if you cannot sort out a license.

Prior Art - To Search or Not to Search (Vidya Atal)

• Alcacer + Gittelman 2006 showed 40% had prior art added by USPTO examiner
• 2/3 citations on an average patent added by USPTO
• Langinier + Marcoul (2003), Lampe (2007) — incentive to disclose prior art
• Issue of bad (non-novel) patents may be because people have poor incentives to search
• Mainly related this to fact that even a bad patent (if it gets past examination) has a +ve payoff

Today I’ll be presenting my paper Forever Minus a Day? Theory and Empirics of Optimal Copyright Term at Stanford in the Social Science and Technology Seminar Series (also here).

This new paper is a heavily revised version of the copyright-term specific portions of my original ‘Forever Minus a Day’ paper (see post from last summer). The rest of the original paper can now be found in Optimal Copyright over Time: Technological Change and the Stock of Works which was published in the December issue of the Review of Economic Research on Copyright Issues (RERCI).

Update post-talk (2008-05-16): the slides are now online at:

http://rufuspollock.org/economics/papers/optimal_copyright_term_talk_stanford.pdf

This January I met Alan Kirman at the Robinson Workshop on Rationality and Emotions. Over lunch we had a brief discussion about the difficulties of modern macroeconomics. I was therefore intrigued to see a new paper of his (co-authored with Peter Howitt, David Colander, Axel Leijonhufvud and Perry Mehrling) entitled Beyond DSGE Models: Towards an Empirically-Based Macroeconomics which was presented in January at the AEA conference (and looks like it will be appearing in the AER ‘Papers and Proceedings’).

The paper has much to say about the current state of macro, in particular the serious problems with DSGE (dynamic stochastic general equilibrium models) and where we should go from here. As the abstract puts it:

This paper argues that macro models should be as simple as possible, but not more so. Existing models are “more so” by far. It is time for the science of macro to step beyond representative agent, DSGE models and focus more on alternative heterogeneous agent macro models that take agent interaction, complexity, coordination problems and endogenous learning seriously. It further argues that as analytic work on these scientific models continues, policy-relevant models should be more empirically based; policy researchers should not approach the data with theoretical blinders on; instead, they should follow an engineering approach to policy analysis and let the data guide their choice of the relevant theory to apply.

It is worth quoting at some length from the paper in order to bring out the full ramifications of the story the authors tell:

Keynesianism Goes Wrong

With the development of macro econometric models in the 1950s, many of the Keynesian models were presented as having formal underpinnings of microeconomic theory and thus as providing a formal model of the macro economy. Specifically, IS/LM type models were too often presented as being “scientific” in this sense, rather than as the ad hoc engineering models that they were. Selective micro foundations were integrated into sectors of the models which give them the illusory appearance of being based on the axiomatic approach of General Equilibrium theory. This led to the economics of Keynes becoming separated from Keynesian economics.

The Reaction and a New Dawn (Rational Expectations and Neoclassical GE Models)

The exaggerated claims for the macro models of the 1960s led to a justifiable reaction by macroeconomists wanting to “do the science of macro right”, which meant bringing it up to the standards of rigor imposed by the General Equilibrium tradition. Thus, in the 1970s the formal modeling of macro in this spirit began, including work on the micro foundations of macroeconomics, construction of an explicit New Classical macroeconomic model, and the rational expectations approach. All of this work rightfully challenged the rigor of the previous work. The aim was to build a general equilibrium model of the macro economy based on explicit and fully formulated micro foundations.

But ‘Technical’ Difficulties Intervene

Given the difficulties inherent in such an approach, researchers started with a simple analytically tractable macro model which they hoped would be a stepping stone toward a more sensible macro model grounded in microfoundations. The problem is that the simple model was not susceptible to generalization, so the profession languished on the first step; and rational expectations representative agent models mysteriously became the only allowable modeling method. Moreover, such models were directly applied to policy even though they had little or no relevance. … [emphasis added]

But There Was a Reason For This: Other Stuff is Hard

The reason researchers clung to the rational expectations representative agent models for so long is not that they did not recognize their problems, but because of the analytical difficulties involved in moving beyond these models. Dropping the standard assumptions about agent rationality would complicate the already complicated models and abandoning the ad hoc representative agent assumption would leave them face to face with the difficulties raised by Sonnenschein, Mantel and Debreu. While the standard DSGE representative models may look daunting, it is the mathematical sophistication of the analysis and not the models themselves which are difficult. Conceptually, their technical difficulty pales in comparison to models with more realistic specifications: heterogeneous agents, statistical dynamics, multiple equilibria (or no equilibria), and endogenous learning. Yet, it is precisely such models that are needed if we are to start to capture the relevant intricacies of the macro economy.

