Resources
Things I share with students and collaborators: how to find a predoc, AI tools for research, useful datasets, and research tools and guides.
Predoc Tips
Sources for finding predoctoral and RA positions
This list is not exhaustive, but it’s a good place to start.
A hub of predoc resources. Predoc.org collects useful materials, including practice coding tasks and a newsletter with advertised positions.
Sites that aggregate many postings
- NBER posts non-NBER RA positions here, and advertises some directly here.
- Professors and others tag relevant RA postings on the @econ_RA Twitter/X page (no account needed to read).
Organized predoc programs with a set hiring cycle
- Stanford: SIEPR predoctoral research fellows program
- Chicago: EPIC (Energy Policy Institute)
- Harvard: EPoD Research Fellow postings here
- Harvard: Opportunity Insights (two-year commitment) here
- Yale: predoctoral fellows program
- Princeton: a cross-department university program
Programs for US citizens and/or permanent residents only
- Harvard Research Scholar Initiative (no visa sponsorship)
- PhD Excellence Initiative (NYU Stern)
Other sources to check
- J-PAL Careers (note: some J-PAL roles do not sponsor visas, but other postings do)
- Innovations for Poverty Action (IPA)
- Brown and Princeton often post in their own career portals, e.g. Predoctoral Research Jobs at Princeton
- The World Bank; look for Research Assistant and ET Consultant roles
Guides
- J-PAL’s 2-page guide on landing an RA position
- My archived predoc search guide (PDF, Nov 2020)
AI Tools
I ran a hands-on workshop for economists at UCSD on Claude Cowork + Code. The session covered using AI tools in economics research: writing and editing Stata, R, and Python, integrating results into Overleaf, and keeping projects and replications organized, with live examples. Slides and the workshop thread:
Claude Desktop Workshop for Economists
A shorter follow-up on permissions: the difference between Manual and Auto mode, and why version controlling with git and/or Dropbox does more of the safety work than the permission prompt.
Permission Modes and Auto: A Short Follow-Up
Data
Datasets you may find useful to explore geo-spatial and/or mental-health-related research ideas.
Geospatial data for development & urban research
Publicly available spatial and satellite data for development and urban research:
| Topic | Dataset | Geographic units | Time coverage | Key variables | Spatial data | Access |
|---|---|---|---|---|---|---|
| Settlement & population | GHS - Global Human Settlement Layer (JRC / Copernicus) | Global grid (100 m - 1 km); GHS-UCDB city polygons | 1975 onward, in epochs | Built-up surface (GHS-BUILT), population (GHS-POP), settlement model / degree of urbanization (GHS-SMOD), consistently delineated cities (GHS-UCDB) | Rasters; UCDB as vector polygons | Free, open |
| Household surveys + GPS | DHS - Demographic and Health Surveys | Survey clusters, country | 1985 onward, repeated cross-sections | Health, fertility, and wealth microdata; geospatial covariates such as night lights | Randomly displaced GPS cluster coordinates | Free, registration; GPS needs a short access request |
| Economic activity proxy | Night lights - VIIRS & DMSP (Earth Observation Group) | Global grid (VIIRS ~500 m; DMSP ~1 km) | DMSP 1992-2013 (an extension series runs to 2019); VIIRS 2012 onward | Nighttime radiance, a standard proxy for local economic activity | Rasters | Free, but bulk downloads now need an EOG account; also on Google Earth Engine |
| Urban extent, from night lights | Annual urban extents - Zhao et al. (2022), paper | Global grid, ~1 km | 1992-2020, annual | Binary urban / non-urban maps built from harmonized night lights (DMSP-OLS through 2013, VIIRS-derived after). Note this is a core-urban-domain concept, not a Landsat impervious-surface product like GAIA, so the two are not drop-in substitutes | Rasters | Free, open (CC BY 4.0) |
| Gridded GDP | Kummu et al. - gridded GDP & HDI | Global grid, 5 arc-min (~10 km); the update also gives 30 arc-sec and admin-level polygons | 1990-2015, annual; the 2025 update runs 1990-2022 | GDP per capita PPP, total GDP, HDI. The 2025 update extends GDP per capita and downscales to admin-2, but drops HDI, so for gridded HDI you still need the 2018 release | NetCDF; update adds gpkg and CSV | Free, open (CC BY 4.0) |
| Gridded GDP | Local GDP estimates - Rossi-Hansberg & Zhang (2026), paper | Global cells at 1, 0.5, and 0.25 degrees | 2012-2022, annual | Predicted local GDP with uncertainty bounds, from a random forest on population, night lights, land use, emissions, and vegetation. Predictors reach beyond night lights, which helps if night lights are your regressor | Shapefiles + CSV | Public; no explicit license stated |
| Administrative boundaries | GADM | Country down to level 2-3 | Current, versioned releases | Administrative names and codes | Vector boundaries | Free for non-commercial use |
| Census microdata + boundaries | IPUMS International | Harmonized GEOLEV1 / GEOLEV2 (consistent over time) and unharmonized year-specific admin units | 1960 onward, census rounds | Census microdata linked to boundaries; IPUMS USA does the same for the US | GIS boundary files | Free, registration |
Working with night lights. Matt Lowe’s Night Lights and ArcGIS: A Brief Guide (2014) walks from the raw downloads to an analysis-ready dataset, covering clipping, gas-flare removal, and projections. The ArcGIS workflow still holds; its data links predate the move to the Earth Observation Group, so pair it with the VIIRS and DMSP links above. (Linked via the Internet Archive: the original MIT URL now 404s.)
