Lukas Püttmann    About    Research    Blog

My favourite books 2019

My reading this year fell mainly into three buckets: Memoirs, novels by Sally Rooney and books by the historian Thomas Madden.

These were my favorites:

  1. Maoism: A Global History”, by Julia Lovell. Explains how maoist thought and actions by the Chinese government influenced movements in other countries. Not much talked today - especially by the Chinese government.
  2. Istanbul: City of Majesty at the Crossroads of the World”, by Thomas Madden. A fascinating start-to-end story of the Eastern Roman and the Osman Empire taking place in the city on the Bosporus.
  3. Aladdin: A New Translation”, by Yasmine Seale. This is the first documented Aladdin story, intriguingly first published in France in the early 18th century. I wonder about the morality of it: If you had some tool (a Jinn in this tale) that could get you whatever you want, I don’t think the acceptable way to behave would be to kill a magician, get the biggest house full of diamonds and force the terrified princess to sleep in your bed.
  4. Writing My Wrongs”, by Shaka Senghor. He describes life on the streets of Detroit, the murder he committed, 19 years in prison of which he spent 7 years in an isolation cell (4 years in a row).
  5. The Education of an Idealist”, by Samantha Power. I liked the start of the book more than the part when she became ambassador. I think the legacy of her administration’s foreign policy doesn’t hold up well and I didn’t learn much from her time at the UN. They seemed to spend a lot of time on symbolic gestures, but got the big things on Syria wrong. I much prefered her very personal story of her long path through war journalism, academia and government.
  6. Open: An Autobiography”, by Andre Agassi. Samantha Power recommended this book in her interview with Tyler Cowen on her memoir.
  7. Why We Sleep: The New Science of Sleep and Dreams ”, by Matthew Walker. This book has influenced how I think about sleep and has made me more protective of mine.
  8. Venice: A New History”, by Thomas Madden. History writes the greatest stories.
  9. Conversations with Friends”, by Sally Rooney. There are lead characters that I would have guessed to relate better to than a female bisexual Irish poet, but this is a book I knew from the first page I would enjoy and read to the end.
  10. Shoe Dog: A Memoir by the Creator of Nike”, by Phil Knight. Fascinating story of the founding of Nike, touching on how the early running subculture and post-war Japan. Knight is a different founder from those we have in our minds today: He first finished his MBA and worked in accounting when he needed funds for his business. He had very supportive parents and also turned to them at the beginning for money. (Olli’s review)


Learning to work in hierarchies

Samantha Power is the former US Ambassador to the UN under Obama, a Harvard professor and a former war journalist reporting from 90s Bosnia. In her recent memoir, “The Education of an Idealist”, she describes her upbringing in Dublin with an alcoholic father, migrating to the US and her path through journalism, academia and government.

I like her explanation of the different modes of working in journalism/academia vs. government.

Mode 1

In her early twenties, Power worked as a journalist in wartime Bosnia. She reported from Sarajevo while it was under siege and published many articles in large newspapers and magazines. She learned to write in an engaging way which helped her later, when she became an ambassador to bring the stories of individuals to her official UN meetings, to make other diplomates relate to them and their suffering.

She enjoyed her independence, the travel and the attention and status from seeing her name published in the New York Times or the Washington Post.

But she missed the opportunity of shaping events herself. When she interviewed politicians, she would often think: “I would rather be on the other side of the table.”

So she decided to join law school:

Jonathan Moore, the former US diplomat and refugee expert I met working at Carnegie, had become someone I turned to at critical moments. With school beginning at the start of September, I needed to make a final decision, so I telephoned him and asked what I should do. Jonathan didn’t hesitate. “Get the hell out of there,” he urged me. “You need to break out of the compulsion for power, glory, ego, relevance, contribution. Get out. Get out before it gets you, and you forget what got you in.” I didn’t think self-consciously about power, glory, and ego, but Jonathan knew I didn’t mind seeing my name in print.

After her graduate studies, she became a professor of human rights law at Harvard Kennedy School. At the same time, she published “A Problem from Hell: America and the Age of Genocide”, for which she won the Pulitzer Prize.

Mode 2

Power learned about Obama, the new Senator from Illinois, from his DNC speech, reached out to him through friends and joined his team. She entered his administration to work at the National Security Council in a role focused on managing contacts with the UN and keeping an eye on human rights issues.

But she was frustrated that colleagues would not listen to her, that she could not travel as she liked and that her initiatives would come to nothing. She felt like she wasted her time when she and her colleagues would spend a long time marking up each other’s word documents. She needed to learn the vocabulary, such as that stepping over a level in the hierarchy was a “process foul”.

