2020 Semiannual Review

Fri Jul 31 2020

Table of Contents

Introduction and overview

NB: I started writing this post in mid-June before results were released, but wrote the bulk of the post-exam thoughts after results were released. So I was in the amusing position of having to "predict" my results in every subject despite already knowing my marks. I have tried to be honest, however, and wrote down my predictions as they were reported on the Google Spreadsheet.

As mentioned in my previous post, I'm now calling the reviews semiannual and annual reviews.

This semiannual review coincides with the end of my undergraduate degree, and marks another change of state. 2014 was the transition from high school student to NSF, 2017 the transition from NSF to undergraduate student, and 2020 the transition from undergraduate student to... graduate student? full-time employed? Who knows at this point. More on this later.

I'll talk about the eight core courses I took, how the exams went, and then finish with some final parting throughts on Oxford and my degree.

What I did this year

Overview

Since the last review post (2019 Annual Review), I have:

  • got accepted to ETH Zurich (Statistics), Harvard, and Oxford;
  • got rejected by ETH Zurich (Data Science), Stanford, and Carnegie Mellon;
  • submitted my Politics thesis;
  • submitted my BEE thesis;
  • baked bread for the first time;
  • weathered lockdown in Oxford;
  • finished Finals;
  • applied and accepted GSoC with MGGG.

Calendar

Month What I did
Jan 20 HT 2020: BEE, Thesis in Politics, Game Theory
Feb 20 HT 2020: BEE, Thesis in Politics, Game Theory
Mar 20 Finished HT 2020. Coronavirus hit. Full lockdown.
Apr 20 Revised for finals
May 20 Revised for finals
Jun 20 TT 2020. Finished finals (which marks---hopefully---the end of my degree).

Results of Master's applications

I was rejected from ETH Zurich's Data Science course, Stanford's Data Science course, and CMU's Computational Data Science course. I was also rejected from the ETH Zurich and Stanford Knight-Hennessy scholarships.

I was accepted to ETH Zurich's Statistics course, Oxford's Statistical Science course, and Harvard's Data Science course.

My top choice would have been Stanford's Data Science course, especially with the scholarship. But I am happy with all of them, and very pleased to have gotten a choice of three. I decided on Harvard in the end because I think the marginal gain to staying in Oxford for another year is quite low, despite the fact that I've enjoyed myself very much. ETH Zurich would also have been really nice, but I chose Harvard over ETH for three reasons: i) I've been told that the brand name of Harvard is stronger; ii) if IMDA is paying for my studies, then the low cost of ETH Zurich is not worth the extra year of bond; iii) studying in Harvard gives an opportunity for me to intern and work in the US. However, there were pros to ETH Zurich as well. I'd be eligible for Swiss employment having studied in a Swiss university (Swiss labour laws mandate that employers scour the entirety of Switzerland and Europe and make sure that no non-foreign worker can fill the role before allowing employers to hire a foreign employee), and I would get to learn German with less friction. It was a difficult decision and I would have been happy with any of these acceptances, so I consider myself lucky. I am very grateful to Bassel, Sergi, Nick/Jonathan, and Andy, who wrote me strong letters of recommendation that certainly helped my admissions.

Finals 2020

Eight papers: three core Econs, two Econs options (BEE/GT), three politics papers (Thesis, Theory, PolSoc). I was allowed to take one less Politics paper and do a thesis, which was very nice indeed---thank you to Sergi and Tia for putting in a good word for me to the Senior Tutor, the Academic Registrar, and the Politics department.

The papers I took were as follows:

  1. Macroeconomics
  2. Microeconomics
  3. Quantitative Economics
  4. Game Theory
  5. Theory of Politics
  6. Political Sociology
  7. Thesis in Politics
  8. Behavioural and Experimental Economics

I'll first write about the new examination arrangements due to the coronavirus before talking about the individual papers.

Coronavirus uncertainty, safety net, and new examination arrangements

In March 2020 the coronavirus pandemic swept across Europe and many countries started the lockdown. All of the undergraduates were unceremoniously kicked out by Merton, given only a weekend's notice. Many of them left their things in their rooms and left. I was allowed to stay because I was from (what was at the time) a "Cat 1 country", ergo a country that was "heavily affected by the pandemic". I think in those days the countries/areas were China, Hong Kong, Taiwan, South Korea, and Italy.

Until now I really cannot understand why Merton reacted in this way. I have heard that Merton was one of only two colleges (the other being Magdalen) to send all its international undergraduates home, and they were very harsh about it. Other colleges like Corpus and even the official university advice said that international students were welcome to stay.