Aggregated catalogs. The UPenn Libraries GIS guide (global and US spatial data) and the geo4.dev data catalog (development-focused) list many more sources.
Mental health data
Publicly available datasets with a validated mental-health or wellbeing measure, plus a few open-replication economics papers and speech/audio depression corpora; free with registration unless a restricted/paid flag is noted. Sleep has its own section below, though the survey tables here also note what sleep each dataset collects.
Development and low- or middle-income country panels
| Dataset | MH / sleep measures | Geographic coverage | Level | Total N | Youth available? | Access |
|---|---|---|---|---|---|---|
| IFLS - Indonesia Family Life Survey | CES-D-10 Sleep: PROMIS quality & disturbance items (wave 5) | Indonesia, 13 provinces; province-level (GPS restricted) | Individual & household | ~30,000 individuals / 7,200+ households | Yes, age 15+ | Free, registration |
| MxFLS - Mexican Family Life Survey | Zung/Calderon depression Sleep: daily hours (time-use, all waves) | Mexico, national; state & municipality | Individual & household | ~35,000 individuals / 8,400 households | Yes, age 15+ | Free, registration |
| Young Lives | SRQ-20, GAD-7/PHQ-8, Cantril Sleep: time-use hours/day (rounds 2-5, 7) | Ethiopia, India, Peru, Vietnam; region/district | Individual (child cohort) | ~12,000 children | Youth cohort (child to young adult) | Free, registration (UK Data Service) |
| NIDS - National Income Dynamics Study | CES-D-10 Sleep: only the CES-D restless-sleep item | South Africa, national; district municipality | Individual | ~28,000 individuals / 7,300 households | Yes, age 15+ | Free, registration |
| CFPS - China Family Panel Studies | CES-D, Kessler K6 Sleep: hours (2014+), bedtime & naps (all waves) | China, 25 provinces (~95% of pop.) | Individual & household | ~42,600 individuals / 14,960 households | Yes, age 10+ | Free, registration + data-use agreement |
| KLPS - Kenya Life Panel Survey | CES-D-10 (KLPS-4) Sleep: bed/wake times, quality, naps (KLPS-4) | Kenya, Busia County cohort (followed nationwide and abroad) | Individual (+ 2nd-gen children) | ~7,500 cohort + ~5,200 children | Yes (young-adult cohort) | Free, open (CC0) |
| DHS - Demographic and Health Surveys | Mental Health module (optional, introduced with DHS-8 in 2022): full PHQ-9 + GAD-7 Released: Nepal 2022, Bangladesh 2022, Mozambique 2022-23, Zimbabwe 2023-24, Zambia 2024. Lesotho 2023-24 fielded PHQ-9 only, no GAD-7. Timor-Leste 2025-26 is fielded but not fully released; Sierra Leone 2026 is planned. Senegal 2023 is tagged “mental health” but uses a different, locally developed instrument, not PHQ-9/GAD-7 Sleep: none | The module is only in the surveys listed, not in most DHS; 63 countries overall, with randomly displaced GPS clusters | Individual & household | ~5,000-30,000 households / survey | Yes, 15-49 (Bangladesh: ever-married women only; Nepal and Lesotho: a subsample) | Free, registration |
United States, UK & Europe
| Dataset | MH / sleep measures | Geographic coverage | Level | Total N | Youth available? | Access |
|---|---|---|---|---|---|---|
| Add Health | CES-D (modified) Sleep: duration, timing & quality items (all waves) | US, national; geocodes restricted | Individual | ~20,000 (in-home) | Yes (adolescent cohort) | Public-use free; full sample restricted |
| NSDUH - Nat. Survey on Drug Use & Health | Kessler K6, MDE module Sleep: MDE insomnia / hypersomnia items only | US, national + state (small-area est.) | Individual | ~67,500 / year | Yes, age 12+ | Free public-use files |
