Stress Testing: The Macro vs. Micro Battle in Risk Management
Financial stress testing has become a cornerstone of risk management, providing financial institutions and regulators with critical insights into potential vulnerabilities under adverse conditions. There are two primary approaches to stress testing: Top-Down (macro) and Bottom-Up (micro). The Top-Down method, driven by regulatory bodies, evaluates systemic stability, whereas the Bottom-Up approach is employed by individual institutions to assess firm-specific risks.
The Macro (Top-Down) Approach to Stress Testing
Regulators like the Federal Reserve, the European Central Bank (ECB), and the International Monetary Fund (IMF) impose top-down stress testing to evaluate the resilience of the financial system under severe yet plausible scenarios. This approach leverages macroeconomic indicators such as GDP decline, rising unemployment, and heightened market volatility.
For example, in the 2023 CCAR (Comprehensive Capital Analysis and Review) results, the Federal Reserve modeled a severe downturn, projecting a 10% unemployment rate, a 6.5% GDP contraction, and a 45% decline in equity markets. Despite significant losses, major banks were required to keep their Common Equity Tier 1 (CET1) ratios above 4.5%.
As the highest quality of regulatory capital, banks are required to maintain a minimum CET1 ratio (CET1 capital as a percentage of risk-weighted assets), typically 4.5%, with additional buffers imposed by regulators. CET1 primarily consists of common stocks, retained earnings, regulatory adjustments (such as deductions for goodwill and intangible assets, and so on), and other comprehensive income (unrealized gains and losses on certain financial assets that are not included in net income but still affect a bank’s equity).
Top-down analysis can also be used to shape risk-adjusted business strategies. Banks that underperform in stress scenarios may opt to raise additional capital, reduce risk-weighted assets, or adjust liquidity buffers.
For example, after the 2020 stress tests, JPMorgan Chase adjusted its share buyback program to preserve capital in anticipation of economic volatility, demonstrating how stress tests influence executive decision-making (2020 Annual Stress Test Disclosure, JPMorgan Chase).
Pros and Cons of Top-Down Approach
Regulators use top-down stress testing to detect systemic risks that may lead to widespread financial instability. By evaluating potential contagion effects, this method uncovers vulnerabilities that individual firm assessments might overlook. It played a crucial role in assessing financial weaknesses before and after the 2008 financial crisis.
Regulators employ standardized models and assumptions, facilitating comparisons across financial institutions. This uniformity enhances transparency, enabling policymakers and market participants to assess financial stability across various banks and jurisdictions.
Top-down stress testing helps ensure that banks uphold sufficient capital reserves in accordance with regulatory frameworks like Basel Norms. By enforcing capital requirements, regulators strengthen confidence in the financial system and reduce risks tied to economic downturns.
The Top-Down Approach does have limitations. Since top-down models take a generalized approach to risk exposure, they may miss institution-specific vulnerabilities. This can result in an incomplete evaluation of a bank’s financial health, especially for those with distinctive asset structures and risk characteristics.
Additionally, since stress test results adopting a top-down approach are published periodically, they may not respond swiftly to sudden market disruptions, reducing their effectiveness in addressing rapidly evolving economic and financial conditions.
The Micro (Bottom-Up) Approach to Stress Testing
Unlike regulatory stress tests that take a macroeconomic perspective, internal stress tests focus on an institution’s unique portfolio composition. Financial institutions conduct bottom-up stress testing using proprietary models tailored to their specific asset portfolios, operational risks, and capital structures to analyze credit risk, market risk, and liquidity exposure at the granular level.
Pros and Cons of Bottom-Up Approach
Bottom-up stress testing enables firms to tailor stress scenarios to their unique business models and exposures. By adapting scenarios to changing market conditions, institutions ensure that risk assessments stay accurate and up to date.
By analyzing stress scenarios at a granular level, firms can identify portfolio vulnerabilities more effectively. This deeper insight supports better capital allocation and strengthens risk management strategies, enabling proactive responses to potential financial distress.
On the downside, Bottom-up stress testing relies on proprietary models, which can create inconsistencies across firms and result in varying risk assessments. Moreover, firms may design models that present their financial position more favorably, raising concerns about bias and potential model manipulation.
Regulators often scrutinize bottom-up stress tests due to concerns that banks may underestimate risks to meet capital adequacy requirements. This can result in regulatory interventions and increased compliance obligations for financial institutions.
