Why Stress Tests Must See the Whole Financial System
The subject of this series is the IMF's systemwide stress-testing approach, especially its effort to integrate risks from nonbank financial intermediaries (NBFIs) with the risks of banks, market-based finance, and their clients. This opening focuses on the question that makes the approach necessary: why can a test that looks carefully at one institution still miss the way financial stress grows across the system? The practical value of answering it is straightforward. A reader who can see the connections can interpret the IMF's two tool families—and later, their use in Financial Sector Assessment Programs (FSAPs)—as instruments for finding amplification channels rather than as a collection of technical labels.
The distinction matters because “stress testing” can sound like a single activity when it is actually a choice about what to include. One exercise may ask whether a bank's capital and liquidity remain adequate after a loss. A systemwide exercise asks an additional question: what do the bank, its counterparties, its clients, investment funds, and the markets they share do next? The IMF paper's public description places this broader question at the center. It describes approaches that cover multiple financial subsectors and their clients, with the purpose of identifying cross-sector and cross-border amplification channels and improving policy responses.
That does not make the institution-level question irrelevant. It changes its meaning. A bank may be able to absorb a direct loss while the financial system around it becomes harder to fund, harder to value, or harder to transact in. Conversely, an apparently vulnerable institution may be less consequential if it has few connections and its assets can be sold without moving prices. The systemwide lens is therefore not simply a larger perimeter. It is a way of asking whether relationships turn a manageable shock into a reinforcing one.
From an institution's loss to a system's reaction
Consider a deliberately simple scenario. A decline in the value of a widely held asset creates a loss for several financial institutions. A bank-only test might calculate the effect on the bank's capital, funding position, and ability to continue lending. That is useful information. But it leaves open what happens if an investment fund faces investor withdrawals at the same time, if a borrower must provide more collateral, or if many holders try to sell the same asset. The first loss is only the beginning of the systemwide question.
This is why the IMF's subject is not “NBFI risk” in isolation. The paper describes shocks spreading through banks, NBFIs, and market-based finance, while also emphasizing their clients. These are distinct parts of finance, with different balance sheets, obligations, and incentives. Yet they can be tied together by a common asset, a funding relationship, a collateral arrangement, or a client whose financing depends on more than one source. A category such as NBFI identifies a broad institutional perimeter; it does not tell us how a particular shock will travel inside or across that perimeter.
The same point applies to the word “market.” Market-based finance is not merely a backdrop against which institutions operate. Prices and trading conditions can change the value of assets, the amount of collateral required, and the cost of raising funds. If prices fall in an orderly market, an institution may adjust gradually. If prices fall while many participants need cash, the act of selling can make the market less liquid and produce further losses. The systemwide approach matters because these reactions can affect the institutions that did not initiate the sale.
Cross-border connections widen the question again. A financial institution may be domestic, but its investors, lenders, assets, collateral, or counterparties may not be. The IMF's abstract explicitly identifies cross-border amplification as a reason to develop systemwide tools. That does not tell us that every country faces the same pathway, or that cross-border transmission is always the dominant risk. It does establish a boundary for the analysis: a national sector table cannot automatically be treated as the perimeter of the financial system if important claims and reactions cross jurisdictions.
A useful mental model is to separate three layers. The first is the shock: a change in prices, funding conditions, confidence, or the ability to meet an obligation. The second is the response: institutions draw liquidity, investors redeem, borrowers post collateral, or holders sell assets. The third is the feedback: those responses alter prices, market functioning, and the conditions faced by other institutions. A conventional test may model the first layer and part of the second. A systemwide test is designed to make the links among all three visible.
The diagram is intentionally modest. It does not claim that the IMF paper uses this exact architecture, nor that every stress event follows every arrow. It translates the paper's stated scope and mechanism list into a way of seeing the analytical problem. Banks, NBFIs, market-based finance, and clients remain separate actors. The arrows matter because a systemwide result depends on more than adding their individual losses together.
Why liquidity makes the perimeter consequential
Liquidity is where the difference between a balance-sheet snapshot and a systemwide process becomes most apparent. Solvency asks whether assets ultimately cover obligations. Liquidity asks whether an institution can meet an obligation when payment, funding, or collateral is needed. The two can interact, but they are not interchangeable. An institution can own valuable assets and still face pressure if those assets cannot be turned into cash quickly without a large price concession.
