Prepping for Placements in 2020

The million dollar question. This comes from a student at one of the colleges I taught at recently.

“How do you think the placement season will be? How can we train ourselves during this time to have better chances?”

  1. Things are going to be really bad this year. There’s no point in beating around the bush. You’ll be lucky to get placed, and even luckier to get a really good, high paying job. I don’t mean to be discouraging, but it is best we go into this season eyes wide open.
  2. That being said, let’s think about the second question more carefully, because there are things that I would recommend:
    1. I say this every placement season, but it is even more important this year. First, beware the streetlight effect. Second, never play cricket with Sachin Tendulkar.
    2. What is the streetlight effect? Here’s Wikipedia on the subject, but the gist is people search in the easiest, most obvious place. In an interview, that place is your CV.
    3. When, in an interview, you hand the interviewer your CV, it is literally the most perfect example there can be of the streetlight effect. The interviewer doesn’t know you from Adam (or Eve), and will therefore begin to ask you questions basis stuff you’ve written in your CV.
    4. Therefore, possess the ability to speak – thoroughly, meaningfully and concisely – about every single word on your CV.
      1. Thoroughly means you should know, and I mean really know, every single project, subject and achievement you have listed. No faffing!
      2. Meaningfully means you should be able to answer how a business might benefit because of the work you have done, or the topic you have learnt, or the internship you did. “I did XYZ in my internship” is a bad answer. “The business was able to achieve ABC, because of  I doing XYZ” is a good answer.
      3. Concisely means you should leave the interviewer with the feeling that you know what you’re talking about, but you shouldn’t overburden the listener with an endless stream of sentences. Practice by saying what you want to about a project, and then try to repeat the explanation in literally half the time. Keep at it!
    5. So if you are going to be subjected to the streetlight effect during an interview, be prepared for it. Or, and even better, give the interviewer something other than your CV to talk about.
    6. Which brings me to the most important thing you can do right now, and it not training yourself. It is training others.
      1. You’re not going to be the only one saying I did a course on Coursera | I learnt MS Excel | I did project XYZ with firm PQR
      2. But you could end up being the only one saying I taught kids in my society math using videos from 3Blue1Brown | I taught my batchmates Excel, and so learnt it better | I mentored a bunch of people online on <insert subject of your choice here>
      3. I’ll say this more concisely. Try and utilize this time to do, not learn. It is vastly underrated.
    7. Which brings me to my final point, about never playing cricket with Sachin Tendulkar. Here’s what I mean: if I ever met Sachin, and got to be in a contest with him, I would have to be stark raving mad to choose a contest involving bat and ball. But I’m fairly confident I can beat him in, say, a quiz on economics.
      What is the point? The point is that in an interview about analytics, for example, good luck trying to be better than the interviewer on machine learning algorithms. That is literally that person’s job! With your experience, learning and age, it’s like playing cricket with Sachin. But what if the interviewer learns that you put up videos on, say, photography, and you have been doing so for three months? And that of the 100 videos you shared, 50 got more than 10,000 views? And that you got to interact with people the world over as a consequence? You didn’t learn a course. Instead, you put yourself out there, you shipped a product that other people could benefit from, you were a mentor to other people. Now Sachin is playing your sport – and your chances just got a whole lot better. (Example: my CV is this blog, not a piece of paper)
  3. So my advice would be to identify a skill at which you are genuinely good, and to teach other people that skill, in public. Blogs, YouTube videos, Zoom sessions, whatever. Make that your CV, and crack that interview.

I hope this helps! Thank you to said student for asking the question, and if anybody has any follow-up queries, don’t hesitate to write in.

Cheers!

 

DNA, RNA, RT-PCR, Testing Methods, Supply Chains… and Politics

What is Reverse Transcription Polymerase Chain Reaction?

Reverse transcription polymerase chain reaction (RT-PCR) is a laboratory technique combining reverse transcription of RNA into DNA (in this context called complementary DNA or cDNA) and amplification of specific DNA targets using polymerase chain reaction (PCR). It is primarily used to measure the amount of a specific RNA. This is achieved by monitoring the amplification reaction using fluorescence, a technique called real-time PCR or quantitative PCR (qPCR). Combined RT-PCR and qPCR are routinely used for analysis of gene expression and quantification of viral RNA in research and clinical settings.

