Chapter 1. Career orientation: what blended-finance research actually is

The job

Blended-finance research is the empirical evaluation of programs that combine public, philanthropic, and private capital to fund development projects in low- and middle-income settings. Rural microcredit. Agricultural insurance. Off-grid solar. School-construction bonds. Pay-for-success contracts. The defining feature is the mix. A development finance institution (the World Bank’s IFC, the European Investment Bank, the African Development Bank) puts in concessional capital that lowers risk for private investors who would otherwise not enter.

The research question is almost always one of two things. First, did the program work? Did rural household consumption rise? Did smallholder yields improve? Did female labor-force participation move? Second, for whom did it work, and at what cost? Heterogeneity, distribution, cost-effectiveness, replicability.

If you have spent the past year reading Banerjee and Duflo, J-PAL working papers, and the World Bank’s Development Impact blog, you already know what the output looks like. Your job for the next decade is to produce that output, faster and cleaner than the people you are competing with.

Who hires

The realistic employer set, ranked roughly by how much methodological discretion you have:

1. Tier-one research labs. J-PAL (MIT), IPA (Yale), CEGA (Berkeley), DIME (World Bank), 3ie. You work on field RCTs with a senior PI. Methodological discretion is medium-high. Salary is modest by US-finance standards (USD 60–90k starting in a US-based role, less in regional offices), but the publication credit and network are the point.

2. Multilateral DFIs. World Bank, IFC, IDB, AfDB, ADB, EIB. You join the impact-evaluation team or the country-program team. Methodological discretion is medium. Salary is generous (USD 90–140k starting for an Economist-track G role, more if you have a PhD). Internal politics is real. Publications happen but on a slower clock than academia.

3. Foundations and philanthropies. Gates Foundation, Open Philanthropy, Hewlett, Rockefeller, IDRC, FCDO (the UK’s former DFID). You design or commission impact evaluations. Methodological discretion is high if you are senior, lower if you are program staff. Salary ranges widely.

4. Implementers and NGOs. BRAC, Pratham, ICRW, Care, Heifer. You sit inside the program and run its monitoring and evaluation. Methodological discretion is low to medium. Salary is lower than the above. The advantage is you see the program from inside, which tends to make you a better researcher later.

5. Consultancies. Mathematica, NORC, Abt, IDinsight, Ideas42. Project-based contract work, often evaluating DFI- or foundation-funded programs. Discretion varies by firm. Salary is competitive (USD 80–110k starting).

6. Academia. Tenure-track economics or public policy. The hardest path, the most autonomy, the longest tail.

For a Portugal \to Brazil \to pan-African trajectory specifically, your most useful early employers are the IFC’s Sub-Saharan Africa team, the African Development Bank in Abidjan, the Gates Foundation Agricultural Development Programme, and any of J-PAL Africa, J-PAL Latin America (Santiago), or IPA’s country offices (Mozambique, Cape Verde, Brazil). The European Investment Bank’s Lisbon office has a growing rural-Portugal portfolio, and the OECD’s Development Centre in Paris does Portuguese-language work on Lusophone Africa.

What they want

Three things, weighted unequally.

(1) Methodological credibility. They want to know that your identification strategy holds, that your standard errors are clustered at the right level, that you have thought about the TWFE crisis and the weak-instrument problem and the McCrary test. They do not want to read a paper and find a junior reviewer can break it. This curriculum is built to give you that floor.

(2) Domain knowledge. They want to know that you understand how a microcredit institution sets its interest rate, why farmers in northeast Brazil do not take up index insurance, what PRONAF eligibility actually requires, how the EU’s NUTS-3 boundaries map onto Portuguese district economies. You learn this by reading and by going to the field. Plan for at least one fieldwork stint in your first two years. There is no real substitute.

(3) Communication. A DFI brief is not an academic paper. A two-page executive summary that a country director can read in the back of a car between meetings is more valuable than a 40-page draft they will never open. Practice writing for the non-economist. The papers that get cited and used are the ones a policymaker can read without a journal subscription and without an econometrics degree.

The conference circuit

In rough priority order:

• NEUDC (Northeast Universities Development Consortium). The main applied-dev conference, rotates among Harvard, Yale, MIT, Cornell, Brown, Tufts. Submission deadline around July, conference November.
• PacDev (Pacific Conference for Development Economics). West Coast counterpart, March.
• AEA / ASSA Annual Meeting. January, the field’s main US conference. Hard to get into the development sessions; easier into related ones (public, labor, agricultural).
• BREAD (Bureau for Research and Economic Analysis of Development). Boutique workshops twice a year, invitation-skewed. Submit if you have a connection.
• CSAE Oxford. March, the main UK-centered dev conference. Strong African focus.
• GLAD (Geography and the Logistics of Aid and Development). Newer, geospatial-development crossover.
• EUDN (European Development Network). Annual EU-based meeting.
• J-PAL / IPA / 3ie methodology workshops. Usually open registration, useful for technique deep-dives.

