The Methodologist's Path · A-Level Curriculum
THE METHODOLOGIST'S PATH
Becoming the "Analyzed, Interpreted, and Wrote"
Person
A cross-domain co-author curriculum for an A-Level student
Curriculum · Books · Courses · Networks · Timeline · Portfolio
Prepared: May 2026
Page 1 of 19
The Methodologist's Path · A-Level Curriculum
What you actually noticed
The author note "Analyzed and interpreted the data and wrote the paper" describes a specific and
powerful role in modern research: the methodologist. This person does not own the domain question —
they own the empirical machinery. Across economics, public health, political science, and biology, a small
number of methodologists end up on a disproportionate number of papers because senior researchers
know they can hand over a dataset and get back a defensible analysis.
This is one of the most strategic paths an ambitious student can take, for three reasons. First,
methodological skills transfer across domains — labour economics, environmental health, agricultural
development, education all reduce to similar estimators applied to different data. Second, the demand far
exceeds supply: there are always more interesting questions than there are people who can credibly
answer them. Third, the skills are objectively learnable from public materials. You do not need a particular
passport or a famous adviser to learn fixed effects or write reproducible code; you only need years.
The role, named correctly
In economics this is the "applied econometrician" or "empirical micro" RA-track. In medicine and
epidemiology, the "biostatistician." In political science, the "quantitative methods" person. In data-
science industry language, the "causal inference scientist." Same job, different vocabulary.
Page 2 of 19
The Methodologist's Path · A-Level Curriculum
People who built careers this way
These are not the household names you read about; they are the people whose CVs senior researchers
actually study when hiring.
• Marcella Alsan (Harvard) — built a career as the empirical co-author on Black health and history
papers before her own programme took off. The methods opened the doors.
• Heidi Williams (Stanford) — became the patent/innovation data person before she had her own
theory of innovation.
• Melissa Dell (Harvard) — the canonical example of a young economist whose value at first was
sheer technical ability to digitise and exploit historical data.
• Stefanie Stantcheva (Harvard) — survey-experiment methodology became her brand.
• In Bangladesh, the diaspora researchers (Mobarak at Yale, Asad Islam at Monash, Mahreen
Mahmud at Oxford, M. Niaz Asadullah at Monash) lead networks that include exactly this kind of
methodologist.
None of them started as the famous co-author. They started as the person who could be trusted with the
data.
Page 3 of 19
The Methodologist's Path · A-Level Curriculum
How long it actually takes
Below is a realistic schedule assuming 15–25 hours per week of focused work in addition to A-Level
coursework. "Focused work" means deliberate practice: replicating published papers, debugging, reading
methods chapters carefully — not watching coding tutorials at 2x speed.
Phase Skill milestones What you can do at the end
Months 0–6 Statistics, linear algebra, Clean an LFS round, replicate a simple regression table from
R fluency, OLS, basic a published paper, present results to your mentor without
panel data panicking
Months 6–12 Causal inference toolkit: Run a complete empirical project end-to-end on public data;
DID, IV, RDD, matching, pass a code review by a PhD-level mentor
synthetic control. LaTeX.
Git.
Months 12–18 Time series OR spatial OR Independently produce the empirical section of a working
machine-learning-for- paper; identify problems in others' code
causal-inference (pick
one). Bootstrap, clustered
SE, multiple testing.
Months 18–30 Specialisation: pick one Be the methodologist on a Q2 paper. First author or strong
domain (development, second author.
health, climate,
education) and one
method niche (Bayesian,
ML-causal, structural,
spatial)
Months 30–48 Production-grade: pre- Trusted on Q1 papers. PhD admissions at top programmes
analysis plans, structural become plausible.
estimation, frontier
methods. Mentor your
own juniors.
