Simplex Plots

In an n-player game with three types or strategies, it is possible to graph "evolutionary trajectories" using replicator dynamics, but this requires a 'simplex': an equilateral triangle in which each point represents a combination of three proportions that sum up to one. The vertices are the points at which 100% of the players choose a given strategy. An example is adding a "Bourgeois" strategy that plays hawk if "home", dove if "away" to the classic Hawk-Dove game:

Using replicator dynamics, we can then evaluate for any given proportion of types where the evolutionary trajectory leads us - and represent this in the simplex using arrows:



Making these graphs "by hand" is way too painful. Luckily Richard McElreath at UC Davis wrote an r-package called baryplot which lets you make simplex plots very easily. It only requires specifying the payoffs for each strategy and starting points for the arrows.

Monty Hall versus Cognitive Dissonance

The New York Times' Tierney Lab blog has a nice summary of a debate between Yale psychologists Egan, Santos and Bloom and Economist Chen:

TierneyLab: Monkeys, Candy and Cognitive Dissonance

To rescue the theory of cognitive dissonance from the Monty Hall Problem, an economist argued that psychologists needed to redesign their experiments. Now the psychologists have done just that.

The psychologists apparently forgot to account for the Monty Hall problem in their prior work.

"Remember, you're sitting in a chair, in the sky!"

A comedian's observation of the endowment effect (cf Kahneman and Tversky):



While I am of course in many respects also often guilty as charged, the fascination of flying doesn't seem to be wearing off. Spotting things from an airplane is great fun! Especially when I can see large geographic structures, or famous landmarks. Things that are usually only visible on a map or satellite images, such as the Lake of Constance and my Alma Mater when flying out of Zurich, or the Øresund including the Bridge, or Long Island... try spotting Stony Brook University when flying into JFK next time, it's relatively easy as the campus has a very distinct shape....

Swine flu anyone?

Swine flue is almost forgotten already. Hans Rosling brings it to the point:

Electoral map of Poland

This post at the very entertaining Strange Maps blog blew my mind - an electoral map of Poland. The divide between PO and PiS voting in the 2007 legislative elections follows tightly the border of Imperial Germany.



Actually looking at the data is often enough stressed as an important and helpful first step when dealing with quantitative information. And electoral results displayed as maps ("Red state - blue state...") of course isn't exactly a new thing. Additionally, ever since John King showed off CNN's magical map on an hourly basis during the last presidential election I got a little tired of seeing the U.S. as a blue and red patchwork. But this electoral map of Poland is really quite interesting. Today's political cleavage echoes the turbulent past. Obviously the potential reasons for this effect are numerous: different political and economic structure, genocidal tyranny in the "Generalgouvernment"-part during the occupation, deportation and re-settlement of the western part after the Second World War. After all, some research reveals that this divide between "Poland A" and "Poland A" seems well known in the Polish public discourse.
Admittedly, I don't know too much about Polish electoral behavior, but it is still amazing to see how even after 45 years of Communist dictatorship and almost 20 years after the fall of the Iron Curtain, this line hasn't disappeared.

Barry Schwartz on the Paradox of Choice

Another fascinating talk from TED.com: Psychologist Barry Schwartz claims that the utility of freedom of choice decreases in the number of alternative options. I am not entirely sure whether he is actually right, and what the optimal number of choice alternatives in a given situation would be. This might have interesting implications when taken together with the use of heuristics, though. Also , my own behavior standing in a Starbucks or choosing cereals strongly supports his hypothesis.

NYU-CESS conference

This weekend, I attended the 2nd NYU-CESS conference on experimental political science. With even a larger crowd and more papers than last year, the interest in experimental methods in the discipline seems stable or increasing. Unfortunately it was spread over two days, which might be a bit too much. I'd rather see the presentation time curbed and have it in a more compact format.


The department had a strong showing as well, in terms of both attendance and a presentation by Leonie Huddy and Lily Mason. Another interesting paper came from Jamie Druckman, who (among other things) suggests that the duration of framing effects depends on a person's processing mode (see graph).

Some recent stuff from the NYT

New York Times on

"R"

"The self defeating mind"

and the possibility of a love potion (or vaccine?)

