From Model Outputs to Financial Statements: Reconciling Financial Narratives in Stress Testing

[#1]

At its core, stress testing represents a forward-looking financial narrative that connects macroeconomic shocks to institutional performance. A recession scenario, for example, may lead to higher levels of unemployment, which increases borrower defaults, resulting in higher credit losses and declining profitability. These outcomes ultimately affect regulatory capital ratios and determine whether a bank remains resilient under stress.

The credibility of this narrative depends on whether the numbers produced by different models align with each other and with the institution’s financial statements. When projections from credit risk models, income forecasts, and capital calculations fail to align, the narrative becomes inconsistent and difficult to interpret. This is where reconciliation plays a crucial role.

Reconciliation

Reconciliation ensures that outputs produced by various analytical components, such as credit risk models, portfolio aggregations, and financial statement projections, ultimately form a coherent financial story. It ensures that the exposures in risk models align with the balance sheet, and that projected losses are properly reflected in provisions, profitability, and capital ratios. Reconciliation ensures that when numbers are placed under economic pressure, they continue to tell a believable and internally consistent story.

Common Sources of Inconsistencies

Despite the structured workflow outlined in the previous section, financial narratives can easily break down if reconciliation is not performed carefully. One common source of inconsistency arises from model fragmentation. Large financial institutions rely on multiple specialized models to estimate credit losses, revenue projections, and balance sheet dynamics. These models often operate with different data sources and aggregation levels. For example, credit models may rely on granular loan-level data, while financial planning models operate at aggregated portfolio levels. If the totals generated by these models do not align, analysts may encounter discrepancies between projected losses and financial statement impacts.

Aggregation challenges also contribute to reconciliation problems. Stress testing exercises often begin with detailed loan-level simulations but must ultimately produce results that align with enterprise-level financial statements. If the exposures used in credit risk models differ from the balances recorded in accounting systems, the resulting projections will diverge. For example, the total loan exposure simulated in risk models may not match the loan portfolio reported on the balance sheet due to differences in segmentation, timing, or data definitions.

Another source of inconsistency arises during scenario translation. Macroeconomic shocks affect multiple aspects of a bank’s operations simultaneously, yet these impacts may be modeled independently. A recession scenario may increase credit defaults while also reducing interest income and affecting funding costs. If these relationships are not properly coordinated across models, the resulting projections may contradict one another. For instance, we may observe situations where projected credit losses increase sharply while capital ratios remain stable, or where profitability improves despite worsening economic conditions. Such outcomes undermine the credibility of the stress testing narrative.

Reconciliation in SAS Stress Testing

SAS Stress Testing Solution ensures all independently modeled components – assets, income, and sometimes partial liabilities – come together into a fully balanced financial statement. The solution defines a sample rulebook that drives how imbalances are resolved when the system detects that one side of the balance sheet has moved without a corresponding offset (see pic below).

 
At its core, reconciliation rulebook is governed by a simple accounting identity:


However, most stress testing models do not explicitly generate all three components in a synchronized way. For example, some models estimate how much loans will grow and how much profit will be kept by the company, but they often don’t fully account for where the money comes from – like customer deposits or borrowed funds. This creates a structural imbalance that must be resolved systematically. The table above is central to this process – it is not just a mapping table, but a set of balancing rules that drives the reconciliation engine.

Each row in the table specifies how a given Delta Segment (an asset category) should be offset through a corresponding Funding Segment. Across all rows, the structure is identical – every asset movement, whether commercial loans (CI Term Loans, CRE Construction, CRE Permanent, LOC), retail exposures (Auto, Credit Cards, HELOC, Residential Mortgages), or non-loan assets (Property & Equipment, Intangibles), is mapped to Equity → Other Reserves, with a Funding Percent of 100%. Interpreted correctly, this means that the entire value of each asset delta is offset through equity.

Despite the terminology Funding Segment, the table is not modeling economic funding behavior (such as deposits or borrowings). Instead, it defines how the system should mechanically restore balance whenever an imbalance arises.

