The jobs an economics and data background leads to, what each one does on a normal day, which skills on this site you need for it, and what AI tools are doing to the work. One page, checked once a year.
Last checked: October 2026An economics degree with data skills trains you to do three things employers pay for: turn a vague business or policy question into something a dataset can answer, tell a causal story apart from a correlation, and explain the result to someone who will act on it. The Python and statistics in the labs are how you get from the question to the answer. The economics is how you know whether the answer means anything.
Most people start in one of the roles below. The titles overlap and change from firm to firm, so read the day-to-day description, not the title. Pay and growth figures are from the US Bureau of Labor Statistics Occupational Outlook Handbook, rounded; each card links to the page it came from. The Handbook has no entry for "data analyst" or "analytics engineer", so for those roles the closest Handbook occupation is named and the figure should be read loosely.
Day to day. Pull data with SQL or pandas, clean it, build the weekly or monthly numbers a team runs on, make charts and dashboards, and answer "why did this move?" when a number changes. Usually inside one business unit: marketing, operations, finance, a product team.
Skills here. Nearly every step on the Labs and skills page: loading and cleaning, groupby, merging, plotting, simple regression. The Python videos cover the pandas you will use daily. Practice the Entry-Level Analyst questions on Interview practice.
No Handbook page of its own. Closest: market research analysts, median pay about $79,000 (2025), growth 7% over 2025–35, and operations research analysts, about $89,000, growth 12%.
With AI. The pulling, cleaning and first-draft charting is what the tools do well, so a day has fewer of those hours in it. What is left is the harder part: knowing which number matters, noticing when the data is wrong, and saying clearly what changed and what to do about it.
Day to day. Build and test predictive models (churn, demand, fraud, credit), design and read A/B tests, and work with engineers to put a model into a product. More statistics and more code than the analyst role, and more time defending a modeling choice to people who will question it.
Skills here. The modeling and evaluation labs on Labs and skills: train and test splits, cross-validation, regularization, ROC curves, bootstrap intervals. Both interview tracks on Interview practice apply.
Data scientists: median pay about $120,000 (2025), growth 35% over 2025–35, bachelor's degree typical.
With AI. The tools write boilerplate model code and suggest features, so the baseline model gets cheap. The value shifts to the parts that need judgment: is the training data representative, is the test leaking, does this metric match the business goal, and is the question causal when the model is only predictive.
Day to day. At a government agency, central bank, think tank or consultancy: track an economy or a market, forecast, estimate the effect of a policy or a merger, and write memos and reports that non-economists will read. Heavy on regression, panel data and careful writing. Often the first job that uses the economics directly.
Skills here. The regression, panel-data and inference labs (confidence intervals, p-values, clustered standard errors) on Labs and skills, plus the writing you practice in lab recommendations. The Product Economist questions on Interview practice are the closest match.
Economists: median pay about $125,000 (2025), growth 5% over 2025–35, master's degree typical. Entry-level analyst posts under the same title usually take a bachelor's.
With AI. Literature summaries, first drafts and routine data updates get faster. The causal question does not: deciding what the counterfactual is, which comparison group is credible, and whether an estimate survives the obvious objections is the work, and it is what the memo gets judged on.
Day to day. At a tech firm, retailer, airline or marketplace: measure how users or customers respond to a change, run experiments, estimate elasticities, and set or test prices. You sit between product managers, engineers and finance, and your job is to be the person in the room who knows what the data says.
Skills here. Experiments and A/B testing, elasticity regressions, groupby and cohort analysis on Labs and skills. The Product Economist track on Interview practice was written for this role.
Closest Handbook page: market research analysts, median pay about $79,000 (2025), growth 7% over 2025–35. Pay at large tech firms runs higher than the Handbook median.
With AI. Dashboards and experiment read-outs are increasingly generated for you. Knowing when an experiment is underpowered, when a price test is contaminated by a competitor's move, and how to explain a trade-off to a product manager is what the firm keeps paying for.