Larry Young's Nature article on it is here

Prediction Models, Strömberg and Norpoth

At Freakonomics, Justin Wolfers blogs about predicting the winner in the upcoming election. In the post (here), he presents a great model by David Strömberg who makes his forecast based on states and their electoral college votes, giving the Dem's a 63 percent chance of winning. The paper can be found here.
To bear the department’s standard here: While I like Strömberg’s model for getting closer at the data-generating process, Helmut Norpoth’s very parsimonious model so far did a good job at predicting the popular vote in every election since 1952 with the exception of JFK’s narrow victory over Nixon. All this based mainly on the New Hampshire primary, and therefore allowing for a prediction months before election day. A nice description can be found in Helmut's own guest post at pollster.com and this Washington Post article.
The only problem this time around being that the prediction is virtually a tie, giving only a 50.1 percent lead to Senator Obama.

Mankiw's 10 principles of economics - Revisited

The best-selling introductory economics textbook is Greg Mankiw's "Principles of economics". Everybody who - like me - worked with it, will remember the 10 principles he proposes. Here comes the fastest and funniest summary of these ten principles I have ever heard...

Bush, the CIA, Giuliani's Girlfriend and Proportions.

A quick comment on this week's news and this morning's meet the press: I must admit there are some things where I cease to understand what is going on in this country.
The government de-classified the latest National Intelligence Estimate saying Iran halted its nuclear weapons program in 2003. That came unexpected, but is basically good news, right? Yet this report has been around the administration (while classified) for quite some time, so the government knew of this when Bush delivered his "World War III"-speech... Also the CIA apparently destroyed interrogation videos, I think everybody can see why this is at least somewhat suspicious.
Yet, watch meet-the-press with Republican candidate Rudy "9-11" Giuliani. The candidate who runs almost exclusively on the security issue and thinks waterboarding is an acceptable interrogation technique. As if there weren't more interesting things to spend time on, the show turns into a ten minute questioning on whether and when a security detail for his girlfriend was justified? This is crazy. The state (sadly enough) has to protect its representatives and those close to them. Period. The "softer and gentler" Swedish democracy thought differently and paid for it by loosing Olof Palme and Anna Lindh. But it goes on and on. Tim Russert asks - with a timbre of indignation in his voice - whether the secret service should protect a president's mistress. Well, I would guess so. It's not at all the police's decision how someone behaves in his private life, but rather to make sure an elected representative is not at risk, and this includes the possibility of blackmail... But to save the chance that all the other hypocrites might start liking (or at least vote for) him at some point, Giuliani says "No"...

Comparative Datasets

Gathering some (partly overlapping) datasets:

• Armingeon's Comparative Political Dataset (n=23, t=1960-2004)
• Huber, Ragin, Stephens: CWS (n=18, t=1960-2003)
  
• Scruggs' Welfare Entitlements Dataset (n=18, t=1971-2002)
• UCLA: Golden-Wallerstein-Lange Project (n=16, 1950-2000)
• Swank: Comparative Parties Dataset (n=21, 1950-2002)
• Göteborg Quality of Government Dataset (n=world, 1946-2006)
• Robert Franzese Macroeconomic Policies (n=21, t=1950-1996)
  
• Thomas Cusack's Datasets @ WZB (~n=20, t=1950-1995)
• Przeworski et al. (~n=world, t=1950-1990)

Washing After Sex and HIV...

The NY Times today had an article stating "Washing After Sex May Raise H.I.V. Risk". It discusses the results of a large scale (N=2552) study in Uganda's Rakai district. According to the NYT and apparently a presentation (which I wasn't able to retrieve) at the International AIDS Society Conference in Sydney, which was held by one of the investigators - Dr. Frederick E. Makumbi, who is affiliated with local Makerere University, but also Johns Hopkins' Bloomberg School of Public Health - the data implies that there is either no or even an adverse effect of post-coital penal washing on HIV-Risk. In short: the earlier you jump out of bed and run to the bathroom, the higher your risk of an HIV infection. No wonder the newspapers jumped at the story...

This not only contradicts intuition but also previous findings in infectious disease research, and you can go and read the whole NY Times article for speculations on possible causal explanations and (after-) sex tips by Johns Hopkins' Dr. Ronald H. Gray.