Time also became more precious:

The aspect of government that I had least appreciated before I joined was the importance—and shortage—of “bandwidth.” So much was going on in the world on any given day that one could easily lose an afternoon editing language in various press releases. Mort, my longtime mentor, urged me to prioritize, helping me understand my days as analogous to my mother’s when she worked in the emergency room.

Suddenly she was in the group making the decisions:

[A mentor] often dispensed wisdom on how government worked, and told me I should not have waited until a high-level meeting had ended to make my point. “Listen,” he said firmly. “If you hear nothing else, hear this. You work at the White House. There is no other room where a bunch of really smart people of sound judgment are getting together and figuring out what to do. It will be the scariest moment of your life when you fully internalize this: There is no other meeting. You’re in the meeting. You are the meeting. If you have a concern, raise it.”

In her words, she “did not yet have the relationships, the clout, or the mastery of bureaucratic processes [she] needed to maximize [her] impact.”

Putting it together

It took her some time and the advice from more experienced colleagues such as Susan Rice to get up to speed and learn to deal with her frustations.

To become effective, she used three strategies:

  • Keep an eye on the big picture: Don’t forget the long run even though short-run attention is crowded out.
  • Build a network: She created a weekly “Wine & Cheese” women group to share stories and get advice. And she also played basketball which helped her to get to know the male colleagues in the White House better.
  • Work on smaller projects: These don’t get top leadership attention and are easier to have an impact on.

I enjoyed reading her story and can relate to it. Working in hierarchies is something you have to learn. I found life as a researcher more free, less structured and more bottom-up. I was expected to search for my own research ideas and I could collaborate freely.

It’s quite different for me as a consultant. Our teams work embedded at the client in large hierarchies, so there are more stakeholder to inform and align with. The target audience usually has limited “bandwith”, as your project is only one of many that concerns them. That’s why it’s important to understand the process and to be concise and accurate in your communication.

While the first mode of working is more conducive for being creative, I now realize I could sometimes have been more effective in academia using some simple strategies. A simple, “end of month” email with what I did the last month and what I planned to do in the next would have gone a long way of keeping my supervisors in the loop, even when we couldn’t meet face to face. It would have provided them with an earlier opportunity to provide guidance, voice concerns or to keep me from starting down some rabbit whole.

In other areas, I think researchers use the same tools successfully already. Most writing guides for papers or presentations already advise to be “top down” - to give the message first and the details later.

In the end, the two modes of working are very different, but I think it’s insightful to change the between the two.

La Ville R(ose)

I just came back from the useR! conference in Toulouse which I enjoyed attending and which I recommend.

Some observations:

  • The conference was good mix of lectures, coding workshops and short presentations of new packages and papers.
  • The tidyverse has captured this community. Most code examples use the pipe and dplyr’s functions like select() or mutate() without explanation.
  • The vibe is closer to what I’m used to from academic conferences and it was less crowded and exuberant than NeurIPS.

I mainly visited talks on time series statistics, movement data and a few others.

Time series statistics

The tidyverts collection of packages for tidy time series analysis is fantastic and should be up on CRAN soon. It contains:

  • tsibble: An evolution of the tibble for time series purposes (itself an update to the dataframe). The tsibble makes it explicit which variable indexes time (e.g. “year”) and which groups the rows (e.g. “country”) and stores the information which frequency the data is running on.
  • feasts: New plotting tools tailored to the specific frequencies (e.g. seasons or weekdays) and decompositions of series into cycle and trends using different methods (STL, X11, …).
  • fable: Time series forecasting with many common methods and new visualization functionalities.

Other time series contributions:

  • A presentation on using random forests with time series data (paper, presentation). There was a lively discussion on how to create block bootstrap samples (overlapping? Moving windows? Size of blocks?) needed for the forest. The solution from the talk was to use a validation sample on which to test for the optimal block size.
  • For economists, timeseriesdb - developed at KOF Zurich - might be of interest. It provides a good way to store different vintages of time series and their metadata in a database.
  • The imputeTS package was presented with an application on sensor data. I could well imagine to use that package in a manufacturing analytics study, such as predictive maintenance. In these cases, you often have high-frequency measurements (such as by minute) for thousands of sensors, but the quality of measurements is hard to judge, outliers are common and sometimes missing data is even implicit (such as returning the last value again and again). The presenter pointed out that in these cases, the missingness is often correlated across variables (e.g. when there’s a factory shutdown, bad weather stopping data transmission, etc).
  • Anomaly detection got good coverage: The package anomaly (presentation), the stray package and a trivago real-world example with an interesting conclusion (presentation).

