Why Collateral Is Rewriting Direct Lending
This chapter asks why a market often associated with unsecured, cash-flow lending has become increasingly secured, and what that change does to private-credit risk. The answer is not that collateral makes a loan safe. It is that collateral can compensate for imperfect information, change the terms a lender is willing to offer, and tie the supply of nonbank credit to asset prices. Understanding that mechanism lets the reader interpret the evidence without treating a descriptive pattern as a universal rule.
When information is weaker, the contract carries more weight
A lender rarely knows as much about a borrower as the borrower does. That gap matters in any credit market, but it can be especially important in direct lending to smaller or less transparent companies. The lender must judge whether the borrower can generate enough cash to service the loan, whether the business is taking risks the lender cannot see, and what would remain if the business failed. Specialization in an industry or region can narrow that information gap. Distance and a broadening portfolio can widen it.
Collateral is one response. A borrower pledges an asset that gives the lender a claim beyond the borrower’s expected operating cash flow. The collateral may be a current asset, real estate, intellectual property, machinery, or another hard asset; the loan-level data do not reliably identify which type secures every loan. The economic function is more general than the asset label. Collateral can reduce the lender’s exposure to informational uncertainty, align incentives, and improve priority or recovery if the borrower defaults. It does not replace underwriting, and it does not prevent the pledged asset from losing value.
The observed expansion of secured direct lending is substantial. In the BIS loan-level data, which cover U.S. deals from the first quarter of 2000 through the fourth quarter of 2024, outstanding secured loans reached about $370 billion. Outstanding unsecured loans exceeded $300 billion. Secured loans therefore represented more than half of outstanding direct lending by 2024, compared with less than 10 percent in the 2000s. The increase was broad across industries rather than the result of one sector suddenly changing its borrowing habits.
That shift changes how “private credit” should be pictured. An unsecured cash-flow loan depends mainly on the lender’s assessment of the borrower’s future operations. A secured loan still depends on those operations, but it adds a second state variable: the value and enforceability of the pledged asset. The lender may have more protection in a default, yet the borrower’s access to credit can now respond to asset-price movements that have little to do with current operating performance. The contract has gained a buffer and a source of cyclicality at the same time.
The loan terms are consistent with selection into secured lending. Compared with unsecured loans, secured loans in the study were somewhat smaller on average, carried higher spreads, and had somewhat longer maturities. These differences remained when the researchers compared borrowers in the same metropolitan statistical area and industry. A higher spread on a secured loan does not mean the collateral failed to help. It may mean that riskier borrowers are more likely to need or be asked to pledge collateral in the first place. For the same borrower, better collateralization can reduce pricing even though the population of secured loans contains more borrowers with difficult risk profiles.
This distinction between selection and protection is central. Imagine two firms with similar financing needs but different access to information. The lender that knows an industry well may rely more on business knowledge and relationship monitoring. A lender farther away or less specialized may ask for collateral because it has less confidence in its own assessment or because recovery rights matter more if the assessment is wrong. The collateral requirement then tells us something about the lender-borrower information relationship, not simply about the borrower’s asset wealth.
A stylized example makes the two comparisons clearer. Suppose a middle-market borrower approaches a specialized local fund and a larger fund whose portfolio spans many industries and regions. The local fund may be able to interpret the borrower’s customers, suppliers, and operating model with less reliance on a pledged asset. The larger or more distant fund may prefer a secured structure. If the larger fund’s loans carry higher spreads on average, that does not show that collateral caused the higher price. The two lenders may be selecting different risks or serving borrowers with different information conditions. To learn what collateral does for pricing, the useful comparison is closer to the same borrower or the same borrower environment with and without the protection—not simply the average secured loan versus the average unsecured loan.
The reverse can also happen in a downturn. A secured lender may have a better claim if the borrower fails, but the value of the asset supporting that claim can fall at the same time that new lending becomes scarce. A collateral requirement can therefore improve the lender’s position relative to an unsecured claim without preventing losses or preserving the borrower’s access to fresh credit. This is why recovery protection and credit supply should be analyzed separately.
The BIS evidence points in that direction. Loans were less likely to be secured when the lender was more specialized in the borrower’s industry or metropolitan area, and more likely to be secured when the lender and borrower were farther apart. Over the same broad period, lenders became less specialized and the average lender-borrower distance increased. The researchers calculate that the roughly 50-percentage-point fall in average industry or metropolitan-area specialization could account for an increase of about 10 to 11 percentage points in the share of secured loans, using their estimated relationships. That is a back-of-the-envelope interpretation, not a complete causal account of the market’s growth.
Scale can therefore create a paradox. Larger funds may have more capital and the ability to diversify across borrowers, but a broader portfolio can reduce the narrow industry or local knowledge that supports relationship lending. If information becomes harder to acquire at the same depth, collateral can become a way to manage the resulting uncertainty. The shift toward secured lending may then reflect the institutional evolution of private credit as it grows, not just a sudden change in borrower quality.
The diagram separates three relationships that are easy to collapse. Specialization and distance are information conditions. Collateral is a contractual response. House prices and monetary policy can change the value of assets used in that contract. The final effect on borrower cash flow or default is not established by the secured-lending relationship alone.