In the midst of the lockdown and all the restrictions Singapore issued an advisory encouraging students abroad to return home. The anxiety of coronavirus swept over everyone, and I was inundated with many calls and emails and WhatsApp messages from IMDA, friends, and family asking me to go home. But I really did not want to, for three reasons:

  1. There were only ten weeks left to finals, and if I were to go back home, I would lose precious time for revision: two weeks in self-quarantine (while being very jet-lagged), then having no conducive space to study (libraries, cafes, etc. were all closed);
  2. Heathrow was packed to the brim, while Oxford was very empty. I was unlikely to get the coronavirus if I stayed put, but the chance of catching it would go up a lot if I started traveling on the crowded trains, going to the crowded airport terminal, sitting in the crowded airplane... and so on.
  3. Celine felt very anxious and unmoored due to the coronavirus situation (the sudden departure of all our friends, the uncertainty of our examination arrangements, having Trinity term snatched away from us, not knowing whether I'd be called to go home), and I wanted to stay with her and keep her company.

At that moment we didn't know if Trinity Term would be conducted remotely, and didn't know what form the examinations would take. There was significant discussion on the PPE cohort group chat. Some universities/courses cancelled finals entirely and graded purely on the basis of tutorial work/coursework. Other universities like UCL gave a 24-hour take-home exam. I helped another PPEist draft an open letter to the examiners appealing for the University to allow for student choice in assessments: that is, to accept either open-book exams under conventional time limits, a submission of coursework with more relaxed time limits, postponing the exams for the start of MT2021 or TT 2021, or even to take a "declared to have deserved honours" with an unclassified degree. The reason for doing so is because no one-size-fits-all solution exists. Some students may not have a quiet environment in their homes to take a timed online exam, and may prefer a take-home exam; others may need to take care of family members and want to defer their exams completely; yet others may simply prefer to take the exams in the format they've prepared for.

There was also a Student Union survey asking for our most preferred choice. In the end, I don't think anything much transpired out of it. The vast majority of students preferred being assessed by take-home exams and/or coursework. Timed 4h open-book exams were actually the least popular option. Nonetheless, that was the option that was implemented---presumably because it required the least amount of extra work on the part of the examiners (a 24h time limit with no other changes would have trivialised many of the papers). The University did indeed implement a "deemed to have deserved honours" (DDH) option, but nobody I know took it up just because it was a bit dodgy. The PPE department also relaxed the requirement for a First (only a 68.5 average needed with no requirement of three papers > 70), and introduced a safety net, dropping the two lowest-scoring papers in the eventual classification calculation.

Revising for Finals

Celine and I studied mainly in 59 until we got rumbled by the porters, after which we decided to move to Liddell (where the porters are much more accommodating).

I only started revision proper around the start of HT Vac. This was the main schedule:

  • HTW7: (Mar 1--7): Started properly writing the Politics thesis (~0 words on 1st March, 7000 words on 7th March)
  • HTW8: (Mar 8--14): Kept plugging on with the thesis, but didn't make much progress
  • HT Vac W1 (Mar 15--21): Kept plugging on the Politics thesis, started working on BEE
  • HT Vac W2: Finished first draft of Politics thesis.
  • HT Vac W3: Worked on both theses
  • HT Vac W4: Started revising Macro and compiling Macro notes
  • HT Vac W5 (Apr 12--18): Continued compiling Macro notes
  • TTW0: Submitted theses, Macro PYPs
  • TTW1: Micro
  • TTW2: Micro, bit of QE
  • TTW3: QE
  • TTW4: QE, bit of Macro
  • TTW5: Start of Finals (Three core papers)
  • TTW6: Revise for Theopol, Game Theory
  • TTW7: GT, revised and compiled notes for PolSoc

My time was mainly focused on my theses and Macro until Thursday TTW0, the 23rd of April (the submission deadline). After that, there was exactly a month from the start of Trinity term to the start of my exams (TTW5 Monday). This meant that I basically had one month (and a bit) to revise six papers.

TTW1 and TTW2 was mainly devoted to Micro, with a bit of QE on the end, and TTW3 and W4 QE, with a bit of Macro re-revision on the end.

I basically did no revision for Theopol, Game Theory and PolSoc whatsoever: I started revising Theopol on the 1st of June (a Monday) for the first time since Hilary last year, and then took the exam on the Saturday, June 6th. Then it was three days of concentrated revision (~8h) for Game Theory on the Tuesday, June 9th. Then I really should have started on PolSoc, but I couldn't muster up any energy until Wednesday, so I did one and a half days of PolSoc revision before the final exams.

