Portfolio Management interview preparation

Questions

100

Traced to a firm

40

Firms

24

Updated

September 2026

Asked at

All firmsBLBlackRock4Vanguard4WMWellington Management4Amundi3ACAQR Capital Management3Neuberger Berman3SCSchroders3Man Group2MSCI2Northern Trust2AllianceBernstein1Apollo Global Management1Blackstone1BMBNY Mellon1Carlyle Group1Fidelity Investments1Goldman Sachs1Invesco1Millennium Management1MSMorgan Stanley1NUNuveen1PIMCO1SSState Street1TPTPG1

Topic

All topicsPortfolio theory5Factor models8Asset allocation11Rebalancing3Portfolio construction7Benchmarks and tracking error5Performance measurement8Risk management6Fixed income and LDI5Currency and global3Implementation and costs5Active versus passive6India markets7Brainteasers5Career and fit16

Level

AnyCoreIntermediateHard

Type

AnyTechnicalCaseMarket viewFitBrainteaser

1. 048What is value at risk, and what are its weaknesses in a portfolio context?Risk managementIntermediatetechnicalBLBlackRockRisk and Quantitative Analysis · New York · 2026

   Say this

   VaR is the loss you would not expect to exceed over a given horizon at a given confidence level, say a 1 percent chance of losing more than 4 percent in a day. Its weaknesses are that it says nothing about how bad the tail is, it is not sub-additive, and it is estimated from a history that may not contain the event you care about.

   Then walk it

   1. Three ways to compute it. Parametric, assuming normality, which is fast and wrong in the tails. Historical simulation, replaying actual past returns on today's holdings, which is the industry default. Monte Carlo, which lets you model non-linear payoffs properly.
   2. Weakness one, it is a threshold not an expectation. A 99 percent VaR of 4 percent is consistent with a worst case of 5 percent or of 40 percent, and for options books the difference is everything. That is why regulators moved to expected shortfall.
   3. Weakness two, it is not sub-additive, so the VaR of a combined portfolio can exceed the sum of the parts. That makes it a mathematically improper risk measure and it breaks risk budgeting, because contributions do not add up.
   4. Weakness three, the history. Historical simulation over two calm years will not produce a stressed number, so VaR was lowest just before both 2008 and 2020. Volatility clustering means the model is most reassuring when it should be most alarming.
   5. Weakness four, it is blind to liquidity and to the path. A ten-day VaR assumes you can hold or exit at marked prices, and in a real stress the exit price is the problem, so I would pair it with a liquidity-adjusted measure and with time-to-liquidate estimates.
   6. So in a portfolio seat I would use VaR as one dial among several: expected shortfall for the tail, scenario and reverse stress tests for the events not in the sample, factor exposures for what the bet actually is, and drawdown limits for the thing clients actually experience. VaR's real virtue is that it aggregates across asset classes into one comparable number, and that is worth keeping.

   Where candidates lose it

   Defining VaR and stopping, or getting the direction of the confidence statement muddled. In a risk and quantitative seat the expected content is the tail blindness, the failure of sub-additivity, and procyclicality, that VaR is lowest right before the event. Also say what you would use alongside it, because 'VaR is bad' is not a risk framework.

   Expect next

   • So explain expected shortfall.
   • Why is VaR not sub-additive?
   • How would you stress test beyond the historical sample?

   Reported by candidates at BlackRock (Risk and Quantitative Analysis, New York, 2026). Source: Wall Street Oasis.

2. 049Why did regulators move from VaR to expected shortfall?Risk managementIntermediatetechnicalRisk managementRisk and quantitative analysis

   Say this

   Because expected shortfall measures the average loss in the tail rather than the entry point to it, and because it is sub-additive so it behaves like a proper risk measure. That makes it harder to game and it makes risk contributions add up.