Spatial methods

Movement data is currently my favourite kind of data. It’s spreads across time and space, every data point is “weighty”, describing the repositioning of a giant machine or a group of people and you can use interesting methods to analyze such data.

I was therefore happy to see that there is a lot of work going on in creating packages for drawing maps and analyses of movement data. Presentations: 1, 2, 3, 4. The sf package was presented in two workshops (1, 2).

However, I’ve often found this topic quite difficult to start out with in R and I don’t think it’s become much easier yet. I’m still not convinced that I would go this route if I just needed to draw a quick map. A tool like Tableau takes care of all the underlying stuff such as guessing correctly that some column in your data describes US zip codes and draws the right map based on that.

Other: Packages building, data cleaning, big files

Jenny Bryan held a good tutorial on package development and she made her point really well that we should be writing packages much more often.

Hadley Wickham explained how the great tidyr package gets a facelift, renaming spread() to the more expressive pivot_wider() and gather() to pivot_longer.

I was quite impressed by the disk.frame package. It allows splitting a too-large-for-memory dataset into smaller chungs on the local machine and it only pulls in the columns you need. It also allows for quick staggered aggregations, such as calculating the sum of a variable for the different chunks and then taking the sum of that. Interestingly, that wouldn’t work for other functions such as the median.

Moving from an economics PhD to the private sector

I left academia for the private sector half a year ago. I receive many questions about how that’s been, so I hope a blog post might answer most of the questions in one place.

What I do

I handed in my dissertation in the summer and started working as an analytics consultant at a large international consultancy. We are management consultants, traveling to clients and working on projects that range from weeks to months. We bring quantitative firepower to teams and work on projects such as quantitative marketing, optimizing route networks, health care analytics, fraud detection or predictive maintenance. We use many different techniques, such as random forests, time series forecasting, optimization, webscraping or deep learning.

Our studies are often full analytics studies from the start, and then using optimization or ML is more fundamental than “sprinkles on the top”. But it’s true that most of our time and many of our highest-value activities are “simple arithmetic”, such as collecting the right data, cleaning and merging datasets, calculating KPIs or writing a dashboard. I also spend a lot of time on tasks such as writing presentations, maintaining Excel sheets and calls and meetings with clients.

Why I switched

I decided to give the private sector a try for several reasons:

  • I really like coding. I’d like to get better at it and it’s something that might be valued more in industry than in academia.
  • I want to apply data science in a real-world setting (not least to find out if its all just a hype).
  • I found research very satisfying, but I don’t want to specialize on one topic. My dissertation touched many areas and the demands of academia would likely have made it necessary for me to be more focused.
  • I wanted to try something new.

Advantages of the private sector

1. People

It’s a cliché, but the thing I like most about my job are the people. My colleagues are inspiring, fun and highly motivated.

People are very friendly and sociable in the business world and consulting attracts people with good “verbal and social” skills.

2. Teamwork

You can make fast progress when teams work well together. And it’s not as bad to sit in a team room at 1 am if you’re in it together.

3. Speed of projects

I like that projects move faster and at some point they are actually done and you move on to the next one. This has a diversification effect, so even if a project is bad, it’ll be over not too far in the future.

4. Diversity of projects

The variance between projects is huge, because they can differ along so many dimensions: Industry, function, colleagues, place.

For me, it’s a good way to learn about the data science landscape and to see where my skills might be most valuable.

5. Being challenged

In academia you pick your own subject, methods and execute on your own speed. But being thrown into new topics is also activating and doesn’t let you get complacent.

6. Feedback

Consulting encourages a strong feedback culture. It’s common to have feedback talks every two weeks. This is very helpful: It’s hard to judge on your own how you’re doing and it means there won’t be any surprises after the project is done.

7. Compensation

And yes, you’re better paid.

Advantages of academia

1. Freedom

In academia, I had the freedom to pursue my own projects and that’s something I miss in the private sector. Projects are initiated at much higher levels in the hierarchy and in junior roles you’re executing them.

In contrast, economics research is not very hierarchical. You are free to pick your topics and you pitch your ideas to anyone.

2. Ownership

At universities, the outputs of your work are public. Almost everywhere else – and especially in consulting – you might have accomplished cool things, but few people will know about it.

Academia is much more like running your own startup. You bear the risks, but you also reap the benefits. The downside is that you’ll have to live with the existential fears: “Will I publish my paper well?”, “Will I get tenure?”, …

3. Following rabbit holes

Management consultants live and breathe opportunity costs. This keeps you on your toes to keep making progress towards the target.

