Performance Measurement, Analytics and Plan Evaluation

Introduction

Wealth management is incomplete unless the adviser can determine whether an investment plan is working, why it produced its results, and whether it remains suitable under changing conditions. Performance measurement links portfolio outcomes to client objectives. Analytics convert raw account data into evidence about return, risk, diversification, efficiency and resilience. Plan evaluation then asks the broader question: can the household still meet its goals under plausible adverse conditions?

Performance analysis is not simply the ranking of portfolios by return. A sound evaluation distinguishes investment performance from investor cash-flow timing, compares results with an appropriate benchmark, measures downside as well as average risk, identifies sources of active return, and tests whether the plan remains viable under uncertainty. This chapter develops those capabilities for banking, finance and insurance professionals.

Learning Objectives

By the end of this chapter, students should be able to:

1. Calculate and interpret holding-period return, time-weighted return, money-weighted return and CAGR.
2. Measure volatility, downside risk, drawdown, beta, Sharpe ratio and related indicators.
3. Select suitable benchmarks and explain the limitations of benchmark comparisons.
4. Apply single-period Brinson attribution and reconcile attribution effects exactly.
5. Use scenario analysis, stress testing, sensitivity analysis and Monte Carlo simulation.
6. Evaluate the resilience of a financial plan and communicate uncertainty ethically.

1. Return Measurement

1.1 Holding-period return

For an asset or portfolio with beginning value V₀, ending value V₁, distributions D received during the period and external contributions or withdrawals excluded from performance, the holding-period return is:

Here, D includes dividends, interest and other distributions received during the measurement period. If distributions are reinvested and V₁ includes their accumulated value, adding D again would double-count income. Prices and distributions must be measured consistently, and transaction costs should be included if the objective is net investor performance.

For example, an account rises from $100,000 to $106,000 and pays $2,000 in income that is not included in the ending value. The return is ($106,000 − $100,000 + $2,000) ÷ $100,000 = 8%.

1.2 Time-weighted and money-weighted returns

External cash flows complicate measurement. A large contribution before a market rise can make an investor’s personal return differ from the manager’s investment performance.

The time-weighted rate of return (TWR) removes the effect of external cash-flow timing by dividing the evaluation period into subperiods at each material external cash flow. If subperiod returns are R₁, R₂, …, Rₙ:

TWR is generally appropriate for evaluating a portfolio manager because the manager does not control the client’s deposits and withdrawals.

The money-weighted rate of return (MWR), commonly calculated as an internal rate of return (IRR), is the discount rate r that sets the net present value of all investor cash flows to zero:

CFₜ is the investor cash flow at time t, with contributions and withdrawals assigned consistent signs. MWR is appropriate for evaluating the investor’s actual experience because it incorporates the timing and size of cash flows. It may be distorted as a measure of manager skill when the investor controls cash-flow timing.

1.3 Annualised return and CAGR

If an investment grows from V₀ to Vₙ over n years, with no intervening external cash flows and with distributed income treated consistently, the compound annual growth rate is:

More generally, if R is the cumulative total return over n years, then:

If income is distributed rather than retained, the calculation must use total-return values: either reinvest distributions in the ending wealth value or add the value of distributions and their reinvestment consistently. A price-only CAGR excludes income and is not comparable with a total-return benchmark.

For example, $100,000 becoming $121,000 over two years produces CAGR = (121,000 ÷ 100,000)^(1/2) − 1 = 10% per year. The cumulative return is 21%, not 20%, because of compounding.

2. Risk and Downside Analytics

Return should never be interpreted without its risk context.

2.1 Volatility

For a sample of periodic returns R₁ through Rₙ, the sample standard deviation is:

R̄ is the arithmetic mean and n is the number of observations. The denominator n − 1 is used when the observations are treated as a sample from a broader return process. For independent, identically distributed monthly returns, a commonly used annualisation is sᵧ = sₘ × √12. This scaling is an approximation and assumes a stable variance structure; serial correlation can make it unreliable.

Volatility treats positive and negative deviations symmetrically. It is therefore not a complete measure of risk for investors who primarily fear losses.

2.2 Downside deviation and maximum drawdown

Downside deviation focuses on returns below a minimum acceptable return (MAR), such as 0% or an inflation-adjusted target. One sample convention is:

The convention must be stated because some systems divide by the total number of observations rather than n − 1.