Some general thoughts before I delve into the specifics:

Theses are a bad deal overall (worst trade deal in the history of trade deals) and take up so much of my time. To be clear, the act of working on a thesis is well worth it and there were a lot of side-rewards (getting to know Nick and Daryl and MGGG and so on), but within the system of the Oxford examinations they are worth much less. You do much, much more work on a thesis compared to a regular paper; I would estimate (at least) three times the amount of work for my politics thesis compared to a regular politics module, and between two to four times the time for BEE compared to a regular economics module. While I don't regret doing either thesis ex ante, I regret taking BEE ex post. (The thesis in politics was, all things considered---an excellent decision with good outcomes). More on this later. Ironically, I thought that PolSoc and Game Theory actually went the best despite the fact that I spent so little time on either of them.

I also felt like I didn't actually revise that hard: I'm not great at putting in the consistent and boring work daily. I work best in short and intense bursts, especially when the content is interesting, and having to plug along re-learning stuff I already know for an exam many weeks later really saps my motivation. I think I only studied around 4 or 5 hours properly a day on average, if that---and often got sidetracked with YouTube videos, computer games, and side projects.

I repurposed my A-level timetable spreadsheet for my Finals revision.

This is the original A-levels timetable:

The original A levels timetable

I drew up a macro-level plan for Finals where I decided how to spend my time, and to check off all the past-year papers I completed. This plan was more aspirational than anything else---I constantly fell behind.

Oxford revision planning

I also used the spreadsheet to record what I did every day in the weeks leading up to Finals:

Open-book exams changed the way that I (and everyone else) revised. I no longer had to do any memorisation for the technical papers, which I was very happy with. I also spent loads more time creating comprehensive notes which I could paste almost verbatim into the exams--- this proved especially helpful for Core Micro/Macro and PolSoc. (More on this later).

Taking finals

Celine and I were incredibly nervous for the first week of Finals (TTW5) for three reasons. Firstly, it was the start of the exams. Secondly, the first week had the most number of exams (my subsequent three papers were spread over a period of two weeks). And last, nobody really knew how online exams would be like and whether or not the papers had been made more difficult because of the open-book format plus the increased time allotted.

We took our exams in the common room of the Liddell Building, a room with a lovely view of the gardens outside. Celine took the nice table and I made do with the broken ping pong table.

Taking online finals was really quite nice. You don't have to dress up in sub fusc and faff around with different-coloured carnations; many times I just sauntered out of bed, made myself a sandwich and did the paper. (Nicole made sure to dress up to give a sense of gravitas to the whole proceedings, but that was too much effort for me).

After the first week, everything was relatively smooth-sailing: I had put a lot of effort into my Core Econ papers and much less for my Game Theory and Politics papers, and hence my stress levels were much lower. Below I give some thoughts on each of the papers:

Thesis in Politics

The thesis has really come a long way from its beginnings in 2019. I wrote a lot about the initial beginnings of my thesis in the 2019 in review post, which was actually the semiannual review conducted at the end of the second academic year. Looking back, I started seriously thinking about my thesis in April 2019, and I submitted my thesis on the 23rd of April, 2020. That's one entire year I've been working on my thesis. And I haven't even taken breaks, because I was thinking hard about my thesis even during the summer! This is one year of effort for one paper.

This is insane. I can't stress how raw of a deal this is wrt the Oxford examination systems. Let's compare it to a difficult but fun course, Game Theory. I spent around 10 hours a week on Game Theory (~2h lectures a week + 2h tute + 5-6 hours writing the tute worksheet), so that's 80 hours during term time, plus another 20 hours for revision (5 PYPs ~4h each), so that's 100 hours in total. With that amount of effort it's "realistic" for me to expect a 78 (ex ante, not ex post). And this was all fun and easy and quite brainless: attend the lectures, do a nice and interesting problem sheet, and when it came to the exams just do all the past year papers, which was also fun.

In contrast, I have spent more than 300 hours working on my thesis. The following estimates are not that reliable but I've tried to be as accurate as possible.

The literature review and idea exploration probably took me around 50 hours (initial idea generation: ~5 hours, discussion with Jarel: 5 hours, discussions with Tak Huen and reading Political Analysis: 10 hours, discussion with Bassel and writing the code for that exploration: 15 hours, discussions with Filip and trying to salvage the eventually unworkable idea: another 10 hours). None of that made it in the final thesis, but it did get me to Eubank and Rodden.

Then I worked with Eubank and Rodden. That was easily another 50 hours. Discussion with Andy + more idea generation: 10 hours. Emails back and forth + hashing out the initial criticisms: 5 hours. Researching a way to generate travel durations: 5 hours. Writing the code to generate travel durations: must have been at least 25 hours.

Still not actually found the thrust of my actual thesis! Had to come up with the human compactness metric: at least 10 hours of thinking. Then writing the code to calculate human compactness, different compactness measures, optimising... : at least 30 hours. Then visualisation, plotting, writing and rewriting first drafts, calculating the metrics, thinking about how to position the paper, and so on.. took at least 30 hours. And finally writing the final version (and editing) the actual paper took almost two weeks of concentrated writing, which was another 40 hours.