But this absence of slack means that what you learn and what you code is determined by what your project needs right now. You don’t just play detective and work on whatever interests you. In academia, new knowledge is the goal and you have to try many ideas and projects to find what has potential. Most private sector jobs are not “unicorn jobs”.

4. Work hours

I don’t think that I worked much less in academia. But work vs. leisure is much more sharply separated in the private sector.

When and how you work is largely out of your control in consulting and that’s a downside. In research, you have self-determined work hours. You might think it’s an advantage to push hard during the week and then to have the weekend off, but - at least for me - that’s not true.

On net

It’s maybe a bit exaggerated, but I would say that the private sector is more conducive to happiness, but research is a better way to find meaning (see here) in life.

How to apply

As economists, we’re very well prepared for data science jobs. We know econometrics, statistics, mathematics and we have internalized most MBA knowledge (e.g. accounting, corporate finance, marketing). We know how to manage long projects, work independently and be convincing in communicating our results.

However, we’re not obvious hires for data science positions. Many people conflate economics students with business majors and might not consider us quantitative enough. It’s not clear to employers that we know ML or that we can program.

I would recommend to learn R or Python. Nobody uses Stata or Matlab (or Julia, EViews, …). You should know one of the two languages reasonably well to be able to start coding on unfamiliar subjects quickly.

You’ll be expected to learn other languages or software soon, too. For example, I’ve learned VBA, Alteryx and Tableau since I started and none of that I’d used before. But I think it’s important to know R or Python, so that you’re productive from the start and that you can show that you can actually code.

On the ML side, the first step is to actually learn about it. An economics program doesn’t teach you machine learning, so you need to look elsewhere. I recommend online courses, checking for courses at your own computer science department or trying Kaggle competitions. A great way to learn about it and to show that you know it is to apply ML in your research.

Given that the outputs of our academic work are public and that you might have time to pursue side projects, it’s a good idea to create public artifacts. It makes sense to take this into account early in your PhD when choosing research projects. I wrote papers with text mining and webscraping, partly because I wanted to learn about these topics. Having written a paper like that is great for your application, because it’s a public output that proves that you know the subject. I’ve explained my patent paper in most job interviews and it was very helpful.

Because outputs in the private sector are not public, the next best proxy is your work experience and years in your PhD often don’t count. Public artifacts are marks of your skill that can compensate for lack of work experience.

When deciding where to apply, it’s important to consider the framework in which a firm operates. So if you work for a consultancy, then you’re a consultant first and a data scientist second.

Conclusion

I’ve enjoyed the whole experience and I’m happy with my choice. But there are also things I miss about academia.

NeurIPS 2018

I attended the NeurIPS (recently renamed from NIPS) conference this year and here are some observations:

  • I find it remarkable how for many people machine learning is equivalent to deep learning and deep learning is equivalent to computer vision.
  • Susan Athey talked about causal inference. She’s been doing a lot of work combining econometrics with machine learning, but it was very interesting to see this presented from the other side of the fence. When she asked who knew about instrumental variables, a reasonable number of hands (maybe 5% of a very large crowd) went up.
  • Edward Felten (Princeton) talked about how experts should talk to decision makers. Basically, don’t come with the “truth” and don’t provide “just the facts”. Structure the decision problem by inquiring about the decision makers preferences and provide information and an option menu of recommendations based on that.
  • Many of the sessions touched on economics: There were papers on market design, value function iteration or - more close to home - on the social effects of automation.

Differences to economics conferences:

  • Many more consumers than producers of research
  • Different demographics: Younger, more nerdy, fewer women, more Asian
  • Fewer parallel sessions, instead “tutorials” (really lectures) by famous people that last 1.5 hours
  • Videos/GIFs on slides. Works suprisingly well and engages people. Examples include simulations of how fruit flies move, 3d shapes circling around, bubbling formations of networks
  • The anthropomorphizing is a bit weird: They say “The algorithm should be able to reason about its own uncertainty”, instead of “We need accurate standard errors”
  • The money involved is different. We don’t have huge recruiting parties by Uber or have Yo-Yo Ma lead a session on the intersection of music and deep learning.

Montreal is the way I wish France was. People are laid back and cosmopolitan. Everything is in French and people switch between languages without effort.


Related posts:

My favourite books 2018

Keeping up the tradition of the last two years, I’m again providing a list of 10 my favorite books this year.