To see how this works, consider that stress testing models produce the following changes over a projection period:

At this stage, assume liabilities have not been explicitly modeled, so that Δ Liabilities = 0 and Δ Assets = +330.

The reconciliation engine now applies your balancing rules table. For each Delta Segment, it looks up the corresponding Funding Segment and applies the 100% funding rule:

So, the system generates Δ Equity = +330, and the balance sheet is now satisfied, and restores the accounting identity:

From a workflow perspective, this process unfolds in a structured sequence. First, individual models generate raw deltas across asset classes and possibly income components. Second, these outputs are aggregated and aligned at the same segmentation level as the “Delta Segment” definitions. Third, the reconciliation engine consults the balancing rules table and applies the specified mappings – in this case, a 100% allocation of every asset delta to equity. Finally, a validation step confirms that the balance sheet is fully reconciled.

Caveats to Consider (for this sample balancing rule)

From a design perspective, the sample balancing rule allows practitioners to isolate asset-side dynamics and validate reconciliation logic without introducing the complexity of liability modeling or circular dependencies.

However, the financial implication must be treated carefully. This configuration implies that all asset growth (or contraction) is fully absorbed by equity, which is not representative of real-world banking behavior. In practice, assets are funded through a mix of deposits, borrowings, and capital. As a result, while the sample rule ensures perfect balance sheet consistency, it does not provide meaningful insights into funding structure, leverage, or liquidity risk.

Interpreting Reconciliation Results from the Sample ST Cycle

We now turn to a set of illustrative reconciliation results from the sample Stress Testing cycle to understand how these dynamics manifest in practice – the first displays a high-level outcome, and the second presents a decomposition of the high-level outcomes.

At the aggregate level, the first table shows the net change in Assets, Liabilities, and Equity across each projection year. Notice these values satisfy the accounting identified earlier.

The presence of both positive and negative values is entirely expected and does not violate this identity. Instead, the sign simply indicates direction of movement – positive values represent increases in balance sheet components, while negative values represent decreases. For instance, in 2024, assets decline while liabilities and equity increase, indicating a contraction in the asset base that is offset elsewhere. In contrast, in 2025 and 2026, assets grow significantly, while liabilities and equity decline, suggesting that asset expansion is being absorbed by reductions in funding sources, particularly equity.

While this first table confirms that reconciliation holds, it does not explain how the system achieves this balance. That explanation emerges from the second, expanded table, where movements are broken down into detailed components. The most critical relationship appears between ASSET → LOAN_CONTR and EQUITY → OTHER_RESERVES, which directly reflects the balancing rules table where every asset delta is mapped 100% to equity.

The interpretation of signs becomes particularly meaningful at this stage. When assets decrease (negative values), equity increases (positive values), and when assets increase (positive values), equity decreases (negative values).

At the same time, the expanded table reveals that equity movements are not driven solely by this balancing mechanism. There are also modeled economic components, particularly visible in retained earnings and taxes.

Retained earnings reflect profitability dynamics – positive values indicate profits that increase equity, while negative values indicate losses that reduce it. Taxes, on the other hand, represent liability adjustments linked to income, with positive values indicating increased obligations and negative values indicating reductions. This introduces a layered interpretation of equity, where total movement is driven by both economic effects (income and provisions) and mechanical reconciliation effects (balancing through other reserves).

Cash acts as a secondary balancing mechanism, particularly from a liquidity perspective. Negative values indicate outflows, while positive values indicate inflows, complementing the reconciliation process when some financial flows are explicitly modeled and others are implicitly balanced.

The key takeaway is that positive and negative values describe direction, and the system functions precisely because these movements are consistently offset. Equity, particularly Other Reserves, acts as the shock absorber, moving in the opposite direction of asset changes to ensure balance. In this way, the first table confirms that reconciliation has been achieved, while the second table explains how it has been achieved.

Additional Information

For more information on SAS Stress Testing 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