Banks need to build numerous models and try to separate bank-specific decisions from macroeconomic effects. Projecting the balance sheet and income statement under the most likely scenarios is certainly no easy task—and banks cannot rely solely on internal performance data.
When stress testing needs arise from an ad hoc situation such as the COVID-19 pandemic, the challenge is even more complex as the process lies outside of business as usual. Timelines are compressed and validated models might not be adapted for the specific stress.
In this paper, we discuss an approach that addresses those issues. We propose an alternative simple, coherent methodology that allows us to forecast and stress test the entire balance sheet and profit and loss statement for all of the roughly 6,000 banks in the United States consistently. This methodology can be used as a primary approach for banks without the means to produce such stress testing exercises, or as a challenger or benchmark to validate the results of a set of primary models.
Our approach is also useful for strategic planning as it allows one bank to compare its balance sheet and income statement to its peers and the industry, and to explore potential mergers and acquisitions.
Introduction
Moody’s Analytics has always been on the front line of quantitative analysis to support loss forecasting and forecasting the other line items of the balance sheet and income statement. A few years ago, at the peak of the Comprehensive Capital Analysis and Review (CCAR) exercise, Moody’s Analytics helped customers develop models and execute simulations to produce regulatory reports such as the FR-Y14. We have harnessed this experience into Capital Risk Analyzer,1 a capital planning and stress testing solution that offers both a cloud-based platform to execute champion and challenger forecasting models, and off-the-shelf models to automate stress testing. In this paper, we will discuss one of the approaches in this solution based on a top-down model, the Call Report Forecast (CRF).2 This model uses publicly available historical data through the FDIC call report to derive forecasts for industry-level aggregates based on asset sizes and geographies. We will describe how this approach can be useful to produce stress testing results with unexpected scenarios such as COVID-19. First, we will examine the methodology to produce industry-level forecasts, then look at COVID-19–based scenarios, and finally combine them to quickly produce bank-level forecasts under these scenarios.
Call Report Forecast methodology
As described in the article “Stress-Testing and Strategic Planning Using Peer Analysis,”3 we assume perfect competition given the large number of banks in the United States so that exogenous actions taken by managers at an individual bank will not affect the trajectory of industry-level aggregates. It allows us to model the data on industry-level aggregate outcomes for each line item on the call report without worrying about the effect of any specific action taken by a manager at an individual bank. As a consequence of our assumption, we can model the behavior of the industry against cyclical economic variables and thus isolate the pure effect of the macroeconomy on the series of interest. This basic principle applies, to a greater or lesser extent, to all the series in the call report. Our assumption of perfect competition vastly simplifies our analysis by allowing us to concentrate on pure macroeconomic factors and thus isolate the internal factors for analysis conducted separately.
The CRF methodology is reductive in nature in the sense that we begin with aggregate industry data and identify macroeconomic forces that affect the entire banking system. We then isolate the trend and cycle from each peer or bank’s market share, leaving us with series that reflect the bank’s own strategic decisions, the effects of its peers’ actions, and other idiosyncratic behavior. This two-layer approach to modeling allows us to disentangle macroeconomic and bank-specific effects. Peer groups can be viewed as an intermediate layer. Doing so lets us differentiate between factors that affect the entire industry and factors that affect only the peer group of interest, whether because of geography or because of particular portfolio concentrations.
For example, a bank runs a large marketing campaign for a particular product during a boom. We can reasonably assume that at least some of the observed increase in volume is the result of the marketing campaign rather than the state of the economy. Other peer banks that did not market themselves as aggressively will therefore not experience the same rate of growth. If the increased marketing is not merely a reflection of economic conditions, our methodology will find the aggressive bank’s market share growth to be idiosyncratic and will not extrapolate that growth through the forecast window.
Current approaches to pre-provision net revenue (PPNR) modeling fail to draw these kinds of distinctions, if all movements in the data are caused by external macroeconomic forces rather than internal manager-initiated actions. These approaches cannot be used to conduct “what-if” analyses that explore how management actions affect performance. In contrast, our forecasts look clearly at expected outcomes based on macroconditions, so that what-if analyses can focus on determining how management actions cause performance to deviate from the performance of the peer group.
We recognize that building models of smooth series exhibiting clear cyclical behavior is generally more conducive to forecast accuracy than trying to correlate economic variables with noisy bank-level data. Our approach therefore begins with fully specified industry-level econometric models for each series (see Mehra [2016]4). Although the actions of an individual manager can have a profound impact on the behavior of the bank, these actions will have an undetectable impact on the industry as a whole. For this reason, we are safe in modeling the effect of macroeconomic variables on industry-level series without worrying about bank-specific management decisions.