The IMF paper describes systemic liquidity risks that can affect numerous institutions simultaneously. That word—simultaneously—is important. If only one institution needs cash, a market may absorb its sale. If many institutions, funds, and clients respond to the same shock at once, the market's capacity to absorb transactions becomes part of the problem. The risk is not just that one balance sheet is weak. It is that the same action, repeated across the system, changes the conditions under which everyone else must act.
The named mechanisms show how this can happen. A run places immediate pressure on a funding source when creditors or depositors seek repayment. A redemption creates a similar need for an investment fund to return money to investors, potentially requiring the fund to sell assets. A margin call or collateral call requires a participant to provide cash or eligible assets after market values or exposures change. These events do not have identical legal or economic forms, but they share a feature that matters for systemwide analysis: they can force a response before an institution has time to wait for orderly conditions.
Forced sales add the market channel. If a fund sells assets to meet redemptions, the sale may reduce the price received, especially when other holders are selling too. Lower prices can weaken the balance sheets of banks, funds, and clients that own the same or related assets. They can also trigger new collateral demands or make lenders more cautious. The process can therefore move from liquidity pressure to fire sales, from fire sales to price dynamics, and from prices back to liquidity pressure. The source's abstract names each of these mechanisms and the possibility of disruption in core markets; the precise strength and ordering of the links depend on the scenario and the model assumptions.
This is also why it would be misleading to treat an NBFI as an external appendage to a bank stress test. An investment fund may not take deposits or make loans in the same way as a bank, but its redemption behavior and asset sales can influence the prices and market liquidity on which banks and borrowers depend. A bank may be exposed through its own holdings, through financing provided to a fund or client, or through collateral whose value changes. The relevant relationship is not that the sectors are identical. It is that their actions can meet in the same funding and market system.
The feedback can also stop. Investors may not redeem; a fund may hold enough liquid assets; a lender may extend funding; a market may absorb sales; or a borrower may replace one source of credit with another. Those possibilities are not evidence that stress is harmless. They are reasons to test alternative scenarios instead of treating the most severe chain as inevitable. A credible systemwide exercise should make clear which reactions are assumed, which are observed in the data, and which are only plausible branches. The available public description does not provide a particular scenario, liquidity metric, or calibration, so it supports the mechanism-level explanation but not a quantitative ranking of outcomes.
Two IMF lenses for seeing the connected system
The IMF paper identifies two base tools that make the systemwide problem more tractable. The first is an Excel-based flow-of-funds framework. Its role can be understood as a broad map of financial relationships: which sectors provide funding, hold claims, owe liabilities, or channel resources to clients. Such a map helps an analyst ask where a shock could enter and which institutions or sectors are connected to the affected claim. It supplies coverage that a single-institution balance sheet cannot provide.
The second is investment-fund liquidity analysis with fire-sale and market-impact dynamics. This lens focuses more closely on what happens when a fund faces liquidity pressure and responds by selling assets. It links redemptions or other demands for cash to portfolio liquidation, then asks how the liquidation changes prices and market conditions. In the systemwide picture, that price effect can become an input into the balance sheets and collateral requirements of other actors.
The tools are therefore complementary in a precise sense. A flow-of-funds view helps answer where are the relationships and exposures? Investment-fund liquidity analysis helps answer how might one stressed portfolio's response affect the market used by others? The first emphasizes breadth and visibility across sectors. The second emphasizes a particular behavioral and market-feedback mechanism. Neither description, by itself, proves that the tools always run as one engine or that either produces a definitive forecast.
That boundary is part of the subject's significance, not a footnote. The public abstract names the tools, their broad purposes, the channels they examine, and their application and enhancement within FSAPs across various countries. It does not, by itself, provide the workbook's worksheet structure, a price-impact equation, a country-specific loss estimate, or a performance ranking. A careful reader should resist filling those gaps with invented precision. The right question is whether a model makes a material channel visible and whether the result is stable enough, under plausible assumptions, to inform further financial-stability work.
The central lesson is consequently narrower and stronger than the claim that “everything is connected.” The IMF's systemwide stress-testing approach treats banks and NBFIs together because their funding needs, asset prices, collateral, clients, and markets can transmit stress across institutional and national boundaries. Its two named tools address different parts of that problem: one helps organize the network of claims and flows, while the other examines how investment-fund liquidity pressure can become fire-sale and market-impact feedback.
The next question is causal rather than merely classificatory: once a shock enters this connected system, how do runs, redemptions, collateral calls, sales, and price changes interact over time? That sequence is what turns a broad perimeter into an explanation of amplification—and it is the basis for judging what a systemwide stress test can reveal.