Blah Blooh Bleeh Blah. Right?

Well, this is the test that will tell us if a person has got the corona virus or not. So listen up!

The corona virus is in the form of RNA:

Coronaviruses, so named because they look like halos (known as coronas) when viewed under the electron microscope, are a large family of RNA viruses. The typical generic coronavirus genome is a single strand of RNA, 32 kilobases long, and is the largest known RNA virus genome. Coronaviruses have the highest known frequency of recombination of any positive-strand RNA virus, promiscuously combining genetic information from different sources when a host is infected with multiple coronaviruses. In other words, these viruses mutate and change at a high rate, which can create havoc for both diagnostic detection as well as therapy (and vaccine) regimens.

But as best as I can tell, detecting the corona virus becomes pretty difficult unless it turns into DNA, which can be done by a process called Reverse Transcription.

With the newly formed DNA, replicate it – have it reproduce a lot, basically. That’s where PCR comes in. And with that (and a fluroscent dye that is added to make detection easier) you have a sample that you can check for the presence of the corona virus.

The first, PCR, or polymerase chain reaction, is a DNA amplification technique that is routinely used in the lab to turn tiny amounts of DNA into large enough quantities that they can be analyzed. Invented in the 1980s by Kary Mullis, the Nobel Prize-winning technique uses cycles of heating and cooling to make millions of copies of a very small amount of DNA. When combined with a fluorescent dye that glows in the presence of DNA, PCR can actually tell scientists how much DNA there is. That’s useful for detecting when a pathogen is present, either circulating in a host’s body or left behind on surfaces.

But if scientists want to detect a virus like SARS-CoV-2, they first have to turn its genome, which is made of single-stranded RNA, into DNA. They do that with a handy enzyme called reverse-transcriptase. Combine the two techniques and you’ve got RT-PCR.

So, here’s how it works, best as I can tell:

Coronavirus Detection Steps

 

That article I linked to from Wired has a more detailed explanation, including more detailed answers about the “how”, if you are interested. Please do read it fully!

Now, which kit to use to extract RNA from a snot sample, which dye to use, which PCR machine to use – all of these and more are variables. Think of it like a recipe – different steps, different ingredients, different cooking methods. Except, because this is so much more important than a recipe, the FDA wags a finger and establishes protocol.

That protocol doesn’t just tell you the steps, but it also tells you whether you are authorized to run the test at all or not. And that was, uh, problematic.

For consistency’s sake, the FDA opted to limit its initial emergency approval to just the CDC test, to ensure accurate surveillance across state, county, and city health departments. “The testing strategy the government picked was very limited. Even if the tests had worked, they wouldn’t have had that much capacity for a while,” says Joshua Sharfstein, a health policy researcher at Johns Hopkins School of Public Health and the coauthor of a recent journal article on how this testing system has gone awry. “They basically were saying, we’re going to use a test not only developed by CDC, but CDC has to wrap it up and send it to the lab, and it’s just going to be state labs doing it.”

The effect was that the nation’s labs could only run tests using the CDC’s kits. They couldn’t order their own primers and probes, even if they were identical to the ones inside the CDC kits. And when the CDC’s kits turned out to be flawed, there was no plan B.

By the way, if you want a full list of the various protocols that are listed by the WHO, they can be found here.

Back to the Wired article:

Another in-demand approach would look for antibodies to the virus in the blood of patients, a so-called serological test. That’d be useful, because in addition to identifying people with Covid-19, it could tell you if someone was once infected but then recovered. “The better your surveillance, the more cases you’re going to catch, but even with perfect surveillance you won’t catch everything,” says Martin Hibberd, an infectious disease researcher at the London School of Hygiene and Tropical Medicine who helped develop one of the first tests for the coronavirus SARS in the early 2000s. “Until we’ve got a full test of this type of assay, we don’t know how many cases we’ve missed.”

A serological test would also probably be cheaper than a PCR-based one, and more suited to automation and high-throughput testing. A researcher in Singapore is testing one now.

Here’s an early paper on the topic, if you are interested.

