US Manufacturing

So, it's the first week of school and I'm waaaay too busy to actually generate new ideas for this blog (or even to recycle old ideas). But I did chime in on old-friend Kevin's post over at Emergent Fool. We'll see where that thread leads.

Smart Firms Run Experiments

First a note about Google Adsense. I must have misread their TOS when I posted about this before. I thought you got a check when you accrued $10 from blog ads, but it turns out you need $100. So, I will be sending a $100 check to the Utah Food Bank sometime in the far-distant future, assuming I live that long. (Current tally: $18.58.)

---

Continuing on the theme of statistics and data mining...

Every year I teach pricing.

I always conclude that optimal prices depend on two things: Demand elasticity and marginal cost.

Every year students ask how they can learn about demand elasticity.

I say that learning about demand elasticity is really pretty easy. Just do some simple pricing experiments. Change your price a bit here and there, and see how quantity demanded varies with price. Use this to construct some estimates of demand elasticity to plug into your pricing formulas.

And every year one or two students argue that this is really completely unrealistic; that there's no way a real firm would vary prices just to see what happens.

Here's a response, courtesy of Monday's WSJ.

Google, it turns out, is constantly experimenting not with price (since their searches are free to users), but with quality. Users told Google they wanted more search results on a page. So the firm tried it.... and found search traffic actually went down. The extra results slowed down the page, and consumers noticed the difference. Here Google isn't measuring the elasticity of demand with respect to price, they're measuring the elasticity of demand with respect to the number of search results listed. Having measured this elasticity carefully in its experiment, it is now able to make better decisions.

It turns out that the smart firms are doing this all the time, and they're changing not only price but also product characteristics. They're carefully measuring the results of their experiments, and using these estimates to make better managerial decisions. It also turns out that the cost of such experimentation is dropping quickly... Which means the pace of experimentation and change is only going to quicken.

Game Theory Applications

Two game-theory-related items sent in by former students (thanks guys!):

Will the Iranian government develop an A-bomb? They will, if doing so is their best strategy in response to what other actors are doing. Or, in other words, if it's a Nash Equilibrium to develop the bomb.

On a lighter note, game theory seems to be useful for understanding whether the Joker should cooperate with other villains to conquer Batman. I didn't follow all the details, but this seems to be a mathematical proof that the Joker is best off acting alone.

The applications of game theory are indeed endless.

Data Mining for Stocks

Recently, I linked to an NYT article about data mining and statistics. I wrote about how valuable these skills are becoming, but indicated that it's possible to make big mistakes if you don't know what you're doing. This is why it's a good idea to get trained by experts who use statistical analysis in their own work all the time (such as our faculty here at the University of Utah's David Eccles School of Business).

And as if by divine command the WSJ posts an article illustrating just this fact:

Data Mining Isn't a Good Bet For Stock-Market Predictions

Cool nugget in the article: Data mining techniques suggest that Bangaldeshi butter production is highly correlated with US stock returns. So, get the butter data and make a fortune in the stock market, right?

This, of course, is completely ridiculous. As the article reminds us, correlation does not imply causation, and you have be both smart and well-trained (see plug to DESB, above) to know the difference.

But there's more to the stock-market/data-mining interface than just this. To illustrate my point, note that we can probably summarize the WSJ article as follows:

Data mining with stock market data might give you spurious correlations.

But the deeper point (which isn't discussed in the WSJ article but should be) is this:

Any correlations you find doing data mining on stock market data are probably spurious correlations.

Why?

It's the profit motive together with the price mechanism.

Here's how it works: Suppose data mining uncovers a real, causal relationship between some variable and future equity prices. As an example, suppose that any time it snows more than 6 inches at Alta on Martin Luther King Jr. Day, then the S&P 500 goes up by 10% in February. Well, the profit motive means there are lots and lots of people out there looking for such patterns in the data. And once these patterns are discovered, people will start to trade using this information.