One feature of our industry-level models is that they exploit adding-up constraints and identities that help anchor forecasts for noisier peripheral elements of the industry-level balance sheet and income statement. For example, we can understand the macroeconomic drivers of the total non-interest expense series better than we can for the subcategories that make up the series. We do construct models for each subcategory, but we adjust the forecasts so that they add up to the previously determined forecast for the smoother, more aggregate series. The individual models for the subcategories help us distinguish the peculiarities of each series, while the more accurate forecasts of the smoother, broader series serve as guide rails.
Suppose we want to forecast a peer group’s stock of commercial real estate loans held on its book. Then, for each quarter we define the peer group’s market share of commercial real estate (CRE) loans as the combined sum of the CRE balance of all the constituent banks in the peer group divided by the industry-level aggregate balance of CRE loans reported in the CRF. In general, forecasting a market share turns out to be easier than forecasting a dollar amount. With an industry-level forecast and a peer-level market share forecast, forecasting the bank’s dollar balance of CRE loans is trivial (Figure 1). This can then be extended to the bank level using a combination of quarterly growth rates on the peer-level forecast and the last historical data point from the bank.
We apply a “share down” process to project the behavior of peer groups. Many peer-level series are modeled using a market share approach, while for other series we use a “beta model” methodology analogous to well-known empirical explorations of the capital asset pricing model (see Poi [2016]5). Our algorithms fit various market share and beta models and use cross-validation to pick the model that forecasts a particular series most accurately. A feature of the market share approach is that it prevents the modeler from forming prior expectations about the signs of macroeconomic factors’ coefficients. For example, although we can reasonably assume that industry-level mortgage credit losses will rise in a recession, we cannot say whether small banks (assets less than $1 billion) will have an increasing or decreasing share of such losses. Determining the effect of the macroeconomy on a peer’s market share for a series is thus a purely statistical exercise that does not require the modeler to justify the signs of any coefficients.
Our observation that peers’ market shares often exhibit trend behavior typically yields few objections. Our claim that market shares are often procyclical or countercyclical is somewhat more controversial. Figure 2 shows the market shares of net loans and leases for some of our peer groups. Typically. market shares exhibit predominantly trending behavior (Peer 1 in Figure 2). Such market shares do not exhibit much relationship to the macroeconomic cycle, so share forecasts show little deviation between the Baseline and Severely Adverse scenarios.
In contrast, the market shares shown for Peer 2 display clear correlation to the macroeconomy. During the Great Recession, loans and leases fell and then grew back as the recovery took hold. Such examples are often encountered. Cyclical changes in market share might reflect conscious decisions made by bank managers or passivity in the face of rivals’ aggression. We typically find that conservatism by banks in a peer group leads to countercyclical market share in most asset categories. Procyclicality is often allied to an upward trend in market presence, reflecting the fact that such banks are taking risks to gain an increased share.
Moody's Analytics has updated its March Baseline forecast and alternate scenarios to reflect the quickly unfolding impact of the COVID-19 pandemic. These scenarios can be used for loss forecasting, stress testing, and other applications. According to Moody’s Analytics Baseline assumptions, ultimately there will be millions of infections across the globe, including in Europe and the United States. COVID-19’s mortality rate is assumed to be 1.5%, consistent with the experience so far, and the percentage of those seriously infected needing some form of hospitalization is expected to reach 10%. The peak of the pandemic is assumed to occur in May, winding down quickly by this summer, with a vaccine in place before next winter. The exact assumptions for the US Baseline, upside (S1), and downside (S3) scenario are listed in Figure 4.
Once the pandemic becomes significant in a country, that economy shuts down similarly to what happened in China. Businesses are disrupted, and many schools and day care centers are closed, making it difficult for parents to get to work. International travel and trade are impaired. Stock prices decline, consistent with declines suffered on average in past recessions. The Fed and other global central banks slash interest rates in response. The Fed pushes short-term rates to the zero lower bound, and 10-year Treasury yields flirt with going negative. Asian and other emerging market central banks also lower rates further, but generally do not push them to the zero lower bound, since doing so would put too much pressure on their currencies. Most governments have so-called automatic stabilizers in their budgets—countercyclical tax and government spending policies designed to support their economies in tough times. These stabilizers are helpful in cushioning the economic blow of the virus in the pandemic scenario. All these events have unfolded at unprecedented speed in the second half of March 2020.
