Don’t Destabilize my Emotions

Funniest term to come from a Philippine government official: “destablizing emotions”

We should adopt a rule in using language: Don’t ever mix two languages in one sentence while speaking.

Attitudes about Trade

The Stolper-Samuelson Theorem says that certain groups are better off with trade and some people are worse off. Grossman and company have used this result in political economy models to predict tariff rates.

I’ve been reading that SWS has a new poll now showing that many pinoys support globalization. I recently read Dani Rodrik’s paper that tries to link political economy models’ predictions on who benefits from trade, and those whom surveys identitfy as supporters of globalization. His results are largely consistent with theory.

Strategic Incompatibility

The fact that ATMs charge a fee for foreign transactions was something that always seemed natural to me. We’re reading a paper by my professor on Strategic Incompatibility, or the incompatibility that comes from the desire to increase market power, to make your deposit products more attractive.

I think they attack this in a novel way, by first estimating how a monopolist would price foreign transactions. If actual foreign fees were higher, the difference would be the premium of the foreign fee which tells you the extent the bank is trying to make its deposit products more attractive.

Indeed, they find that large banks have a premium. We expect this, because they have most to gain from network effects.

For small banks, we might expect another effect, a negative premium, to attract more business into their deposit base. They don’t disscuss it however.

Great Photographs

Very nice photography. It is interesting that these are news photos and yet they look staged… One can only imagine the thousands of photos that didn’t make it…

Degrees of Onscreen fear

There is an interesting article today on inquirer.net about Marilyn Reynes being interviewed by British film critic Pete Tombs. Marilyn is (in?)famous for being in 5 of the 8 Shake Rattle and Roll movies.

Read the whole thing. Enlightening. My favorite part of the interview :
“The actress pointed out to Tombs that there were different degrees of onscreen fear. “It’s hard to act scared if you don’t know where the fear is coming from.”

TRIPs without Drugs II

Now to the gory details. The authors, who i now collectively christen CGJ, calculate a structural model of the Quinolone market. This essentially means estimating the demand and supply side of the model, i.e. elasticity parameters, etc. The parameters are used to construct counterfactuals, a what-if situation. In particular, they ask the question, what if domestic patented goods are withdrawn from the market as a results of TRIPs? — how much is the welfare damage (short-run)?

The actual numbers they get are beside the point (mainly because, when the TRIPS safeguard comes offline in 2005, all quinolone molecules will be off-patent anyways, rendering the numbers moot). The key finding is that domestic goods are not perfectly substitutable for foreign goods, even for the same molecule/Active ingredient. They conjecture (altho they can’t prove) that it must have something to do with better distribution facilities/marketing by local firms. Hence even in the presence of lots of substitutes (made by foreigners), the welfare loss will be large.

The authors conclude that compulsory licensing is a great idea, and price controls, while seemingly similar to licensing, will have deleterious effects on long run R&D in the market.

This is interesting, i have to research more on compulsory licensing. I wonder if there are models for this that exist?

TRIPs without Drugs I

I’m working on slides for my presentation in international investment. The idea is an intriguing counterfactual in the Indian pharmaceutical market: what happens if the next day, all patented drugs are taken-off the market?

Counterfactuals are tough to do in economics using historical data. To do it, you need to write down a fully formed structural model of behavior. Usually, data is hard to come by. The real problem is that the results you get are influenced alot by the assumptions and model you use.

Before the model and results, i’ll sketch the background of the paper in this post. Here, Chaudri, P Goldberg (yes, the same as the Price Discriminiation P Goldberg), and Jia think about the effects of the expiration of TRIPs developing country patent safeguard on welfare and profits in India’s Pharmaceutical industry. India has a well-developed domestic industry and is famous for it.

In 2005, TRIPs stipulated that developing countries had to re-tool their intellectual property laws to recognize product patents. I don’t know whats happened in the industry since then and i’ll have to research that to round up my presentation. They look at the Quinolones, a class of anti-bacterial drugs.

The skinny is, after estimating their model, they find that the expiration of the safeguard will mean heavy static loses to the tune of $305 million, about half the size of the the entire systemic antibacterial drugs in 2000. Most of that is lost consumer welfare. Wow. More on the model and results soon.

New Shows

In grad school, the emphasis is on learning the techniques — but at the end of the day, economics is all about telling good stories. Good economists tell good stories. I’m in economics, in part, because of its knack for telling interesting stories about how the world works.

A model is ’story’. Its a collection of agents (characters), who have objective functions (motivations), that exhibit certain regularity over time (story arcs) or characteristics (genre expectations). Moreover, models are designed to focus on important characteristics in a given situation.

Same for movies and TV shows — they only have half an hour to an hour worth of screen time, so they focus on the important characteristics of behavior. These lead to plot, and if the show is good, it can lead to important insights into life :Lost, Ugly Betty, Battlestar Gallactica, Sex and the City, Deal or No Deal, Gilmore Girls… the list goes on. all surprising and interesting…

Speaking of new TV shows, NBC and the other TV networks have released their new Fall line-ups. Some of them look promising…

A Tale of Two Studies

This week, we’re reading P. Goldberg’s JPE article “Dealer Price Discrimination in New Car Purchases: Evidence from the Consumer Expenditure Survey”. Don’t let the droll title fool you; it’s a paper that wants to tackle an empirical mystery. I also think its implications are more important than it lets on, as you’ll see.

The mystery unfolds: one day, a paper that sought to study price discrimiation was written by Ayres and Siegelman (1985, AER). They had what one would imagine to be a perfect scientific study. They formed a group of buyers, women and men, blacks and whites. They gave them the same bargaining strategy, and asked them to go looking for a new car in the Chicago area. The results are interpreted as follows: a black man and a white man with the same script goes to the same dealer – if the dealer offers a black man a higher initial (and/or final price), then this is evidence of discrimination.