Comparative Analysis
Numerical Examples
The examples presented here are simplified representations, designed to provide non-banking and non-financial professionals with a foundational understanding of financial stress testing, rather than replicating the intricate interdependencies within the real-world banking system.
Example I – Top-Down Stress Test
An illustration of a top-down stress test within the Comprehensive Capital Analysis and Review (CCAR) framework.
Scenario 1 (Baseline) – Top-Down ST with Projected Loan and Trading Loss of $6 billion
Macroeconomic Shock
- GDP contracts by 6%
- Unemployment rises to 10%
- Stock market declines by 40%
- Corporate bond spreads widen by 300 basis points (3%)
Bank Financials (Before Stress)
- Risk-Weighted Assets (RWA) = $500 billion
- Tier 1 Capital Ratio = 12%
- Common Equity Tier 1 (CET1) Ratio = 10%
- Regulatory Minimum Tier 1 Capital Ratio = 8%
- Regulatory Minimum CET1 Ratio = 4.5%
Projected Loss
- Projected Loan & Trading Losses = $6 billion
Computations
- Initial Capital Levels
- Initial Tier 1 Capital = 12% x 500 billion = 60 billion
- Initial CET1 Capital = 10% × 500 billion = 50 billion
- Capital Depletion Due to Losses
- New Tier 1 Capital = 60 billion - 6 billion = 54 billion
- New CET1 Capital = 50 billion - 6 billion = 44 billion
- Post-Stress Ratios
- New Tier 1 Capital Ratio = (54/500) x 100 = 10.8%
- New CET1 Ratio = (44/500) x 100 = 8.8%
- Stress Test Results (Baseline)
Since the bank maintains capital above regulatory minimums, it is deemed resilient under stress.
Scenario 2 (Severe) – Extended Top-Down ST with Liquidity & Additional Risk Factors
Macroeconomic Shock
- GDP contracts by 6%
- Unemployment rises to 10%
- Stock market declines by 40%
- Corporate bond spreads widen by 300 basis points (3%)
- Funding markets tighten, leading to higher borrowing costs
Bank Financials (Before Stress)
- Risk-Weighted Assets (RWA) = $500 billion
- Tier 1 Capital Ratio = 12% ($60 billion)
- Common Equity Tier 1 (CET1) Ratio = 10% ($50 billion)
- Regulatory Minimum Tier 1 Capital Ratio = 8%
- Regulatory Minimum CET1 Ratio = 4.5%
- Liquidity Coverage Ratio (LCR) Before Stress = 120%
- High-Quality Liquid Assets (HQLA) Before Stress = $80 billion
Liquidity Coverage Ratio (LCR) is a key measure in financial stress testing, ensuring that banks hold sufficient high-quality liquid assets (HQLA) to endure a 30-day liquidity crisis. Regulators generally mandate a minimum LCR of 100%, requiring banks to maintain enough liquidity to navigate short-term financial disruptions.
HQLA comprises assets that can be readily converted into cash with minimal depreciation, including government bonds and treasury securities (which retain full value), as well as highly rated corporate bonds (which may be subject to some value reduction).
Projected Loss
- Projected Loan & Trading Losses = $6 billion
Liquidity Stress Scenario
Due to tightening funding conditions:
- Depositor withdrawals increase by 20%
- Short-term funding costs increase by 100 basis points (1%)
- HQLA declines by 15% due to market volatility
Market Risk Impact
Due to equity market declines and bond spread widening, the bank faces:
- Trading portfolio losses of $2 billion
- Fair value losses on held-for-sale bonds of $1.5 billion
Held-for-sale (HFS) bonds are debt securities that a company or financial institution intends to sell before maturity rather than hold until they mature. These are often used to maintain flexibility in investment portfolios while avoiding frequent trading. For example, a bank buys government bonds but plans to sell them within a year based on interest rate movements.
Computations
- Post-Stress Capital Ratios These remain the same as in the baseline scenario (see Capital Ratios above).
