Why Zombies Persist: The Bank Incentive Behind a Firm That Will Not Exit
Zombie firms persist because a weak operating position can be matched by a lender’s incentive to postpone recognizing a loss. This chapter focuses on that bank–firm mechanism: how a company whose operating earnings do not cover interest can keep borrowing or avoid immediate repayment, why weak bank health intensifies the pattern, and how relationships among several creditors can delay exit. The point is not to assume that every distressed firm is being subsidized. It is to identify the conditions under which credit support becomes a persistence mechanism.
Evergreening turns a payment problem into a survival strategy
A firm with an interest coverage ratio below one faces a straightforward arithmetic problem. Its EBIT is smaller than the interest owed on its debt. If the lender demands full payment, the firm must use cash, sell assets, find another source of finance, restructure the loan, or default. A single year of weak coverage can be temporary. A three-year sequence is harder to reconcile with ordinary operating recovery, especially when the firm remains in the same measured state year after year.
Evergreening is the set of lending practices that can postpone that reckoning. A bank may lend enough to cover interest while leaving principal outstanding. It may extend the maturity, capitalize unpaid interest, or temporarily relax repayment requirements. These actions can give a firm time to recover, but they can also keep an economically weak borrower operating without resolving the underlying mismatch between earnings and debt service. The distinction depends on whether the accommodation leads to viable restructuring or simply delays recognition of an eventual loss.
The BIS evidence points to bank dependence rather than a durable shift into independent non-bank finance. Within country–year and sector–year comparisons, zombie firms have a 13.8 percentage-point lower share of non-bank liabilities and a 9.0 percentage-point lower trade-credit share than comparable non-zombies. Their liabilities are, if anything, more concentrated in bank debt. This result is not a complete map of every funding source: off-balance-sheet borrowing is not observed, and the classification of liabilities relies on balance-sheet categories. It nevertheless makes a bank-centered mechanism more plausible than the idea that zombies simply survive by finding abundant alternative finance.
Why would a bank accept this arrangement? The immediate reason is that recognizing a borrower’s failure can reduce the value of the bank’s assets and expose earlier lending decisions as losses. A rollover preserves the appearance of continuity and may give the borrower a chance to pay later. That incentive is strongest when the lender is already weak or when the relationship is important enough that withdrawal would force a difficult restructuring. The bank may not expect the firm to become a high-productivity competitor again; it may prefer postponement to a visible loss today.
The mechanism does not require the bank to make a conscious decision to keep a permanently nonviable firm alive. It can arise from repeated small decisions: extend the maturity, renew a facility, or wait for an improvement in prices. Over several reporting periods, those decisions create the durable zombie state observed in the data. What looks at the firm level like unusual resilience can therefore be the accumulated effect of creditor forbearance.
The two basic forms of accommodation have slightly different appearances but the same central effect on timing. In one, the bank provides enough new credit for the firm to meet interest while the principal remains outstanding. In the other, the bank temporarily requires less from the borrower through a restructuring or repayment holiday. Neither action automatically creates new operating earnings. It changes when the bank must decide whether the original claim can be recovered. That is why the interest-coverage measure and the lending relationship need to be read together: the firm’s operating capacity remains weak while the financing arrangement prevents that weakness from becoming an immediate exit.
This also explains why a rising zombie-lending share is informative even though it is not a direct record of every loan decision. The measure asks how much of a bank’s observed lending is connected to firms classified as zombies. If that share remains high from one year to the next, it indicates persistence in the bank’s portfolio as well as persistence in the firms’ status. The interaction with bank health then asks whether the pattern is stronger precisely where recognizing losses would be more difficult. The evidence does not claim that the ratio itself identifies a single act of evergreening; it shows that the portfolio and the borrower state move together in the direction the mechanism predicts.
The sequence has two possible endpoints. If the firm’s operations improve enough to cover debt service, the extension can be a bridge to recovery. If they do not, the same extension can make the eventual restructuring more costly by preserving excess capacity, tying up productive resources, and allowing claims to accumulate. A mechanism explanation must therefore include termination conditions, not only the reason support begins.
Weak bank health protects the wrong borrowers
The study measures zombie lending at the bank level as the share of a bank’s loans directed to zombie firms. That share is persistent: the baseline coefficient on its lagged value is 0.574. More importantly, the persistence becomes stronger when the bank’s asset quality is weak. The interaction between lagged zombie lending and a high-NPL indicator is 0.085. When the bank’s Tier 1 capital ratio is used instead, the interaction with zombie lending is −0.020. Because a higher capital ratio is associated with less persistence, the negative sign has the same economic interpretation: thinner capital is linked to stronger continuation of zombie exposure. The two interactions remain jointly significant when included together.
These coefficients do not mean that every weak bank increases every zombie loan. They describe a conditional pattern across banks and years, with bank–country and year fixed effects and lagged explanatory variables. The result is consistent with loss avoidance: when asset quality is poor or capital is thinner, the cost of recognizing a bad loan is more salient, so the incentive to maintain an existing exposure increases. The bank can appear to preserve a relationship while the quality of its portfolio quietly deteriorates.
The direction of this relationship should still be interpreted carefully. Weak banks may evergreen more, but zombie lending can also make a bank weaker by leaving it with low-quality assets. Lagged variables and fixed effects reduce some timing and compositional concerns; they do not turn the regression into a controlled experiment in which a bank is randomly assigned a distressed borrower. The strongest conclusion is therefore about a mutually reinforcing pattern: weak asset quality and thin capital are associated with more persistent zombie exposure, and that persistence is consistent with a lender trying to postpone loss recognition.