When you have a paper draft you would like feedback on, NEUDC and PacDev are the bread and butter. The trick is to submit even when you do not think the paper is ready. The rejection is data; the acceptance is leverage.

Reading habit

A working research economist reads four to eight papers per week, most of them skimmed in eight to fifteen minutes. The papers that get read end to end are the ones whose abstract has all three of: identification you do not immediately understand, results that surprise you, and a setting you might use yourself. Chapter 11 has a curated 50-paper list to get you started.

Subscribe to the NBER Development Economics weekly digest, the J-PAL working-paper RSS, the World Bank’s Development Impact blog, VoxDev, and the Center for Global Development blog. Spend thirty minutes on Monday morning triaging the week’s drops. Most of what arrives is not for you. The point is to know what is in the air.

Software fluency

Three tools, in this rough priority order.

Stata is the default at the World Bank, IFC, IDB, and most J-PAL / IPA RCT pipelines. You will be expected to write clean .do files that another team member can run a year from now. Master reghdfe, csdid, rdrobust, ivreghdfe, eventdd. Know how to manage Stata projects with iefolder (the World Bank’s project-structure ado) and iebaltab (balance tables).

R has been dominant in academic development economics since around 2018, and dominates in causal ML and modern reproducibility tooling. You will write papers in R or co-author with people who do. Master fixest, did, rdrobust, grf, sf and spdep, and the tidyverse for data wrangling. Learn quarto for reproducible documents. It is the right tool for the next ten years of academic-side output.

Python is the third leg, mostly for data engineering (pandas, geopandas) and any machine-learning work that goes beyond grf. Less central for headline-result econometrics but required for anything that involves satellite imagery, large administrative datasets, or text. PyTorch, scikit-learn, and econml for causal ML.

You do not need to be a software engineer. You need to be the strongest applied econometrician on the team, with code that runs reproducibly. That is the bar.

Languages

For your trajectory specifically: keep your Portuguese sharp (Lusophone Africa, Brazil), build at least basic French (Francophone West Africa, AfDB working language), and continue with English at native-research level. Spanish is a strong third for Latin American DFI work. Many Africa-side researchers also pick up working German for KfW and BMZ collaborations. Do not try to do all of these at once. Pick the next one based on your next field site.

Geography

The thing nobody tells junior development researchers: where you live affects what you publish. If you base yourself in Lisbon you will do a lot of EU-funded Lusophone Africa work and not much SADC-region work. If you base yourself in Maputo, the reverse. The IFC’s Washington headquarters and the World Bank’s main office are still the gravitational centers, but increasingly the action is in regional hubs (Nairobi, Abidjan, and Lagos for Africa; São Paulo for LatAm; Manila for Southeast Asia).

If you can afford the income hit, spend six to eighteen months in your primary region of interest at some point in your first five years. The papers you write will be qualitatively different, and DFIs hire for that fieldwork credential.

A realistic 5-year plan

Year 1. Finish this curriculum. Re-implement Helfrich (2026) on US NMTC for your Portugal extension. Submit the Portugal paper to a working-paper series (IZA, CESifo, BSE Working Papers). Attend NEUDC even if only as audience. Apply for a J-PAL or IPA RA or pre-doctoral fellowship if you do not already have one.

Year 2. First fieldwork. Either a J-PAL country office or an IFC summer rotation. Write a second paper from the data you gather. Submit to NEUDC or PacDev. Build a personal website with your research, code, and CV.

Year 3. PhD application cycle if you are heading academic, or full-time DFI / J-PAL hire if you are heading institutional. Either way, have two working papers and one published or R&R by the end of this year. This is the year that decides whether you go academic or institutional.

Years 4 and 5. Specialize. By the end of year five you should be the person in your DFI or department who is called when a rural-credit evaluation comes up. That specialization is what generates the rest of the career.

A note on imposter syndrome

You will at some point convince yourself that every other person in the room knows something fundamental that you do not. They do not. They have a different combination of gaps than you do, and most of them have spent some afternoon last week quietly Googling whatever they were too embarrassed to ask in the meeting. The work, in any given week, is showing up and running the regression cleanly. That is the whole game. The feeling that you are not enough is mostly your mind telling you that you care about doing the thing well, which is a feature.

This curriculum is the floor of the field. Everyone you will work with knows what is in these chapters. If you know them too, you are a member of the conversation. After that, the differentiation comes from the questions you choose to ask.