Drunk Cycling

In this video, Steve Levitt and Stephen Dubner talk about their finding that you are 8 times more likely to die walking drunk than driving drunk.

Levitt says this

“anybody could have done it, it took us about 5 minutes on the internet trying to figure out what some of the statistics were… and yet no one has every talked or thought about it and I think that’s the power of ideas… ways of thinking about the world differently that we are trying to cultivate with our approach to economics.”

Dubner cites the various ways a person could die walking drunk

1. step off the curb into traffic.
2. mad dash across the highway.
3. lie down and take a nap in the road.

Which leads him to see how obvious it is ex post that drunk walking is so much more dangerous than drunk driving.

I thought a little about this and it struck me that riding a bike while drunk should be even more dangerous than walking drunk. I could

1. roll or ride off a curb into traffic.
2. try to make a mad dash across an intersection.
3. get off my bike so that i can lie down in the road to take a nap.

plus so many other dangerous things that i can do on my bike but could not do on foot. And what the hell, I have 5 minutes of time and the internet so I thought I would do a little homegrown freakonomics to test this out. Here is an excerpt from their book explaining how they calculated the risk of death by drunk walking.

Let’s look at some numbers, Each year, more than 1,000 drunk pedestrians die in traffic accidents. They step off sidewalks into city streets; they lie down to rest on country roads; they make mad dashes across busy highways. Compared with the total number of people killed in alcohol-related traffic accidents each year–about 13,000–the number of drunk pedestrians is relatively small. But when you’re choosing whether to walk or drive, the overall number isn’t what counts. Here’s the relevant question: on a per-mile basis, is it more dangerous to drive drunk or walk drunk?

The average American walks about a half-mile per day outside the home or workplace. There are some 237 million Americans sixteen and older; all told, that’s 43 billion miles walked each year by people of driving age. If we assume that 1 of every 140 of those miles are walked drunk–the same proportion of miles that are driven drunk–then 307 million miles are walked drunk each year.

Doing the math, you find that on a per-mile basis, a drunk walker is eight times more likely to get killed than a drunk driver.

I found the relevant statistics for cycling here, on the internet. I calculate as follows. Estimates range between 6 and 21 billion miles traveled by bike in a year. Lets call it 13 billion. If we assume that 1 out of every 140 of these miles are cycled drunk, then that gives about 92 million drunk-cycling miles. There are about 688 cycling related deaths per year (average for the years 200-2004.) Nearly 1/5 of these involve a drunk cyclist (this is for the year 1996, the only year the data mentions.) So that’s about 137 dead drunk cyclists per year.

When you do the math you find that there are about 1.5 deaths per every million miles cycled drunk. By contrast, Levitt and Dubner calculate about 3.3 deaths per every million miles walked drunk.

Is walking drunk more dangerous than biking drunk?

Here is another piece of data. Overall (drunk or not) the fatality rate (on a per-mile basis) is estimated to be between 3.4 and 11 times higher for cyclists than motorists. From Levitt and Dubner’s conclusion that drunk walking is 8 times more dangerous than drunk driving we can infer that there are about .4 deaths per million miles driven drunk. That means that the fatality rate for drunk cyclists is only about 3.8 times higher than for drunk motorists.

That is, the relative riskiness of biking versus driving is unaffected (or possibly attenuated) by being drunk. But while walking is much safer than driving overall, according to Levitt and Dubner’s method, being drunk reverses that and makes walking much more dangerous than both biking and driving.

There are a few other ways to interpret these data which do not require you to believe the implication in the previous paragraph.

1. There was no good reason to extrapolate the drunk rate of 1 out of every 140 miles traveled from driving (where its documented) to walking and biking (where we are just making things up.)
2. Someone who is drunk and chooses to walk is systematically different than someone who is drunk and chooses to drive. They are probably not going to and from the same places. They probably have different incomes and different occupations. Their level of intoxication is probably not the same. This means in particular that the fatality rate of drunk walkers is not the rate that would be faced by you and me if we were drunk and decided to walk instead of drive. To put it yet another way, it is not drunk walking that is dangerous. What is dangerous is having the characteristics that lead you to choose to walk drunk.

These ideas, especially the one behind #2 were the hallmark of Levitt’s academic work and even the work documented in Freakonomics. His reputation was built on carefully applying ideas like these to uncover exciting and surprising truths in data. But he didn’t apply these ideas to his study of drunk walking. Of course, my analysis is no better. I just copied some numbers off a page I found on the internet and applied the Levitt Dubner calculation. It only took me 5 minutes. (And I would appreciate if someone can check my math.) But then again, I am not trying to support a highly dubious and dangerous claim:

So as you leave your friend’s party, the decision should be clear: driving is safer than walking. (It would be even safer, obviously , to drink less, or to call a cab.) The next time you put away four glasses of wine at a party, maybe you’ll think through your decision a bit differently. Or, if you’re too far gone, maybe your friend will help sort things out. Because friends don’t let friends walk drunk.

Lowest Unique Bid

Auction sites are popping up all over the place with new ideas about how to attract bidders with the appearance of huge bargains.  The latest format I have discovered is the “lowest unique bid” auction.  It works like this.  A car is on the auction block.  Bidders can submit as many bids as they wish ranging from one penny possibly to some upper bound, in penny increments.  The bids are sealed until the auction is over.  The winning bid is the lowest among all unique bids.  That is, if you bid 2 cents and nobody else bids 2 cents, but more than one person bid 1 cent, then you win the car for 2 cents.

In some cases you pay for each bid but in some cases bids are free and you pay only if you win.  Here is a site with free bidding.  An iPod shuffle sold for $0.04.  Here is a site where you pay to bid.  The top item up for sale is a new house.  In that auction you pay ~$7 per bid and you are not allowed to bid more than $2,000.  A house for no more than $2,000, what a deal!