Serological assays are of critical importance to determine seroprevalence in a given
population, define previous exposure and identify highly reactive human donors for the generation of convalescent serum as therapeutic. Sensitive and specific identification of Coronavirus SARS-Cov-2 antibody titers will also support screening of health care workers to identify those who are already immune and can be deployed to care for infected patients minimizing the risk of viral spread to colleagues and other patients.

As far as I can tell, this method has not been deployed at all thus far, and that applies to India as well. Here’s a Wikipedia article about the different methods of detecting Covid-19 – it’s about more than that, the first section applies here. Here’s an article from Science about a potential breakthrough.

But whether you use any variant of the RT-PCR or the serological test, given the sheer number of kits required, there is going to be crazy high demandand a massive supply chain problem.

Along with, what else, politics, and bureaucracy:

 


The Wired article is based on reporting in the US, obviously, but there are important lessons to be learned here for all countries, including India.

Here are some links about where India stands in this regard:

 

I’ll be updating the blog at a higher frequency for the time being – certainly more than once a day. Also (duh) all posts will be about the coronavirus for the foreseeable future.

If you are receiving these posts by email, and would rather not, please do unsubscribe.

Thanks for reading!

 

Econ101: Policy Responses to a Pandemic

If you haven’t played it already, go ahead and give this game a try: The Fed Chairman Game. I have a lot of fun playing this game in class, especially with students who have been taught monetary policy. It usually turns out to be the case that they haven’t understood it quite as well as they think the have! (To be clear, that’s the fault of our educational system, not the students.)

But the reason I started with that is because the game always throws up a scenario that mimics a crisis, and asks you what you would do if you were the Chair of the Fed.

In this case, policymakers the world over are now staring at a very real crisis, and they need to be asking themselves: what should we do?


 

There are two broad answers, of course: monetary policy, and fiscal policy.

The Federal Reserve has cut interest rates to zero, and while it has other tools to stimulate the economy, a crisis like this requires fiscal as well as monetary responses. The legislation passed thus far has been important, but another round of fiscal policy will be required immediately to fully address this crisis.

A robust fiscal response can provide income support to households, ensure broad and continuous access to safety net programs, provide incentives for employers to avoid layoffs, provide loans to small businesses, give liquidity cushions to households and firms, and otherwise stimulate the economy.

That’s a write-up from Brookings. The specifics follow in that article, but the article makes the point that more of the lifting will need to  be done by fiscal, rather than monetary policy. And that is true for a variety of reasons,  which the article does not get into, but long story short – fiscal, more than monetary.

But, ok, fiscal policy of what kind? Should we give money to firms or to workers? Here’s Paul Krugman with his take…

And here’s Alex Tabarrok with his response:

So what’s the correct answer? Well, as we’ve learnt before, and will learn again, macro is hard! In an ideal world, all of the above, but as is manifestly clear, we are not in an ideal world. If we must choose between giving money to firms or to people, to whom should we give it? My opinion? People first, businesses second. This is, of course, a US centric discussion, what’s up with India?


 

Here’s, to begin with, a round-up from around the world – you can search within it for India’s response thus far.

Calls are getting louder for governments to support people and businesses until the new coronavirus is contained. The only questions are how much money to shovel into the economy, how to go about doing it, and whether it will be enough.

Already, officials from Paris to Washington DC are pulling out the playbook used in Asia for slowing the spread of Covid-19: they’re restricting travel and cracking down on public gatherings. While those measures have the potential to reduce deaths and infections, they will also damage business prospects for many companies and cause a synchronized worldwide disruption.

Here’s the FT from two weeks ago about the impending slow down:

Venu Srinivasan, whose company TVS is one of India’s largest makers of motorcycles and scooters, said the business had lost about 10 per cent of production in February owing to a lack of Chinese-made parts for the vehicles’ fuel injection system. He added that TVS has now managed to find a new supplier.

But Mr Srinivasan said he was bracing for India’s recovery to take longer than anticipated. “One would have expected a V-shaped recovery, but instead you have an L shaped recovery,” he said. “It’s been the long haul.”