And maybe (at the highest) I should hope for a 78. I predict either 73, 75 or 78.

Behavioural and Experimental Economics

I didn't like BEE that much because of the way the course was taught.

I like the idea of BEE and I think Johannes is doing something very laudable to teach this very different course within the conservative Oxford system. The course is very new: I think this is the second or third year it's being run? But as a consequence he's quite conservative with his marking and the spread of marks is low, and the mean is exactly 65.

I didn't learn much from the lectures. Also, it's quite difficult for him to give so many tutes to so many groups. I think there were eight groups and he gave three tutes to each of the groups --> twenty four tutes in total. I also felt that the tutorials weren't that helpful because i) one hour is not enough time to really dig into ii) three tutes are not enough to get proper feedback about our experiment design/data analysis. But again, cannot really blame Johannes because 8x3 is already quite a large teaching load on top of giving essentially all of the lectures.

I spent too much time building the very complicated experiment on NodeGame, and running in on Amazon Mechanical Turk. Most other groups did simple survey experiments using Qualtrics, but I went all out trying to build something real-time and what not. I guess I was trying to be ambitious and do an experiment that was very nontrivial and very impressive. But in the end I don't think the very technically impressive experiment was that impressive at all. It was certainly difficult and I learned quite a bit, but it really wasn't worth the additional trouble.

My behavioural economics experiment tried to investigate how communication fosters cooperation in the context of the Stag Hunt game. We know that the Stag Hunt game has two pure equilibria, one where both players hunt Hare and one where both players coordinate on Stag. The {Hare, Hare} equilibrium is risk dominant (it gives the higher expected payoffs if you assume the player plays 50-50), while the {Stag, Stag} equilibrium is Pareto-dominant (it's a Pareto-improvement over the other equilibrium). Which equilibrium will people play? Opinions differ. But what is clear, and has been proven in experiments, is that

  1. communication helps coordination on the Pareto-efficient payoff;
  2. free-form communication (as opposed to "restricted" communication where participants can only say "I intend to play Stag" or "I intend to play Hare") helps even more.

When participants were given the opportunity to send an enhanced message like "Let's both play Stag because it gives us both the most money which is $10", this message performed much better than giving participants the option to send the message "I intend to play Stag".[1]

But why might this be the case? You might think that the enhanced message helps because it displays an understanding of the game (i.e I understand that if both players play Stag, both players get more money). Alternatively, you might think that the message helps because it displays some sort of altruism (it gives us both the most money --> I care about you).

And so I investigate how the enhanced message improves upon the shorter message. I developed a theory that could separate these different mechanisms of action, then built a real-time game with a library called NodeGame [2] with the help of the library's creator, Stefano Baletti. Unfortunately, I got a null result: even in the control condition (when participants were not allowed to send any message), coordination rates were >90%. This meant that I couldn't detect any result in any of my treatments.

Overall, I do not recommend this module. I predict a mark of 71.

Macroeconomics

I spent the most time revising for this, which in hindsight is probably not that smart because it tends to be a low-variance subject. I made rather comprehensive notes which I plan to clean up and upload onto this website soon. Even then, I actually studied no debt and fiscal policy whatsoever, instead focusing on the first seven weeks. The problem with Macro (and Micro to a smaller extent) is that it covers way too much in too little time. I spoke to Bassel and he concurs 100%. For instance, the IS-PC-MR model plus the NKPC extension alone can take up eight weeks, but we were taught it all in three weeks. Ditto for open economy: nobody understands these topics because it's all rushed through in a week. Similarly for Micro, general equilibrium alone (consumer optimisation + trade theory + welfare theorem results) could have filled an eight-week course. As it stands Alex (not his fault) spends a slide or two about the FWT and SWT and the Liberal Paradox and that---in Bassel's words---is "fucking bullshit".

I thought the paper went well, but not spectacularly so. Section A was hard (answer all three questions). The first question was a relatively standard IS-PC-MR question, and I answered that easily. The second question was a tricky one about the Solow model and how the optimal steady-state ratio of consumption depends on technology, savings rate and the rate of capital depreciation. The third question was really hard and quite mathematically rigorous. I doubt they'd give such an question in a regular closed-book exam. The first part asked us to derive the Hall random walk result and it just got harder from there. The second part asked us to derive a nontrivial mathematical expression for permanent income. I still don't quite know what the answer is for the third part. I believe the questions in section A were made progressively harder to distinguish between candidates.

Section B was quite straightforward (choose two essay questions from six). Most questions were textbook, and my prepared notes were very helpful. I answered a question on Acemoglu's directed technological change and a question on RBC. I chose these two questions because a lot of the marks/analysis can be gotten just from setting up the model correctly, and I had already prepared good explanations of the models.