I suppose the principle here is that people are attracted by these extreme bargains and ignore the rest of the distribution.  So you want to find a format which has a high frequency of low winning bids.  On this dimesion the lowest unique bid seems even better than penny auctions.

Caubeen curl: Antonio Merlo.

Sniping and Squatting

Via kottke, Clusterflock gives five simple rules for effective bidding on eBay:

Step One:Find the product you want.

Step Two:

Save the product to your watch list.

Step Three:

Wait.

Step Four:

Just before the item ends, enter the maximum amount you are willing to pay for the item.

Step Five:

Click submit.

This is called sniping.  That’s a pejorative label for what is actually a sensible and perfectly straightforward way to bid.  eBay is essentially an open second-price auction and sniping is a way to submit a “sealed” bid.  It’s a popular strategy and advocated by many eBay “experts.”  But does it really pay off?

Tanjim Hossain and I did an experiment (ungated version.)  We compared sniping to another straightforward strategy we call squatting. As the name suggests, squatting means bidding your value on an object at the very first opportunity, essentially staking a claim on that object.  We bid on DVDs and randomly divided auctions into two groups, sniping on the auctions in the first group and squatting on the other.

The two strategies were almost indistinguishable in terms of their payoff.  But for an interesting reason.  A lot of eBay bidders use a strategy of incremental bidding.  That’s where you act as if you are involved in an ascending auction (like an English auction) and you bid the minimum amount needed to become the high biddder.  Once you are the high bidder you stop there and wait to see if you are outbid, then you raise your bid again.  You do this until either you win or the price goes above the maximum amount you are willing to pay.

Against incremental bidders, sniping has a benefit and a cost (relative to squatting.)  You benefit when incremental bidders stop at a price below their value.  You swoop in at the end, the incremental bidders have no time to respond, and you win at the low price.

The cost has to do with competition across auctions for similar objects.  If I squat on auction A and you are sniping in auction B, our opponents think there is one fewer competitor in auction B and more opponents enter auction B than A.  This tends to raise the price in your auction relative to mine.  In other words, squatting scares opponents away, sniping does not.

We found that these two effects almost exactly canceled each other out for auctions of DVDs.  We expect that this would be true for similar objects that are homogeneous and sold in many simultaneous auctions.  So the next time you are bidding in such an auction, don’t think too hard and just bid your value.

Now, I am still trying to figure out what I am going to do with all these copies of 50 First Dates we won in the experiment.

Ask Carole

Tongue-in-cheek relationship advice from evolutionary psychology (via Mindhacks.)

Some Darwinists might say your optimal strategy would be to pair-bond with the older male but surreptitiously allow the younger, sexy male to fertilise you. But be careful, most men consider being cuckolded the greatest of betrayals.

And how about this?

You should have your husband medically assessed. It may be that some form of genetic disorder underlies his erratic behaviour, in which case he will need counselling and support. But you will also need to inform your daughters so that, if they are carriers, they do not themselves mate with men suffering from the same condition.

The Happy Contract

Big events at Northwestern this weekend, including Paul Milgrom’s Nemmers’ Prize lecture and a conference in his honor.  (My relative status was microscopic.)   A major theme of the conference was market design and I heard a story repeated a few times by participants connected with research and implementation of online ad auctions.

Ads served by Yahoo!, Google and others are sold to advertisers using auctions.  These auctions are run at very high frequencies.  Advertisers bid for space on specific pages at specific times and served to users which are carefully profiled by their search behavior.  This enables advertisers to target users by location, revealed interests, and other characteristics.

Not content with these instruments, McDonalds is alleged to have proposed to Yahoo! a unique way to target their ads and their proposal has come to be known as The Happy Contract.  Instead of linking their bids to personal profiles of users, they asked to link their bids to weather reports.  McDonalds would bid for ad space only when and where the sun was shining.  That way sunshine-induced good moods would be associated with impressions of Big Macs, and (here’s the winner’s curse) the foul-weather moods would get lumped with the Whopper.

The Tyranny of Memory

I can relate to this.

For most of history, almost everything people did was forgotten because it was so hard to record and retrieve things. But this had a benefit: “social forgetting” allowed us to move on from embarrassing moments. Digital tools have eliminated this: Google caches copies of blog posts; networking sites thrive by archiving our daily dish. Society defaults to a relentless Proustian remembrance of all things past.

From an article in Wired.

 

Palindromic Video

I once linked to something like this.  But that didnt hold a candle to this one:

Did you notice that when the song starts to go in the right direction his voice has an Eastern European accent?  I have no idea whether this guy is a native English speaker.  If he is then this is an artifact of singing backwards.  If he really is Eastern European then it says something about language accents that they appear even when singing a foreign language backwards.

NAJ Ain’t a Journal of Economics

Its one of the many novel ideas from David K. Levine:  the non-journal.  You write your papers and you put them on your web site.  Congratulations, you just published!  Ah, but you want peer review.  The editors of NAJ just might read your self-published paper and review it.  We supply the peer-review, you supply the publication.  Peer-review + publication = peer-reviewed publication. That was easy.

(NAJ is an acronym that stands for NAJ Ain’t a Journal.)

Its been around for a few years with pretty much the same set of editors.  Its gone through some very active phases and some slow periods.  David is trying to breathe some new life into NAJ by rotating in some new editors.  So far so good.  Arthur Robson is a new editor and he just reviewed a very cool paper by Emir Kamenica and Matthew Gentzkow called “Bayesian Persuasion.”