The honest co-authorship arithmetic
Q3 South-Asia or regional journal: realistic at the 12–18 month mark. Q2 specialist field journal:
realistic at 24–30 months with good mentorship. Q1 specialist journal as second author: 30–48
months and requires network. Top-five (AER, QJE, Econometrica, JPE, RES): not on a high-school
timeline regardless of skill. Anyone who tells you otherwise is selling something.
Page 4 of 19
The Methodologist's Path · A-Level Curriculum
Ivy League admissions — short version
Top US universities (Harvard, MIT, Stanford, Princeton, Chicago, Yale) admit roughly 3–7% of international
applicants. Among rejected applicants every cycle are students with peer-reviewed publications. So the
credential by itself is not decisive. What this path gives you, however, is several things admissions officers
care about deeply.
What the path produces that matters for admissions
• Letters of recommendation from researchers at named institutions. The single highest-leverage
item in your file. A line like "X performed the empirical analysis at the level of my second-year
doctoral students" from a Harvard or Yale faculty member moves the needle far more than the
publication itself.
• A clear intellectual story. Admissions officers read tens of thousands of files. "International
student with strong grades and one research project" is common. "International student who
became the empirical analyst on three papers across labour and climate and can explain every line
of their code" is rare and memorable.
• Interview material. If you reach an interview, being able to discuss your econometrics deeply
makes you unforgettable. Most 17-year-olds cannot.
• Demonstrated fit. Harvard's Economics department, MIT's Course 14, Princeton's BSE programme
— these admit students with revealed preferences for technical economics work.
What it does not do
• Compensate for weak grades or scores. You still need top A-Level grades, top SATs, and a strong
school record.
• Substitute for genuine intellectual depth. Admissions officers can tell when a student name-tagged
onto papers vs. drove the work. Be the latter.
• Replace the rest of the application. The essays, extracurriculars, and your character story still
matter.
Realistic positioning
If you execute this path well — strong letters from named researchers, demonstrable technical depth, two
to three substantive papers, top academic record — your admit probability at top-10 US economics
programmes moves from roughly 1–3% (baseline international applicant pool) to perhaps 10–20%. Not
certainty, but a meaningful multiple. Compare this to the alternative path of generic extracurriculars and
competitions, which leaves you at the baseline.
Page 5 of 19
The Methodologist's Path · A-Level Curriculum
The curriculum
This is what you actually study, in order. Treat the books as the spine and the courses as muscle around it.
Skipping the foundations to get to the exciting parts is the most common reason students stall in month 9
— they cannot debug their own results because they never learned why OLS works.
Phase 1 (Months 0–6): Foundations
Mathematics
• Linear algebra: Strang, Introduction to Linear Algebra (5th ed.). Also watch his 18.06 lectures on
MIT OCW. You need vectors, matrices, eigenvalues, rank, projections, and least squares —
properly.
• Calculus: Stewart's Calculus is fine; what you actually need are gradients, Hessians, and Lagrange
multipliers. If you've done A-Level Maths and Further Maths, this is mostly review.
• Probability and statistics: Casella & Berger, Statistical Inference (chapters 1–10). This is the
foundation for everything else. Slow and deliberate.
• Math for economists (optional but useful): Simon & Blume, Mathematics for Economists. Strong
on optimisation and dynamics.
First exposure to econometrics
• Primary text: Stock & Watson, Introduction to Econometrics (4th ed.). Read every chapter. Do
every exercise.
• Companion: Hanck, Arnold, Gerber & Schmelzer, Introduction to Econometrics with R (free at
[Link]). Replicate every Stock-Watson example in R yourself.
Programming
• R as primary language: Wickham & Grolemund, R for Data Science (2nd ed., free at
[Link]). This is non-negotiable; learn tidyverse properly.
• Stata as second language: If your university gives access. Cameron & Trivedi's Microeconometrics
Using Stata is the reference; mostly learn by reading other people's code.
• LaTeX: Overleaf has tutorials. Learn enough to typeset a 20-page paper with cross-references,
tables (with booktabs), and a bibliography (BibTeX).