Several policy actions have already been put in place and are fully incorporated into our Baseline forecast assumptions. This includes three rounds of fiscal stimulus, the latest installment coming in at $2.2 trillion. The stimulus will be dispersed through direct payments to individuals; a boost to unemployment insurance; loans and guarantees for large firms; small business loans/grants; aid targeted to specific sectors such as transportation, healthcare, and education; business tax cuts; and aid to states among others. Currently, there is strong potential for a fourth round of fiscal stimulus. On the monetary policy front, the Fed has lowered interest rates and started quantitative easing as well as other lending facilities. The federal funds rate—a peg for both short-term and long-term rate setting as well as consumer rates—has been slashed to the 0%-0.25% range. Further, the Fed will buy back $500 billion of Treasury securities and $200 billion of asset-backed mortgage securities to instill more liquidity in the markets. Poor implementation of these policy actions and lack of timely reinforcement rounds pose the biggest risks to the Baseline forecast.
Because of the COVID-19 pandemic, a US recession is now part of the Baseline scenario. Real GDP decreases by an almost 6% peak to trough and declines for 2020. Unemployment will peak at 9% in 2020Q2. The struggling manufacturing, transportation, agriculture, and energy industries are hit hard, but so too are the travel and tourism industries and the construction trades. However, there are significant layoffs across nearly all industries, with healthcare and government being the notable exceptions. Figure 5 lists the details of how the recession and a subsequent rebound may play out.
The most important risk to the forecasts is how long new coronavirus cases continue to increase in the United States and the timing of the end of shutdowns. Other risks include the effectiveness of the stimulus, given that consumer sentiment remains weak, and lack of clarity around return-to-work guidance. Finally, if demand and supply-side bottlenecks take too long to dissipate, we could see a surge in bankruptcies that will have an impact on credit markets.
Forecasting a bank-specific balance sheet and income statement
The strength of the Call Report Forecast model at the industry level resides in the large set of data that removes idiosyncrasies at the bank level. While the methodology employed to produce an industry-level forecast can be used at the bank level, we also must consider the trade-off between forecast precision and the time and efforts to produce these forecasts for a large number of banks at the same time. Thus, we propose a very simple approach to quickly and efficiently generate forecasts under different stress scenarios that can be produced on a large scale.
We combine the industry-level forecast, deriving the period-over-period growth of the different line items of the balance sheet and income statement, and apply it to the bank’s current call report data. Without any input from our customers, we can then generate a bank-level forecast under many scenarios, including the COVID-19 scenarios, for any institution publishing call reports. When our customers have already developed a budget or a Baseline forecast, we can also apply the sensitivities of our forecast to our customers’ Baseline forecasts and derive the alternative forecasts for COVID-19 scenarios (Figures 6 and 7).
Take the example of a large financial institution with a sizable retail portfolio. We can use an industry-level forecast such as a CCAR peer group. For smaller banks with a localized presence in specific states, we can use peer groups that are a combination of asset size and geography as well.
Conclusion
The events of the COVID-19 crisis will continue to affect our economy over the months to come. As a result, many banks will struggle to make informed decisions in the face of unprecedented uncertainty. The first step toward planning in this environment is to produce credible and reliable projections of their balance sheet and income statement under a range of possible scenarios. The potential issues banks are facing include:
- Capital planning and stress testing framework: banks need robust infrastructure and a process in place to simulate a new economic environment and the impact of potential remediation strategies
- Quantitative analytics: models must provide credible, reliable estimates under a wide range of conditions
- Response time: banks must be able to create estimates before the environment changes and those estimates become irrelevant
With this paper, we presented a methodology and a solution that banks can easily adopt to address these issues. As a primary or challenger approach, the framework presented can help management augment their existing planning infrastructure or develop a new process to make business decisions. Each new crisis often comes with unforeseen levels of stress to specific drivers of our economy and/or financial risks. When such situations arise, the more tools finance and risk professionals have to formulate their decisions, the better positioned they will be to address these events. The speed and robustness of this approach makes it an invaluable addition to any bank’s tool set.
1 Capital Risk Analyzer: https://www.moodysanalytics.com/product-list/capital-risk-analyzer
2 Call Report Forecast (CRF): https://www.economy.com/products/data/forecast-bank-call-reports
3 A. Hughes and B. Poi, “Stress-Testing and Strategic Planning Using Peer Analysis,” Risk Perspectives, Vol. 8 (November 2016): 59-68.
4 S. Mehra, Bank Call Report Forecast Database, Moody’s Analytics product literature (2016).
5 B. Poi, Peer-Group Analysis in Bank Call Report Forecasts, Moody’s Analytics product literature (2016).
6 J.J. Groen and G. Kapetanios, “Revisiting Useful Approaches to Data-Rich Macroeconomic Forecasting,” Federal Reserve Bank of New York Staff Report #327 (May 2008, revised October 2015).