Drawdown measures decline from a previous peak. If Wₜ is wealth at time t and Pₜ = max(W₀, …, Wₜ), then:

Maximum drawdown is the most negative drawdown over the observation period. It describes an experienced loss path, not merely dispersion around an average. It is sensitive to the chosen start and end dates.

2.3 Sharpe ratio

The Sharpe ratio measures excess return per unit of total volatility:

Rₚ is the portfolio return, R_f is the risk-free or appropriate low-risk reference return for the same period and currency, and σₚ is portfolio return standard deviation. The numerator and denominator must use the same periodicity. An annualised Sharpe ratio requires compatible annualised excess return and volatility; simply mixing monthly and annual figures is invalid.

The ratio is most informative when returns are reasonably comparable across periods and volatility is an acceptable risk proxy. It does not distinguish upside from downside volatility and can be unstable for short samples or non-normal returns.

2.4 Other risk-adjusted measures

The Sortino ratio replaces total volatility with downside deviation. The Treynor ratio divides excess return by beta and is most meaningful for a well-diversified portfolio where systematic risk is the relevant risk. Information ratio is active return divided by tracking error:

R_b is benchmark return and TE is the standard deviation of active returns. None of these ratios is universally superior. The choice should reflect the client’s risk, benchmark and decision problem.

3. Benchmarks and Evaluation Standards

A benchmark is a reference portfolio or target used to interpret results. A valid benchmark should be:

• specified in advance;
• investable or at least replicable;
• consistent with the portfolio’s asset classes, currency and risk exposure;
• measured on a total-return basis when the portfolio return includes income;
• transparent, independently calculated and rebalanced according to known rules; and
• aligned with the mandate rather than selected after observing results.

A strategic policy benchmark may combine indices according to target asset-class weights. A peer benchmark compares with similar funds but may be affected by survivorship bias, style differences and changing constituents. A liability benchmark may be more relevant for an insurer or pension plan because success depends on meeting obligations, not merely beating a market index.

Benchmark mismatch creates misleading conclusions. Comparing a globally diversified portfolio in local-currency total-return terms with a domestic price index can overstate or understate skill. Benchmark weights should also be documented: beginning-of-period, average or rebalanced weights can produce different comparisons.

4. Return Attribution

Attribution explains active return, defined as portfolio return minus benchmark return. It is different from contribution analysis, which explains how each holding contributed to the portfolio’s absolute return.

4.1 Single-period Brinson attribution

Under a common single-period Brinson framework, let wᵖᵢ and wᵇᵢ be portfolio and benchmark weights for sector or asset class i, and Rᵖᵢ and Rᵇᵢ their respective returns. Portfolio and benchmark returns are:

A simple allocation effect is:

A selection effect is:

The interaction effect is:

Expanding the portfolio return proves the exact identity:

This equality holds when the same categories, weights and returns are used consistently. Under this stated single-period framework, effects sum exactly, not approximately. Alternative Brinson conventions may allocate the interaction differently, but the convention must be disclosed.

4.2 Worked attribution example

Consider two asset classes:

Benchmark return = 0.60×8% + 0.40×4% = 6.40%.

Portfolio return = 0.70×9% + 0.30×3% = 7.20%.

Active return = 7.20% − 6.40% = 0.80 percentage points.

Allocation = (0.70−0.60)×8% + (0.30−0.40)×4% = 0.40 percentage points.

Selection = 0.60×(9%−8%) + 0.40×(3%−4%) = 0.20 percentage points.

Interaction = (0.70−0.60)×(9%−8%) + (0.30−0.40)×(3%−4%) = 0.20 percentage points.

Thus, allocation + selection + interaction = 0.40 + 0.20 + 0.20 = 0.80 percentage points, exactly equal to active return. If management and transaction costs were 0.10 percentage points, net active return would be 0.70 percentage points; costs should be shown as a separate deduction unless the return inputs are already net of costs.

Attribution is informative but not causal proof. A positive allocation effect may reflect a deliberate decision, an unintended drift or a temporary exposure. Small samples and correlated decisions also limit inference about skill.

5. Evaluating a Financial Plan

Portfolio performance is only one component of plan success. A financial plan should be evaluated against goals, constraints and resources.