The paper tells you how a prosecutor manages to convict the innocent.  Suppose that a judge will convict a defendant if he is more than 50% likely to be guilty and suppose that only 30% of all defendants brought to trial are actually guilty.  A prosecutor can selectively search for evidence but cannot manufacture evidence and must disclose all the evidence he collects.  The judge interprets the evidence as a fully rational Bayesian.  What is the maximum conviction rate he can achieve?

The answer is 60%.  This is accomplished with an investigation strategy that has two possible outcomes.  One outcome is a conclusive signal that the defendant is innocent.  Since the judge is Bayesian, the innocent signal occurs with probability zero when the defendant is actually guilty.  The other outcome is a partially informative signal.  If the prosecutor designs his investigation so that this signal occurs with probability 3/7 when the defendant is innocent (and with probability 1 when guilty) then

1. conditional on this signal, the defendant is 50% likely to be guilty (we can make it strictly higher than 50% if you like by changing the numbers slightly)
2. 3/7 of the innocent and all of the guilty will get this signal.  (3/7 times 70%) + 30% = 60%.

The paper studies the optimal investigation scheme in a general model and uses it in a few applications.

 

 

Coordinating Attention

Despite my vast legion of Twitter followers, every one of my attempts to start a new trending topic has failed to catch on.  Now I think I understand why.

Suppose that your goal is to coordinate attention on a topic that seems to be on a lot of minds.  Attention is a scarce resource and you have only a limited number of topics you can highlight.  But suppose, as with Twitter, you see what everyone is talking about.  How do you decide which topics to point to?

You probably shouldn’t just count the total number of people talking on a given subject, counting everyone equally.  You might think that you would instead give extra weight to the few people that everyone is listening to.  Because whatever they say is more likely to be interesting to many, and will soon be on many minds.  On Twitter, those would be the people with the most followers.  But there is a strong case for doing the opposite and giving extra weight to people with few followers, especially people who are relatively isolated in the social network.  This is not out of fairness (or pity) but actually as the efficient way to use your scarce resource.

Efficient coordination means making information public so that not just everyone knows it, but everyone knows that everyone knows it (etc.) If we all have to choose simultaneously what to focus our attention on and we want to be part of the larger conversation, then it matters what we think others are going to focus their attention on.  Coordinating attention thus requires making it public what people are talking about.

Suppose we have two topics that are getting a lot of attention, but topic A is being discussed by well-connected individuals and topic B is being discussed more by a diverse group of isolated individuals.  Topic A is already public because when you see it discussed by a central figure you know that all other of her followers are seeing it to.  Topic B therefore has more to gain from elevating it to the status of trending topic, which immediately makes it public.

I always knew that my Twitter followers were among the wisest. Now I see the true depth of their wisdom.  By adding to my follower numbers, they reduce the weight of my comments in the optimal weighting scheme thus ensuring that the crazy things I say will be ingored by the larger network.  Join the cause.

Penny Auctions

Swoopo.com has been called “the crack cocaine of auction sites.” Numerous bloggers have commented on its “penny auction” format wherein each successive bid has an immediate cost to the bidder (whether or not that bidder is the eventual winner) and also raises the final price by a penny. The anecdotal evidence is that, while sometimes auctions close at bargain prices, often the total cost to the winning bidder far exceeds the market price of the good up for sale. The usual diagnosis is that Swoopo bidders fall prey to sunk-cost fallacies: they keep bidding in a misguided attempt to recoup their (sunk) losses.

Do the high prices necessarily mean that penny auctions are a bad deal? And do the outcomes necessarily reveal that Swoopo bidders are irrational in some way? Toomas Hinnosaar has done an equilibrium analysis of penny auctions and related formats and he has shown that the huge volatility in prices is in fact implied by fully rational bidders who are not prone to any sunk-cost fallacy. In fact, it is precisely the sunk nature of swoopo bidding costs that leads a rational bidder to ignore them and to continue bidding if there remains a good chance of winning.

This effect is most dramatic in “free” auctions where the final price of the good is fixed (say at zero, why not?) Then bidding resembles a pure war of attrition: every bid costs a penny and whoever is the last standing gets the good for free. Losers go home with many fewer pennies. (By contrast to a war of attrition, you can sit on the sidelines as long as you want and jump in on the bidding at any time.) Toomas shows that when rational bidders bid according to equilibrium strategies in free auctions, the auction ends with positive probability at any point between zero bids and infinitely many bids.

So the volatility is exactly what you would expect from fully rational bidders. However, Toomas shows that there is a smoking gun in the data that shows that real-world swoopo bidders are not the fully rational players in his model. In any equilibrium, sellers cannot be making positive profits otherwise bidders are making losses on average. Rational bidders would not enter a competition which gives them losses on average.

In the following graph you will see the actual distribution of seller profits from penny auctions and free auctions. The volatility matches the model very well but the average profit margin (as a percentage of the object’s value) is clearly positive in both cases. This could not happen in equilibrium.

Net Neutrality Myth

Net neutrality refers to a range of principles ensuring non-discriminatory access to the internet.  A particularly contentious principle urges prohibition of “managed” or “tiered” internet service wherein your internet service provider is permitted to restrict or degrade service.  ISPs argue that without such permission they are unable to earn sufficient return on investment in network capacity and would be deterred from making such improvements.

One argument is based on congestion.  Managed service controls congestion, raising the value to users and allowing providers to capture some of this value with access fees.  This is a logical argument and one I will take up in a later post, but here I want to discuss another aspect of managed service:  price discrimination.

Enabling providers to limit access, say by bandwidth caps, opens the door to “tiered” service where users can buy additional bandwidth at higher prices.  This generally raises profits and so we should expect tiered service if net neutrality is abandoned.  What effect does the ability to price discriminate have on an ISP’s incentive to invest in capacity?