| NHANES | PHQ-9 Sleep: SLQ items (2005+); wrist accelerometry 2011-14 | US, national only in public file | Individual | ~5,000 / year | Yes (12-17 restricted) | Free (18+); 12-17 file restricted |
| NCS-R / NCS-A | CIDI diagnostic Sleep: insomnia items; NCS-A adds bedtime & hours | US, national | Individual | 9,282 / 10,123 | Yes (NCS-A, 13-18) | NCS-R free; NCS-A restricted |
| HRS - Health & Retirement Study | CES-D-8 Sleep: Jenkins insomnia items (2002+); time-use hours | US, national | Individual | ~20,000 / wave | No (50+) | Free, registration |
| Healthy Minds Study | PHQ-9, GAD-7, flourishing Sleep: duration items; ISI module (some waves) | US colleges; Census region only, individual colleges blinded | Individual (student) | ~935,000 (675+ colleges) | Students only (college) | Free, de-identified; short data-request form |
| Understanding Society (UKHLS) | GHQ-12, SWEMWBS; youth SDQ Sleep: PSQI-derived items (waves 1, 4, 7, 10, 13) | UK; region public, finer restricted | Individual & household | ~40,000 households / ~100,000 individuals | Yes (youth panel 10-15) | Free, registration (UK Data Service) |
| ELSA - English Longitudinal Study of Ageing | CES-D-8 Sleep: items in waves 4/6/8; wrist actigraphy (wave 10) | England; region-level public | Individual | ~11,400 (Wave 1 core) | No (50+) | Free, registration |
| SHARE | EURO-D Sleep: trouble-sleeping & medication items; hours (waves 8-9) | 28 European countries + Israel; country-level | Individual | ~160,000 respondents | No (50+) | Free (scientific use), registration |
| UK Biobank | PHQ-9, GAD-7, CIDI-SF Sleep: duration, chronotype & insomnia items; actigraphy (~103,000) | UK; location restricted (1 km grid) | Individual | ~500,000 | No (40-69) | Application + fee + agreement |
Cross-national and global
| Dataset | MH / sleep measures | Geographic coverage | Level | Total N | Youth available? | Access |
|---|---|---|---|---|---|---|
| WHO World Mental Health | CIDI diagnostic Sleep: insomnia items (chronic-conditions section) | 28+ countries; country-level | Individual | >200,000 interviews | No (adults) | Restricted (consortium agreement) |
| HBSC - Health Behaviour in School-aged Children | Psychosomatic scale, Cantril Sleep: sleep-onset difficulties (all rounds); bedtimes (optional) | 45+ countries; country/region | Individual (student) | ~220,000+ / round | Students only (ages 11/13/15) | Aggregate public; microdata by request (embargo) |
| Global Burden of Disease | Modeled prevalence & burden (not survey items) Sleep: none | 204 countries + some subnational | Country-year (aggregate) | Aggregate (not respondents) | Yes (age bands, incl. 10-19) | Free, registration |
Subjective wellbeing (life satisfaction and happiness, not clinical mental health)
| Dataset | MH / sleep measures | Geographic coverage | Level | Total N | Youth available? | Access |
|---|---|---|---|---|---|---|
| Gallup World Poll | Cantril ladder, daily affect Sleep: “well-rested yesterday” item only | 160+ countries; country-level | Individual | ~1,000 / country / year | Yes, age 15+ | Paid microdata; some free aggregates |
| World Values Survey | Life satisfaction, happiness Sleep: none | 64 countries (Wave 7); country-level | Individual | ~95,000 / wave | No (18+) | Free, registration |
Columns in the four tables above: MH / sleep measures = each dataset's mental-health instruments and any sleep data it collects (from the questionnaires / codebooks). Youth available? = whether adolescents / young adults are covered (by age in a filterable general sample, by student status, or as a youth-only sample).
