← Uber Interview Insights

Uber·Data Scientist·Technical Phone Screen·Senior

Senior
May 2026

Summary

Uber data science interview that threw a full econometrics case at me. One question, but it had about seven layers to it and I was not prepared for how deep they wanted to go on the causal inference side.

Questions Asked (1)

Q1

How would you estimate the causal effect of price on ETA (or ETA on price) given that the two are determined simultaneously? Walk through an econometric strategy including instrument selection, a 2SLS spec with fixed effects and clustered errors, exclusion restrictions, elasticity computation, and diagnostic tests. How would the results feed into surge pricing decisions?

A/B Testing & ExperimentationPricing & MonetizationProduct Analytics & Metrics
Author's notes

This wrecked me a little.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Frame the simultaneity as an endogeneity problem and propose an instrumental variables strategy using a valid instrument that shifts price but not ETA directly. Walk through a 2SLS specification with fixed effects and clustered standard errors, then discuss how to compute elasticities and run diagnostic tests. Finally, connect the estimated causal effect to surge pricing decisions, emphasizing practical implications.

Pro tip: Emphasize that the instrument must be exogenous and relevant, and that you would validate it with first-stage F-statistics and overidentification tests if possible. Also, mention that in practice, Uber often uses natural experiments or randomized pricing tests to complement IV estimates.

1. Define the causal question and endogeneity

Clearly state the simultaneity between price and ETA: price affects demand and supply, which in turn affect ETA, and ETA affects price through surge algorithms. Explain why OLS is biased.

2. Propose an instrumental variable strategy

Suggest a valid instrument that affects price but not ETA directly, such as cost shifters (e.g., fuel prices, driver incentives) or exogenous demand shocks (e.g., weather, events) that are uncorrelated with ETA conditional on controls.

3. Specify the 2SLS model with fixed effects and clustered errors

Outline the two equations: first stage regresses price on the instrument and controls; second stage regresses ETA on instrumented price. Include fixed effects (e.g., time, location) and cluster standard errors at the appropriate level (e.g., city or driver).

4. Discuss exclusion restriction and diagnostic tests

Explain why the instrument satisfies the exclusion restriction (only affects ETA through price) and list diagnostic tests: first-stage F-statistic for weak instruments, overidentification test if multiple instruments, and possibly a test for endogeneity (Durbin-Wu-Hausman).

5. Compute elasticity and link to surge pricing

Show how to compute the elasticity of ETA with respect to price from the 2SLS coefficient, and discuss how this elasticity informs optimal surge multipliers to balance supply and demand, considering trade-offs between wait times and affordability.

Key Points to Mention

  • Simultaneity bias and the need for instrumental variables
  • Choice of instrument: relevance and exogeneity (e.g., cost shifters, weather shocks)
  • 2SLS specification with fixed effects (time, location) and clustered standard errors
  • Exclusion restriction justification and diagnostic tests (weak instruments, overidentification)
  • Elasticity computation and interpretation (e.g., percentage change in ETA for a 1% change in price)
  • Practical implications for surge pricing: setting multipliers to optimize ETA and revenue

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.