I predict a mark between 68 to 73.

Microeconomics

I didn't really learn Micro properly in MT2018 as I was focused on PolSoc. This meant that I had to basically re-learn the whole General Equilibrium section all over again. I also skipped most of Willemens' slides about game theory.

This paper was very easy, and I felt really bad because I (thought) I fucked this paper. The first question was a GE question with a twist (exchange economy + social planner) similar to a past-year question. The second question was a GT question, which I misread and fucked up. I probably got like 20% on that question, which I wanted very much to impress the examiners with.

First mistake (stupid error): forgot to export my latest .xopp (Xournal++ file) into PDF, which meant I didn't submit the bit where they ask me to find the NE of the game. I also didn't read the part where they asked me to calculate the expected payoffs in the equilibria.

Second mistake: in the game, an SPE can be sustained by the threat of reversion to the risk-dominant NE. But I was very stupid and drew the finite state machine to revert to the minmax payoff, which is not sustainable because both players want to deviate! Super careless.

The third question was a nice straightforward question about risk and expected utility. I solved it perfectly.

The fourth question was a signalling question, also pretty straightforward.

Part B was a gimme---the long question was a Cournot game that was almost exactly identical to the long question in the 2018 PYP. And so I solved it perfectly. It was pretty funny that they actually took down/blocked access to the PYP and answers for the duration of the exam. I had of course downloaded the PYP and the solutions long ago.

I chose to do a very textbook question on the principal agent problem:

Explain the term “agency cost” in the context of Principal-Agent problems. Why does it arise? There are a number of situations under which the agency cost is absent (or zero). What are these situations? For each one, describe the outcome and how the Principal achieves it.

I was able to write a very standard textbook answer, easily setting up the model with reference to my notes.

I'm very glad that this paper did not have any trade in it. This paper was so easy that I suspect it will be scaled down. I predict anywhere between 65 to 73, with mean 68.

Quantitative Economics

In contrast with Macro, which was reasonable, and Micro---which was too easy, this paper was incredibly hard. At least part of the reason is because James Duffy revamped the syllabus this year making it more mathematical.

For QE I basically didn't revise any time series. I only watched several lectures once by Sophocles. This turned out to be an excellent idea because I could have done the entire paper without time series knowledge.

The first question asked us to derive the mean and variance of the sample average of a Bernoulli variable, as well as how to standardise it. It wasn't easy but it wasn't too hard either.

I quite liked the second question: it was a clever question that looked like a time series question, but was in reality a question about the CEF. The question asked for the MSFE-minimising forecast of Yt+1Y_{t+1} using the information available only in period tt. The key insight is that the MSFE-minising forecast using YtY_t alone is exactly the CEF. And thus I was able to give a proof that the CEF is the best predictor of Yt+1Y_{t+1} using YtY_{t} alone with the help of the CEF proof in James's notes. They then asked under what conditions would this MSFE-minimising forecast be a constant plus y: this is basically asking when the CEF is linear.

The third question was unique and has never been asked before in a PYP. It was pretty hard as well. It was a question with codetermination of dummy variables, and proving that no matter which dummy variables you use the OLS regression gives comparable estimates (as would be expected).

For section B I chose question 4, which was a rather standard hypo testing question with heterogeneity. The interpretation part asked about selection bias and how the TOT can or cannot estimate the ACE. That part was rather straightforward, but there was an interesting additional bit where you had to formally show that the value of the t-statistic is greatest when the two groups in the trial are of equal size: one concludes that, for maximal power, both treatment and nontreatment groups should be the same size.

I spent quite a long time agonising over whether to attempt question 5 or question 7. Question 7 was a normal-ish regression interpretation question but I didn't seem to have anything to say. Question 5 was a maths-heavy question that involved a regression of YY on XX and X2X^2, and asked us first about the plausibility of the OR assumption in the "short" regression of YY on XX only. Part b) was really strange and I am still not sure how to solve it: it was about two estimates of the parameters of the long and the short regression, and how to account for the different parameter estimates. Part c) was a nonstandard but doable derivation if you used the OVB formula. Part d) was an interesting question that asked whether the linear model could be used as an approximation to the quadratic one, and whether a fifth-order polynomial could also work. Finally, part e) asked us to extend the proof of instrumental variables for the quadratic regression (show that β1,β2\beta_1, \beta_2 can be recovered from a population reg. of YY on ZZ and Z2Z^2). I would have solved it given more time, but this was the last question in this paper, so I sketched out a partial proof.

I thought I did relatively well on this paper given the extremely high difficulty of the paper, and I predict a grade of 68 with upper bound 73.