Not having access to the study makes it a little difficult to challenge the findings on a sound basis, but I am nevertheless inclined to voice some doubts about the results. While assuming the setup and conduct of the study itself are proper, there are at least three points that immediately come to mind:

First of all it's not a clinical study, but a field study involving a survey. So there is obviously no randomization in terms of after sex washing behavior (funny idea btw. but somewhat unethical should there be really some connection between washing and HIV-infection-risk), which makes it a little difficult to establish causality.

Secondly, whereas the existence of an HIV infection can be measured relatively accurate, the long-term-reported-after-sex-washing- behavior is a different story. Not only from the point of view of reported behavior in general! There might also be reason to think about social desirability in the subjects... In other words: Wouldn't you prefer to say you wash yourself a lot when the epidemiologists come?

And thirdly and in my eyes most obviously: the relation between washing behavior and the cicumstances of the sexual contact: I would bet that the usual immediate post coital washing behavior (if any) in a long term relationship differs massively from say the one after visiting a prostitute... Bearing these aspects in mind, the "don't wash" results might very well be not as valid as the data seems to suggest.

Israel's chances to disarm Iran unilaterally

At Henry Farrell's Political Science Weblog, I found this article (published in International Security). Whitney Raas and Austin Long assess Israel's capabilities to destroy Iran's Nuclear Facilities - in quite some detail.
Although I found it quite an interesting piece of reading, I am wondering whether such a study in a public journal is a good idea (assuming Iran has a "International Security" subscription ;) ). Granted, the facts it is based on are public, and the Israelis probably have better intelligence and risk-assessments. On the other hand - insights and conclusions by independent experts might be an interesting read for the Iranian leadership, as their internal assessments might be biased. In addition, the paper provides some hints where weaknesses are - on both sides. Being no military expert, I cannot assess the real relvance of the study. From a political point of view however, more uncertainty i.e. incomplete information on the Iranian side of the chessboard seems desirable if one is interested in a solution that leaves Iran "nuke-less"...

GoogleScholar vs. SSCI

vs.

Searching literature and identifying the relevant papers for your academic work is - in my eyes - one of the most time-consuming and sometimes annoying activities. Databases that do not feature sorting in terms of "relevance" may be helpful for very specific searches, but are quite useless when you want an overview of a large field. A good alternative are always literature reviews, however - you need to find one and most importantly: it has to be your topic.
Most social scientists rely heavily on Thomson's SSCI. Over the last few years Google Scholar has emerged as a potential rival or at least a supplement to SSCI-based approaches. Apart from being free - which is important as not all universities and certainly no laymen can afford the SSCI access - Google Scholar has some other advantages over the SSCI, but there are drawbacks as well. Finding recent working papers and also several versions of published papers is a great advantage in my eyes. As Google Scholar works somewhat non-systematic, one might have doubts about the overall coverage and about the reliability of the citation numbers and its relevance sorting. Walters (2007) addresses these issues and provides an interesting benchmark of Google Scholar, the SSCI and some other databases using the field of later-life migration:

Because Google Scholar’s search, retrieval, and record management mechanisms are relatively unsophisticated, the current version of GS is unlikely to replace conventional social science databases for serious scholarly work. Google Scholar is nonetheless valuable due to its comprehensive coverage. Within the field of later-life migration, GS indexes 27% more core articles than the second-ranked database (SSCI) and 2.4 times as many core articles as the lowest-ranked database (GEOBASE). Moreover, it covers 88–100% of the core articles found in each of the other seven databases. read more...

While I always had doubts about using Google Scholar when searching literature, this sounds like a good legitimization for using it. For quantification however I would still stick with the SSCI - merely as it is the quasi-standard for citation data.

Philipp Mayr and Anne-Kathrin Walter of GESIS (German Social Science Infrastructure Services) offer another benchmark: in a (german-language) paper and a presentation.

Prediction Markets and the French Elections [UPDATE]

Okay, so the race is over, here are the results, taken from TF1, the forecasts of the FTD prediction market and the last Ipsos poll (absolute deviation in brackets):

	Official Result	FTD forecast	Ipsos
Nicolas Sarkozy	31.18	32.00 (0.82)	30.0 (1.18)
Ségolène Royal	25.87	26.36 (0.49)	23.5 (2.37)
François Bayrou	18.57	18.10 (0.47)	17.0 (1.57)
Jean-Marie Le Pen	10.44	14.00 (3.56)	13.5 (3.06)
Others	13.94	9.54 (4.4)	16.0 (2.06)
Root MSE		2.58	2.15

Thus overall performance is quite good, but the high turnout apparently came unexpected and Le Pen's share was overestimated. If i find the time, I will collect the last pre-election survey results for comparison...