A useful evaluation sequence is:

1. Define goals, dates, amounts and priority levels.
2. Identify cash flows, taxes, inflation, insurance benefits, liabilities and fees.
3. Establish assumptions for returns, volatility, correlations, salary growth, longevity and withdrawals.
4. Project assets and liabilities through time.
5. Test adverse scenarios and parameter sensitivity.
6. Estimate the probability or range of goal achievement.
7. Recommend actions and specify monitoring triggers.

The adviser should distinguish nominal values from real values. If nominal growth is g and inflation is π, the exact real growth rate is:

The approximation g − π is useful only when rates are modest. Taxes and fees should be modelled explicitly or clearly identified as embedded in net returns.

5.1 Scenario analysis

Scenario analysis evaluates a small number of coherent narratives, such as base case, high inflation, recession, prolonged low returns, early retirement or longevity extension. Each scenario changes several related assumptions. It is useful for communication because clients can understand the story and proposed response.

5.2 Stress testing

Stress testing applies unusually adverse but plausible shocks, such as a 25% equity decline, a sharp interest-rate increase, property illiquidity or an unexpected medical expense. A stress test is not a forecast. Its purpose is to identify vulnerabilities, liquidity needs, concentration risk and decision rules.

5.3 Sensitivity analysis

Sensitivity analysis changes one assumption at a time, or a controlled subset, to identify influential variables. For example, the analyst may vary retirement age, savings rate, inflation, expected return or longevity. It helps prioritise data quality and advice. One-at-a-time sensitivity can understate risk when variables move together, so it should not replace joint simulation.

6. Monte Carlo Simulation

Monte Carlo simulation generates many possible future paths by repeatedly sampling from specified distributions and applying the plan’s cash-flow rules. For a simplified annual wealth model:

Wₜ is wealth at the beginning of year t, Rₜ is the portfolio return, Cₜ is contribution and Xₜ is withdrawal. Contributions and withdrawals must use the same currency and timing convention as W. A return applied before a year-end withdrawal is not equivalent to a return applied after it.

A common stylised return model is:

where μ is the assumed mean return, σ is standard deviation and Zₜ is a standard normal random variable. This is a simplifying model, not a law of markets. More realistic models may incorporate correlations among asset classes, regime changes, fat tails, inflation-linked returns and serial dependence.

The output should include a distribution of terminal wealth, minimum wealth, funded status or probability of meeting a goal. A “success probability” is conditional on model assumptions; it is not a guarantee. A model producing 85% success does not mean that the client has an objective 85% chance in the real world unless the assumptions adequately represent the relevant uncertainty.

Simulation quality depends on input quality, number of trials, treatment of fees and taxes, cash-flow timing, correlation assumptions and the definition of success. Increasing trials reduces simulation noise but cannot correct a biased model. Results should therefore be shown as ranges and accompanied by sensitivity and scenario analysis.

7. Integrated Illustrative Case

Background

The following fictional case illustrates plan evaluation. A 45-year-old client has $500,000 invested, contributes $30,000 at each year-end, and plans to retire at age 65. The strategic allocation is 70% equities and 30% bonds. The client’s priority is maintaining purchasing power while limiting a severe early-retirement loss.

Performance evidence

During the latest year, the portfolio returned 7.20% gross while its policy benchmark returned 6.40%. The attribution analysis above gives 0.40 percentage points from allocation, 0.20 from selection and 0.20 from interaction. The portfolio incurred 0.10 percentage points of management and transaction costs, so net portfolio return is 7.10% and net active return is 0.70 percentage points if the benchmark is stated gross of those costs. The report must state whether benchmark returns are gross or net and avoid comparing unlike bases.

The portfolio’s annualised volatility was 10.5%, the risk-free return was 3.0%, and the observed maximum drawdown was −12%. Its Sharpe ratio was (7.10% − 3.0%) ÷ 10.5% = 0.39. This is a descriptive statistic for the period, not a prediction.

Resilience analysis

The adviser tests three deterministic scenarios:

The one-year portfolio return under the recession shock is 0.70×(−25%) + 0.30×2% = −16.90%. On $500,000, ignoring contributions and taxes, wealth falls to $415,500. The loss is material but not necessarily plan-ending because the client is still working and contributing.