• Git: Bryan, Happy Git and GitHub for the useR (free at [Link]). Get a private GitHub
repo running by month 2.
Courses (free)
• MIT OCW 18.06 (Linear Algebra) — Gilbert Strang. 35 lectures.
• MIT OCW 14.32 (Econometrics) — pair with Stock-Watson.
• Khan Academy AP Statistics if you want to fill probability gaps.
• DataCamp / Coursera "Statistics with R" specialisation (Duke) if you need structured R practice.
Page 6 of 19
The Methodologist's Path · A-Level Curriculum
Month-6 milestone
• You can clean a BBS LFS round end-to-end in R.
• You can derive OLS and explain Gauss-Markov in your own words.
• You have replicated one published paper using public data and your replication runs in one
command.
• You have a GitHub portfolio with at least two clean projects.
Page 7 of 19
The Methodologist's Path · A-Level Curriculum
Phase 2 (Months 6–12): Causal inference
Modern empirical work in social science is built on a small set of identification strategies. You need each
one in your hands. The good news is that the literature here is exceptionally well-taught — there are
arguably better free resources for causal inference than for any other technical field.
Primary books
• Angrist & Pischke, Mostly Harmless Econometrics. Short, punchy, the foundation. Re-read it after
your first project; it makes more sense the second time.
• Cunningham, Causal Inference: The Mixtape (free at [Link]). More accessible
than Angrist-Pischke. Has R and Stata code throughout. The DID and synthetic control chapters are
the best free introduction available.
• Huntington-Klein, The Effect (free at [Link]). Even more accessible; strong on the
philosophy of identification before the algebra.
• Wooldridge, Econometric Analysis of Cross Section and Panel Data (2nd ed.). The reference text.
Do not read cover-to-cover; use as encyclopedia.
• Imbens & Rubin, Causal Inference for Statistics, Social, and Biomedical Sciences. Heavier; for
when you want to go deeper into potential outcomes.
Methods to master
Difference-in-differences (DID)
• Classic 2x2 DID and parallel-trends testing.
• Two-way fixed effects (TWFE) and its pitfalls — read Goodman-Bacon (2021) and de Chaisemartin-
D'Haultfœuille (2020).
• Modern staggered DID estimators: Callaway-Sant'Anna (2021), Sun-Abraham (2021), Borusyak-
Jaravel-Spiess (2024). R packages: did, fixest, didimputation.
• Event-study graphs and inference.
Instrumental variables (IV)
• Two-stage least squares, the LATE theorem (Imbens-Angrist 1994).
• Weak instruments — Stock-Yogo critical values, Olea-Pflueger F-statistic.
• Shift-share instruments (Bartik); Borusyak-Hull-Jaravel (2022) and Goldsmith-Pinkham-Sorkin-Swift
(2020).
Regression discontinuity (RDD)
• Sharp and fuzzy RDD.
• Bandwidth selection: Calonico-Cattaneo-Titiunik (2014); rdrobust package.
• McCrary density test, validity checks.
Page 8 of 19
The Methodologist's Path · A-Level Curriculum
Matching and weighting
• Propensity score matching and its limitations (King-Nielsen 2019 — "why PSM should not be
used").
• Entropy balancing (Hainmueller 2012), inverse probability weighting.
• Doubly robust estimation.
Synthetic control
• Abadie-Diamond-Hainmueller (2010, 2015) — read both.
• Permutation inference, placebo tests.
• Generalised synthetic control (Xu 2017), matrix completion methods.
Inference
• Clustered standard errors — when to cluster, at which level (Abadie-Athey-Imbens-Wooldridge
2023).
• Bootstrap, wild cluster bootstrap (Cameron-Gelbach-Miller 2008).
• Multiple hypothesis testing — Romano-Wolf, list-wise corrections.
• Randomisation inference.
Courses
• Scott Cunningham's Causal Inference workshops on YouTube — the entire Mixtape course is free
in video form.