R Jagannathan in the LiveMint suggests this:

This is how it could be designed. Any unemployed urban youth in the 20-30 age group could be promised 100 days of employment and/or skilling options paid for by the government at a fixed daily rate of ₹300 (or thereabouts, depending on city). At an outlay of ₹30,000 per person annually, the unemployed can be put to work in municipal conservancy services, healthcare support, traffic management, and other duties, with the money also being made available for any skill-acquiring activity chosen by the beneficiary (driver training for Ola-Uber, logistics operations, etc). All companies could be given an opportunity to use the provisions of the Apprentices Act to take on more trainees, with the apprenticeship period subsidized to the limit of ₹30,000 per person in 2020-21. If the pilot works, it could be rolled out as a regular annual scheme for jobs and skills. Skilling works best in an actual jobs environment.

 

He also mentions making the GST simpler, which the Business Standard agrees with:

Certainly, the rationalisation of GST will also affect government revenues. However, a simpler and more transparent system would allow greater collection and reduce evasion. The government will receive a windfall this year from lower crude oil prices. The moment to move on the structural reform agenda is now. The GST Council has done well to address the inverted duty structure in mobile phones. Further rationalisation will give confidence to the market that the government is serious about reforms. It was promised that GST would remain a work in progress, and that the GST Council would act often to improve it. So far, however, the changes have been marginal and haphazard. A more structured and rational approach, which outlines a quick path to a single rate, would pay dividends for the economy in the longer run. It would also be an effective way to manage the immediate effects of a supply shock such as is being caused by the pandemic.

Also from the Business Standard, a report on the government now considering (not happened yet) relaxing bad loan classification rules for sectors hit by the corona virus. That’s pretty soon going to be every sector!


 

Assorted Links about the topic – there’s more to read than usual, please note.

Here is Tyler Cowen on mitigating the economic impacts from the coronavirus crisis.

Here’s Bill Dupor, via MR, about the topic:

First, incentivize behavior to align with recognized public health objectives during the outbreak.

Second, avoid concentrating the individual financial burden of the outbreak or the policy response to the outbreak.

Third, implement these fiscal policies as quickly as possible, subject to some efficiency considerations.

Again, via MR, New Zealand’s macro response.

Arnold Kling is running a series on the macro response to the crisis.

Claudia Sahm proposes direct payment to individuals:

This chapter proposes a direct payment to individuals that would
automatically be paid out early in a recession and then continue annually
when the recession is severe. Research shows that stimulus payments that
were broadly disbursed on an ad hoc (or discretionary) basis in the 2001 and
2008–9 recessions raised consumer spending and helped counteract weak
demand. Making the payments automatic by tying their disbursement to
recent changes in the unemployment rate would ensure that the stimulus
reaches the economy as quickly as possible. A rapid, vigorous response to
the next recession in the form of direct payments to individuals would help
limit employment losses and the economic damage from the recession.

Here are the concrete proposals, the entire paper is worth a read:

Automatic lump-sum stimulus payments would be made to individuals
when the three-month average national unemployment rate rises by
at least 0.50 percentage points relative to its low in the previous 12
months.
• The total amount of stimulus payments in the first year is set to
0.7 percent of GDP.
• After the first year, any second (or subsequent) year payments would
depend on the path of the unemployment rate.

 

Macroeconomics IS HARD!

Economics in the times of COVID-19, there is already a book. I learnt about it from Tim Taylor’s blogpost. I have not read the book, but will soon.

The NYT, two weeks ago, on the scale of the problem facing policymakers.

 

Technology and the Coronvirus

How can one use technology to learn more about the coronavirus and its effect on society?

This first example is based on a blogpost on Marginal Revolution:

There’s – let’s assume – 3 million people in Pune city. Let’s assume that you work in a firm that has a 1000 employees. Let’s assume that 500 people in your city have the corona virus.

Then the math says that all other things equal, there is a 15% chance that one of those 1000 employees is infected with the corona virus. Click on the link and play with the numbers to generate different scenarios.

Ah, you say, but Pune has only 16 cases so far, so there’s no need to worry.

And I agree with you, so long as we get the social distancing right. The more the virus spreads, the likelier it is that it will spread. And the more people will get it, the more people will transmit it – remember, R0 is at least 2.

Put another way, social distancing is to ensure that we don’t reach 500, let alone a higher number.

By the way, if you change 500 to 5000, the probability that a person in a 1000 employee firm has the virus shoots to 81%.