Game Theory

Apart from doing the tutorial questions, I must have spent twenty-five hours---at most---revising for this module. There wasn't really much to do. Originally I wanted to learn everything properly (getting the definitions absolutely bang-on clear and formal, and learning all the topics lectures including bargaining, ESS, and auctions), but realised it wasn't worth it in the end. And so I focused on my favourite/strongest topics: core game theory, repeated games, sequential/perfect Bayesian equilibrium, and so on. I really couldn't have expected it to go any better. I spent a couple of hours really grokking PBEs and how to solve for them. Professor Munoz-Garcia's notes on a systematic procedure for finding PBEs was incredibly useful here. I also spent about an hour or two revising the rest of the definitions such as rationalisability, mixed extensions of a NE, and so on, and then just started doing past year questions as far back as was feasible. (Interestingly enough, the game theory paper before 2013 actually involved writing essays---one question quoted a Platonic dialogue and asked candidates to comment on its application to reputation games).

I was pleasantly surprised to find that the ESS question was basically a complete copy of the 2017 past year paper's. However, I was sad that there was no 2x2 reputation game this year, and I believe that the core game theory questions were definitely made (much) more difficult to compensate for the open book exam. I think the repeated Bertrand game question was absolutely brutal, and I shouldn't have done it. (But then again none of the other questions looked good).

The repeated games question was definitely made more difficult by Peter Eso to compensate. I am almost certain he added a part (g) because all similar questions in previous years stopped at what was part (f) this year.

Nonetheless, I believe I did well, and I predict a 73 or above.

Theory of Politics

As mentioned, I basically did no revision for Theopol at all. I started revising Theopol on the 1st of June (a Monday) for the first time since Hilary last year, and then took the exam on the Saturday, June 6th. This was for three reasons: mainly because of not-great time management, but also because Theopol is quite a low-variance subject, and I wanted to focus on my core econs papers. I was prepared to drop it as a safety net subject, and because I only had about four days, I just revised four topics: libertarianism, egalitarianism, nationalism/cosmopolitanism, and rights/justice. I gave special attention to libertarianism because I like the theory a lot, which is not something I would recommend as good exam prep strategy (do as I say, not as I do)...

The exam absolutely blindsided me (although quite predictably). The questions for my topics were very weird, blending multiple topics together. For instance "Does taxation reduce citizens' freedom?" blends libertarianism with liberty, and "Are you obligated to any greater degree to the states of which you are a citizen than you are to a state in which you live?" blends political obligation with cosmopolitanism/nationalism. My questions were chosen for me, having revised basically nothing. I would have to do the libertarian + liberty question, the rights question, and the egalitarianism question.

The libertarianism + liberty question was very off-the-the cuff. I looked at the reading list for liberty, whereupon I quickly skimmed an essay on the reading list that looked relevant (it was Cohen's "Freedom and Money") and started writing an essay on it. Then I smoked a bit of what I knew from Prelims (positive+negative freedom, triadic conception) + some SEP notes and started writing the essay. I think this was my strongest essay nonetheless, (which speaks more to the quality of my other essays) because I think I made an original point. This essay would be a high 2:1 or low 1st.

"What is the value of rights that are not legally enforced"? I had absolutely NO idea, and basically just smoked this (it was a very, very short essay --- 800 words?) I first wrote about the benefit theory account: rights are valuable insofar as they benefit the people who possess those rights, and if they're not enforced, people don't benefit and thus they're not valuable. Then I sketched out very flimsily three ways in which rights can be valuable even if they are not legally enforced: firstly, rights can be valuable as a recognition of our respect for persons' autonomy. Secondly, rights can exist even if there are, or can be, no legal sanctions. Last, there are rights that seem to be valuable because they are not legally enforced---for instance, rights that arise from promise-keeping or reciprocity.

The final question was "Should egalitarians pay more attention to individuals’ suffering and disadvantage that are the effects of bad luck, or to suffering and disadvantage that are the effects of power?" I interpreted this question as trying to adjudicate between luck egalitarianism and relational egalitarianism. The key contention here of the relational egalitarian is that many types of suffering and disadvantage that would seem to be the effects of bad luck are in fact due to the oppressive social order (aka power); for instance, being born paraplegic or talentless is only a suffering in a society where the capitalist order means that one "needs" to be able-bodied and able-minded to make a living. I argued that point taken---a lot of suffering and disadvantage is mediated through our unequal society---but not everything is about power. People with a disability may suffer from physical pain independent of any sort of societal structures. And therefore it's inappropriate to focus on power alone. Then, because I thought I didn't have enough words (was at ~1100), I trotted out the list of standard criticisms I have against LE concluding that we should be sensitive to both types of suffering, not just power.

I'm not sure these criticisms were relevant to the question: I tried to justify them by saying "Finally, at least part of the reason relational egalitarians claim that we should focus more on power is that focusing on luck can be paternalistic and demeaning. I show that this isn't necessaily true by rebutting two of her criticisms". Hopefully this convinced the examiner--- it certainly wouldn't have convinced me.