The results were spectacular. For a white woman, the premium (over white males) is $200. For black males, it was most stark: $1,100 over the price initially offered to a white man. A&S also reports the same magnitudes for final transaction prices (of which only 20% of total negotiations ended up with a sale).

Enter P Goldberg. Instead of a controlled experiment, she uses survey data, the CES. She runs a regression, and finds that gender or minority status — in fact, none of the demographic or financial status variables are statistically significant. Intuitively, we can interpret this as: being a minority does not affect the average final price, contrary to that found by A&S.

She now has a problem. A&S’s study is much more compelling, because it’s a controlled experiment. Without reconciling her results with A&S, she won’t be able to publish at the JPE – a tragedy indeed!

At this point, theory comes to the rescue.

What theory tells her is that buying a new car is a negotiation. The longer you are willing to wait, the more leverage you have. The more you are willing to wait, the lower your reservation price must be – defined as the highest price a buyer is willing to buy the new car. The other insight is that seller strategy in the face of consumer differences in leverage, must be to offer higher prices to the group with a higher variance of reservation prices. What her regression tells her is that means of reservation prices are the same across gender and race. However one also needs to take into account variance, or the spread of the reservation prices.

She does this by running quantile regressions, basically the same regression, expect now focusing on the top 10% and the bottom 10% in dealer discounts (her measure of the price discrimination premium). She finds what theory tells her to find: that for the top 10% in discounts, being a minority gives you 453 dollars more, while in the bottom 10% (the high reservation price individuals), being black costs you about 784 dollars more.

So why is this important? Smaller, controlled experiments are usually the gold standard in the natural sciences. But in economics, its results may be misleading. The A&S results may just be sellers reacting to their knowledge of the reservation price dispersion of minorities versus non-minorities. Acting ‘similarly’ doesn’t change this fact — i.e. a black man will act the same as a white buyer, but will be given a higher offer because of seller beliefs in the distribution of black reservation prices. They offer prices based on Hence, what we really need is the whole distribution, not just the means. Small controlled studies don’t/can’t give you the variances.

Can you tell if consumers stockpile yogurt?

Today, we talked about Nevo and Hendel’s paper on consumer inventory decisions. The idea is the following. If you only had data on purchases and prices, but wanted to say something about whether inventories played a role in buying behavior, what would you do?

They did what any red-blooded economist would do. You posit a theory. They develop a dynamic model of consumer behavior and the game is find an implication that can only come from the dynamic model, and not the static one. For example, an empirical result that says purchases rise when prices are low won’t cut it as that is also predicted by static models.

So, what is this unique prediction? Well, assuming that people keep stores of goods, then that must mean that the duration until next purchase is longer during a sale. Why? A sale means a lower than usual price, so you are tempted to buy more and keep it in inventory. Since you have more in inventory, it’ll take longer till you need to buy again.

The key issue, as I mentioned earlier, is to infer stockpiling behavior separate from consumption behavior when you only have consumer purchases. The claim is duration till next purchase does exactly that.

I should say something about the kinds of goods they used. They used laundry detergent, an obvious candidate for stock-piling effects. After all, who cleans more clothes just because the price is lower than usual? The other products they use are softdrinks and yogurt. I’ll talk about my thoughts on that later. So here is the data:

Days to next purchase
Detergent — 1.95/28.91
Yogurt — 2.78/21.64
Soft Drinks — 2.5/29.74

The first number is the within effect. Interpret this as the per-HH difference between sale and non-sale periods (recall that duration is longer during sale periods, so this ‘confirms’ our prediction because it’s a positive number). The second number is the between effect. Technically, is the coefficient of the cross-section regression of mean days to next purchase per HH on the proportion of purchases that HH bought on sale. Take detergent — for a HH that bought detergent on sale an ‘average number of times’ (as calculated from the sample), you can expect this ‘average’ HH to wait 29 days more to purchase again than if he buys it non-sale. So it’s a reflection of differences between HH’s to sale prices, hence between effect.

Remember the motivation for this article. Can we infer stockpiling from this? The authors do. As I also said, stock-piling is such a natural conclusion for detergents. What else will you do with it? But for softdrinks and yogurt, the story is harder to tell.

For instance, we can’t rule out the case that most of the new purchases were in fact consumed immediately. We can’t coz, we don’t have consumption data. In fact, its as likely that when a lot of yogurt is bought, most of it is consumed immediately, you get sick of yogurt, and don’t buy it until 22 days later. Technically speaking, this is consistent with diminishing utility over a given time.

The big question then is, how can we infer stock-piling? We need an assumption on consumption, there’s no getting around it! Usually, these models would also predict a form of consumption smoothing where it predicts most people would like to consume a steady amount per day. This is a result of a model, and NOT an empirical finding (not that I know of at least). However, if we accept a strict mode of smoothing, i.e. zero response to price changes at all, then we’ve assumed away the problem of sorting the consumption versus stockpiling effect –i.e. its all stockpiling.

Lastly, to cast even more doubt on the paper, here is their data on the changes in quantities bought.

Quantity
Detergent — 1.14/2.22
Yogurt — .2/.22
Soft Drinks — 3.01/6.44

The interpretation of the numbers is the same as before. The between effects are bigger, but NOT that big. The extra purchases wrought by a sale are small, so its anyone’s guess if it goes to stockpiling or not. In fact, I’d expect a small number to be evidence of a consumption effect.

All in all, I think the claim of separating inventory from consumption changes only from duration is I think premature. I need more to convince me.