- Post-Stress HQL New HQLA = 80 billion - (15% × 80 billion) = 68 billion
- Post-Stress Liquidity Coverage Ratio (LCR) New LCR = (New HQLA) / (Net Cash Outflows) x 100% If we assume that net cash outflows increase to $65 billion due to stress, then New LCR = (68 billion) / (65 billion} x 100 = 104.6%
- New CET1 Capital after including Market Risk Losses New CET1 Capital = 50 billion - 6 billion - (2 billion + 1.5 billion) = 40.5 billion
- Post-Stress CET1 Ratio after including Market Risk Losses = (40.5 billion) / (500 billion) ×100 = 8.1%
- Stress Test Results (With Liquidity and Additional Risks):
- The bank maintains capital and liquidity above regulatory thresholds, demonstrating resilience.
- Market risk losses impact CET1 but do not push it below minimum requirements.
- Liquidity stress causes an LCR drop, but it remains compliant.
Scenario 3 (Severe) –Top-Down ST with Counterparty & Interest Rate Risk Impact
Macroeconomic Shock
- GDP contracts by 6%
- Unemployment rises to 10%
- Stock market declines by 40%
- Corporate bond spreads widen by 300 basis points (3%)
- Funding markets tighten, increasing short-term borrowing costs
- Key counterparty defaults, leading to additional credit losses
- Interest rates drop by 200 basis points (2%), impacting net interest margin
Bank Financials (Before Stress)
- Risk-Weighted Assets (RWA) = $500 billion
- Tier 1 Capital Ratio = 12% ($60 billion)
- Common Equity Tier 1 (CET1) Ratio = 10% ($50 billion)
- Liquidity Coverage Ratio (LCR) Before Stress = 120%
- High-Quality Liquid Assets (HQLA) Before Stress = $80 billion
- Projected Loan & Trading Losses = $6 billion
- Projected Market & Trading Losses = $3.5 billion
- Regulatory Minimum Tier 1 Capital Ratio = 8%
- Regulatory Minimum CET1 Ratio = 4.5%
Counterparty Default Scenario
In banking, a counterparty refers to the other party involved in a financial transaction. This could be an individual, corporation, financial institution, or government entity. Counterparty Risk refers to the risk that the other party in a transaction fails to meet its financial obligations, potentially leading to losses for the bank.
In the present situation, let us assume a major counterparty (a hedge fund or large corporate borrower) defaults, resulting in:
- $4 billion in additional credit losses
- $2 billion in derivative valuation adjustments
Falling Interest Rates (2% drop)
- Net interest income declines as loan yields fall faster than deposit rates
- $3 billion in lost net interest income over the stress horizon
- Market value of fixed-income securities rises, offsetting losses slightly
Computations
- Revised CET1 Capital After Counterparty Losses
= 40.5 billion - (4 billion + 2 billion) = 34.5 billion - Revised CET1 Ratio After Counterparty Losses
= (34.5 billion) / (500 billion) x 100 = 6.9% - Revised CET1 Capital After Interest Rate Impact
= 34.5 billion - 3 billion = 31.5 billion - Revised CET1 Ratio After Interest Rate Impact
= (31.5 billion) / (500 billion) x 100 = 6.3%
Final Comprehensive Top-Down Stress Test Results
Example II – Bottom-Up Stress Test
Scenario 1 - Stress Testing a Bank’s Mortgage Loan Portfolio
A bank holds a $2 billion mortgage loan portfolio and performs a bottom-up stress test to evaluate how an economic downturn would affect loan losses at the individual loan level.
Stress Scenarios
- Baseline Scenario (No stress):
- Default rate = 2%
- Loss Given Default (LGD) = 30%
- Severe Recession Scenario (Stressed):
- Unemployment rises to 10%
- Housing prices drop by 20%
- Default rate increases to 8%
- LGD increases to 50%
Computations
Expected Loss (EL) = Exposure at Default (EAD) × Default Rate × Loss Given Default (LGD)
-
- Baseline Scenario:
EL = 2 billion × 2% × 30% = 12 million - Severe Recession Scenario:
EL = 2 billion × 8% × 50% = 80 million
- Baseline Scenario:
Therefore, under a severe recession, expected losses rise to $80 million, a 566% increase. The bank can use these results to determine if it has enough capital reserves.
Scenario 2 - Stress Testing a Bank’s Loan Portfolio Across Sectors
Let's extend our bottom-up stress test with sector-wise analysis and Monte Carlo simulation to make it more realistic.