Social distancing matters.

 


 

Kartik Shashidhar, a guy you should follow on Twitter, came up with this app yesterday:

 

For the scenario outlined above, it takes about 10 days for 100% of the population to get infected with the virus, assuming the average connections per individual are 20.

If the connections per individual are 10, you get about 8 more days. Play around with the numbers, again, to generate different scenarios.

Here’s the original Twitter thread:

You can read more about small world networks here.

 


 

This is a great chart, and it comes from a great website. Thank you to Anish Parulekar for sharing the link.

Click through to the site to take a look at a whole host of other neat visualizations – in particular, take a look at how the Case Fatality Rate (CFR) varies by country.

Now, this particular chart is interesting because it helps you understand how to think about COVID-19. The best of all worlds is the origin, of course, where the R0 and the CFR are both zero.

Chicken pox is irritating, because it spreads quickly. But the good news is it is essentially never going to kill anybody. Bird flu is dangerous, because it kills a lot of people. But the good news is that it doesn’t spread to a whole lot of people.

Diseases that kill a lot of people and spread quickly, those are really problematic. That’s why smallpox, polio and SARS are (were) so problematic. And the Spanish flu, that roamed the whole world over, was one such disease. I hope to post a book review this Thursday about it.

COVID-19 is worse than the Spanish flu.


 

I’d rather not excerpt from this superb visualization – and accompanying article – from the Washington Post. Also, WaPo, thank you for not putting this behind a paywall!

Side note: Harry Stevens does data visualizations better than most folks.


 

And finally, a shout out to MRU for all of these resources.


 

Tomorrow, I’ll try and share with you stories from Spain, Italy, France, Iran, South Korea and China about how they fought, and are fighting, with the virus. Any links you can send my way are most appreciated! My email address is ashish at econforeverybody dot com.


Thank you for reading: stay safe!

 

 

 

 

 

Information about the Coronavirus in India

What is the corona virus?

There isn’t one specific coronvirus.

Coronaviruses are a large family of viruses which may cause illness in animals or humans. In humans, several coronaviruses are known to cause respiratory infections ranging from the common cold to more severe diseases such as Middle East Respiratory Syndrome (MERS) and Severe Acute Respiratory Syndrome (SARS).

By the way, the’re called corona viruses because they look like coronas, or crowns.

Coronaviruses are named for their appearance: Under the microscope, the viruses look like they are covered with pointed structures that surround them like a corona, or crown.

The most recently discovered virus causes the disease COVID-19.

How does it spread?

People can catch COVID-19 from others who have the virus. The disease can spread from person to person through small droplets from the nose or mouth which are spread when a person with COVID-19 coughs or exhales. These droplets land on objects and surfaces around the person. Other people then catch COVID-19 by touching these objects or surfaces, then touching their eyes, nose or mouth. People can also catch COVID-19 if they breathe in droplets from a person with COVID-19 who coughs out or exhales droplets. This is why it is important to stay more than 1 meter (3 feet) away from a person who is sick.

WHO is assessing ongoing research on the ways COVID-19 is spread and will continue to share updated findings.

 


 

When and how did it reach India?

As best we can tell, 30th January, 2020. This chart below shows you the spread since then.

Data and visualization can be tricky, and later on in this post, keep an eye out for another visualization about the corona virus.

Where does one get official data from in India?

The Ministry of Health and Family Welfare (MoHFW) is the official source that you should begin with. This is their website, this is their Twitter ID. At the moment of writing, the website reports 110 confirmed cases in India.

How is testing being done in India?

On the MoHFW website, there is a link about when to get tested for the corona virus in India. Shown below is a screen grab of that link.

Two things stand out:

  1. The second bullet point uses the word “and“. Having the symptoms is not enough, you must necessarily have traveled to any of the countries listed above. Other than whatever has been said in the previous sentence, you qualify for testing if you are a contact of a laboratory confirmed positive case. Unfortunately, “contact” isn’t clearly defined, at least on this page.
  2. Testing will currently be done by government laboratories only.

Which immediately leads to the next question:

What is our capacity to test for the coronavirus?

India has activated 67 laboratories for conducting the first test, and 51 of those are equipped to conduct confirmatory tests, which is not even one lab per district. India has 732 districts.