The clearest evidence comes from comparing firms that are always classified as zombies with firms that are never classified as zombies. For never-zombie firms, a one-percentage-point increase in the relationship banks’ NPL ratio is associated with lower log debt, lower leverage, and lower debt growth. For always-zombie firms, the same bank-health measure is associated with higher log debt and higher leverage, while the debt-growth estimate is not statistically significant. In other words, when banks become less healthy, credit sensitivity differs by borrower type. The firms that appear chronically distressed are not responding like ordinary borrowers whose lenders are tightening terms.
That contrast is a form of credit reallocation. A healthy firm with a productive investment opportunity may lose financing because its bank is conserving capital or reducing risk. A zombie firm may retain or even increase the debt recorded on its balance sheet because the bank is rolling over the existing exposure. The per-firm coefficients should not be converted directly into aggregate dollar flows: always-zombie firms are much fewer than never-zombie firms, and their debt balances differ. The relevant finding is the direction of the lending response, not a claim that all credit moves from healthy firms to zombies in equal amounts.
This is why the zombie problem is not contained within the borrower. A bank that refuses to recognize a loss can also fail to price new credit according to productive opportunity. The result is a weaker selection process: resources remain attached to capacity that produces low returns, while healthier firms face a financing environment shaped by the bank’s past mistakes. The bank’s private incentive to smooth recognition can therefore create a broader allocation cost.
The effect on a healthy borrower is not necessarily an immediate default or a visible refusal of credit. It may be a smaller facility, a shorter maturity, a higher hurdle for investment, or a decision to postpone hiring. Across many firms, those margins can reduce the expansion of businesses that would otherwise take market share from a zombie. The contrast between the always-zombie and never-zombie groups is valuable for precisely this reason: it shows that the lending response is not simply a uniform reaction to a bad macroeconomic year. Borrowers with different operating conditions receive different treatment when the same relationship banks become less healthy.
Relationships can lock a firm into the zombie state
The bank’s health is only one part of the mechanism. The structure of the relationship between a firm and its lenders also matters. Among firms already classified as zombies, a higher lagged share of relationship-bank loans going to zombies is associated with a greater probability of remaining a zombie. The interaction with relationship-bank Tier 1 capital is negative and significant, indicating that the lock-in relationship is stronger when the connected banks are less well capitalized.
A second measure uses the number of relationship banks. Zombies with more banks are more likely to remain zombies; the coefficient on the number of banks is 0.0078. The interpretation is not that diversification is always harmful. Multiple lenders can provide resilience to a healthy business. In distress, however, each lender has only a partial view of the firm and an incentive to protect its own claim. A restructuring plan that imposes losses on one bank may preserve value for all creditors, but reaching that plan requires coordination. If no lender wants to move first, the firm can remain in a holding pattern.
This is a coordination problem rather than a simple story of one irresponsible bank. Each lender may prefer a rollover because immediate exit crystallizes its own loss, while the collective result is continued operation by a firm that is not recovering. More creditors can make negotiations harder, especially when weak bank capital narrows the willingness to absorb a write-down. The evidence is consistent with that interpretation, but it does not observe every negotiation or prove that the number of banks causes persistence by itself.
The coordination problem also creates a difference between maintaining a claim and solving the borrower’s problem. A lender can monitor a firm, keep its facility open, and still have no agreed plan for reducing debt to a sustainable level. Monitoring may limit the bank’s own exposure without restoring the firm’s productivity. Conversely, a coordinated restructuring can preserve a viable operation while changing ownership, maturities, or creditor priority. The data establish the association between relationship structure and persistence; the institutional question is which relationships can produce a credible recovery decision rather than another cycle of delay.
The domestic–foreign distinction adds another qualification. In the broader evidence, foreign banks have higher persistence in some zombie-lending relationships, which is consistent with maintaining and monitoring an exposure rather than abruptly withdrawing. Yet the balance-sheet cost of zombie lending is concentrated in domestic Asian bank lending: domestic loans are associated with higher NPLs and lower ROA, while comparable foreign-bank lending does not show the same impairment. A bank can therefore stay connected to a zombie firm without carrying a large enough exposure, or without using the same lending practices, to move its consolidated home-country balance sheet.
That difference does not make foreign lending irrelevant. It helps explain why a cross-border bank channel may be weak even when foreign banks do not immediately abandon the relationship. The international consequences can instead appear in the goods produced by the zombie firm and in the prices faced by its trading partners. The lender relationship remains the condition that keeps capacity operating; it need not be the route through which the price effect crosses a border.
What would break the cycle?
The evidence supports a bank-centered persistence mechanism, but it does not establish that every zombie firm is sustained by evergreening. Firms can draw on cash, sell assets, receive government support, borrow from non-bank entities, or benefit from trade-credit arrangements. The study cannot observe all off-balance-sheet borrowing, and the non-bank comparison depends on assumptions about which liability categories are bank-sourced. Those alternatives should remain visible without diluting the main pattern.
A genuine exit path requires more than a lender deciding to stop renewing a loan. A viable firm may need coordinated restructuring that reduces debt service while preserving its operating business. A nonviable firm may need an orderly liquidation or sale so that workers, equipment, and market opportunities can move elsewhere. If the legal and institutional process cannot distinguish those cases, a bank may face a choice between indefinite extension and a disorderly failure. That is one reason insolvency capacity belongs in the same analysis as bank capital and NPL recognition.
The practical conclusion is narrower than “banks should never support distressed firms.” Temporary support can preserve a business that has a recoverable problem. The danger arises when support is repeatedly renewed without a credible recovery test, while healthy borrowers absorb the financing cost and the distressed firm’s capacity continues to influence prices. Once the loan rollover has become a stable operating condition, the next question is no longer only why the zombie survives. It is what that survival does to the firms, resources, and prices around it.