Monte Carlo simulation is employed here because it enables stochastic modeling of expected losses under varying economic conditions. Unlike deterministic models that rely on fixed assumptions, Monte Carlo simulations incorporate uncertainty and variability in key risk factors, offering a more comprehensive assessment of potential losses.
For instance, rather than using fixed Default rates, we simulate them using normal distributions to generate thousands of possible outcomes across different economic scenarios.
Similarly, Loss Given Default (LGD) tends to exhibit non-linear behavior in stress conditions, making it essential to model various possible values using the Beta distribution to better capture tail risks.
Finally, the simulated results are used to estimate the 95th percentile worst-case losses (Value at Risk (VaR)), helping determine the regulatory capital required to mitigate potential losses in a severe downturn.
Let us assume that the bank’s total loan portfolio of $2 billion is divided into three sectors as displayed below. The fixed Default and LGD rates are as follows:
Computations
Expected Loss (EL) for Each Sector with Fixed Default and LGD Rates.
- Baseline Expected Loss
- Mortgages: 1 billion × 2% × 30% = 6 million
- Corporate: 700 million × 3% × 40% = 8.4 million
- SME Loans: 300 million × 5% × 50% = 7.5 millionTotal Baseline Loss: $21.9 million
- Stressed Expected Loss
- Mortgages: 1 billion × 8% × 50% = 40 million
- Corporate: 700 million × 10% × 60% = 42 million
- SME Loans: 300 million x 15% × 70% = 31.5 millionTotal Stressed Loss: $113.5 million
Monte Carlo Simulation for More Realistic Estimations
Step 1 - We first Simulate Default Rates and LGDs for three sectors using appropriate distributions.
-
- Default Rate Simulation Distribution (in %)
- Mortgages: Normal(2, 1) in Baseline, Normal(8, 2) in Stress
- Corporate: Normal(3, 1.5) in Baseline, Normal(10, 3) in Stress
- SME Loans: Normal(5, 2) in Baseline, Normal(15, 5) in Stress
- Default Rate Simulation Distribution (in %)
The settings for simulating default rates could be justified depending on the context.
For instance, when simulating mortgage default rates, we assume a normal distribution with the following parameters: Normal(2,1) for Baseline and Normal(8,2) for Stress.
Does a 2% mean default rate appropriately represent baseline conditions? Under normal economic circumstances, mortgage default rates are typically low - often below 2% for prime borrowers but higher for subprime loans.
Likewise, is an 8% mean default rate a suitable assumption for stress scenarios? In severe economic downturns, such as the 2008 financial crisis, mortgage default rates have surged, though the extent depends on the portfolio's risk profile.
A fourfold increase (from 2% to 8%) might be reasonable in a deep recession, but analyzing past crises is crucial to assess their impact on default rates for similar portfolios.
Additionally, a standard deviation of 1 in Baseline implies a relatively narrow distribution, with 68% of cases falling between 1% and 3% (given the mean of 2%). This may be appropriate if default rates remain stable.
In contrast, a standard deviation of 2 in Stress suggests greater variability, with 68% of cases ranging from 6% to 10%, reflecting increased uncertainty under adverse conditions.
- LGD Simulation Distribution (in %)
- Mortgages: Beta(2,5) in Baseline, Beta(3,4) in Stress
- Corporate: Beta(3,4) in Baseline, Beta(5,3) in Stress
- SME Loans: Beta(4,3) in Baseline, Beta(6,2) in Stress
The Beta(α, β) distribution has a mean of α / (α + β). For mortgages under baseline conditions, Beta(2,5) results in a mean LGD of 2 / (2+5) = 0.286 (28.6%), indicating relatively low loss severity. This aligns with the lower LGDs typically observed in stable economic environments. The right-skewed distribution suggests that most values will cluster toward the lower end.
In contrast, under stress conditions, Beta(3,4) produces a mean LGD of 3 / (3+4) = 0.429 (42.9%), reflecting increased loss severity. This is consistent with higher LGDs observed during downturns, driven by declining property values and liquidity constraints. The shift to Beta(3,4) reduces right skewness and increases the likelihood of higher LGDs.
Step 2 - Calculate Expected Loss (EL) across 10,000 iterations.
Step 3 - Plot the distribution of Expected Losses.
Step 4 - Compute the 95th percentile worst-case loss for risk assessment.