• Joshua Angrist, Mastering Econometrics, on MarginalRevolutionUniversity (free).
• Susan Athey & Guido Imbens, Machine Learning and Causal Inference, Stanford — lecture videos
on YouTube.
• Coursera: Susan Athey's "Causal Inference" specialisation if you want structure.
Month-12 milestone
• You can produce a clean DID event-study plot from raw data in under a day.
• You can defend an identification strategy to a sceptical PhD student.
• You have written your first working paper draft on a public dataset.
• You have at least one R package or substantial code repository on GitHub that someone other
than you has used.
Page 9 of 19
The Methodologist's Path · A-Level Curriculum
Phase 3 (Months 12–24): Depth and a chosen specialisation
In months 12–24 you must specialise. Generalists are not valuable; specialists with adjacent fluency are.
Choose one methods track and one domain track and read everything in it. The combinations below are
the ones with the most active demand right now.
Methods specialisation tracks (pick one)
Track A: Time-series and macroeconometrics
• Enders, Applied Econometric Time Series (4th ed.) — the textbook.
• Lütkepohl, New Introduction to Multiple Time Series Analysis — for VARs.
• Kilian & Lütkepohl, Structural Vector Autoregressive Analysis (2017) — current frontier on SVAR
identification.
• Hamilton, Time Series Analysis — the classical reference.
• R packages: vars, urca, fGarch, bvarsv, BVAR, dynamac. Master vars and bvarsv.
• Useful for: macro shocks, exchange rates, monetary policy, remittances, commodity prices.
Track B: Spatial econometrics and geospatial methods
• LeSage & Pace, Introduction to Spatial Econometrics (2009).
• Anselin, Spatial Econometrics: Methods and Models (1988).
• Pebesma & Bivand, Spatial Data Science with R (free at [Link]).
• Lovelace, Nowosad & Muenchow, Geocomputation with R (free at [Link]).
• R packages: sf, terra, spdep, spatialreg, splm. Python: rasterio, xarray, geopandas, earthengine-
api.
• Useful for: agriculture, climate, urban economics, public health, deforestation, conflict.
Track C: Machine learning for causal inference
• Athey & Imbens (2019), "Machine Learning Methods Economists Should Know About," Annual
Review of Economics — read first.
• Chernozhukov et al., "Double/Debiased Machine Learning" (2018, Econometrics Journal).
• Wager & Athey (2018) and Athey, Tibshirani & Wager (2019) on causal forests.
• Hastie, Tibshirani & Friedman, Elements of Statistical Learning (free PDF).
• R packages: grf, DoubleML, EconML (Microsoft). Python: econml, dowhy, causalml.
• Useful for: heterogeneous treatment effects, policy targeting, prediction-aided causal work.
Currently the hottest methods space.
Track D: Bayesian methods
• Gelman et al., Bayesian Data Analysis (3rd ed., free PDF on Gelman's site).
• McElreath, Statistical Rethinking (2nd ed.) — the most pedagogical Bayesian text ever written.
• Koop, Bayesian Econometrics — for the econometrics specifically.
Page 10 of 19
The Methodologist's Path · A-Level Curriculum
• Tools: Stan, brms, rstanarm in R. PyMC in Python.
• Useful for: small-sample problems, hierarchical models, time-varying parameters, structural
estimation.
Track E: Structural estimation
• Train, Discrete Choice Methods with Simulation (free PDF on Train's site).
• Adda & Cooper, Dynamic Economics: Quantitative Methods and Applications.
• Aguirregabiria & Mira (2010) on dynamic discrete choice — the survey.
• More demanding; less in fashion than ML-causal right now but very deep.
Domain breadth tracks (pick one as primary, one as secondary)
Development economics
• Banerjee & Duflo, Poor Economics; What the Difference a Day Makes; Good Economics for Hard
Times.
