Data and analytics innovations to address emerging challenges in credit portfolio management

A survey of financial institutions shows that firms have made significant progress during the past several years in using new data and techniques for credit portfolio management, but also demonstrates that challenges remain around technology, talent, and integration of new use cases like climate and environmental, social, and governance (ESG) risk. In recent years, many financial institutions have increased their adoption of data and new technologies to manage credit portfolios. McKinsey and the International Association of Credit Portfolio Managers (IACPM) surveyed 44 financial institutions globally on the latest developments in data and analytics for credit portfolio management.

The objectives of the survey were to understand the use of traditional and alternative data sources for credit risk information, to determine how financial institutions use analytical approaches across portfolio segments, and to inform the path forward to incorporate next-generation data and analytics across the corporate and commercial real estate (CRE) portfolios of small and medium-size enterprises (SMEs).

More than 60 percent of respondents said that they have increased, over the past two years, their use of new types of data and deployment of advanced analytical techniques, like machine learning, for advanced credit portfolio management. An even larger portion of respondents (more than 75 percent) expect these trends to continue over the next two years.

Over the past several years, financial institutions have made significant progress in using new data and techniques for credit portfolio management.

As they look to deploy new analytics, companies are obtaining data from sources like automated client financials; internal credit behavior data and cross-product data from internal sources; and credit bureau, economic forecasts, and news data from external providers. This includes alternative data as well; for example, in the corporate portfolio, more than half of respondents are currently using, piloting, or considering news media, social media, or third-party account data. Relatively fewer banks are using internal cross-portfolio data on consumer-to-wholesale crossover accounts, with 44 percent of banks evaluating whether to use this data for their corporate portfolio and 38 percent for their SME portfolio (Exhibit 1).

Innovative external data sources are more used for corporate segment while small and medium-sized enterprises use more innovative internal data sources.

Given segment characteristics, implementation of new, data-intensive techniques like machine learning are focused on specific asset classes and specific use cases. Adoption of machine learning models is higher in SME segments than large corporates for a good reason. In SME portfolios, these models have found their use in credit scoring, early-warning-signal development, and credit pricing (Exhibit 2). In corporate asset classes, however, their usage is largely confined to early-warning-indicator development.

Machine learning models are primarily gaining traction for risk scoring of small and medium-sized enterprises and early warning across the board.

Banks that have fully automated decisions for a majority of the portfolio (more than 50 percent) are still relatively rare (about 11 percent for SME portfolios and about 4 percent for midmarket). However, for SME portfolios specifically, about 30 percent of respondents reported that they have automated more than 30 percent of their decisions. In addition, in the SME space, respondents report a significant benefit in turnaround time, with 37 percent of participants reporting a more than 10 percent decrease (Exhibit 3).

Where implemented, use of innovative data and/or advanced analytics improved turn-around-time for small and medium-sized enterprises.

The benefits of new data and analytics in the midmarket, corporate, and CRE spaces have not translated to a reduction in turnaround time to such an extent: only 13 percent, 3 percent, and 12 percent of banks that have automated some of their credit decisions across midmarket, corporate, and CRE portfolios, respectively, have seen more than a 10 percent decrease in turnaround time.

At the same time, the deployment of machine learning and other analytical techniques has brought several challenges to the forefront. Survey respondents noted various barriers to increased adoption of innovative data solutions and advanced analytical methods in credit portfolio management. This includes data quality assessment, talent availability, and difficulty in validating and explaining new techniques (Exhibit 4).

Data quality assessment and talent management are the top challenges for use of both advanced analytics and innovative data solution.

Climate and ESG risks are emerging as the next biggest challenges for credit portfolio management

Every survey participant was asked about the biggest challenges facing credit risk and portfolio management analytics in the next two to three years. Notable challenges included capital, provisioning and regulatory requirements for stress-testing models (58 percent), challenges posed by model uncertainties after the COVID-19 pandemic (51 percent ), and the incorporation of machine learning models within regulatory and risk constraints (42 percent). However, an overwhelming majority (86 percent) cited climate risk and ESG as the next big challenge (Exhibit 5).

86 percent of the participants believe that the climate, and environment, social and governance, is the next biggest challenge for credit assessment.

ESG risks are composed of environmental risks arising from operations and consumption of the output (that is, services and products) of the organization; social risks arising from how the organization treats people, including employees, customers, and the communities in which it operates; and governance risks arising from poor practices in the organization’s interactions with its shareholders, board, and management. These risk factors may have a positive or negative impact on the financial performance or solvency of an entity, sovereign, or individual.

Within this overall taxonomy, climate risk falls within the category of environmental risk, and is connected to both the direct and indirect effects of physical hazards associated with climate change. (Direct damages are caused by, for example, hazards like floods, wildfires, and hurricanes; indirect damages are consequences such as potential increases in insurance premiums and the effect on at-risk communities’ living standards.) It is also connected to transition risk: the policy, technology, and regulatory risk inherent in transitioning away from an economy overly reliant on activities that produce greenhouse gases.