   Then walk it

   1. Definition: expected shortfall, also called conditional VaR, is the expected loss given that you are beyond the VaR threshold. If 97.5 percent VaR is 3 percent and the average loss in the worst 2.5 percent of cases is 5 percent, expected shortfall is 5 percent.
   2. Reason one, tail sensitivity. VaR is indifferent to what happens past the threshold, so a book can be restructured to look identical on VaR while having a far worse tail. Selling deep out-of-the-money options is the textbook way to do exactly that.
   3. Reason two, coherence. Expected shortfall is sub-additive, so combining two portfolios never increases measured risk. That matters practically because it means you can decompose total risk into additive contributions per position or per desk, which is how a risk budget is actually run.
   4. Basel's Fundamental Review of the Trading Book replaced 99 percent VaR with 97.5 percent expected shortfall, calibrated to a period of stress, which was deliberately chosen so the capital number does not collapse in calm markets.
   5. The trade-off, which I would name: expected shortfall is harder to backtest. You can count VaR exceptions against an expected frequency, but testing the average size of tail losses needs far more data, so validation is weaker exactly where the measure is most demanding.
   6. For an asset manager rather than a bank, the practical consequence is that I would report both, plus named scenarios. Expected shortfall for the tail, VaR for comparability and history, and scenarios because no statistical measure covers an event outside its sample.

   Where candidates lose it

   Describing expected shortfall as just a bigger VaR. The two substantive reasons are that it looks inside the tail and that it is sub-additive so risk contributions add. Concede the backtesting weakness; candidates who present it as strictly superior have not thought about validation.

   Expect next

   • How would you backtest expected shortfall?
   • How do you use sub-additivity in a risk budget?
   • What confidence level would you run for an equity fund?
3. 050If you already measure volatility, why do you care about maximum drawdown?Risk managementIntermediatetechnicalAsset managementHedge funds

   Say this

   Because drawdown is what clients experience and what triggers redemptions, while volatility is a statistical average that has no memory of the path. Two funds with identical volatility and identical annual returns can have drawdowns of 12 percent and 40 percent, and only one of them still has a business.

   Then walk it

   1. Volatility is a dispersion measure with no ordering. Reshuffle the same monthly returns and volatility is unchanged while maximum drawdown changes completely, because drawdown depends on the sequence.
   2. Drawdown is where autocorrelation shows up. A trend-following or momentum strategy has long strings of same-sign returns, so it produces deeper drawdowns than its volatility implies. A mean-reverting strategy produces shallower ones.
   3. The business consequence is the real answer. Redemptions cluster after drawdowns, so a manager in a 35 percent hole faces outflows precisely when the opportunity is best, and is forced to sell to fund them. Drawdown risk becomes forced-seller risk.
   4. The arithmetic of recovery is the other half. A 50 percent drawdown needs a 100 percent gain to get back, so the compounding cost of a deep hole is not symmetric with the gain that preceded it.
   5. The statistical caveat I would volunteer: maximum drawdown is a single realised event from one path, so it is a very noisy statistic. A longer track record mechanically shows a deeper maximum drawdown, which makes cross-manager comparison unfair unless you standardise the window.
   6. So I would use both, and add expected drawdown or the distribution of drawdowns from a simulation rather than relying on the one realised maximum. And I would set client-facing limits in drawdown terms, because that is the number that actually governs behaviour.

   Where candidates lose it

   Saying drawdown is just another way of measuring risk. The specific insight is path dependency: volatility is order-independent and drawdown is not, so autocorrelation in returns drives the difference. And do not quote maximum drawdown across managers with different history lengths without saying that longer records mechanically show deeper drawdowns.

   Expect next

   • Which strategies have drawdowns worse than their volatility implies?
   • How would you set a drawdown limit?
   • What is the Calmar ratio?
4. 051Explain path dependency and why the sequence of returns matters so much to a real portfolio.Risk managementHardsuperdayWealth managementPension and endowment investing

   Say this

   Because compounding is multiplicative and because real portfolios have cash flows. With no flows, order does not change the terminal value. Add contributions or withdrawals and the order changes everything, because a loss suffered when the balance is largest is a far bigger loss in money terms.