The essay has a strong central thesis but meanders halfway because I ran out of things to write. In hindsight it would probably have been better to cut out the criticisms of LE, or contextualise them in a better way. I predict a grade of 62 or 65--- max 67 only if the examiners are generous.

I predict an overall grade of 63.

Political Sociology

After my Game Theory exam on Tuesday, I really should have started on PolSoc immediately. But I couldn't muster up any energy until Wednesday, so I did only one and a half days of PolSoc revision before the final exams.

However, I definitely put more work into PolSoc overall than Theopol. During MT2018 I put a lot of hours reading all the readings thoroughly and thinking very hard to come up with sophisticated methodological criticisms. I remember spending many hours in OWL drawing diagrams to fit everything together and find the big picture of all the twenty or so readings I read that week. I was diligently Anki-ing all the readings and I can still remember working through the enormous backlog of cards on the bus ride to Colby College during the Christmas vacation. I also spent some time during Hilary term basically re-reading the four topics for Karina's revision classes. I guess you could say the work I did for this module was front-loaded: there was very little PolSoc revision done during the Finals period.

I revised class, welfare, social attitudes, and turnout. This was admittedly not ideal, but I only had one and a half days of reading anyway. Luckily, the paper was very nice and straightforward, a nice contrast to the Theopol questions. And I was very lucky to have all four of my topics come up. I didn't do the question on Class (although it was quite easy) because I felt that wasn't my comparative advantage: let the British folks ace that question.

I wrote two strong essays and one "meh" one. IMO, the inequality one was very strong, the social attitudes one slightly worse, and the rational choice model one very meh.

The first question "Does inequality affect support for the welfare state" was very very similar to a PYP I had already done/my tutorial essay/my revision essay, so I spent my time copy-and-pasting paragraphs in and writing very clear structure + glue words. I'm quite proud of the "synthesis of models" I previously came up with in my revision essay, and it was able to be 派上用场 in this question. I think this was an excellent essay and predict a score of 73/75.

The second question "Are the forces that have driven increasing social liberalism in Western countries in decline?" should be answered in two parts: i) what are these forces, ii) are they in decline. As is my usual style I lay out a clear dichotomy of two possible forces: in this case it was Inglehart's formative affluence vs. elite cueing via education. Again, many of the criticisms had already been written, and so I used the time to think about how to structure the essay and craft very good intro + links. I think this is a great essay and predict a 73.

The meh question "Does the rational choice model of turnout help explain why turnout levels vary so much across countries?" wasn't bad per se. It had all the content and cited the right folks: it simply needed a bit more time (~10--15 minutes) and additional words to smooth over some of the awkward transitions from point to point. But due to time constraints I didn't get to do so. So this is the weakest one: very jumpy, doesn't quite flow and make a lot of sense. I would give this a 66/67.

Overall, this paper went well (ask Celine about this --- I was basking when she came in to check on me), and I predict a grade of 71.

Final thoughts about Finals

I really enjoyed the open book format for three reasons.

First, it reduced the memorisation burden, which is great for a lazy guy like me. Second, I could use my existing workflow for writing essays, which is to write in Markdown + run pandoc to typeset with LaTeX\LaTeX and export to PDF. This had the advantage of my essays looking very nice and neat, with especially beautiful equations in Micro and Macro.

Here's an example from Macro:

Macro essay screenshot

And for Theopol and PolSoc I was able to generate a nice table of contents page, which helped very much to structure my thinking, and signpost my arguments for the examiners:

PolSoc table of contents

Lastly, the open book format allowed me to prepare and make use of many of my strongest arguments and references. This is not just a memory issue: some of my methodological criticisms can be quite nuanced, and so I spent some time rewriting the arguments before the exam to make them very clear. This can take quite many words, and you wouldn't have the ability to write so many words in the exam itself.

The Anki flashcards which I spent so much time crafting and memorising turned out in the end to be a waste of time. Also, I realise that Anki flashcards are not a replacement for writing proper written notes: they really pale in comparison to a traditional document wrt knowing what information you have already and the ability to access said info. This was a mistake I made for PolSoc because I should have written proper written notes instead of just Anki-ing my work. However, writing notes structured in a more "Q&A" style is very useful, and I tried to do that with my Macro notes in particular.

I was a bit too relaxed studying for Finals, as Celine will attest (we were studying together for the vast majority of the time). I remember the same thing happening during my A-levels revision. I'm just not very good at making myself consistent with my revision. I wasted several days before Macro I playing Starsector from morning till night, which is honestly a pretty unconscionable move (fun game, though). I'm not sure how I should fix this. I tried lots of stuff: planning out the day on the whiteboard, written timetables, Google Calendar, Complice etc... but none of them could stop me from occasionally just binging YouTube. But I was very very lucky to have Celine to keep me on track.

