RBI Grade B DSIM Interview

RBI Grade B selection has three stages: Phase I (qualifying), Phase II of 300 marks, and an interview of 75 marks, so final merit is out of 375 and the interview is 20 per cent of it, as per the latest advertisement. A psychometric test is held before the interview and is not counted in the score. The board typically has four to five members: a retired or serving RBI Executive Director or Chief General Manager in the chair, two or three senior RBI officers, and an external expert, often an economist or academic. Interviews run 20 to 40 minutes and may be taken in Hindi or English. The board starts from your biodata and your graduation or post-graduation subject, moves to RBI's functions, the latest monetary policy and inflation data, banking regulation and financial stability, and then to current economic developments; DEPR and DSIM candidates are pushed much deeper into economics and statistics respectively, including their dissertation or research work. The ExamPilot mock compresses this to fifteen minutes with the same arc: biodata and motivation, a demanding economics and regulation round with an RBI-level member, two regulatory or departmental situations, and a short close on posting and career.

1. 1

   Survey methods

   The RBI's household inflation expectations survey covers respondents in major cities. What biases can creep into such a survey, and how would you correct for them?

   What a strong answer covers

   Frame and coverage limits since only urban households are asked, non-response, recall and rounding to round numbers, and anchoring on recent food prices. Corrections include design weights, careful questionnaire wording, trimming outliers and reporting the median alongside the mean. Avoid claiming that a large number of respondents removes bias.

2. 2

   Index numbers

   Walk us through how the all-India CPI is constructed, and explain why a base year revision changes more than just the reference year.

   What a strong answer covers

   A Laspeyres-type index with weights from a household consumption expenditure survey, prices collected from rural and urban markets for a fixed basket, and aggregation from item to group to all-India. A base revision updates weights and the basket to current spending, for example a lower food share. Avoid confusing CPI with WPI construction.

3. 3

   Time series

   You must forecast next quarter's monthly CPI inflation. Which model would you start with, and how would you handle seasonality and sudden vegetable price shocks?

   What a strong answer covers

   Starts with a seasonal ARIMA after testing for stationarity, adds seasonal dummies or differencing, considers exogenous drivers such as rainfall, fuel prices or minimum support prices, and judges models on out-of-period forecast errors. Treats vegetable spikes as outliers or separate components. Avoid jumping to a complex model without diagnostics.

4. 4

   Research work

   Tell us your Master's dissertation in two minutes: the research question, data, method and main result. Then name its biggest weakness.

   What a strong answer covers

   A clear, structured summary a non-specialist member can follow, followed by an honest limitation such as data quality, endogeneity, small number of observations or an untested assumption, and how you would fix it with more time. Avoid jargon-heavy recitation or claiming the work has no limitations.

5. 5

   Inference

   A regression shows that a bank's credit growth significantly affects district output, but the coefficient is tiny. How would you interpret it for a policy note?

   What a strong answer covers

   Distinguishes statistical significance, meaning the effect is unlikely to be zero, from economic significance, meaning whether its size matters for policy. Explains the magnitude in plain units, checks confidence intervals and specification. Avoid reporting only the p-value or treating significance as importance.

6. 6

   Communication

   How would you explain a confidence interval for quarterly GDP growth to a policymaker who only wants one number?

   What a strong answer covers

   Gives the point estimate along with a range showing uncertainty, explains in plain words that the method captures the true value in most repeated estimates, and why the width matters for decisions such as rate changes. Avoid the common misstatement that there is a 95 percent probability the true value lies in this particular interval.

7. 7

   Econometrics

   What goes wrong if you simply regress inflation on money supply growth using thirty years of Indian data, and how would you fix it?

   What a strong answer covers

   Non-stationarity and spurious regression, structural breaks such as the shift to inflation targeting, reverse causality, omitted supply shocks and changing velocity. Fixes include unit-root and cointegration tests, break tests, instrumental variables and sub-period analysis. Avoid interpreting a raw correlation as a causal policy lever.

8. 8

   RBI statistics

   Name the RBI's regular surveys that you know, and explain how one of them feeds into the Monetary Policy Committee's decisions.

   What a strong answer covers

   Correctly names surveys such as the consumer confidence survey, the household inflation expectations survey, the industrial outlook survey, the order books, inventories and capacity utilisation survey and the survey of professional forecasters, then explains how one informs the MPC's view of demand or expectations. Avoid inventing survey names.

9. 9

   Data handling

   A branch-level transactions dataset has many missing values concentrated in rural branches. How would you handle them before analysis?

   What a strong answer covers

   First studies whether values are missing at random or systematically, since rural concentration suggests a pattern, then chooses suitable imputation such as multiple imputation, flags imputed records and runs sensitivity checks. Documents every assumption. Avoid simply deleting incomplete records, which would bias results against rural branches.

• English
• Hinglish (Hindi + English)Hinglish