Dollar strengthening against yuan

In the market for foreign exchange, the price of a dollar is appreciating relative to the Chinese yuan, nearing 7 yuan/dollar.  Either:

• the demand for dollars is increasing (more Chinese consumers want to buy US goods and services or invest in the US); or 
• the supply of dollars is decreasing (fewer US consumers want to buy Chinese goods and services or invest in China).   

According to the article, this could be due to fears of a trade war with the US:

• fewer US consumers would buy Chinese goods due to tariffs on imported Chinese goods; or 
• fewer US companies would build factories in China or source parts from China.

Both of these would lead to a decreased supply of dollars and a higher "price" of dollars.  

In terms of a broader strategy, a stronger yuan would hurt Chinese producers with plants in China and could lead to unemployment.  This could put pressure on Chinese politicians to loosen up unfair restrictions on foreign firms and/or strengthen foreign intellectual property rights.

Are we in a housing bubble?

The term "bubble" means that prices are too high.  But the question is difficult to understand unless we know what to compare them to.  We can use the ideas of Chapter 9 to measure housing prices relative to the cost of renting.  Because renters can buy, and owners can sell and rent, in the long run the prices should be comparable (a mobile asset should be indifferent about where it is used).  Here is the Fed data on the relative costs of owning relative to renting.  

Many people (some economists) think that there was a housing bubble in 2006 whose bursting lead to the subsequent recession (shaded blue area).  Currently owning is 30% more expensive than renting.  In 2006, owning was 70% more expensive than renting.  

Google measures ad effectiveness with "Ghost Ads"

PROBLEM:  How to measure Ad Effectiveness (paper link)

To measure advertising effectiveness, we want to make a simple comparison, "Did showing the ads change users' behaviour, relative to not showing them?"

This is especially hard to do when algorithms choose ads to show to consumers who are most likely to click on them and purchase.  This leads to Selection Bias, i.e., those who see the ads are those most likely to click on them.  One cannot tell whether people purchase because they were shown the ad (the "treatment effect," what we want to measure) or whether people who were going to purchase anyway were shown the ad ("selection bias," what we want to rule out)

To eliminate selection bias, Google uses Ghost Ads to make the experimental ads visible to the ad platform and experimenter but invisible to the control group, exactly as the algorithm would have chosen.  Using the algorithm to select the number and order of ads, but then "ghosting" the ads in the control group, eliminates selection bias.

Figure 4 illustrates how it work.

The treatment group (upper left) sees three Louboutin shoe ads and three ads from other advertisers. The ad platform "selects" the control group (lower right) as those consumers who would have seen an identical quantity and order of the other ads.  To create the control, the researcher simply "ghosts" the ads that would have appeared in the same order and location as the Louboutin shoe ads.

Google also can select users to match the demographics of the treatment and control groups. 

MORAL:  Carefully designed experiments can accurately measure causal relationships.  In this case, the treatment ads lifted website visits by 17% and purchases by 11%.

See simple exposition or the paper.(Ungated: https://courses.cit.cornell.edu/jl2545/adpapers/Randall%20Lewis.pdf)

HT:  Hal

REPOST: can I take credit for this?

Former student Ford Scudder (NJ state treasurer) must have been paying attention in class, as he reduced the discount rates for NJ defined benefits pension system from 7.9% down to 7%.  The move is controversial as it increases the amount that NJ must set aside for its generous retirement benefits:

The decision will likely increase what the state will have to pay into the pension system next year by $234 million, according to the Treasury Department. Instead of a $3 billion pension contribution in his first budget, Murphy [the new Governor] would likely have to make a $3.2 billion contribution under that estimate.

But it is better in the long run:

“Given the current elevated level of asset values across the board, long-run expected returns have diminished, so it is appropriate to lower the assumed rate of return,” Rijksen said. “Our actuaries have suggested doing so, and it is the unmistakable trend in public pension plans across the country, with some other 20 state pension plans having adopted or being in the process of moving to an assumed rate of return at 7 percent or below.”

Readers of this blog will know that I motivate the topics of discounting and its inverse, compounding, with use our under-funded defined-benefits pensions.  Here is a recent post about another fund, from another under-funded state, doing the same thing:

Thursday, February 2, 2017

Another pension fund lowers discount rate to 7%

If a pension fund has to pay out $100 in 30 years, and earns 7.5% on its investments, it must save 100/(1.075)^30=13.14 today.  If it earns only 7.0%, the amount that it much save increases by 15%.

Calstrs, the second biggest pension fund in the world, just admitted that it is reducing its target rate of return (also its discount rate) from 7.5% to 7.0%.  The increase in payments is split between the teachers and the State of California, the employer of the teachers.