Building more realistic models along these lines involves enormous work with little immediate payoff; one must either move beyond the extremely restrictive class of economic models to far more complicated analytic macro models, or one must replace the analytic modeling approach with virtual modeling. Happily, both changes are occurring; researchers are beginning to move on to models that attempt to deal with heterogeneous interacting agents, potential emergent macro properties, and behaviorally more varied and more realistic opportunistic agents. The papers in this session describe some of these new approaches. [emphasis added]

Some Closing Comments of My Own

So there you go: plenty of tough challenges and a big dose of humility. To some extent here it seems thing run on 30-40 years cycles: Keynesianism from 1945-1975, Rational Expectations DSGE from 1975-2005 and now we’re into the era of complexity and ‘loose’ tools with emphasis on empirics and heuristics rather than formal models. Whether this new approach will deliver more than the old is yet to be seen. After all, one reason that there are so many physicists getting interested in Economics and Finance is that the going is so hard in, e.g., condensed matter physics (superconductivity anyone …). If the economy really is so complex will we ever do any better at the macro scale than we do for the weather and if so will it not rely on some conceptual breakthrough rather than just doing using more hard-core dynamical systems theory and running more agent-based simulations?

That said, as the authors argue, the ’simple’ route isn’t working and the hardness of the path is no reason not to attempt it — an argument in many ways directly inverse to the traditional ‘drunkard-and-the-lamp’ approach in which we restrict our models, often beyond the point in which they remain relevant, in order to maintain analytical tractability. Thus, though cautious regarding what more ‘complexity-oriented’ methods can deliver, I am in wholehearted agreement with the authors that they justify much greater exploration.

Over the last couple of months for the purpose of my research on happiness/subjective-well-being I’ve been putting together some notes on theories of contextual judgement. The first part of these is now in a form suitable for public consumption and I’ve posted them at:

http://www.rufuspollock.org/economics/notes/theories-of-contextual-judgement/

Following on from the third workshop a month ago, yesterday saw the third in the series of “Workshops on Well-being” take place at the LSE. This time the presentations were given by Mat White of Plymouth University and Andrew Steptoe of UCL. Below are some (very) impressionistic notes.

Presentation by Mat White (+ Paul Dolan): Accounting for the richness of our daily activities

1. Social psychologist: started out on risk perception, trust etc. (Fear of crime)

2. General problems with life satisfaction data

   • lots of it deals with attributes which are beyond realm of govt intervention (e.g. race, gender)
   • Response/cross person comparisons issue: same externals result in different reported happiness levels across individuals (e.g. old, poor people are happiest in Dolan’s Welsh data, perhaps because of a “Don’t grumble” attitude). [ed: essence is the qualia problem: can we compare different people’s report of their internal states, both across people and across time. Or more pithily: is my ‘Good’ or ‘OK’ the same as your ‘Good’ or ‘OK’?]
   • Subjective well-being isn’t one thing but a composite: SWB = Feelings + Thoughts + Time

3. Solutions

   • Experience Sampling Method: ask people during day
   • Problems: costly, only points in time, no duration etc
   • Day Reconstruction Method (DRM): solve duration issues
   • Can now base utility as integral of well-being function over time (ed: what utility always was but we just didn’t have the moment by data)
   • Find what one might expect re. what activities are nice
   • However no/v. weak correlation with e.g. income
     • But maybe because those payoffs are in the future
     • Or maybe because there are rewards in terms of thoughts, feelings about themselves etc (Eudamonia)

4. This project: add thoughts (about activities) to DRM

   • 625 Germans
   • 5815 Episodes (3057 single activities)
   • Online panel
   • Have 12 adjectives they can use which break down into ‘pleasurable’ and ‘rewarding’

5. Adding in Eudamonia makes a big difference!

   • Nice graph contrasting the DRM with ‘pleasure’ vs. ‘rewarding’ (at least partially inversely correlated).
   • Argue that we should sum both ‘eudaimonic’ and ‘hedonic’ evaluations over whole day.
   • Can now plot activities on x-y graph with x=hedonia, y=eudamonia (normalized about the mean values)
   • Get a slight -ve correlation
   • ed: this makes sense due to selection effects. Let w be total well-being and h hedonia score, e is eudamonia score. Suppose w is a linear combination of these underlying factors: w = h + k e. Now we would generally choose only to do activites with w > w0 (some outside option) => h+ke > w0 which gives the -ve correlation.
   • If reweight with duration [ed: equivalent to doing integral] then get a slight +ve correlation

   • ed: this reweighting by duration causes major changes to the form of the data. In particular all longer activities receive a positive shift while short ones receives a negative shift (explanation below). Whether this is what could/should do with the data was not entirely clear.

