What does an economist at Facebook do?

Michael Bailey, who is an economist at Facebook, reports on Quora:

I currently (Feb 2014) manage the economics research group on the Core Data Science team. We are a small group of engineer researchers (all PhDs) who study economics, business, and operations problems. As Eric Mayefsky mentioned, there are various folks with formal economics training spread across the company, usually in quantitative or product management roles.

The economics research group focuses on four research areas:

Core Economics – modeling supply and demand, operations research, pricing, forecasting, macroeconomics, econometrics, structural modeling.

Market Design – ad auctions, algorithmic game theory, mechanism design, simulation modeling, crowdsourcing.

Ads and Monetization – ads product and frontend research, advertiser experimentation, social advertising, new products and data, advertising effectiveness, marketing.

Behavioral Economics – user and advertiser behavior, economic networks, incentives, externalities, and decision making under risk and uncertainty.

I think a more interesting question is “what *could* an economist at Facebook do?” because there is a LOT of opportunity. There are incredibly important problems that only people who think carefully about causal analysis and model selection could tackle.  Facebook’s engineer to economist ratio is enormous. Software engineers are great at typical machine learning problems (given a set of parameters and data, make a prediction), but notoriously bad at answering questions out of sample or for which there’s no data. Economists spend a lot of time with observational data since we often don’t have the luxury of running experiments and we’ve honed our tools and techniques for that environment (instrumental variables for example). The most important strategic and business questions often rely on counterfactuals which require some sort of model (structural or otherwise) and that is where the economists step in.

tl;dr economists at Facebook compute counterfactuals.


Facebook is one of the most important companies on Earth right now and it's creating an absolutely unprecidented amount of value. I think there would be no better place for an economist to make a huge contribution to world growth and the economic AND psychological wellbeing of humanity.

Let's not get carried away here. I don't think "Facebook is one of the most important companies on Earth" Pump your brakes. I've talked to more and more people that are deactivating their accounts.

Facebook just bought out the world's most popular messenging service at an absolute steal compared to its future value. A few ancedotes from you mean nothing. Facebook is connecting humanity like nothing ever before and the sky is literally the limit for the stock price. The importance of these great new technologies like Facebook and Twitter cannot be understated.

Yep, by pumping more cash into that money losing business it will become worth $50 billion.

I remember Compaq's CEO defending buying DEC by pointing out the market cap in the billions of the share DEC had retained when the roughly $100 million investment in Altavista spun off and bought by one of the Facebooks of 1999. A deal to create or unlock wealth or value or Altavista which was Google before Google was Google by way of finally letting Wall Street trade shares in it - Google knew better that to give Wall Street any say in how it is run, unlike Altavista.

If Facebook is the future of humanity, then it is a very sad day for humanity.

Connecting with people is a sad day? Increasing the knowledge of consumers of interesting, and relevant new products available is a sad day? I think the great thing about Facebook is it's helping us move into a more social era. Where in the past people just used to sit around watching TV, the new generation is much more social and interested in people and this is due to Facebook enriching people's lives.

Connecting, in the sense that it only works when people spend a lot of time talking about themselves. It's a push, not a pull model. Very different from the way we connect with others in meatspace.
If facebook is enriching people's lives, is by making them far more self centered.

Hmm, you sound like you work there. Maybe you are Michael Bailey. Or maybe you work in the PR department?:

There are incredibly important problems that only people who think carefully about causal analysis and model selection could tackle.

Sounds like a grad student taking himself a bit too seriously.

It's hard detecting sarcasm on the web.

You do know 'Commentor' is not a word, right?

Yeah - most people's last names aren't actual words, I'm no exception.

# "The economics research group focuses on four research areas..."

And just what do they actually produce from all that PhD focus ?

The boastful emphasis on their input/activity versus any productive output... strongly suggests absence of the latter.

Lack of productive output at FACEBOOK? I assume you're joking.

...So how much of your net worth is tied in up fb stock, anyway?

I smell 'bubble.'

On the contrary Facebook has nowhere to go but up, the recent purchase of WhatsApp shows that there is still plenty of fresh thinking going on in Zuckerberg's inner circle. Expect more revolutionary ideas, not less. We haven't even started to see Facebook's true rise. Those naysayers who are calling "bubble bubble bubble" simply lack imagination and believe in economic fallacies which suppose there is some limit to capitalism.

Trolling level: boring.

Yeah - change it up. We know how you feel about FB. What about those busses giving the San Fran meat puppets fits?

What do you want me to say? Silicon Valley companies have the right to provide bussing for their employees what's he problem here?

Yeah, this guy illustrates how hard it is to maintain an interesting level of trolling.

Easier said than done. It reminds me of this great article about talk radio:


"What's amazing is that when you get new people who think that they can do a talk-radio program, you watch them for the first time. By three minutes into it, they have that look on their face like, 'Oh my God, I've got ten minutes left. What am I going to say?'"

So yeah it seems like JaMRC has run its course and is now just looping.

The exposure of their click farms is going to be their undoing. It increasing looks like a whole lot of fraud is a big part of their business model.