It can easily reduce that incentive and this undermines the industry argument against net neutrality.  Here is a simple example to illustrate why.  Suppose there is a small subset of users who have a high willingness to pay for additional bandwidth.  Under net neutrality, all users are charged the same price for access, and none have bandwidth restrictions.  An ISP then has only two choices.  Set a high price and sell only to the high-end users, or set a low price and sell to all users.  When the high-end users are relatively few, profits are maximized with low prices and wide access.  It is reasonable to think of this as describing the present situation.

Suppose tiered access is now allowed.  This gives the ISP a new range of pricing schemes.  The ISP can offer a low-price service plan with a bandwidth cap alongside a high-priced unrestricted plan.  As we vary the cap associated with the low-end plan, we can move along a continuum from no cap at all to a 100% cap.  These two extremes are equivalent to the two price systems available under net neutrality.

Often one of these in-between solutions will be more profitable than either of the two extremes.  The reason is simple.  The bandwidth cap makes the low-end plan less attractive to high-end users and as a result the ISP can raise the price of un-capped access to high-end users.  It’s true that low-end users will pay less for capped service but often the trade-off is favorable to the ISP and total profits increase.

The upshot of this is that total bandwidth is lower, not higher, when an ISP unconstrained by net-neutrality uses the profit-maximizing tiered-service plan.  Couched in the industry’s usual terms, the ISP’s incentive to increase network capacity is in fact reduced by moving away from net neutrality.

(Of course it can just as easily go the other way.  For example, it may be that presently only the high-end users are being served because to lower price enough to attract the low end users, the ISP would lose too much profit from the high-end.  In that case, allowing tiered service would induce the ISP to raise capacity and offer a capped service to previously excluded low-end users without significantly reducing profits from the high-end.  Note however, this is not typically how industry lobbyists frame their argument.)

Using Games for Research

Not game theory, but research about and even within games.  Hit or miss:

One of the most high-profile projects (and most obvious recent failures) was Indiana University’s Arden: The World Of William Shakespeare, which reportedly had a grant of $250,000. It was an experimental MMO which came about via the work of Professor Ed Castronova, author of Synthetic Worlds. Castronova wondered whether the creation of a genuinely educational MMO was possible, and set up the student development project to find out. Having spent thousands of dollars on Arden it was shut down. Castronova cited “a lack of fun”.

via BoingBoing.

Tons of Twitter Stats

Here, via Michael Nielsen.  For example:

• Twitter’s user growth is no longer accelerating. The rate of new user acquisition has plateaued at around 8 million per month.
• Over 14% of users don’t have a single follower, and over 75% of users have 10 or fewer followers.
• 38% of users have never sent a single tweet, and over 75% of users have sent fewer than 10 tweets.
• 1 in 4 registered users tweets in any given month.
• Once a user has tweeted once, there is a 65% chance that they will tweet again. After that second tweet, however, the chance of a third tweet goes up to 81%.
• If someone is still tweeting in their second week as a user, it is extremely likely that they will remain on Twitter as a long-term user.
• Users who joined in more recent months are less likely to stop using the service and more likely to tweet more often than users from the past.

Here is a pie-chart:

Have I mentioned that you should be following me on twitter? (I am talking to you Sandeep.)

The Circular Concept of Opportunity Cost

Since I am willing to pay $X that means my opportunity cost of not buying is -$X, thus my willingness to pay is indeed $X.   That appears to be what Google CEO Eric Schmidt is saying in the following deposition transcript talking about Google paying X=$1.65 billion for YouTube, a $1billion premium over what he estimated YouTube to be worth.  From an article at cnet.

Baskin: So you orally communicated to your board during the course of the board meeting that you thought a more correct valuation for YouTube was $600 million to $700 million; is that what you said, sir?

Mancini objects to characterization of the testimony.

Schmidt: Again, to help you along, I believe that they were worth $600 million to $700 million.

Baskin: And am I correct that you were asking your board to approve an acquisition price of $1.65 billion; correct?

Schmidt: I did.

Mancini objects.

Baskin: I’m not very good at math, but I think that would be $1 billion or so more than you thought the company was, in fact, worth.

Mancini objects.

Schmidt: That is correct.

Later…

Baskin: Can you tell us what reasoning you explained?

Schmidt: Sure, this is a company with very little revenue, growing quickly with user adoption, growing much faster than Google Video, which was the product that Google had. And they had indicated to us that they would be sold, and we believed that there would be a competing offer–because of who Google was–paying much more than they were worth. In the deal dynamics, the price, remember, is not set by my judgment or by financial model or discounted cash flow. It’s set by what people are willing to pay. And we ultimately concluded that $1.65 billion included a premium for moving quickly and making sure that we could participate in the user success in YouTube.

Star Inflation and Star Polarization

There is a story in the Wall Street Journal about user ratings on web sites such as Amazon or eBay.  It seems that raters are unduly generous with their stars.

One of the Web’s little secrets is that when consumers write online reviews, they tend to leave positive ratings: The average grade for things online is about 4.3 stars out of five.

And some users are fighting back:

That’s why Amazon reviewer Marc Schenker in Vancouver has become a Web-ratings vigilante. For the past several years, he has left nothing but one-star reviews for products. He has called men’s magazine Maxim a “bacchanalia of hedonism,” and described “The Diary of Anne Frank” as “very, very, very disappointing.”

I have noticed that Amazon reviewers are highly polarized with 5 stars being the most common with 1 star reiews coming in second.  And in fact it makes a lot of sense.  Say you think that a product is over-rated at 4.3 stars and you think that 4 stars is more appropriate.  If there are more than just a few ratings, then to bring the average down to 4 you would have to give the lowest possible rating.

Once enough ratings have already been counted, subsequent raters will be effectively engaging in a tug of war.  Those that want to raise the average will give 5 stars and those that want to reduce it will give 1.