• Ray, Development Economics — the textbook.
• Journals to read regularly: Journal of Development Economics, World Development, EDCC, Journal
of Development Studies, AEJ: Applied.
• VoxDev ([Link]) — weekly summaries.
Labour economics
• Borjas, Labor Economics — textbook.
• Cahuc, Carcillo & Zylberberg, Labor Economics — advanced reference.
• Journals: Journal of Labor Economics, Labour Economics, ILR Review, AEJ: Applied, Journal of
Human Resources.
Environmental and climate economics
• Tietenberg & Lewis, Environmental and Natural Resource Economics.
• Phaneuf & Requate, A Course in Environmental Economics.
• Read everything from Climate Impact Lab ([Link]); follow Hsiang, Burke, Carleton.
• Journals: JEEM, ERE, Climate Change Economics, NCC, Energy Economics.
Health economics
• Bhattacharya, Hyde & Tu, Health Economics.
• Currie & Madrian, Handbook of Labor Economics chapter on health.
• Journals: Journal of Health Economics, Health Economics, AJHE, Demography.
Public economics
• Saez et al. — the entire NBER Public Economics WP series.
• Hindriks & Myles, Intermediate Public Economics.
Page 11 of 19
The Methodologist's Path · A-Level Curriculum
Month-24 milestone
• You have one publishable working paper as first author or strong co-author.
• You have read 200+ papers in your specialisation.
• You can write a methods appendix that survives PhD-level scrutiny.
• You have presented at one conference.
• You have replicated at least three Q1 papers and posted replication packages on your GitHub.
Page 12 of 19
The Methodologist's Path · A-Level Curriculum
Phase 4 (Months 24–48): Production and network
By now your bottleneck is no longer skill — it is access. The world has thousands of people who can run a
fixed-effects panel; what makes you valuable is being in the room when the question is being defined. This
phase is about becoming the person senior researchers want in that room.
Production-grade practices
• Pre-analysis plans: register them on AEA RCT Registry or OSF before you see the data. Read 10
PAPs from published work and write three of your own.
• Replication packages: every paper you write should produce a one-command end-to-end
replication. Use Snakemake or a Makefile.
• Containerisation: learn Docker. Reviewers in 2027 will run your code; you need it to still work.
• Pre-registration culture: read Christensen & Miguel (2018, JEL) on transparency in economics.
• Power calculations: learn ex-ante power for the designs in your toolkit. R packages: pwr, simr.
Building the network
• Cold-email Bangladeshi-origin economists at top universities — Mobarak (Yale), Asad Islam
(Monash), M. Niaz Asadullah (Monash), Mahreen Mahmud (Oxford), Asim Khwaja (Harvard) —
with a one-paragraph email containing one specific question after you have a draft to attach.
Without a draft, do not email.
• Domestic mentors: Selim Raihan (SANEM), Mohammad Yunus (BIDS), Sayema Bidisha, Atonu
Rabbani at Dhaka University.
• RA applications: many top researchers post RA positions on Twitter and on the NBER and J-PAL
websites. Apply for remote RA roles starting at month 18.
• Conferences: SANEM, BEA, Y-RISE Annual Conference (when in Bangladesh), Asian Meeting of the
Econometric Society. Apply for travel grants.
• EconTwitter / EconBluesky: follow the researchers in your field, engage with their working papers,
respond thoughtfully to their threads. This is how a generation of methodologists got noticed.
Portfolio assets you should have by month 36
• At least one peer-reviewed publication (Q2 or Q1) — first author or strong co-author.
• One working paper under review at a Q1 journal.
• A GitHub profile with 5+ substantive projects, including code accompanying each paper.
• A personal academic website (Hugo + Wowchemy is the standard) with CV, papers, code, contact.
• Conference presentation history — at least 3 talks at meaningful venues.
• Two strong letters of recommendation from researchers at named institutions.