Weathering the coronavirus

Celine and I weathered the coronavirus quite well. While it was sad that many of our friends left, we still had each other and the Liddell flatmates. The coronavirus didn't really affect us too much because we were basically spending our days in the house, with occasional trips to the grocery store or around the parks. It helped a lot for us to have a daily routine: waking up, studying, having lunch, having a nap, studying, cooking dinner, watching some videos, and then going off to bed. It was eerie but also quite nice that the streets of Oxford were so empty. That was really nice peace and quiet.

Thoughts about my degree

The three years in Oxford have made me grow very much as a person. I've done, and continue to do, a plethora of embarrassing, cringey, thirsty things. Sometimes recalling these things keep me up at night. It is a slow and painful process of learning not to say/do these things. But I am slowly learning, and I am grateful to all my friends for being patient with me.

I had a lot of firsts in Oxford:

  • first time visiting Europe,
  • first time matriculating,
  • first time living independently,
  • first time cooking dinners for friends,
  • first time with a woman,
  • first time taking Prelims and getting trashed,
  • first time cooking for a crowd,
  • first time joining an acapella group,
  • first time joining a gin + ironing party,
  • first time punting,
  • first time skiing and snowboarding,
  • first time playing board games with friends,
  • and many many more.

I really enjoyed my time in Oxford. I am incredibly grateful for all the people I've met, all the friends I've made, and the excellent tutors I've had the pleasure to learn from. Bassel and Sergi have made the most impact in my three years, but every tutor in Merton has been wonderful: a big thank you to Tia, Karina, Simon and Ralf. My friends have helped me immensely as well. I will write a proper review about my three years in Oxford after finishing this post.

My plan for the summer

The original idea was to take this summer off, seeing as I have been working (kind of) hard studying for Finals. But it seems like lots of different projects have started creeping up on me.

Before the pandemic, the plan was to take the summer off, travel Europe with Celine (visiting friends along the way), catch my family for graduation on the 7th of August, then bring them around Europe before I fly to the US at the end August and start my master's at Harvard.

Now everything has been turned topsy-turvy. Harvard is going full remote, which means that I'm no longer eligible for an F1 visa, which means I cannot enter the country, which means I'm not eligible for any CPT or OPT internship clearance, which means I don't want to go to Harvard anymore.

Some projects I'm currently working on include:

  • writing a path tracer in Julia;
  • building an online board game engine;
  • learning Classical Chinese;
  • doing some clustering project for Inzura;
  • adding new features to the Districtr web app for MGGG GSoC;
  • cleaning up my thesis to submit as a working paper;
  • cleaning up my work with Rodden and Eubank to submit as a working paper;
  • self-studying probability, statistics and data science.

I have written an (aspirational) plan for this summer, but I haven't been following it very well. I think this is because I am not able to say no to interesting projects.

Addendum on blogs being taken down

Two of my favourite blogs, Philip Guo and Slate Star Codex were taken down. This makes me somewhat worried. Prof Guo wrote that one of the reasons he took down his public blog was because he was afraid that some of the things he wrote many years ago would be taken out of context, with bad outcomes. This happened recently to Stephen Hsu, who lost his place as the vice president for research and graduate studies after a petition by MSU's Graduate Employees Union. If two people I respect and admire very much have taken down their blogs because they are worried about being doxxed by media, should I also take note and be more careful about what I post?

Final thoughts

This is what I wrote at the end of 2019:

Despite my poor mood towards the end of the year, I'm optimistic about 2020. As long as I put in the effort, I'm confident that I will get a First, and I am cautiously optimistic that I'll get a place in a Master's program.

I have enjoyed my time in Oxford very much and will be sad to leave the place and all my friends. I very much look forward to squeezing out all I can in my last 6 months in Oxford ("sucking the marrow out of Oxford", lol).

I have mixed feelings about these six months. On the one hand, I'm very pleased that I fulfilled my academic and career goals: I got a First (N.B: spoiler alert), and I got several Master's offers. But on the other hand, all my friends had to leave 59, there was the whole coronavirus thing, and all the post-exams celebrations (punting, trashing, schools dinner) and graduation are no longer present. It saddens me to have our last moments in Merton snatched away so suddenly, without even a chance to bid goodbye to my friends.

As lovely as these three years have been, and as much as I loved living the idyllic Oxford student life, we must always move on eventually. In this respect, then, I am grateful to have finished walking this section of road, and look forward to the journey ahead with excitement and optimism.


  1. see Dugar and Shahriar 2018. ↩︎

  2. https://github.com/nodeGame/nodegame ↩︎