We'll end up with a lot of people watching the MLK-day Alta snowfall. When it snows, they'll all buy stock, hoping to cash in on that 10% return in February. And what will their purchases do to market prices? Drive prices up, of course. How far? Prices will be driven up the full 10% on the day right after MLK day. And this means that prices won't go up by 10% in February; because that increase has already been priced into the market.

The profit motive together with the price mechanism will tend to knock out any "real" data-mining/stock-market connections, pretty much immediately.

So the question to ask about data-mining for stocks is this: If the connection is real, then why haven't traders found it (and exploited it) already?

This isn't to say that one could never make money doing data-mining type stuff on stocks. I do know people who do just this, and they seem to make money (sometimes, at least). But these guys have really big computers, and really big data sets, and they're doing really complicated stuff, and on top of that they're always asking themselves why the correlations they've found haven't been found (and exploited) by others. It's an extremely, extremely competitive area. So you gotta be careful.

Cherries

Here's a nice illustration of how market power allows producers to destroy social value (courtesy of Saturday's SLTrib):

Utah growers: No longer a bowl of cherries

Utah's cherry growers plan to let 10 million pounds of cherries rot on the ground this year, rather than putting them on the market and allowing consumers to buy them. Why? To keep prices high.

A few weeks ago, I wrote something about the MBA Oath, and included a made-up example of how market power in the market for fresh daisies destroyed social value. This real-world story about cherries fits that example perfectly.

This cartel behavior --- which would be illegal in most industries --- is coordinated by the Cherry Industry Administrative Board. The US government gives agricultural producers the right to organize to limit production. And we all pay higher prices for food as a result.

Do Physicians Respond to Financial Incentives?

Interesting letter to the editor from Dr. James R. Fowler in the SLTrib on last Saturday. Here's a snippet:

In 40 years of practicing medicine, neither I, nor any physician I have known, has ever based a patient's treatment on financial remuneration. It is an insult to me and the entire medical profession to imply that physicians base their treatments on anything but what is best for their patients.

I'm going to post a response, but first I want to put a big caveat in. I don't know Dr. Fowler, and I probably don't know any of the doctors that he's referring to. It's entirely possible that the first sentence in the excerpt above is the literal truth.

But the general statement that physicians don't respond to financial remuneration is not supported by the data. Health economists have done a lot of work studying this question, and the broad picture is that financial incentives do seem to influence physicians' choices regarding patient care.

Here are a couple of examples:

In the 1989 New England Journal of Medicine, Alan Hillman, Mark Pauly and Joseph Kerstein compared physicians who were paid either salaries or on "capitation" to those paid by fee-for-service. Salaries mean that doctor pay doesn't vary at all with how they provide care to patients. Capitation means that doctors are paid a fixed amount per patient, regardless of the level of care provided to that patient. Fee-for-service means that the doctor gets paid more when the patient "buys" more health-care. The results? Patients of salary or capitation docs had a lower rate of hospitalization than patients of fee-for-service doctors.

In 1999, Jon Gruber, John Kim, and Dina Mayzlin reported (in the Journal of Health Economics) a study of rates of Caesarian Sections for childbirth. Medicaid, it turns out, offers much much lower reimbursement rates for C-Sections (relative to normal childbirth) than does private insurance. That is, a doc faced with a Medicaid patient might get $130 more for doing a C-Section than for a normal childbirth. But that same doc with a patient covered by private insurance might get $550 more from a C-Section. Rates of C-Section are much lower for Medicaid patients than for the regular population, and this remains true even after controlling (statistically) for factors such as breech, fetal distress and maternal distress. Further, the authors show that changes in the Medicaid differential (say, from to $130 to $400) are associated with changes in the relative rates of C-Sections in the Medicaid and private-insurance populations.

These are just two studies out of dozens. I could go on and on.

Two caveats are important here: First, evidence like this doesn't mean that all doctors make all decisions in response to financial incentives. It's likely that there are many physicians, like Dr. Fowler, who don't respond to financial incentives. But the evidence suggests that at least some docs respond to financial incentives at least some of the time. And this means that that policy makers should think about physicians' incentives as part of the broader health-care-reform picture.