Shoot Your Food

I have a simple system for organizing recipes.  I try out recipes I find in cookbooks, blogs, magazines, whatever.  When one hits I do the following.

1. Take a picture of it.
2. Write down a list of the ingredients I wouldn’t typically have stocked.
3. Email the above plus a link to the recipe (or what page in what cookbook) to myself.

Because the time you really need recipes is when you are shopping and you see, say some really good looking okra and you need to know what else to get.  You pull out your iPhone, you search for okra in your mail folders, you get a picture and a list of ingredients.  You go home and cook.

The picture is absolutely key.  Think of your cookbooks at home.  Which recipes do you most often cook?  Its the ones with the beautiful photos in the middle of the book.  The photo reminds you how yummy its going to be.  Wouldn’t you love to cook this tonight?

The Browser Wars Ended By The Browser Trojan Horse

Remember the browser wars?  Resistance to open web standards, and “best viewed in Internet Explorer.”  Remember “polluted java?”  Here are paragraphs that caught my eye from ars technica’s overview of Google Wave.

In September, Google released Chrome Frame, a plugin for Internet Explorer that makes it possible for Microsoft’s browser to use Chrome’s rendering engine. Microsoft was not happy about this sudden but inevitable betrayal. Google later revealed that Wave was one of the catalysts that compelled them to launch the Chrome Frame project.

The developers behind the Wave project struggled to make Wave work properly in Microsoft’s browsers, but eventually determined that the effort was futile. Internet Explorer’s mediocre JavaScript engine and lack of support for emerging standards simply made the browser impossible to accommodate. In order to use Wave, Internet Explorer users will need to install Chrome Frame.

While we are on the subject I highly recommend the ars technica piece on Google Wave.  In addition to lots of detail on the technology and implementation, it talks about Google’s commitment to open standards, open source, and decentralization.  I came away less worried.

I have not been invited yet to try the beta.

The True Number May Be Higher

Mindhacks discusses a surprising asymmetry.  Journalists discussing sampling error almost always emphasize the possibility that the variable in question has been under-estimated.

For any individual study you can validly say that you think the estimate is too low, or indeed, too high, and give reasons for that… But when we look at reporting as a whole, it almost always says the condition is likely to be much more common than the estimate.

For example, have a look at the results of this Google search:

“the true number may be higher” 20,300 hits

“the true number may be lower” 3 hits

There are two parts to this.  First, the reporter is trying to sell her story.  So she is going to emphasize the direction of error that makes for the most interesting story. But that by itself doesn’t explain the asymmetry.

Let’s say we are talking about stories that report “condition X occurs Y% of the time.”  There is always an equivalent way to say the same thing: “condition Z occurs (100-Y)% of the time” (Z is the negation of X.)   If the selling point of the story is that X is more common than you might have thought, then the author could just as well say “The true frequency of Z may be lower” than the estimate.

So the big puzzle is why stories are always framed in one of two completely equivalent ways.  I assume that a large part of this is

1. News is usually about rare things/events.
2. If you are writing about X and X is rare, then you make the story more interesting by pointing out that X might be less rare than the reader thought.
3. It is more natural to frame a story about the rareness of X by saying “X is rare, but less rare than you think” rather than “the lack of X is common, but less common than you think.”

But the more I think about symmetry the less convinced I am by this argument.  Anyway I am still amazed at the numbers from the google searches.

The Future of Facebook

The most important development in the way we interact on the web will come when a system of micropayments is in place.  The big difficulties are coordination problems and security.  The strongest incentive to build and control a massive social network is that it will enable Facebook to host a micropayments economy within its closed environment, solving both the coordination problem and a big part of the security problem.

Here’s the future of Facebook.  You will subscribe to your friends.  A subscription costs you a flow of micropayments.  Your friends will include the likes of Tyler Cowen, The Wall Street Journal, gmail, Jay-Z, Harry Potter and the Deathly Hallows, etc.

Remember that the next time you hear somebody say that there is no way to monetize Facebook or Twitter.

Google Acquires reCAPTCHA

We all work for google now.  Previous posts on reCAPTCHA here and here.  beanie bow:  lance fortnow.

Do Spammers Give to Charity Too?

Suppose I want to send you an email and be sure that it will not be caught in your spam filter. What signal can I use to prove to you that my message is not spam? It must satisfy (at least) two requirements.

1. It should be cheaper/easier for legitimate senders to use than for spammers.
2. It should be cheap overall in absolute terms.

The first is necessary if the signal is going to effectively separate the spam from the ham. The second is necessary if the signal is going to be cheap enough for people to actually use it.

It is easy to think of systems that meet the first requirement but very hard to think of one that also satisfies the second. Now researchers at Yahoo! have an intriguing new idea that has received a great deal of attention, CentMail. According to this article, Yahoo! is planning to roll it out soon.

The sender pays a penny to have a trusted server to affix an electronic “stamp” to the message. Given that spammers could not afford to pay even one cent per message given the massive volume of spam, the receiver can safely accept any stamped message without running it through his spam filter.

Now here is the key idea. The penny is paid to charity. How could this matter? Because most people already make sizable donations to charity every year, they can simply route these donations through CentMail making the stamps effectively free. Thus, condition 2 is satisfied.

The first question that comes to mind is the titular one. (Settle down Beavis.) Remember, we still have to worry about condition 1 and whatever magic we use to make it cheap for legitimate email better not have the same effect on spam. But just like you, any spammer who makes donations to charity will be able to send a volume of spam for free. Apparently the assumption is that spam=evil and evildoers do not also contribute to charity. And we must also assume that Centmail doesn’t encourage entry into the spamming business by those marginal spammers for whom the gift to charity is enough to assuage their previous misgivings.