Key mental-health economics papers (with public replication data)
| Paper | Year | Journal | Population | Intervention | Replication | Mental-health variables |
|---|---|---|---|---|---|---|
| Haushofer & Shapiro | 2016 | QJE | Poor households, Kenya | RCT: unconditional cash transfers | Harvard Dataverse | Psychological-wellbeing index: CES-D, Cohen stress, WVS happiness & life satisfaction, salivary cortisol |
| Baranov, Bhalotra, Biroli & Maselko | 2020 | AER | Perinatal mothers, rural Pakistan | RCT: perinatal CBT (Thinking Healthy) | openICPSR | SCID (major-depression diagnosis), Hamilton scale, disability, GAF, social support |
| Bessone, Rao, Schilbach, Schofield & Toma | 2021 | QJE | Low-income adults, Chennai (India) | RCT: night-sleep devices / incentives; workplace naps | Harvard Dataverse | Psychological-wellbeing index: depression, stress, happiness, life satisfaction, Cantril ladder |
| Banerjee, Duflo, McKelway, Schilbach et al. | 2023 | Ann. Intern. Med. | Elderly living alone, Tamil Nadu (India) | RCT: phone-based CBT; one-time cash transfer | Harvard Dataverse | Geriatric Depression Scale, WHODAS, single-item loneliness |
| Angelucci & Bennett | 2024 | AER | Adults with depression, Karnataka (India) | RCT: antidepressant pharmacotherapy; livelihood support | openICPSR | PHQ-9 (screening + severity) |
Detecting depression from speech / audio - public corpora, mostly from the speech-ML and clinical communities (I found no economics study that has released audio-based depression data):
- DAIC-WOZ / E-DAIC (USC). Field-standard English corpus: clinical interviews, audio + transcripts, PHQ-8 labels (the AVEC benchmark). Free but restricted (signed application, institutional email).
- Androids Corpus. Italian speech (reading + interview), clinician diagnoses; open direct download (academic terms).
- EATD-Corpus. Chinese speech + text with SDS depression labels; open download.
- MODMA (Lanzhou). Audio (+ EEG), clinical MDD diagnosis; free but account + agreement.
Sleep data
Sleep measured objectively (actigraphy, wearables, polysomnography) alongside self-report, plus the economics field experiments that have collected it. For sleep items inside the large mental-health panels (IFLS, MxFLS, Add Health, HRS, Understanding Society, and others), see the survey tables in the mental health section above.
Sleep in economics field experiments - objective wearable/actigraphy sleep alongside self-report and economic outcomes:
| Study | Year | Population | Sleep measure | Data |
|---|---|---|---|---|
| Bessone, Rao, Schilbach, Schofield & Toma, QJE | 2021 | Low-income adults, Chennai (India) | Actigraphy + self-report | Harvard Dataverse (public) |
| Giuntella, Saccardo & Sadoff, JPE (forthcoming) | 2025 | ~1,150 US university students | Fitbit + self-report | NBER w32550; replication not yet public |
| Avery, Giuntella & Jiao, REStat | 2025 | US college students | Wearable + self-report | paper; no public package located |
Objective sleep-data repositories (polysomnography and actigraphy):
| Source | Sleep measure | Coverage | Access |
|---|---|---|---|
| NSRR - National Sleep Research Resource (NHLBI) | PSG + actigraphy + questionnaires | 13+ cohorts, 26,000+ people (SHHS, MESA, MrOS, CHAT, …) | Free, per-dataset data-use agreement |
| UK Biobank accelerometer sub-study | Wrist actigraphy (7-day); derived sleep duration/efficiency/timing | ~104,000 participants | Approved application + fee |
| NHANES accelerometry (2011-2014) | Wrist accelerometry (minute-level) | US, nationally representative | Fully public, no application |
| PhysioNet (Sleep-EDF, MMASH, Apple-Watch+PSG) | PSG and/or consumer wearable with PSG labels | Small validation cohorts | Mostly open access |
Research Tools
Useful Stata tools with self-help guides, including making maps (SPMAP / GRMAP) and sample do-files.
Digitizing historical maps with QGIS and Python
A step-by-step guide to georeferencing and digitizing historical maps (co-authored with a lab student).
SHRUG: open geospatial data for India
The Socioeconomic High-resolution Rural-Urban Geographic Platform: open data covering roughly 600,000 villages and 8,000 towns in India, from the Development Data Lab.