Second, it remains unclear what the "right" level of care is. That is, the second study shows that we get more C-Sections under private insurance than under Medicaid, and that this is related to reimbursements. But do we have too many C-Sections under private insurance, or too few under Medicaid? It's simply not clear from studies like this what the right level of care is, so it's hard to determine the right way to structure physician incentives.

All this commentary so far has been directed at the first sentence of Dr. Fowler's snippet, so let me comment a bit on the second. Is it an insult to doctors to suggest that they, too, are at times influenced by financial considerations? Maybe so, but to me this just means that doctors are people too.

Cash for Clunkers, Part 2

Ryan posted a great question about my recent Cash-for-Clunkers post:

I'm curious -- does it make economic sense to pay people money to destroy cars that still have value? I realize their goal is to stimulate the economy and save the environment, but it seems to me that the $3 billion in the cash-for-clunkers program could be better spent. Why ruin working cars?

I think the "stimulate the economy" argument for Cash-for-Clunkers is a bit overblown. (So does the WSJ editorial page.) Even if the government spends $3 billion on this program (and causes consumers to spend tens of billions more of their own money) this is really small potatotes compared to the US GDP of about 13 trillion.

On top of that, it's pretty clear that destroying valuable things isn't a good idea.

Unless, that is, the things are privately valuable but not socially valuable.

Suppose, for example, that a clunker-driver values his clunker at $500, but the rest of society bears $600 of costs because the clunker is being driven. Why might a driver impose costs on others just from the mere act of driving an old car? I can think of three possible reasons: dependence on "foreign oil", local pollution, and global warming. With the numbers in my example, the private value of the clunker is $500, but the social value is -$100. Society overall would be better off if the clunker didn't exist.

An example policy response to foreign oil, local air pollution or global warming would be to raise emissions and fuel economy standards for new vehicles. However, all that new technology costs money, so raising standards for new vehicles will make new cars more expensive. And if new cars are more expensive, consumers will naturally avoid them and hold on to their clunkers a bit longer. And this defeats the whole purpose of raising emissions and fuel economy standards, since these old cars are dirty and low-MPG.

So what can be done about this? One option is to raise the price of old vehicles as well --- if consumers see that the prices of both old and new vehicles rise, then we won't see (as much of) the substitution away from new to old. Instead, we'll see more substitution from new and old cars to things that aren't cars at all --- bikes, buses, scooters.

How can we raise the price of old vehicles? There are two things that change prices --- demand and supply. And that's where Cash-for-Clunkers comes in. Destroying the old cars decreases supply and will raise prices. And this could be good for the world if there are negative externalities associated with dependence on foreign oil, local pollution, and global warming.

Here's an article outlining more of this reasoning.

I've seem some suggestions in the blogosphere that instead of putting sodium silicate in clunker engines, we should give the cars to poor people. It's true that Cash-for-Clunkers will make used cars more expensive, but, as I noted above, that's the whole point. And the best way to help the poor isn't to distort downward the prices they pay or distort upward the wages they receive --- it's to bolster their incomes directly.

Be a Data-Driven Manager

Awesome article in today's New York Times:

For Today’s Graduate, Just One Word: Statistics

The article is about how firms --- faced with mounds and mounds of data --- now find they need people who know how to organize and analyze it. If you have the skills to answer real-world questions using real-world data, you're going to do well.

But getting those skills requires you to do some serious book learnin'. Econometrics and statistics are tricky stuff, and it's easy to make big mistakes.

This is, I think, one reason why the overall rates of graduate school attendance have been rising. The world is an increasingly complicated place, and so there are a lot of valuable, real world skills that are so complicated that you really need a college degree before you can even start to study them carefully. And you need enough experience with the world to see how you'd use the skills, so you can appreciate the relevance.

This is great (for me) because I've been trying to move my MBA classes this direction anyway. So, I'll invite first-year MBA students to ask me how one would use data to answer the questions that come up in class. I'll try to have answers for you.