But these seem like reasonable assumptions. The more tricky issue is whether the 1 penny will actually deter spammers. It is certainly true that at current volume levels, the marginal piece of spam is not worth 1 penny. But for sure there is still a very large quantity of spam that is worth significantly more than 1 penny. For proof, just take a look in your snailbox. Even at bulk rates the cost of junk-mail advertising is several pennies per piece. With Centmail your Inbox would have at least as much stamped spam as the amount of junk mail in your snailbox.

This leads to the crucial questions. Any system of screening by monetary payments should be viewed with the following model in mind. First, ask how many pieces of spam you would expect to receive per day at the specified price. Next, ask how many spam you are willing to receive before you turn on your spam filter again. If the first number is larger than the second, then the system is not going to substitute for spam filtering and this undermines the reason to opt-in in the first place. For Centmail and me these numbers are 50 and 1.

Now continued spam filtering won’t necessarily destroy the system’s effectiveness. The stamp can be used in conjunction with standard filtering rules to reduce the chance your ham gets classified as spam. Then the question will be whether this reduction is enough to induce senders to adopt the setup costs of opting in.

Finally there is no reason theoretically that the total volume of spam would be reduced. Providing spammers with a second, higher class of service might only add to their demand.

The Unbundled Economy: Self-Promotion and Street Teams

The NY Times has an article about a new wave of independent films and their marketing.

When “The Age of Stupid,” a climate change movie, “opens” across the United States in September, it will play on some 400 screens in a one-night event, with a video performance by Thom Yorke of Radiohead, all paid for by the filmmakers themselves and their backers. In Britain, meanwhile, the film has been showing via an Internet service that lets anyone pay to license a copy, set up a screening and keep the profit.

The article is about the variety of roll-your-own distribution and marketing campaigns employed by filmmakers who lack studio backing.  But the lede is buried:

Famous fans like Courtney Love were soon chattering online about the film. And an army of “virtual street teamers” — Internet advocates who flood social networks with admiring comments, sometimes for a fee, sometimes not — were recruited by a Web consultant, Sarah Lewitinn, who usually works the music scene.

Here is wikipedia on street teams.  The origin is traced back to the KISS army, a grass-roots fan club that aggressively promoted the band KISS and later “vertically integrated” with the KISS marketing machine where they had access to exclusive promotional merchandise.

Today you can hire a consultant to assemble a street team to promote your band, movie, (hmmm… blog?), … A good consultant will find (or make) fans with a selected personality type, street-cred, and social network and organize them into a guerilla marketing squad armed with swag.

Virtual street teams operate in online social networks.  Presumably then, actual people are no longer required.  A good consultant can manufacture online identities, position them in a social network, getting Twitter followers and Facebook friends and cultivate the marketing opportunities from there.  You can imagine the pitch:  “We can mobilize 10,000 follower-tweets per day…”

Here is the web site of ForTheWin.com, the agency of Sara Lewitinn who coordinated the virtual street team for the film Anvil! The Story of Anvil.

For The Win! is an cooperative of club urchins and nightlife denizens charged with the task of defending the best of pop culture from the daily onslaught of the whack. At night we comb the streets in search of the best fashion, art, music, and movies New York City has to offer. By day we make sure we spread the word to the world by any and all means necessary of the internet to it’s biggest platforms without skipping a step or taking anything for granted. Each of our campaigns is as unique as the artist it represents.

Note they also count Slighly Stoopid, Electrocute, and The Pet Shop Boys (!) as clients.

Wagging the Long Tail

Developer Kalid Shaikh has been banned from the iPhone App Store.  By conventional welfare measures this would seem to be a big blow to efficiency:

As the MobileCrunch article points out, a search at AppShopper.com shows 854 apps by Shaikh. The majority of Shaikh’s apps seemed to be data on a specific subject simply pulled from the web without providing any other original or unique content. Most apps were priced at $4.99 and this banishment could represent lost sales of thousands of dollars per day. Shaikh reportedly has admitted that the goal was not to produce valuable apps but to focus on monetization instead. All of Shaikh’s apps have already been removed from the App Store and can no longer be purchased.

Perhaps conventional welfare measures would need to be amended in this case.  Note however, that removing one large supplier of what is essentially spam from the App Store will not affect the equilibrium quantity of spam.  (And this is not Apple’s stated reason for removing him.)

Got Skype? The Holdup Problem.

Millions of internet users who use Skype could be forced to find other ways to make phone calls after parent company eBay said it did not own the underlying technology that powers the service, prompting fears of a shutdown.

Why are there firms?  A more flexible way to manage transactions would be through a system of specific contracts detailing what each individual should produce, to whom it should be delivered and what he should be paid.  It would also be more efficient:  a traditional firm makes some group of individuals the owners and a separate group of individuals the workers.  The firm is saddled with the problem of motivating workers when the profits from their efforts go to the owners.

The problem of course is that most of these contracts would be far too complicated to spell out and enforce.  And without an airtight contract, disputes occur.  Because disputes are inefficient, the disputants almost always find some settlement which supplants the terms of the contract.  Knowing all of this in advance, the contracts would usually turn out to be worthless.  The strategy of bringing spurious objections to existing contracts in order to trigger renegotiation at more favorable terms is called holdup. The holdup problem is considered by some economic theorists to be the fundamental friction that shapes most of economic organization.

Case in point, Skype and eBay.  eBay acquired the Skype brand and much of the software from the founders, JoltId, but did not take full ownership of the core technology, instead entering a licensing agreement which grants Skype exclusive use.  Since that time, Skype has become increasingly popular and a strong source of revenue for eBay.  Now eBay is being held up.  JoltId claims that eBay has violated the licensing agreement, citing a few obscure and relatively minor details in the contract.  Litigation is pending.