Approximately 80,000 current members of Calstrs could see an increase in their yearly pension contributions of $200 or more as a result of Thursday’s move, Calstrs said. The state of California has already budgeted an extra $153 million for its pension contribution to cover the rate change, bringing the total contribution to $2.8 billion.

Randomized trial to measure Facebook ad effectiveness

In general, selection bias results in over-estimating the effectiveness of ads because ads are shown to those most likely to buy, e.g., as advertisers employ techniques like machine learning to maximize the effectiveness of ad campaigns. 

However, a new study from facebook finds a way to cleverly construct a control group that is not subject selection bias.  By comparing "conversion" rates for the winner of an ad auction, e.g., page viewing, registering, or buying, to what would have happened had the second-best bidder won the ad auction, facebook estimates the causal effect of ads (free from selection bias).  

Because facebook has "single user login" they can:

• make sure that no one in the control group has seen the ad; and 
• track conversions across different devices and for several weeks after ad exposure by embedding a "conversion pixel" on the checkout confirmation page.  

The study concludes that targeted ads a 73% increased likelihood of conversion.  

UPDATE FROM READER:  I think that this is the same idea as "ghost ads".   See simple exposition or the paper.(Ungated: https://courses.cit.cornell.edu/jl2545/adpapers/Randall%20Lewis.pdf)

Book recommendation: Algorithms to Live By

Good book written by Computer Scientists.  Each chapter begins by describing how algorithms solve a class of problems, like Optimal Stopping, Sorting, Caching, Scheduling, and the Multi-arm Bandit problem. Then, the authors draw advice from the algorithms to solve analogous problems for humans.

For example, in the scheduling chapter, the authors describe early attempts to split CPU (Central Processing Unit) time among various tasks.  Scientists found that as the number of tasks increased, the CPU would start spending more and more time on the meta-task of scheduling.  When too many tasks where adding, the computer would spend all its time "thrashing," switching between tasks without getting any work done.

The metaphor to our own lives is obvious.  Set aside a certain amount of time for uninterrupted work, especially when writing, and avoid switching between lots of small tasks, e.g., answer e mail only once per day.  

Do Boys have a Comparative advantage in Math?

Nice article from Alex Tabborock suggesting that the male preference for STEM majors is driven by their comparative--not absolute--advantage in math.  Here is the difference:

... consider what happens when students are told "Do what you are good at!" Loosely speaking the situation will be something like this: females will say I got As in history and English and B’s in Science and Math, therefore, I should follow my strengthens and specialize in ... history and English. Boys will say I got B’s in Science and Math and C’s in history and English, therefore, I should follow my strengths and do ... Science and Math.

This explanation can be understood in terms of marginal analysis of two tasks, i.e., the opportunity cost of doing one is the foregone benefit of doing the other.  For boys,

MR(math)>MC(math)=MR(English),  

which drives their revealed preference for math (or STEM) over English (non STEM).

Bribing Amazon Employees

The WSJ first reported that Amazon employees have taken bribes to provide information on product reviews and even to have negative reviews simply removed.

According to the report, middlemen use social media sites like WeChat to track down Amazon employees, offering them cash to turn over internal information or to delete negative reviews. The WSJ also reports that it costs roughly $300 to take down a bad review, with brokers “[demanding] a five-review minimum” per transaction. Amazon employees have also been asked to provide e-mail addresses of customers who left negative reviews, or to provide sales information to give sellers an edge against their competitors. To combat the behavior, an Amazon spokesperson told the WSJ that it has implemented “systems to restrict and audit what employees can access.”

This is not too different from payola in which recording studio reps paid DJ's to play their songs and thus give a boost to record sales. The problem for the station is that it undermines whatever relationship that it might have with a competing studio. Agents are not working in the principal's best interest.

$300 huh. I know that I am naive because I would have overpaid to remove the bad review.

Investing in Theranos

Our chapter on adverse selection uses IPOs as an example where sellers are more informed than buyers and so there is an opportunity for adverse selection and, perhaps, misinformation. But rarely does it blow up as big is being reported for Theranos.

The concept was irresistible: Theranos said it could take a few drops of blood from a simple finger prick to detect everything from H.I.V. to a diabetic’s A1C level. Relying on a proprietary technology to analyze the small quantities of blood, the private company offered a wide array of tests much more cheaply than existing blood tests.

Would that it had worked. Instead, the tests never were able to prove to work as advertised. It appears that the founder, Elizabeth Holmes, was able to fool the likes of George P. Schultz, Henry Kissinger, and Jim Mattis.

In addition to misleading investors about the promise of the company, federal officials charged the two with encouraging patients and doctors to use the company’s blood tests in spite of knowing they “were likely to contain inaccurate and unreliable results.”

HT: CY

Prisoners' dilemma video