Page 13 of 19
The Methodologist's Path · A-Level Curriculum
How to actually break into co-authorships
The skills-portfolio is necessary but not sufficient. The hardest step is the one nobody writes about: turning
competence into invitations. The pattern that works is roughly:
Step 1: The replication portfolio (months 6–12)
Pick three recent papers in your target field from Q1 journals that have public data. Replicate them. Post
the replication on your GitHub with a clean README and a one-paragraph summary of what you learned.
This is your interview material. When you email a researcher, you are not asking them to take a chance —
you are showing them you have already done their job once.
Step 2: The extension paper (months 12–18)
Take one of your replications and extend it: different country, different period, additional robustness, an
interaction the original authors did not run. Write it up as a 15-page comment paper. Send to the original
authors before posting. Many will engage; a few will offer to co-author or refer you onward.
Step 3: The cold email that works (months 18–24)
Cold emails to senior researchers work when they meet four criteria: short, specific, attached, useful.
• Short: three paragraphs maximum.
• Specific: reference one paper of theirs and one question it raises that you have already started
working on.
• Attached: a 10-page draft. Not a CV, not a transcript — a draft.
• Useful: offer something concrete ("I have built a cleaned dataset of X you might find useful").
Expected response rate: 10–20% if the four criteria are met. 0–2% otherwise. Persistence: send 30 such
emails over six months; one or two will produce a collaboration.
Step 4: Junior RA work (months 12–30)
Many Q1 papers list two or three RAs in the acknowledgements; the best of those RAs end up as co-
authors on the next paper. Apply to remote RA positions through:
• NBER RA postings ([Link] maintains an excellent list).
• J-PAL RA postings ([Link]).
• Y-RISE ([Link]/yrise) — Bangladesh-focused.
• IZA, IFPRI, World Bank DECRG short-term RA roles.
• Direct emails to researchers whose work you have replicated.
Page 14 of 19
The Methodologist's Path · A-Level Curriculum
Concrete year-one weekly plan
A defensible weekly schedule for the first 12 months, assuming you are a full-time A-Level student. The
goal is consistency, not heroic weeks.
Weekday evenings (2 hours × 5 = 10 hours)
• 60 minutes: textbook reading with notes (Stock-Watson → Wooldridge → Cunningham).
• 60 minutes: coding practice in R (replicating chapter examples, then solving exercises).
Saturday (5 hours)
• Deep work block: replication project or your own analysis. No phone.
• Code review of your own previous week's work.
Sunday (3 hours)
• 90 minutes: read one paper end-to-end. Write a one-page summary.
• 90 minutes: review notes, plan the coming week, write to your mentor.
Quarterly check-ins
• Q1: Replicate one Stock-Watson empirical example fully from raw data.
• Q2: Replicate one published Q2 paper. Post on GitHub.
• Q3: Begin your own working paper on public data. Draft methodology section.
• Q4: First full draft of working paper. Internal presentation to mentor.
Page 15 of 19
The Methodologist's Path · A-Level Curriculum
Master resource list
The five books to actually finish
• Stock & Watson, Introduction to Econometrics (4th ed.)