Not coincidentally, eBay has publicly stated its intention to spinoff Skype and take it public, a sale that would bring a huge infusion of capital to eBay at a time when it is reinventing its core business.  That sale is in turn being heldup because Skype is worthless without the license from JoltId.  This puts JoltId in an excellent bargaining position to renegotiate for a better share of those spoils. (On the other hand, had Skype not done as well as it did, JoltId would not have such a large share of the downside.)

Whatever were the long-run total expected payments eBay was going to make to JoltId in return for exclusive use of the technology, it should have paid that much to own the technology outright, become an integrated firm, and avoided the holdup problem.

And don’t worry.  You got your Skype.  Holdup may change the terms of trade, but it is in neither party’s interest to destroy a valuable asset.

reCAPTCHA Roulette

To remind you, reCAPTCHA asks you to decipher two smeared words before you can register for, say, a gmail account.  One of the words is being used to test whether you are a human and not a computer.  The reCAPTCHA system knows the right answer for that word and checks whether you get it right.  The reCAPTCHA system doesn’t know the other word and is hoping you will help figure it out.  If you get the test word right, then your answer on the unkown word is assumed to be correct and used in a massive parallel process of digitizing books.  The words are randomly ordered so you cannot know which is the test word.

Once you know this, you many wonder whether you can save yourself time by just filling in the first word and hoping that one is the test word.  You will be right with 50% probability.  And if so, you will cut your time in half.  If you are unlucky, you try again, and you keep on guessing one word until you get lucky.  What is the expected time from using this strategy?

Let’s assume it takes 1 second to type in one word.  If you answer both words you are sure to get through at the cost of 2 seconds of your time.  If you answer one word each time then with probability 1/2 you will pass in 1 second, with probability 1/4 you will pass in 2 seconds, probability 1/8 you pass in 3 seconds, etc.    Then your expected time to pass is

Is this more or less than 2?  Answer after the jump.

The Economics of reCAPTCHA

CAPTCHAs are everywhere on the web now.  They are the distorted text that you are asked to identify before being allowed to register for an account.  The purpose is to prevent computer programs from gaining quick access to many accounts for nefarious purposes (spam for example.)

reCAPTCHA piggy-backs on CAPTCHA.  You are asked to identify two words. The first is a standard CAPTCHA.  If you enter the correct word you identify yourself as a human.  The second is a word that has been optically scanned from a book that is being digitized.  It has found its way into this reCAPTCHA because the computer doing the optical character recognition was not able to identify it.  If you have identified yourself as a human via the first CAPTCHA, your answer to the second word is assumed to be correct and used in the digital translation.  You are digitizing the book.

According to Wikipedia 20 years of the New York Times archive has been digitized with the help of reCAPTCHA.  And, “provides about the equivalent of 160 books per day, or 12,000 manhours per day of free labor.”

The first reaction to this is obvious.  The labor is not free.  In fact it costs exactly 12,000 man hours.  Lots of things can be produced with 12,000 man hours. Lots of leisure can be consumed in 12,000 hours.  Is digitizing the New York Times the best use of this people-time?  On top of that the reCAPTCHA is a tax which reduces the quantity of online accounts transacted and that is a deadweight loss.

But it is just a few seconds of your time right?  Something about that seems to change the calculation.  I bet most people would say that they don’t mind giving away two seconds of their time.  Part of this is due to an illusion of marginal vs total.  People are tempted to treat the act as a gift of two seconds of their time in return for a whole digitized library.  But in fact they are giving away two seconds of their time for one digitized word.

A second part of this is due to a scale illusion. You may successfully convince said reCAPTHArer that she is just getting a tiny fraction of the book for her two seconds but she will probably still say that she is happy with that.  But if you ask her whether she is willing to contribute 1000 seconds for 500 words, probably not.  And, to take increasing marginal costs out of the question, if you asked her whether she thought digitizing the New York Times is worth how many thousands of woman-hours of (dispersed) ucompensated labor she again might start to see the point.

But still, not everybody.  And I think there must be some sound rationale underneath this.  I would not argue that digitizing books is the necessarily the highest priority public good, but the mechanism is inherently linked to deciphering words.  True, we could require everyone who signs up at Facebook to donate 1 penny to fight global warming but A) it is never possible to know exactly what “1 penny toward fighting global warming” means whereas there is no way to redirect my contribution if I decipher a word.  That is not a liquid asset.  And B) two seconds of most people’s time is worth less than 1 penny (we are talking about Facebook users remember) and we don’t have a micro-payments system in place to go down to fractions of pennies.

Perhaps what we have here is a unique opportunity to utilize a public-goods contribution mechanism that transparent and non-manipulable and guarantees to each contributor that he will not be free-ridden on:  everyone else is committed to the same contribution.

Embezzlement Leads to Bank Run in Virtual World

There has been a run on one of the largest banks in an economics-themed online role-playing game called Eve.  The event merited an article at the BBC.  The run was triggered when Ricdic, an executive of the bank made off with a large sum of virtual lucre and exchanged it for real-world cash.

Eve Online has about 300,000 players all of whom inhabit the same online universe. The game revolves around trade, mining asteroids and the efforts of different player-controlled corporations to take control of swathes of virtual space.

It has now emerged that Ricdic used the cash to put down a deposit on a house and to pay medical bills.

“I’m not proud of it at all, that’s why I didn’t brag about it,” Ricdic told Reuters. “But you know, if I had to do it again, I probably would’ve chosen the same path based on the same situation.”

Apparently, the bank had tremendous reserves and has so far withstood the run.  Here is more information.  Either real-world bank regulators have something to learn from Eve or the other way around because here is Ricdic’s comeuppance:

Ricdic has now been thrown out of the game as trading in-game cash for real money is against Eve Online’s terms and conditions.