SAS Code for Monte Carlo Simulation of Default Rates and LGDs
ods graphics / imagemap=on;
/* Set up parameters */
%let num_simulations = 10000;
%let exposures_mortgage = 1000;
%let exposures_corporate = 700;
%let exposures_sme = 300;
data monte_carlo;
call streaminit(12345); /* Set seed for reproducibility */
do sim = 1 to &num_simulations;
/* Generate Default Rates */
dr_mortgage_baseline = rand(“NORMAL”, 2, 1);
dr_corporate_baseline = rand(“NORMAL”, 3, 1.5);
dr_sme_baseline = rand(“NORMAL”, 5, 2);
dr_mortgage_stress = rand(“NORMAL”, 8, 2);
dr_corporate_stress = rand(“NORMAL”, 10, 3);
dr_sme_stress = rand(“NORMAL”, 15, 5);
/* Generate LGDs */
lgd_mortgage_baseline = rand(“BETA”, 2, 5) * 100;
lgd_corporate_baseline = rand(“BETA”, 3, 4) * 100;
lgd_sme_baseline = rand(“BETA”, 4, 3) * 100;
lgd_mortgage_stress = rand(“BETA”, 3, 4) * 100;
lgd_corporate_stress = rand(“BETA”, 5, 3) * 100;
lgd_sme_stress = rand(“BETA”, 6, 2) * 100;
/* Calculate Expected Loss (EL) */
el_baseline = (&exposures_mortgage * dr_mortgage_baseline * lgd_mortgage_baseline +
&exposures_corporate * dr_corporate_baseline * lgd_corporate_baseline +
&exposures_sme * dr_sme_baseline * lgd_sme_baseline) / 100;
el_stress = (&exposures_mortgage * dr_mortgage_stress * lgd_mortgage_stress +
&exposures_corporate * dr_corporate_stress * lgd_corporate_stress +
&exposures_sme * dr_sme_stress * lgd_sme_stress) / 100;
output;
end;
run;
/* Compute 95th percentile worst-case loss */
proc univariate data=monte_carlo cipctldf alpha=0.05;
var el_stress;
output out=percentiles p95=worst_case_loss;
run;
proc print data=percentiles label;
var worst_case_loss;
label worst_case_loss = “95% Worst-Case Expected Loss ($M)”;
run;
/* Capturing the value of worst_case_total in the macro variable worstcase_totalloss */
proc sql;
select worst_case_loss into :worstcase_loss from work.percentiles;
quit;
/* Plot the distribution of Expected Loss */
proc sgplot data=monte_carlo;
histogram el_baseline / transparency=0.5 fillattrs=(color=’#40E0D0′);
histogram el_stress / transparency=0.5 fillattrs=(color=’#6495ED’);
refline &worstcase_loss / axis=x lineattrs=(color=’#DE3163′) label=”95% Worst Case”;
keylegend / location=inside position=topright;
xaxis label=”Total Expected Loss ($M)”;
yaxis label=”Frequency”;
title “Overall Expected Loss Distribution – Baseline vs. Stress”;
run;
Results
Select any image to see a larger version.
Mobile users: To view the images, select the "Full" version at the bottom of the page.
- Baseline vs. Stressed Loss Distribution
- The EL distribution under stress shifts rightward, indicating higher potential losses.
- Most likely loss in stress scenario is $6.858 billion - $15.907 billion.
- 95% worst-case loss: $15.907 billion.
- Regulatory Implications
- The bank may need additional capital buffers.
Scenario 3 - Monte Carlo Simulation with Sectoral Breakdown & Scenario-Based Risk Capital Calculation
- In this situation, we first generate separate distributions for Mortgages, Corporate, and SME Loans. This allows us to compute 95th percentile loss for Mortgages, Corporate, SME Loans, and Total Portfolio.
- Next, we plot separate histograms for each sector's expected loss distribution in stress scenarios.
- Scenario-Based Risk Capital Calculation - Computes Capital at Risk (CaR) based on 95th percentile loss.
- Finally, we aggregate the total loss distribution and also apply the worst-case scenario overlay. This Provides clear visualization of stress impact.
Results
Additional Information
For more information on SAS Stress Testing solution visit the software information page here.
For more information on curated learnings paths on SAS Solutions and SAS Viya, visit the SAS Training page. You can also browse the catalog of SAS courses here.
Find more articles from SAS Global Enablement and Learning here.