     Why does this shift occur. Results are plotted as ‘relative’ values (i.e. normalized about the mean). Thus if original value (x,y) it is plotted at (x-m1,y-m2) where m1 is the overall x-mean and m2 is overall y-mean. Adding duration means original values are now (dx,dy) and these are plotted relative to n1,n2 where n1,n2 are new duration weighted means.

     Letting dbar be the mean duration we could make the rough approximation that n1 = dbar m1, n2 = dbar y1. Then the new x position is: dx-n1 = dx - dbar m1 = d(x-m1) + (d-dbar)m1. Hence the new x-position will be a combination of a linear scaling out from the origin by d plus some offset of (d-dbar)m1. Since m1 is always positive this offset is positive (negative) as the duration of the activity is greater (less) than the mean duration of an activity.

6. After discussion

   • pop-ups (thoughts either +ve or -ve) have a big impact
   • in a regression on day-satisfaction number of +ve and-ve popups explained more than hedonic or eudamonic variables (total value for whole day)
   • could be useful to look at something more than a simple integral [ed: e.g. use contextual judgment stuff]
   • Eudamonia: enters day satisfaction regressions negatively. This is what we would expect given association of ’satisfaction’ with ‘pleasurable’ activites and slight negative correlation of ‘rewarding’ (eudamonic) activities with ‘pleasurable’ (hedonic) ones.
   • ed: could interpret eudamonic value as discounted future value coming from associated payoffs. I.e. if I work hard now this might not be pleasurable but it has high eudamonic content reflecting the future hedonic payoffs (nice garden, good holidays etc) of doing that work (NB: this is intentionally putting things very crudely).

Andrew Steptoe: DRM Analyses

1. Primarily interested in ‘positive affect’ and health outcomes

2. Questions:

   • how accurate is DRM
   • what does DRM tell us about activities and feelings of depressed people

3. Data: Daytracker study

   • 200 healthy women in full-time work
   • 2 x 24hr starting @ 5pm (one work day and one non work-day)
   • International dimension
   • EMA and DRM

4. Comparing EMA and DRM

   • Across aggregate data already see some differences (DRM shows noticeable rise towards end day while EMA does not really show this)
   • Per individual: similar differences but also fairly close correlation
   • Doing actual correlation looking at 4 different aspects (happy, tired, stress, anger) find medium correlations (0.2-0.7) which is reasonable but not great
   • also noticeable that timepoints are important: worst correlations are generally 12noon and 3pm

5. How accurate is the DRM for estimating feelings (esp. in relation to depression)

   • Do depressed people: have diminished pleasure in all activities or is reduced exposure to good stuff?
   • Depressed people are less happy across most interactions (except with Grandchildren) though effect (of depression) does vary and is strongest for being Alone or with your Partner
   • Looking at time: depressed people seem to spend more time (compared to non-depressed) doing things they don’t like
   • Similarly, looking across activities, depressed people are less happy doing most stuff
   • Again looking at time, seem to find depressed people spending more time on things that they particularly dislike (relative to others)
   • [ed: Not sure what this is telling us. After all the activities depressed people spend more time on may still be better than other options even if those options do not get as large a negative ‘hit’ from being depressed — e.g. house-work may not be much worse when depressed than non-depressed but it still might be worse than everything else]

Several years ago I read Michael Kremer’s article entitled “Randomized Evaluations of Educational Programs in Developing Countries: Some Lessons” in the 2003 AER Papers and Proceedings issue (jstor link). This brief article reviewed some of the recent results of evaluating the effects of various different programs on educational outcomes in the developing world. What particularly caught my eye was this paragraph summarizing a teacher incentive program in Kenya:

Some parent-run school committees in the area provide gifts to teachers whose students perform well. Glewwe et al (2002a) evaluate a program that provided prizes to teachers in schools that performed well on exams and had low dropout rates. In theory, this type of incentive could lead teachers either to increase effort or, alternatively, to teach to the test. Empirically, teachers responded to the program not by increasing attendance, but by increasing prep sessions designed to prepare students for exams. Consistent with a model in which teachers responded to the program primarily by increasing effort devoted to manipulating test scores, rather than by increasing effort at stimulating long-term learning, test scores for pupils who had been part of the program initially increased but then fell back to levels similar to the comparison group at the end of the program. [p. 104, emphasis added]

This provides a nice ‘real-world’ example of exactly what can go wrong when providing incentives in a multi-task situation — that is one where the ‘agent’ (here the teacher) performs multiple tasks not all of which can be monitored equally. As such it should make us wary of the current trend to ever more performance-based reward structures in everything from schooling to health-care.