Day to day. One to two years working for a professor or a research group (a university, a Federal Reserve bank, a research institute) before a PhD or a research career. Clean messy administrative data, replicate results, run specifications, make the figures and tables for a paper. The best training there is in careful empirical work.
Skills here. Everything on Labs and skills, especially merging, reshaping, regression and reproducible code; the pandas and plotting videos on Python for the Labs. Many groups also want Stata or R, which you can pick up quickly once pandas is solid.
Not a Handbook occupation; see economists for the career it leads into. Pre-doc pay is typically a salaried research post well below the figures on this page.
With AI. Code assistants speed up the cleaning and the replication. They do not know which specification the referee will ask for, or whether a coefficient that changed sign is a bug or a finding. Research groups hire for that judgment and for care with data.
Day to day. At a bank, insurer, asset manager or fintech: build and validate the models that price risk and set capital (credit scoring, loss forecasting, stress tests), monitor portfolios, and write up model documentation for regulators. More mathematics and more formal validation than the other roles. Plain "quant" trading roles want more math and programming than an economics degree alone gives.
Skills here. Regression with ROC evaluation, regularization, bootstrap and simulation on Labs and skills; the Entry-Level Analyst questions on Interview practice for the data-handling parts.
Closest Handbook page: financial analysts, median pay about $104,000 (2025), growth 7% over 2025–35.
With AI. Model code and documentation drafts come faster. Regulators still require a human to explain and stand behind the model, and much of the job is now checking models, including ones built with AI, for data problems, unstable behaviour and unfair outcomes.
Day to day. Build and maintain the data pipelines and tables that analysts and data scientists use: SQL transformations, data models, tests that catch bad data before it reaches a dashboard. Less economics, more software practice. A common move for analysts who find they like the plumbing.
Skills here. The loading, cleaning, merging and reshaping steps on Labs and skills, taken further into SQL, version control and testing. The Python videos on functions and reading data are the start of it.
No Handbook page. The nearest statistics are under data scientists and database and software roles.
With AI. Writing transformations gets faster; deciding what the tables should mean, and keeping them correct as the business changes, is the job that remains.
Two things are true at once, and students tend to hear only one of them.
First, the tools are used and they save time, but less of it than the headlines suggest. A St. Louis Fed study by Bick, Blandin and Deming, The Impact of Generative AI on Work Productivity (27 February 2025), found that about 22% of US workers used generative AI in a given week in late 2024, and that users saved about 5% of their hours, which works out to roughly 1% of hours across all workers. Analytical jobs are among the heavier users.
Second, the effect is falling on entry-level hiring. Brynjolfsson, Chandar and Chen at the Stanford Digital Economy Lab, Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence (first released August 2025, revised August 2026), track payroll data and find that employment of 22-to-25-year-olds in the most AI-exposed occupations is well below where it would have been had it kept pace with less exposed peers (about 19% below in the revised version). Experienced workers in the same occupations show no such gap. The gap comes from firms hiring fewer young people, not from firing them.
Put together: the routine first-year tasks that used to be how juniors learned the job, the data pulls, the standard charts, the first draft, are the tasks the tools now do. The way in is to arrive already able to do the parts the tools cannot.
What gets automated. Writing boilerplate code. Cleaning data that follows a known pattern. Standard charts and dashboards. Summaries of documents. First drafts of memos. Routine model fitting.
What gets more valuable. Judgment about data: knowing when a dataset is wrong, biased or not the one the question needs. Causal questions: knowing the difference between a prediction and an effect, and which comparison makes an effect believable. Communication: saying what the result means and what to do, to someone who will not read the code. Checking AI output: every role above now includes reviewing work a tool produced.
How to use this site for it. Do the labs yourself before asking a tool, so you know what correct looks like. Use the Python videos to get unstuck fast. Use Interview practice to rehearse explaining a result out loud, because that is the skill interviews test. For books and courses beyond this site, see Good outside resources.