• Wickham & Grolemund, R for Data Science (2nd ed., free online)
• Cunningham, Causal Inference: The Mixtape (free online)
• Wooldridge, Econometric Analysis of Cross Section and Panel Data
• Angrist & Pischke, Mostly Harmless Econometrics
Reference books (use as encyclopedia)
• Cameron & Trivedi, Microeconometrics: Methods and Applications
• Hayashi, Econometrics
• Greene, Econometric Analysis
• Hansen, Econometrics (free online at Bruce Hansen's UWisc page)
• Imbens & Rubin, Causal Inference for Statistics, Social, and Biomedical Sciences
Free university courses worth doing
• MIT 18.06 Linear Algebra (Strang)
• MIT 14.32 Econometrics
• MIT 14.382 Econometrics (Chernozhukov, graduate level — aspirational)
• Stanford EconStats Lab YouTube videos
• Cunningham's Causal Inference Mixtape — YouTube playlist
• MarginalRevolutionUniversity (Angrist's Mastering Econometrics)
• QuantEcon ([Link]) for dynamic programming
• OpenIntro Statistics for foundations
Blogs and newsletters to subscribe to
• World Bank Development Impact blog ([Link]/impactevaluations)
• VoxDev, VoxEU, The IGC blog
• Andrew Gelman's blog ([Link])
• Marc Bellemare's blog ([Link]/wordpress)
• NBER Digest weekly email
• Susan Athey's Stanford CIPS newsletter
R packages to know inside out (in order of priority)
• tidyverse (dplyr, ggplot2, tidyr, readr, purrr, forcats)
• fixest — fastest fixed effects estimator in any language. Author: Laurent Bergé.
Page 16 of 19
The Methodologist's Path · A-Level Curriculum
• did, didimputation, bacondecomp — modern DID.
• rdrobust, rddensity — RDD.
• AER, ivreg — IV.
• synth, Synth, gsynth, augsynth — synthetic control.
• grf, DoubleML — ML-causal.
• brms, rstanarm — Bayesian.
• sf, terra, spdep — spatial.
• modelsummary, stargazer, fixest's etable — table production.
Datasets you should have downloaded and explored before month 12
• BBS Labour Force Survey 2022, 2024 (request via mentor).
• World Bank WDI (entire database, via WDI R package).
• Penn World Table 10.01 (via the pwt10 R package).
• Polity V dataset on regime types.
• V-Dem (varieties of democracy).
• ICPSR studies relevant to your domain.
• ERA5-Land monthly aggregates for South Asia.
Page 17 of 19
The Methodologist's Path · A-Level Curriculum
Pitfalls that derail this path
Skipping the foundations
Students who jump from R for Data Science straight to Causal Forests without finishing Stock-Watson
cannot debug their own results. Their papers fall apart in revisions. There is no shortcut around
derivations.
Collecting credentials instead of skills
Twelve online certificates and one PhD-level skill set are very different. Admissions officers and senior
researchers can tell. Spend your time on real work, not Coursera badges.
Joining "research mills"
Some Bangladeshi and Indian organisations promise A-Level students publications in pay-to-publish
journals. This will actively hurt your application — admissions committees and serious researchers spot
these instantly. Stick to peer-reviewed journals with editorial boards you can verify.
Working in isolation
A mentor is not optional. A peer group is not optional. Find both. The students who succeed have someone
who reads their code weekly.
Not writing
Writing is harder and more important than econometrics. A great regression in a poorly-written paper gets
rejected. Read Deirdre McCloskey's Economical Writing and write 500 words a day, every day.
Mistaking productivity for progress
Running 30 specifications a day feels productive. Reading one paper and understanding it deeply is
progress. Calibrate yourself accordingly.
Page 18 of 19
The Methodologist's Path · A-Level Curriculum
The thing nobody will tell you
In 2026, more high-school students worldwide can run a regression than at any time in history. The bar to
be "useful" has moved sharply upward. What still distinguishes someone is depth: the ability to defend
every line of code, every modelling choice, every interpretation, to a hostile reviewer.
This is a craft. Crafts take years. The students who clear the bar are the ones who treat the work as a craft
— slow, deliberate, occasionally repetitive, deeply satisfying. They do not look for shortcuts because they
understand that the route is the apprenticeship, not the credential at the end of it.
If you can commit to that — three to four years of steady work, a real mentor, a portfolio of replicated and
original work, a few peer-reviewed publications by year three — you will have built something that opens
doors at Harvard, MIT, Princeton, Yale, Stanford, LSE, and any of their economics departments. More
importantly, you will have built something that opens doors for the rest of your career, whether you go to
an Ivy or not.
Start tonight. Open Stock-Watson, page 1.
Page 19 of 19