Sensitivity testing shows that postponing retirement by two years increases the accumulation period and may materially improve resilience, while a higher savings rate reduces dependence on investment returns. The adviser should also assess emergency liquidity, disability and long-term-care insurance, tax location, sequence-of-returns risk and the possibility that equity and credit losses occur together.

A Monte Carlo model could estimate the distribution of wealth at age 65 using correlated equity and bond returns, inflation, contributions, taxes and withdrawal rules. The adviser should report the assumptions, success definition and sensitivity results, then recommend monitoring triggers such as a funding ratio below a specified threshold, a sustained increase in withdrawals or a drawdown that changes the client’s risk capacity.

Learning points

The case demonstrates that strong relative performance does not automatically prove suitability. Attribution explains the latest active return; risk statistics describe the realised path; stress tests expose vulnerability; and simulation examines a range of futures. The final recommendation must integrate all four perspectives with the client’s objectives and capacity for loss.

8. Governance, Ethics and Technology

Performance reports should be reproducible, version-controlled and reviewed for data errors. Advisers must disclose calculation methodology, benchmark basis, fees, taxes, missing data and material assumptions. Selective presentation of favourable periods, survivorship-biased peer comparisons and unsupported claims about simulation probabilities are misleading.

Regulatory obligations vary by jurisdiction, but common principles include fair presentation, truthful performance advertising, suitability or best-interest duties, protection of personal data and retention of auditable records. Client risk tolerance should not be inferred solely from a questionnaire score; capacity for loss, time horizon and essential spending must also be considered.

Portfolio-management systems and AI tools can automate data cleaning, attribution, scenario generation and anomaly detection. They do not remove model risk. Human review is required for input validation, explanation of outputs, detection of biased training data and assessment of whether an apparently precise result is decision-useful. AI-generated recommendations should be traceable to approved assumptions and governed through access controls and independent testing.

9. Best-Practice Checklist

• Use total returns when comparing investments that distribute income.
• Separate external cash-flow effects from manager performance.
• Match return periodicity, risk-free rate and volatility units.
• Define the benchmark before evaluating results.
• Reconcile attribution algebraically and disclose the convention.
• Treat costs, taxes and inflation consistently.
• Combine scenarios, stress tests, sensitivities and simulation.
• Explain that model probabilities are conditional, not guarantees.
• Test liquidity and sequence risk, not only terminal wealth.
• Document assumptions, data sources, version changes and review dates.

Chapter Summary

Performance measurement begins with a correctly defined return and consistent treatment of distributions, cash flows, fees and taxes. TWR evaluates manager performance by removing external cash-flow timing, whereas MWR reflects the investor’s actual cash-flow experience. CAGR annualises cumulative total return and requires consistent treatment of distributed income.

Volatility, downside deviation, maximum drawdown and risk-adjusted ratios provide different views of risk. A benchmark must be relevant, investable or replicable, transparent and measured on a comparable basis. Single-period Brinson attribution decomposes active return into allocation, selection and interaction; with consistent inputs, these effects sum exactly to active return.

Plan evaluation extends beyond historical portfolio statistics. Scenario analysis communicates coherent futures, stress testing reveals vulnerabilities under severe conditions, sensitivity analysis identifies influential assumptions, and Monte Carlo simulation estimates distributions conditional on a model. Ethical evaluation requires transparency, reproducibility, appropriate disclosures and careful communication of uncertainty.

Key Terms

Holding-period return: Total gain or loss over a specified period, including relevant income.

Time-weighted return: Return measure designed to remove the effect of external cash-flow timing.

Money-weighted return: Cash-flow-sensitive return, commonly calculated as IRR.

CAGR: Annual compound rate corresponding to a cumulative return over multiple years.

Total return: Price appreciation plus income and distributions, with reinvestment treated consistently.

Volatility: Dispersion of periodic returns, commonly measured by standard deviation.

Downside deviation: Dispersion of returns below a specified minimum acceptable return.

Drawdown: Decline in wealth from a previous peak.

Maximum drawdown: Largest observed peak-to-trough decline.

Sharpe ratio: Excess return divided by total volatility.

Information ratio: Active return divided by tracking error.

Benchmark: Reference portfolio or index used for comparison.

Active return: Portfolio return minus benchmark return.

Attribution: Analysis explaining sources of active or absolute return.

Allocation effect: Active result associated with differing category weights.

Selection effect: Active result associated with differing category returns.