UPDATE: Ipsos performed better in terms of MSE, however the accuracy for the two top candidates and Bayrou remains impressive.

My own portfolio performed ok (winning 3%). Some time before the market closed I sold Sarko at 32, and Sego at 27 assuming both to be overvalued (that was close...). I overestimated Le Pen, but that wasn't too costly. However I expected a little more of Bayrou... , didn't lose money, though (bought quite a lot at 17.73, real share: 18.57).

Prediction Markets and the French Elections

France and Europe are tautly waiting for the results of the first round of the French presidential elections. Last surveys see Sarkozy and Royal head-to-head - and in the second round. However with 40 percent undecideds, anything remains possible. With the polls closing at 20:00, we will see whether there will be a surprise candidate making it to the second round - like right-wing Le Pen last time. My personal guess is that the centrist Bayrou has indeed some chances of kicking Royal out. But that might also be wishful thinking.

Not only pre-election polls try to give us an idea of the results. There are also prediction markets like the Iowa Electronic Market that tried to exploit the "wisdom of crowds". While the IEM does not have a French presidential election market, the German edition of the Financial Times does. The "strategists stock market" can be found here. Last trades (expected percentages) are as follows:

• Nicolas Sarkozy............... 32,00
• Segolene Royal................ 26,36
• Francois Bayrou............... 18,10
• Jean-Marie Le Pen............ 14,00
• Others.......................... 10,00

It will be interesting to see how it performs compared to classical polls. At Germany's last general election in 2005, although being far from accurate, the "Wahlbörse" prediction market run by the weekly DIE ZEIT outperformed the research institutes. More on this can be found in the article "The predictive quality of political stock exchanges and polls" by Lena-Maria Schaffer, who is now a PhD student at ETH Zurich.

Both The Economist and TIME magazine had nice pieces on the use and usefulness of prediction markets a while ago. They can be found here: The Economist "Information markets", or (no login required): here. TIME asks The End Of Management? (or here).
As a good starting point to learn more, you can read some of the (NBER) papers by Justin Wolfers at UPenn and Eric Zitzewitz of Stanford.

Now I gotta go back to my portfolio and see how my 2120 undervalued Bayrou shares perform...

Bomb, bomb, bomb, bomb bomb Iran...

Republican presidential candidate, Senator John McCain's re-interpretation of the Beach Boys:



Generally I would support the notion that it is acceptable to make fun of almost anything. However, if you are running for this kind of office, you should be a little more careful about your jokes. The tensions between the West and Iran are mounting continuously so that - in the end - if elected, McCain might end up having to make his joke happen. Wouldn't be so funny then, huh?

Virgina Tech University

No need to comment on the event itself, but watch this piece from "The West Wing":



President Bartlet's speech from "20 hours in America". Maybe a little to much pathos, but seeing a tragedy like this at a university, a place of youth and learning and freedom well justifies some pathos...

been there, seen that... [somewhat offtopic]

SPIEGEL online, the online spinoff of the political magazine DER SPIEGEL and one of the fastest and most read german-language news portals published an article on Thailand's Chumphon region today. I agree with almost everything they wrote, this is one of the "undiscovered diamonds" in Thailand:

Die Strände Thailands scheinen mancherorts endlos zu sein. Am Golf von Thailand zwischen Hua Hin und Chumphon kann ein Urlauber lange am Wasser laufen, bevor er ein Dorf oder gar einen anderen Touristen sieht. Weiter südlich bei Surat Thani, Nakhon Si Thammarath und Songkhla haben Besucher ebenfalls viel Platz. Die Infrastruktur für Reisende ist überall gut: Ein schöner Strand, ein sauberes Zimmer, preiswerte Busse, Bahnen, Fähren und Restaurants, dazu exotische Märkte und freundliche Menschen. more

This was one of the most intriguing experiences there: Traveling in a country that welcomes 13.8 million vistors a year usually means never being "alone" at a place, whatever remote. The province of Chumphon (Chumporn being definitely the funnier transcription, but it loses at google: 162,000 versus 684,000) marks an exception: although easily accesible by train from Bangkok, you can find yourself walking endless sand beaches completely alone, maybe with some fishermen's kids playing in the water. Being the only foreigner makes you an attraction and you can experience by yourself what earlier visitors of Thailand meant when writing about the Thai's legendary hospitality and friendliness...