At present, cases are being reported from 13 States and Union Territories. In a country with a population of 1.3 billion, till now, only 6,500 throat swab samples from 5,900 individuals have been sent to these labs; at least 107 have tested positive.

That is from the Hindu Business Line, in a report that came out yesterday.

This is the COVID-19 page on the ICMR website. These are the locations of the testing laboratories in India. These are the locations of the sample collection laboratories.

 


 

OK, but 110 cases, how bad can that be?

Exponentials are hard. Anybody who has taught math or statistics will tell you that. Look at the graph(s) below:

A golden rule that I always teach my students in statistics: first look at the axes! On the horizontal axes here, we have the lag in days behind Italy. But the vertical axis is the more important thing to look at, because it is not linear. We go from 1 to 10, from 10 to 100, and from a 100 to a 1000 (and so on). Each tick on the vertical axis is a 10x increase.

In English? Every country where the virus has spread has seen a 10x increase. If you ask a data scientist to take a look at these data points, and then ask the about the trajectory in India, there’s only one possible answer: we probably go from a 100 to a 1000, and from a 1000 to 10,000. I hope not, of course, and mitigation is possible – social distancing is key!

By the way, if you want to play around with the data, click here to go to the Github page.

OK, so the numbers will go up rapidly, maybe. But the fatality rates are low, right?

Two important things to note:

Two numbers that you need to keep in mind when you think about the corona virus. The R0 and the fatality rate. The R0 for the coronavirus seems to be about two, although of course that number can vary because of a lot of factors. But a baseline R0 of 2 seems to be a reasonable estimate.

In English? If you get it, you will on average spread it to two other people. That’s why the quarantine and the social distancing measures are so very important. It’s not just because you shouldn’t get it yourself – it is more because you shouldn’t be giving it to others.

Now, the answer to the question itself: are the fatality rates low?

The Case Fatality Rate (CFR) for COVID-19 is 3.48 percent.

But as an statistician, it is important to state that the correct answer is it depends!

Unfortunately, it is common to report the CFR as a single value. But the CFR is not a biological constant. The CFR is not a value which is tied to the given disease, but is instead reflective of the severity of the disease in a particular context, at a particular time, in a particular population.

The probability that someone dies from a disease is not only dependent on the disease itself, but also the social and individual response to it: the level and timing of treatment they receive, and the ability of the given individual to recover from it.

This means that the CFR can decrease or increase over time, and that it can vary by location and by the characteristics of the infected population (age, sex, pre-existing conditions).

The real problem is rapidly overwhelmed medical facilities

Read the entire thread, not just the tweet quoted above. The point of sharing that tweet is to help you realize that opportunity costs will come into play very, very quickly at medical centers in India. Whom do I treat – patients with the coronvirus or other patients? And soon enough, it’ll be whom do I treat, this coronavirus patient or that one?

Worst of all, there is no treatment per se, yet. There’s encouraging news on the front from all over the world, India included, but there’s time for a recognized cure to be acknowledged and made widely available. Best to proceed on the assumption that there won’t be one, and prepare accordingly. That’s just good strategy in times like these: budget for the worst case scenario.

OK, so what can we do?

Follow government instructions! We’re all in this together, so whatever your local/state/national government is telling you ought to be followed, no questions asked.

Social distancing is key, and that’s fancy English for avoiding going out. Stay at home as much as possible over the coming days, and cooperate with local authorities. Classes, colleges, schools, clubs, restaurants, malls, gyms – anything in the nature of a public gathering ought to be avoided as much as possible.

Panic is not going to solve anything, but precautions will go a long way towards helping.

Stay home, stay safe!

Where can I learn more?

Here’s a list of resources:

Myth-busters from the WHO.

WHO’s advice for the public.

The Situation Update Report from the WHO (I have posted the latest update at the time of writing, but keep checking for more up to date ones as the days go by)

The WHO dashboard.

Read the Wikipedia article on the Epidemic Diseases Act.

A useful article about the how to think about exponentials.

A request: please email me at ashish at econforeverybody dot com with any resources that you think may prove useful. I’ll do my best to share the more useful ones with everybody.


 

Coming up tomorrow: technology in the times of COVID-19.