Cash for Clunkers and Yahoo/Microsoft

There were two big news stories last week that will for sure be discussed in MBA classes this fall.

Story #1 is the Cash-for-Clunkers program; you know, the one where the government throws in $4500 if you trade your gas guzzler for a high-MPG car.

This one has a personal angle for me... My parents, it turns out, are trading the Ford Ranger pickup truck that I bought in 1990 (and sold to them in 1998) for a Toyota Corolla. Had some good times in that truck... sniff.

Anyway it's a great example for teaching microeconomics, for two reasons. One is that it's a great illustration of the economics of a subsidy. Does the program benefit auto companies or consumers more? It depends, in part, on the relative elasticities of demand and supply. Probably I'll write a homework problem about this for the fall, so I shouldn't discuss it in much detail here. Reason two is that it illustrates how hard it can be to assess demand and supply without actually doing some pricing experiments. The reason Cash-for-Clunkers has been in the news so much is the overwhelming consumer response to the program. The government really didn't know how consumers would respond to the offer of $4500 in exchange for clunkers, and it turned out that supply was quite a bit more elastic (that is, responsive to price) than they thought. It's a good reminder to all managers trying to understand demand for their products--- in the absence of data, it's really hard to guestimate demand. So get yourself some data!

Story #2 is the announcement (finally!) of a search deal between Microsoft and Yahoo.

Personal angle here as well: I met Jerry Yang and Dave Filo (Yahoo's founders) a couple times when I was in grad school at the Stanford GSB and they were in engineering grad school. Ask me to tell you my Yang/Filo/engineering-league-softball story sometime. (Probably I drove my Ford Ranger over to the big field behind the Terman Engineering building to play softball --- how's that for a connection?)

This one's nice because it points out that writing a contract really is an alternative to doing a merger. Back when I used to teach strategy, I'd point to, say, the Disney/ABC or Time-Warner/AOL merger and ask students where the value creation was coming from. Students would typically come up with a list of potential synergies, usually involving various cross-selling plans. One justification, for example, for Time-Warner/AOL was the ability to stuff TW's magazines with AOL CDs. But then I'd ask why a merger was necessary to realize the synergy. Firms write cross-selling contracts all the time, and it was just never clear (to me, at least) why TW/AOL couldn't have done exactly the same thing.

And here's a case where MSFT and YHOO clearly thought hard about doing a full-on merger, but in the end decided they could realize synergies with a cross-selling contract. Yahoo is going to shift all search activity to Microsoft, but continue to sell its own advertisements and offer other forms of content. The parties have split the resulting gains using a detailed revenue-sharing contract.

So Yahoo/Microsoft is one example, but could all mergers be replaced by contracts? If not, then when do we need mergers and when do we need contracts? These questions naturally lead into a discussion of the limits of contracting --- limits coming mostly from problems with observability, verifiability, and enforcement of contract terms. And that's where the economics of contracting come in handy for understanding Wall Street financial transactions.

Do Study Groups Matter?

A couple of information items, and then a link.

(1) I'm now an Associate Editor at the Journal of Labor Economics, which is the leading field journal for studies of labor markets and human resource management. Before you get too impressed, the title basically means that I referee a lot of papers for them. (Academic publishing works like this: An author sends a paper to the journal. The journal's editor sends the paper to a referee, who reads the paper and sends some comments along with a publish/don't-publish recommendation back to editor. The editor decides whether to publish the paper or not, and then forwards the comments to the author.)

Anyway, since I'm on the Journal of Labor Economics team, I figured I'd start blogging periodically about interesting papers published in the journal. The July issue just came out, so I'll get to that in a bit.

(2) I'm branching out on you. I'm going to start putting a bit of content on the U's Red Thread blog. As I noted in my recent iPhone post, I've recently developed an interest in the economics of higher ed, so I'll probably put some education-related thoughts there.

My first post there is on the question of whether study groups --- you know, that group of people you meet with to talk over the upcoming marketing exam --- matter for academic achievement.

You can read it here.