Please, you sound like some anti-globalization protester - pick up the phone 1999 is calling. In the past developed countries leveraged their low wages by exporting textiles, cheap electronics, etc. Now countries like Bangeldesh are leapfrogging development by investing in Facebook-Centered Brand Empowerment - these firms are the wave of the future and are lifting tens of thousands out of poverty. It's another example of the massive economic value being created by Facebook. My only criticism is Facebook somewhat tries cracking down on this - I advocate free trade in Likes.

Calling them "Click Farms" is extremely demeaning to the entrepreneurs and workers engaged in this vibrant new industry.

Hahahahahaha, no please say "I'm not a robot" =)

There are incredibly important problems that only people who think carefully about causal analysis and model selection could tackle

You should use smilies. It's hard to sense sarcasm in text otherwise.

I wonder how effective a paid advertising position called "Econ blog commentor" would be?

How many actual software engineers have you worked with Tyler? No, professors of software engineering do not count. I'd say that, as far as what they are good and bad at, you are far from the mark. A large majority of software engineers out there have absolutely no idea about machine learning. As far as whether they are good or bad when having no data, you results will vary wildly depending on the kind of discipline they work on. Someone working on distributed databases will have very little to do with someone designing a blog engine, embedded programming in a car, or a scientific visualization package.

Yeah, that part of the post generated a big "huh?!"

When all you have is regression, everything looks like panel data.

The person who made that statement is the guy at Facebook, not Tyler.

It seems that the main point of machine learning is actually forecasting. So it is a bit odd that to say that CS is bad at answering question out of sample. What economist mostly do is put a "structure" into place that makes or doesn't make sense and the counterfactuals simply reflect the structure. Generally, forecasting is hard and to say that only economist can do this is a bit laughable. Most complex econ model (the structural kind) with the "causal" effects (at least what is claimed) take a long time to run and are generally so far away from reality due to all the technical requirements for them to run that they are relatively useless. And doing causal interference via experiments, it is not sure whether one needs an econ PhD do set up an experiment. It would seem every person who took psych 101 can do this. But we all know, economist like to consider them the king of all things ever.

I don't think it's that economists are the only ones who can do structural modeling, so much as it is that economics enjoys a unique history of being highly concerned with causal inference versus correlation due to its behavioral models (with their equilibrium conditions) usually focusing on macro-level phenomena where historically experiments were impossible. So economists worked with observational data and structural behavioral models side-by-side. It seems like other disciplines, like computer science, aren't always forced through that kind of constraining, but ultimately enlightening, experience. I wonder if Big Data will have its own Lucas Critique one day.

"...a unique history of being highly concerned with causal inference versus correlation..."

You surely are joking?? "unique history"??

"...where historically experiments were impossible..."

As if cosmologists create universes and solar systems in their labs every day.

Economists really lag the sciences in reasoning, and daily fail to reason from a systems approach, treating the economy as both unconstrained by nature and also an open system. Unemployed people never need food, clothing, housing, transport. And workers are not consumers, and consumers are not workers.

Thus paying workers more will not result in an increase in the number of jobs for workers who are handling the extra demand from increased spending by workers. Hiking taxes taxes on gasoline will kill jobs because using the taxes to fix the roads and bridges will not create any jobs.

Nope not joking. Not sure what specifically I said sounded incorrect to you. Also don't understand why you're saying workers aren't consumers to the economist.

"Economists spend a lot of time with observational data since we often don’t have the luxury of running experiments and we’ve honed our tools and techniques for that environment (instrumental variables for example)."

Replace "Economists" with




Evolutionary biologists


That economists make such a point of this is they are incredibly lazy and just unwilling to go out and get their hands dirty collecting data. Shipping records are very detailed for British shipping back centuries. Church records from which on can draw vital statistics go back centuries. Court records go back centuries with records of common law decisions which are mostly economic matters. The records of Egyptian and Chinese eras going back thousands of years provide a lot of economic data - how did the great projects of thousands of years ago get done? Engineers have been drawn into to figure out the building part, but economists seem to find the political economists of building these great works too hard, handwaving them as done by slaves without considering the high skills involved.

I wouldn't be able to sleep at night

What you say is true mulp, but it does not falsify “Economists spend a lot of time with observational data since we often don’t have the luxury of running experiments and we’ve honed our tools and techniques for that environment (instrumental variables for example)", which is also true.

I doubt Facebook have many, if any, Cosmologists, Geologists, Climatologists, Evolutionary Biologists or Archaeologists on hand when this kind of modelling needs to get done. And even though these people use observational data in similar kinds of ways, they are trying to solve very different problems. In the same way an Economist would be useless when tasked with making some useful progress on problems in Cosmology, a Geologist would be useless at developing useful models of economic behavior derived from observational data for Facebook. It is not just the methods that matter but the theory behind them, obviously. That said, the methods economics has developed suit it's purpose well and are there to topple for anyone who can do better.

Liu was one of the 31 migrant workers elected to the National People's Congress last year. She is a foot masseuse in southeast China's Xiamen City. A native of south China's Anhui Province, Liu was a school dropout at the age of 14 and worked to support the schooling of her siblings.

Comments for this post are closed