Interaction effect: Joint effect of differing weights and differing returns.

Scenario analysis: Evaluation of outcomes under coherent assumed conditions.

Stress testing: Evaluation under severe adverse but plausible conditions.

Monte Carlo simulation: Repeated random modelling of possible future paths.

Review Questions

Short Answer Questions

1. Why can TWR and MWR differ substantially for the same investor?
2. What is the difference between volatility and maximum drawdown?
3. State two characteristics of a suitable benchmark.
4. Why must distributed income be treated consistently in CAGR calculations?
5. Why is a Monte Carlo success probability not a guarantee?

Long Answer Questions

1. Explain the calculation, interpretation and limitations of the Sharpe ratio, Sortino ratio and information ratio.
2. Describe how an adviser should design a complete performance-reporting process for a multi-asset client.
3. Explain the single-period Brinson attribution framework and prove why its three effects reconcile to active return.
4. Compare scenario analysis, stress testing, sensitivity analysis and Monte Carlo simulation.
5. Discuss ethical and governance risks in presenting investment performance and financial-plan projections.

Analytical / Application Questions

1. A portfolio earns 9% before fees, pays 0.6% in fees and has a benchmark return of 8.0%. Calculate net active return and explain what additional information is needed before claiming manager skill.
2. A portfolio has 60% equities and 40% bonds. Under a stress scenario equities fall 30% and bonds rise 5%. Calculate the portfolio return and identify two further plan variables that should be tested.
3. A client’s wealth rises from $200,000 to $242,000 over three years, with no external cash flows and all income reinvested. Calculate CAGR and explain why it should be compared with a total-return benchmark.

Multiple Choice Questions with Answers

Question 1

A portfolio begins at $100,000 and ends at $106,000 after receiving $2,000 of income that is not included in the ending value. What is the holding-period return?

A. 6%

B. 8%

C. 10%

D. 12%

Correct Answer: B. 8%

Question 2

Which return measure is generally most appropriate for evaluating a portfolio manager while removing the effect of client deposits and withdrawals?

A. Money-weighted return

B. Time-weighted return

C. Simple dividend yield

D. Maximum drawdown

Correct Answer: B. Time-weighted return

Question 3

An investment grows from $100,000 to $121,000 in two years, with consistent treatment of income. What is its CAGR?

A. 8%

B. 10%

C. 11%

D. 21%

Correct Answer: B. 10%

Question 4

A portfolio earns 9%, the risk-free return is 3%, and volatility is 12%. What is the Sharpe ratio?

A. 0.25

B. 0.50

C. 0.75

D. 1.50

Correct Answer: B. 0.50

Question 5

Which measure records the largest decline from a previous wealth peak?

A. Arithmetic mean

B. Tracking error

C. Maximum drawdown

D. Money-weighted return

Correct Answer: C. Maximum drawdown

Question 6

Under the stated single-period Brinson framework, active return equals:

A. Allocation effect minus selection effect

B. Selection effect divided by interaction effect

C. Allocation effect plus selection effect plus interaction effect

D. Benchmark return plus portfolio volatility

Correct Answer: C. Allocation effect plus selection effect plus interaction effect

Question 7

A portfolio has 70% equities and 30% bonds. Equities return −25% and bonds return 2%. What is the portfolio return before fees?

A. −16.9%

B. −17.5%

C. −23.0%

D. −7.5%

Correct Answer: A. −16.9%

Question 8

Which technique changes one assumption, such as inflation, while holding other assumptions constant to identify influence?

A. Sensitivity analysis

B. Benchmark construction

C. Attribution reconciliation

D. Time weighting

Correct Answer: A. Sensitivity analysis

Question 9

In the wealth equation Wₜ₊₁ = Wₜ(1 + Rₜ) + Cₜ − Xₜ, what does Xₜ represent?

A. Portfolio volatility

B. Withdrawal

C. Benchmark weight

D. Inflation rate

Correct Answer: B. Withdrawal

Question 10

Which statement about Monte Carlo simulation is most accurate?

A. It guarantees the reported success probability.

B. It eliminates model risk through a large number of trials.

C. Its results are conditional on assumptions about returns, cash flows and other variables.

D. It measures only historical maximum drawdown.

Correct Answer: C. Its results are conditional on assumptions about returns, cash flows and other variables.