DFAST and CCAR require bank holding companies to report, in detail, how they would respond to hypothetical market scenarios that represent macroeconomic shocks like a housing meltdown or a stock market crash. The data used by each company to predict losses and create a response plan must be actual data, not approximated. Using the commercial real estate market as an example, this post explains how a location-based BI on Hadoop solution provides analysts and portfolio managers with a flexible way to use microeconomic real estate market factors as the foundation to a robust and actionable response to regulators.
Some Background on DFAST and CCAR
On June 21, 2018, the Board of Governors of the Federal Reserve System (the Fed) released the 2018 results of the Dodd-Frank Act Stress Test (DFAST). DFAST is a quantitative exercise, testing whether a bank holding company (BHC) can maintain the minimum capital ratio of 4.5% deemed sufficient to absorb losses, support operations, and maintain lending during an adverse economic scenario like a housing or stock market crash. All BHCs passed 2018 DFAST (see accompanying chart).
A week later, June 28th, the Fed followed up with the results of the Comprehensive Capital Analysis and Review (CCAR). CCAR is a qualitative evaluation of the capital plans put forth by the largest BHCs. The capital plan could include a payout of dividends, for example, but the Fed could nix that idea if they determine that the plan won’t retain enough capital to address an economic shock. The accompanying headline demonstrates this.
Hypothetical Market Scenarios
DFAST and CCAR are based on two market scenarios: an “adverse” scenario and a “severely adverse” scenario.
Each scenario tracks the trajectories of hypothetical macroeconomic variables that capture economic activity, asset prices, and interest rates in the financial markets and the overall U.S. economy. They also track GDP growth, inflation, and currency exchange rates against the dollar for four global regions. The accompanying chart of assumed unemployment rates shows the stark difference between the “adverse” and “severely adverse” scenarios.
The effect of the commercial real estate (CRE) market on our economy is considerable and this is reflected in the DFAST/CCAR market scenarios. CRE price drops are significant factors in the scenario models while the projected CRE losses for each BHC is a required reporting item.
Link the Hypothetical to the Actionable through CRE Market Factors
As noted above, the DFAST/CCAR market scenarios are hypothetical and based on macroeconomic variables. However, even though the scenarios are “made up,” the data used by each BHC to predict losses and create a response plan must be actual and defensible. A key Fed guideline states (my emphasis), “Hypothetical behavioral responses by BHC management should not be considered as mitigating factors for the purposes of this analysis.”
The link between hypothetical scenarios and actionable data is critical for BHCs and any financial institution for that matter. Most, if not all regulatory regimes, require that firms be able to defend actions, assumptions, and plans with complete, and accurate data, often at the most granular level. Evidence is presented by exact, not approximated data, whether it be the qualitative aspects of best trade execution, understanding the intent of a trade as it pertains to trade surveillance, or proving capital adequacy plans.
A 2011 Moody’s Analytics report describes a methodology that links stress test macroeconomic assumptions to real estate market factors. The accompanying table shows an example of how they linked macroeconomic GDP growth to net operating income (NOI), a CRE market factor. This process was applied to estimating CRE stressed loss measures under the 2017 CCAR.
The link to real estate market factors is important because it is at that point you can start applying actionable data. For example, NOI is impacted by several factors, including economic vitality, changes in demographics, crime, location desirability, and several other aspects. Below we describe those types of data.
Apply Location-Based Data
Tying geo-demographic data sets such as these to individual properties is an effective way for practitioners to supplement their quantitative analysis and to gain a better understanding of what’s going on around their properties. The diagram below exemplifies this.
The red box represents a bank’s CRE portfolio and internal analytics. The orange boxes represent data sets that can enrich your analysis, down to a single property level if necessary. Below describes a few.
- Financial stress scores: Predict the likelihood that a business will cease operations without paying all creditors, go into receivership, reorganize, and other scenarios.
- Delinquency scores: Predict the likelihood that a company will pay in a severely delinquent manner, seek legal relief from creditors, and other actions.
- Points of interest: Facilitate a thorough review of business information, from analysis to action with location and detailed company information for a diverse set of entities.
- Consumer segmentation: Link address information to demographic, lifestyle and socio-economic insight to better understand who lives where and how they behave.
- Consumer vitality: Understand the desirability of neighborhoods, their economic performance, and how they retain and attract consumers.
- Demographics: provides comprehensive coverage of global demographics such as population, age, sex, and race as they relate to local geographies.
The grey boxes are just as interesting as the colored boxes. They represent the flexibility of integrating additional data sets with your portfolio on an as-needed basis because not every scenario requires every data set. The example below shows how a CRE analyst visualizes a combination of crime data, customer vitality indices, financial stress, delinquency scores, and predictive population growth/decline in one view.
Move beyond Traditional Data Platforms and Business Intelligence Tools
Navigating the process above requires a combination of human intuition, subject matter expertise, data processing, and advanced analytics. It also requires you to move beyond traditional data platforms and business intelligence tools. Driving insights from legacy systems have proven to be slow, cumbersome, and limited to static or stale property data, making it difficult to respond to regulatory requests and to win in a highly competitive market.
The data flexibility referred to above is economically feasible with modern big data environments such as “data lakes”, which are simply data storage and processing systems, typically built on the technology known as Apache Hadoop. Data lakes process more varieties and higher volumes of data at a significantly lower cost than traditional data platforms such as relational databases or data warehouses. To make the most of these advantages, the visual analytics engines of business intelligence tools must be native to the lake, installed directly on the data, otherwise the data needs to be moved to external visualization servers, which is the case with traditional business intelligence tools.
A modern big data environment with native visual analytics enables business users to economically put all the data to use (structured, unstructured, historical, real-time, and/or data generated by machine learning and AI). Data integrity is ensured and time-to-insight accelerated because there is no data transformation overhead and teams collaborate on the same version.
See this four-minute primer that demonstrates how a portfolio manager uses location-based risk management for CRE. Arcadia Data native visual analytics on Pitney Bowes Spectrum for Big Data accelerates the ability to aggregate, geospatially organize, and enrich property information to enhance quantitative analysis. Through these next-generation technologies, your portfolio managers and analysts can quickly visualize and take action on risks and opportunities across their entire CRE portfolio while establishing a robust and defensible process that satisfies regulators.