Blogging in political science

Dan Drezner, one of the most prominent bloggers in the political science corner of the blogosphere was asked to write a discussion paper for APSA on academic blogging in political science. A sneak preview can be found... of course here, in his blog. He discusses chances and downsides of blogging in academia and provides 10 pieces of advice to blogging newbies like me.

CiteULike

This is a short endorsement of a great web-based tool named CiteULike. It was created and is run by Richard Cameron, of Manchester Metropolitan University's Department of Computing & Mathematics. It is basically an online database for storing citations. As the citations can be exported to BibTex and Endnote-Format it can replace storing everything in your citation manager. The killer-app-advantage is a function that allows you to post an article from JSTOR or similar databases directly to your citation database. This is done via a bookmark in your browser... So it's basically a one click functionality - in my eyes superior to all citation import functions offered by the journal databases. Another feature is that you can browse other user's libraries so it can be quite useful for drawing on other peoples collection efforts to find articles related to a certain field. Here is my library. The registration and usage are extremly straightforward.... Great Tool !
Here again the link: CiteULike

Explosive means of communication

So six Israeli fighter planes overflew a vessel of the German navy which is part of the UNIFIL mission to control Lebanon's waters and apparently fired two shots and/or flares. First the IDF denied the incident but by now the first reports seem to be valid and: As the ship "Alster" is a so called "fleet service vessel", the Bundeswehr-Dubbing for a spy ship, there is a video of the situation which has not been published so far.

But apart from the media impact which is somewhat dampened by public attention being drawn to pictures of German soldiers posing with skulls in Afghanistan, the reasons of this incident remain interesting.

So... What the hell was this supposed to mean?

My guess is that someone in the IDF wanted to show UNIFIL who is really in charge, but this sabre-rattling gesture backfired massively, although the German and Israeli governments tried to downplay the situation. And viewed in combination with the French force commander's complaints about Israeli fighters patroling southern Lebanon and his threat to station anti-aircraft-systems to defend his ground troops, the situation becomes somewhat to explosive. And this is definitely not the best way of communication between allies.

Simulation Workshop

ZUMA Simulation Workshop - Koblenz, Germany

From September 18-22, I am going to participate in ZUMA's (German Centre for Survey Research and Methodology) workshop on simulation in the social sciences, to be held at the University of Koblenz. The main goal of the workshop will be to present simulation techniques and get some hands-on practice by modelling own simulations in smaller groups. I am really looking forward to meeting people with similar research interests.

The participants are asked to have an idea of a model in mind - present it - and build/improve it during the workshop. As I recently wrote a term paper building on an existing spatial model of party competition by Mike Laver ("Policy and the Dynamics of Political Competition" APSR, 2005), which I extended by adding behaviour of party members, I already have a more or less fully implemented model to present. I hope the workshop participants, and the organizers Klaus Troitzsch, University of Koblenz and Nigel Gilbert, University of Surrey will provide helpful comments and advice.

Despite the PISA results of Germany in general and Schools in North Rhine-Westphalia in particular:

You Passed 8th Grade Math

Congratulations, you got 9/10 correct!

Could You Pass 8th Grade Math?

Algorithmic Decision-Makers

Artificial Intelligence and Computerbased Decision Making

I recently read two interesting articles on AI and the real life use of computerbased decision making:
Douglas Heingartner of the New York Times asks "Maybe We Should Leave That Up to the Computer?" and describes the intensified use of computerbased decision making especially in the field of finance and routine situations.
50th anniversary of the Dartmouth Conference, the Big Bang of the whole field of A.I:
Marvin Minsky and John McCarthy meet again (with others) to discuss the state of the discipline. Find the conference website here, and a good (but german language) article on the reunion by the german weekly 'DIE ZEIT': Blitzrechnen ohne Geist