Many financial institutions are now assessing their portfolio’s exposure to climate risk, either due to regulatory requirement, or to test the hypotheses that hurricanes, floods, blizzards, tornadoes, wildfires, and other natural hazards can inflict billions in damages across loan portfolios in any given year (Exhibit 6).

More than 50 percent of participants have implemented/or are planning to implement climate stress loss analyses.

Banks that have started climate stress testing are now considering whether to build new credit models or to tailor existing ones for stress testing. Respondents were evenly split into thirds, saying they are developing new loss models, using current models, or are not yet exploring. In addition, analyses on loss scenarios due to climate stress were more concentrated on midmarket, corporate, and CRE portfolios (more than 50 percent of banks for each), with fewer banks (less than 40 percent) conducting these analyses on SME portfolios.

The majority of institutions in Europe, the Middle East, and Africa are developing models internally or subscribing to vendor models to assess climate risk. Institutions in the Asia–Pacific region are the least advanced, and North American institutions fall somewhere in the middle.

Addressing climate risk will require a coherent framework

Our survey indicates that portfolio managers have only recently started to consider how climate and ESG risks affect risk identification and risk measurement, including obligor credit ratings. They now need new tools and processes to analyze climate stress loss and climate scenarios; they must also evaluate how climate risk assessment can be integrated with existing credit processes.

We have identified, based on discussions with survey participants, as well as our extensive work with banks, several important factors to consider about the material impacts of climate risk on credit:

  • Climate risk is typically concentrated in ‘pockets.’ We found both physical and transition risks lie in very targeted areas of the portfolio. To identify the pockets with a high concentration of climate risk impact, financial institutions need to perform a detailed heat mapping to focus their efforts on prioritized hazards for each of the high-risk portfolios. For example, it is commonly observed that most of the credit impact (around 70 to 80 percent of incremental impact) for real-estate-related asset classes comes from 10 to 20 percent of the obligors in the portfolio. This understanding is also reflected in the priorities identified by survey respondents.
  • The average credit impact can be moderate in the near term, but there is likely to be a high degree of obligor-level variability. We found that even in industries exposed to high physical and transition risk, the aggregate/average impact to the portfolio can be moderate. For example, in a portfolio of upstream oil and gas companies, the median impact might be an approximately 7 percent reduction in EBITDA. However, the difference between borrowers with maximum and minimum impact can be stark. In this example of the upstream oil and gas industry, there are several companies with up to a 40 percent negative impact on EBITDA, while others experience a positive impact on EBITDA in some scenarios due to reallocation of oil and gas demand. Financial institutions have started evaluating these impacts and plan to explore them further, indicating there is still a long way to go.
  • For industries exposed to physical hazards, most risk is in knock-on impacts, not direct damages. The near-term credit impact of direct damages is typically covered through insurance in industries like real estate (both commercial and retail). However, the knock-on effects can dwarf direct impacts, and any assessment of material risk drivers would include higher insurance payments and the deterioration in living standards in a community, even though the property itself might not be damaged.
  • Unmanaged climate risk can have a tangible impact on returns and economic profit. Comprehensive capital analysis and review (CCAR), stress tests mandated by the European Central Bank (ECB), or methodologies driven by regulatory capital might not be appropriate for climate risk assessment. These are focused on capital risks and can underestimate the credit impact on single obligors. Climate risk assessment requires understanding returns from new climate-oriented businesses and obligor-specific scenario analyses. If these are not done well, the impact can be significant. At one North American bank, we identified a 35 percent potential erosion of profits by 2030 in the absence of action on key pockets of climate risk exposure.
  • Before addressing and mitigating climate risk, financial institutions must address several barriers related to capabilities, data, and analytics. First, financial institutions need internal alignment on their climate ambitions and aspirations to gain stakeholders’ buy-in and to ensure collaboration with relevant board committees. Financial institutions must also acquire technical capabilities and educate themselves by getting familiar with topics related to climate science, risk-assessment methodologies, and the complex design choices related to net-zero targets and their impact on credit assessment. As many existing risk-assessment tools were not built for the requirements of climate assessment, financial institutions will need an open architecture that can support new methodologies for data quality, standardization, and collection. Finally, to capture and address the holistic impact of climate risk on the portfolio, financial institutions need to increase their focus on interdisciplinary skills and mobilization across credit, front-line, and model risk management.

Before addressing and mitigating climate risk, financial institutions must address several barriers related to capabilities, data, and analytics.

To overcome these barriers, financial institutions need to make significant progress in two important approaches to climate risk assessment: climate scenario analyses and integration of climate into credit processes. Evaluation of data sources that can be used in scenario analyses and credit assessment, as well as analytics that help provide transparency on the effects of potential climate risk, will also help financial institutions understand the evolving data and vendor landscape.