The rules governing play within Eve would not have sanctioned Ricdic if he had simply stolen the cash and used it in the game, nor if he had bought kredits with real dollars.

Fedora Flourish:  BoingBoing

Free, A Model

In case you have not been following the catfight, let me get you up to speed.  Chris Anderson wrote a book called Free.  I haven’t read it, but it apparently says “all your ideas are belong to us” because the price of ideas is crashing to zero.  Malcom Gladwell says “please don’t let my employer read that”…I mean, “No its not.”

Let’s have a model.  There are tiny ideas and big ideas.  The tiny ideas are more like facts, or observations or experiences.  They are costless to produce but costly to communicate.  They are highly decentralized in that everybody produces their own heterogenous tiny ideas.  The big ideas are assembled from a large quantity of tiny ideas.  Different people have different production technologies for producing big ideas from small ones.  These could differ just in cost, or also in terms of the quality of big ideas that are produced, it changes the story a little but doesn’t change the economics.

Start with a world where the marginal cost of communicating a tiny idea to another individual is large.  Then the equilibrium market structure has big-idea producers who incur the high cost of acquiring tiny ideas, assemble them into big ideas and communicate the big ideas to the masses for a price.  This market structure sustains high prices for big ideas and sustains entry by big-idea specialists.

Now suppose the marginal cost of communicating the tiny ideas shrinks to zero.  Then an alternative for end users is to assemble their own big ideas for their own consumption out of the tiny ideas they acquire themselves for close to nothing.  The cost disadvantage that the typical end user has is compensated by his ability to customize his palette of tiny ideas and resulting big ideas to complement his idiosyncratic endowment of other ideas, tastes, etc.   The price of big ideas crashes.  Former producers of big ideas exit the market.  This is all efficient.

An important implication of this model is that the products that Anderson expects to be free are not the products Gladwell produces.  So when Gladwell says that this is absurd because the economics do not support big ideas being sold at a price of zero, he is right.  But that is because the big ideas are not being sold at all, and this is all efficient.

Less Than a Decade Later, the Microsoft Case is Irrelevant

Apparently we have arrived at the long run and we are not dead.

Do you remember the Microsoft anti-trust case?  The anti-trust division of the US Department of Justice sought the breakup of Microsoft for anti-competitive practices mostly centering around integrating Internet Explorer into the Windows operating system.  In fact, an initial ruling found Microsoft in violation of an agreement not to tie new software products into Windows and mandated a breakup, separating the operating systems business from the software applications business.  This ruling was overturned on appeal and evnetually the case was settled with an agreement that imposed no further restrictions on Microsoft’s ability to bundle software but did require Microsoft to share APIs with third-party developers for a 5 year period.

Today, all of the players in that case are mostly irrelevant.  AOL, Netscape, Redhat.  Java.  Indeed, Microsoft itself is close to irrelevance in the sense that any attempt today at exploiting its operating system market power to extend its monopoly would cause at most a short-run adjustment period before it would be ignored.

Microsoft was arguing at the time that it was constantly innovating to maintain its market position and it was impossible to predict from where the next threat to its dominance would appear.  Whether or not the first part of their claim was true, the second part certainly turned out to be so.  It is hard to see a credible case that the Microsoft anti-trust investigation, trial, and settlement played anything more than a negligible role in bringing us to this point.  Indeed the considerations there, focusing on the internals of the operating system and contracts with hardware manufacturers, are orthogonal to developments in the market since then.  The operating system is a client and today clients are perfect substitutes.  The rents go to servers and servers live on the internet unconstrained by any “platform” or “network effects”, indeed creating their own.

The lesson of this experience is that in a rapidly changing landscape, intervention can wait.  Even intervention that looks urgent at the time.  Almost certainly the unexpected will happen that will change everything.

Google’s Market Power

I read news mostly through an rss reader.  The Wall Street Journal syndicates only short excerpts of their articles and if I click through I get a truncated version of the article follwed by a friendly invitation to subscribe to the journal in order to view the rest of the article.  It looks like this.

But its not hard to get the full text of the article.  I just use google and type in the title of the article.  The first link I get is a link to the full text, no subscription required.  I always explained this to myself using a simple market-segmentation idea.  WSJ will not give their content away to someone who is browsing their site directly because that person has revealed a high value for WSJ content.  Someone who is googling has revealed that they are looking for relevant content, without regard to source.  There is more competition for such a user so the price is lower.

But today I noticed that bing, Microsoft’s new search engine, does not get the same special treatment.  If I bing “At Chicken Plant, A Recession Battle,” the link provided leads to the same truncated article as my rss reader.  Since users have free entry across search platforms I can’t see any reason why bing-searchers (bingers?) would be systematically different than googlers in terms of the economics above.  Therefore I am giving up on my theory.  What are the alternatives?

1. Google has a contract with WSJ?
2. WSJ would like to shut out googlers too but finds it hard to shut off a service that users have come to expect. Knowing this, they are keeping bingers out from the outset.
3. The game between content providers has multiple equilibria.  On the google platform they are playing the users’ preferred equilibrium.  On the bing platform they have coordinated on their preferred equilibrium.
4. Google has figured out a secret back-door that bing hasn’t found and WSJ just hasn’t gotten around to closing.

Ok the ideas are gettng more and more lame.  I am stumped.

Incidentally, there was an article in the New York Times about DOJ investigations of Google, and a Google PR offensive:

“Competition is a click away,” Mr. Wagner says. It’s part of a stump speech he has given in Silicon Valley, New York and Washington for the last few months to reporters, legal scholars, Congressional staff members, industry groups and anybody else who might influence public opinion about Google.

“We are in an industry that is subject to disruption and we can’t take anything for granted,” he adds.

Rings a bell.