In between finishing my dissertation, starting a new job and moving to a new city, I read less this year. I was worried that I would have almost no time for reading in my new job, but that hasn’t turned out to be the case. The weekend, late nights in hotel rooms and downtimes while traveling still provide many moments to read.

So here are the books I liked most this year (but which were not necessary published this year) in reverse order:

  1. Prohibition: A Concise History”, by W. J. Rorabaugh. This book shows that alcohol has not always been socially acceptable.
  2. The Invisibility Cloak”, by Ge Fei. Cool and short story, set in Beijing.
  3. Bad Blood: Secrets and Lies in Silicon Valley”, by John Carreyrou. Hard to stop reading.
  4. Leonardo Da Vinci”, by Walter Isaacson. Get the print version, the book has beautiful visuals.
  5. Educated: A Memoir”, by Tara Westover. Very well written.
  6. Lenin: The Man, the Dictator, and the Master of Terror”, by Victor Sebestyen. A great way to learn about this slice of history.
  7. Enlightenment Now: The Case for Reason, Science, Humanism, and Progress”, by Steven Pinker. I tried hard to poke holes in his arguments.
  8. Steve Jobs”, by Walter Isaacson. It’s incredible how intimately Isaacson is covering his life. I also enjoy it as a history of the 90s and 00s, a time I lived through.
  9. The Story of Art”, by E.H. Gombrich. I love this book. This is how to write about art. (Here, too, get the print version.)
  10. The Internationalists”, by Oona Hathaway and Scott Shapiro. This builds on years of scholarship, but the authors explain concepts in an accessible manner. Which other book would discuss the ideas of Grotius, Carlo Schmitt and Sayidd Qutb? This is a book that has changed how I view the world. A piece of their conclusion:

    The example of the Internationalists offers a hopeful message: If law shapes real power, and ideas shape the law, then we control our fate. We can choose to recognize certain actions and not others. We can cooperate with those who follow the rules and outcast those who do not. And when the rules no longer work, we can change them.


Collected links

  1. Some Facts About Indian Polling Stations”:

    Of the 748,584 polling stations about which we have data on building conditions, nearly 24% report having Internet and a similar number report having “Landline Telephone/Fax Connection.” 97.7% report having toilets for men and women. 2.6% report being in a “dilapidated or dangerous” building.

  2. Dave Giles: “Controlling My Heating Bill Using Bayesian Model Averaging” (cool!)

  3. Francis Diebold on a new paper by Art Howen and Hal Varian.

  4. Jane Street: “How to shuffle a big dataset

  5. Kaiser Fung:

    The current scandal demonstrates the value of the business intelligence/business analytics functions within companies. We were told that Facebook first realized that certain metrics were showing unusual trends, and upon investigation, they discovered the bugs.

    This is entirely believable. That’s what happens when you have good data reports. They surface anomalies.

Cherry blossoms in Kyoto, 801-2015

Cosma Shalizi offers a course with great name “Data over Space and Time”. As part of this course he uses data on the first day in the year that the cherry trees in Kyoto started blossoming. This data was collected by Yasuyuki Aono and coauthors.

Let’s check it out:

library(tidyverse)

kyoto <- read_csv("http://www.stat.cmu.edu/~cshalizi/dst/18/data/kyoto.csv")

ggplot(kyoto, aes(Year.AD, Flowering.DOY)) +
  geom_line() +
  labs(title = "Cherry Blossoms in Kyoto",
       subtitle = "801-2015",
       x = "Year",
       y = "Day in year of full flowering") +
  geom_smooth()
Cherry Blossoms in Kyoto by day in year since year 801 until year 2015

Note the earlier blossoming of trees at the end of the sample which is a sign of rising temperatures.

Collected links

  1. Against home cooking, by Rachel Laudan
  2. Every graph in the AER since 1911 and paper. I’ve also found that economists heavy use of linecharts is quite atypical. Outside academia, other formats such as barcharts are much more common.
  3. I was very impressed by this podcast episode of 80,000 hours interviewing Eva Vivalt. I didn’t expect to find meta-analyses interesting, but they are! One of my favorite bits of insight: If we asked researchers about their priors about what the results of a research project will be (DellaVigna-Pope style), then null results suddenly become interesting and worthy to publish. I also liked this blog post by Ricardo Dahis discussing the framework for meta analyses.
  4. Simon Kucinskas: “How to Measure Biased Reaction to News
  5. A short radio review (in German) of Konrad Biermann’s Konrad Adeanuer biography. I liked the book, but found it speculative and thin on facts.
  6. 1979
  7. Movie reviews
  8. Loriot (in German)