   Then walk it

   1. Start with the pure case: the same set of returns in any order gives the same compounded total. So path dependency is not about arithmetic on the returns themselves.
   2. It bites through cash flows. A retiree drawing 5 percent a year who meets a 30 percent fall in years one and two sells units at the bottom and may never recover, while the same returns arriving in years nine and ten leave them comfortable. Identical average return, completely different outcome.
   3. Same mechanism on the accumulation side but with the opposite sign: a young saver with a small balance benefits from an early crash, because most of their contributions buy in cheaply. Sequence risk is largest when the pot is largest relative to remaining contributions, which is the decade around retirement.
   4. Then the volatility drag, which is the second channel. Geometric return is below arithmetic return by roughly half the variance, so plus 50 then minus 50 leaves you at 75. Higher volatility mechanically lowers terminal wealth even with the same average return, which is the real argument for risk control rather than return maximisation.
   5. Institutionally the same thing appears as forced selling: a pension paying benefits, an endowment funding a spending rule, a fund meeting redemptions. All of them convert a paper drawdown into a permanent loss because units are sold at the bottom.
   6. So the design responses are all about the path: a liquidity bucket or bond ladder covering a few years of outflows, glide paths that de-risk into the drawdown phase, flexible spending rules, and managing to drawdown rather than to volatility. Reporting an expected return without the path is close to useless for anyone with obligations.

   Where candidates lose it

   Answering only with the volatility drag arithmetic. That is one channel. The larger one is cash flows: order does not matter without flows and dominates with them. Naming the pre-retirement decade as the point of maximum sequence risk shows you understand why glide paths exist, rather than just repeating that they do.

   Expect next

   • How would you protect a client in the five years before retirement?
   • How does this affect a pension's liquidity policy?
   • What is volatility drag and how big is it?
5. 052How do you think about portfolio risk and transaction cost together, rather than separately?Risk managementHardsuperdayMan GroupInvestment Management · Boston · 2022

   Say this

   You put them in the same objective function. The portfolio you want and the portfolio you can afford to get to are different, so the right target is the one that maximises expected return minus a risk penalty minus the cost of trading there from where you actually are.

   Then walk it

   1. The naive process runs in sequence: optimise for risk and return, hand the trade list to the desk, discover that the turnover costs more than the expected edge. Anything with fast-decaying signals dies this way.
   2. The integrated version maximises alpha minus lambda times variance minus the trading cost of moving from current to target weights. Because market impact is roughly proportional to the three-halves or square of size, the cost term is convex, which naturally produces partial rather than complete trades.
   3. That gives the no-trade region. For each position there is a band around the ideal weight where the expected improvement does not cover the cost of getting there, so you leave it alone. That single idea cuts turnover enormously with almost no loss of expected return.
   4. It also changes what a risk limit means. If reducing an exposure costs 60 basis points in impact, a hard limit breach is a genuine trade-off rather than an automatic trade, and the correct response might be to hedge with a liquid proxy today and unwind the physical slowly.
   5. Liquidity becomes a risk input rather than an operational detail. I would hold days-to-liquidate per position, size illiquid names accordingly, and treat capacity as part of the risk model. A portfolio that takes 15 days to exit has a risk profile that no covariance matrix captures.
   6. And the cost estimate has to be the firm's own. Vendor models are a starting point, but the only credible input is your own realised slippage by name, size and market condition, fed back into the optimiser. Otherwise you are optimising against a fiction.

   Where candidates lose it

   Treating trading cost as the execution desk's problem that arrives after portfolio construction. In a systematic seat the expected answer is a single objective function with a convex cost term, and the concept to name is the no-trade band. Saying you would calibrate the cost model on the firm's own realised slippage rather than a vendor default is what makes it sound like experience.

   Expect next

   • How would you estimate market impact?
   • What does the no-trade band do to turnover?
   • How would you handle a risk limit breach in an illiquid name?

   Reported by candidates at Man Group (Investment Management, Boston, 2022). Source: Wall Street Oasis.

Firm tags come from public, anonymous candidate reports on Wall Street Oasis: strong signal, not sworn testimony. Firms are named as the places a question was reported, not as partners of Fin Maverick. Answers are written for this page to show how to think out loud; they are not scripts to recite.