Scenario analyses to start the climate-risk-assessment journey

Scenario analyses help financial institutions understand and quantify pockets of climate risk exposure. Results from scenario analyses can include a portfolio’s expected loss range under different transition- and physical-risk climate scenarios in the short, medium, and long term. Typically, they would also bolster the financial institution’s understanding of climate drivers and transmission channels and how they interact with obligor-level credit risk factors.

Scenario analyses consist of three major steps (Exhibit 7):

Using a phased framework to develop bespoke scenario impact assessment tool for CRE portfolios.
  1. Identify risks across portfolios. The first phase is to identify the hazards, such as floods, hurricanes, or wildfires, that are most relevant to the portfolio. In addition to physical hazards, the potential macroeconomic impact to different regions (for example, the loss of economic output in counties with high fossil-fuel dependencies), and regional exposure can also feed into the risk-identification process. One of the best ways to illustrate this risk is by using a “climate risk heat map” or an “exposure at risk” metric to understand how much of a portfolio is exposed to climate-related vulnerabilities. For example, a financial institution might determine through this exercise that it is exposed to wildfire and coastal flooding due to its geographical concentration in California.
  2. Assess impact on obligor financials. The next phase is to connect these risks to how they can affect financial ratios (for example, net operating income or property values for a CRE portfolio) or additional drivers of creditworthiness (for example, additional sponsor risk for CRE). At this point, it is important to begin testing detailed scenario analyses for each portfolio while understanding different climate scenarios and their effect on hazards and macroeconomic factors. The transmission mechanism of translating this impact to obligor financials can be complex and must factor in, among other things, uninsured damages, business interruption, and increased insurance costs. As empirical validations may not be possible at this stage, financial institutions may also want to leave ample room for sensitivity analyses of assumptions.
  3. Quantify the effect of change in obligor financials on the portfolio. The final phase integrates obligor-level analysis into credit loss assessment for portfolios. It is important to develop a transparent framework to link changes in obligor financials to changes in credit rating or probability-of-default (PD) and loss-given-default (LGD) parameters. We have found that banks in particular can use existing underwriting or loss-forecasting models (for example, CCAR, FASB’s current expected credit loss model; International Financial Reporting Standards 9; and ECB stress tests) with some modifications to inform this phase.

Integration of climate with credit assessment process

As financial institutions develop their climate-risk-assessment capability through risk identification and climate scenario analyses, the next step includes developing an approach to credit decision making that ensures climate risks are appropriately and sufficiently considered in credit portfolio construction and management. To achieve this, process changes must be implemented and methodology gaps closed to incorporate climate risk quantitative analysis into the credit adjudication process.

We believe that designing and piloting a “climate risk scorecard” that uses knowledge gathered during the risk identification and scenario analyses phase will be critical to this effort. For example, a climate risk scorecard for a high-priority portfolio like CRE would have the following modules:

  • Module 1: Prescreening filter. This is an initial heuristic-based assessment of an area’s vulnerability to climate events to identify properties that require further assessment. Metrics to assess physical and transition risk include flood inundation depth (property level), fire risk zone (property level), and the percent of oil and gas and utilities-related sectors in local GDP (for example, at the county level in the United States). Properties are sorted into high, medium, and low risk, and only properties identified as highly vulnerable go through the next two modules.
  • Module 2: Scenario analyses tool. This tool provides a quantitative estimate of change in credit risk under different climate scenarios for obligors that have been flagged as high risk in the previous prescreening module. The change in credit risk parameters can be reflected in a separate “climate score.”
  • Module 3: Property-level climate scorecard. In this module, a detailed client questionnaire, designed for each sector, enables the qualitative assessment of factors not captured in module two to adjust the climate score (through client questionnaire and a first- and second-line assessment). A detailed client questionnaire is designed for each sector with graded response options and potential source of information. The output is the adjusted climate risk score.

While practices in climate risk assessment are still very much evolving, we believe it will track the methodology laid out in Exhibit 8.

There are 8 steps for banks to take to integrate climate into credit processes.

The research summarized in this article highlights the benefits and challenges of incorporating new data sources and analytical techniques into the various aspects of credit risk and credit portfolio management. The potential upside should motivate institutions to maintain and intensify their efforts, as most expect to do in the near term.

Institutions are rapidly implementing methodologies to assess the credit implications of climate risk. This article describes the benefits of developing a detailed, use-case-driven understanding of the necessary climate data, the technology infrastructure requirements for storage and processing, the reporting requirement for risk assessment, and the best orchestration model across functions. It also outlines the three immediate steps those institutions should take. The lessons learned in implementing data- and analytics-driven approaches to address credit risk assessment, as captured in this survey, inform what credit institutions must do to meet the challenges of today’s risk landscape.

Explore a career with us