The collateral channel runs through house prices
If a material share of secured loans relies on real estate or other assets whose value changes over time, then the private-credit supply can move with those values. The BIS paper cannot identify the type of collateral for every loan, so it does not offer a loan-by-loan inventory of property-backed lending. Instead, it tests whether secured and unsecured direct lending respond differently to local house prices. The comparison is useful because a stronger response in the secured segment is consistent with collateral values affecting the amount lenders are willing to provide.
The empirical design tries to separate the house-price relationship from a simple story about a booming local economy. A growing metropolitan area might raise both property prices and demand for loans. To reduce that concern, the researchers compare secured and unsecured lending within the same areas and use detailed time controls. They also include lender-level time variation, so the comparison can ask whether the same fund changes lending differently across places as their house-price paths diverge. These controls do not make every confounder disappear, but they make the measured contrast more informative than a raw correlation between housing and credit.
The study also addresses reverse causality. It is possible, in principle, that a large expansion of secured lending raises local investment and employment, which then pushes up house prices. The researchers instrument local house prices using housing-supply elasticity interacted with the federal funds rate. The intuition is that a change in interest rates can create housing demand, but the resulting price response is larger where new housing is difficult to supply. That variation can predict house prices without relying entirely on the local credit expansion that the study is trying to explain.
The result is economically meaningful but precisely bounded. A one-standard-deviation increase in the local house-price index was associated with an insignificant 10.7 percent increase in unsecured lending and an additional statistically significant 21 percent increase in secured lending. The instrumental-variable estimates were close to the ordinary least-squares estimates, and the paper reports strong first-stage statistics. The safest reading is not that every house-price increase creates a 21-percent private-credit boom. It is that, in this U.S. sample and identification design, secured lending responded materially more than unsecured lending when local property values rose.
The comparator matters. The 21 percent is an additional response for secured lending relative to the unsecured segment; it is not a claim that secured lending always grew by exactly 21 percent, nor that unsecured lending was unaffected in every specification. The 10.7 percent unsecured estimate was statistically insignificant, while the additional secured response was statistically significant. Saying that the secured channel was stronger preserves both facts: lending can move with local conditions in more than one segment, but the difference between the segments is the main result.
The study’s design also explains why the finding is informative without making it definitive. Fixed effects and the housing-supply instrument address important sources of omitted-variable and reverse-causality concern, but no design turns an imperfect collateral measure into a complete account of every loan. The evidence identifies a relationship between local asset values and lending behavior under the study’s assumptions. It does not identify the exact collateral behind each loan, the full profitability of that lending, or the eventual default rate of the borrowers who received it.
The relationship was also stronger in industries that generally relied more on collateral. That additional pattern strengthens the interpretation that the result is tied to the collateral channel rather than merely to a general local expansion in all kinds of credit. At the same time, the evidence does not establish that the collateral was always real estate, or that the same response holds for every private-credit strategy, country, fund, or point in the cycle.
The result changes the way monetary policy can reach private credit. A rate change affects borrowers directly through financing costs, but it can also affect asset prices. If higher rates weaken house prices, the collateral value supporting secured borrowing may fall even before a borrower misses a payment. If lower rates lift house prices, lenders may be more willing to extend secured credit. This is an indirect channel: the loan is originated by a nonbank, yet its availability can depend on an asset market influenced by policy and broader financial conditions.
That sensitivity can help borrowers in an expansion. Higher collateral values may let firms borrow more or obtain financing on terms that would otherwise be unavailable. But the same mechanism can tighten credit in a downturn. A lender that relies on collateral may reduce new lending when asset values fall, even if the borrower’s current operations have not yet collapsed. The borrower then faces a more difficult refinancing environment, and the reduction in credit can reinforce the original slowdown. The evidence supports this possibility as a mechanism; it does not quantify the size of a full downturn loop.
Collateral also changes the distribution of risk among the parties. A borrower may gain access to funds but put a valuable asset at risk. A lender may gain a recovery claim but still face uncertain valuation, enforcement, and timing. An investor in the fund may see a portfolio that appears better protected while remaining exposed to the fund’s broader liquidity and valuation process. The existence of collateral therefore answers one question—what may support recovery—not every question about who can sell, refinance, or absorb a loss.
The study’s limits are part of its result. Its secured and unsecured classifications are informative, but the collateral type is not reliably observed. Its estimates describe direct lending in a particular U.S. loan-level sample from 2000 through 2024, not the entire private-credit universe. The house-price evidence is strong enough to support a collateral-channel interpretation, but it does not prove that collateral is the only reason secured lending expanded or that all nonbank credit will become equally cyclical.
The essential judgment is therefore two-sided. Collateral can help a lender manage information problems and improve recovery prospects, which can expand financing for borrowers that lack a clean cash-flow story. Yet the more credit depends on collateral values, the more its supply can respond to asset-price conditions. Private credit may be outside traditional bank balance sheets, but this mechanism still connects it to the broader credit cycle. The next task is to combine this loan-level channel with funding design, disclosure, and bank connectivity to see where the system’s vulnerabilities accumulate.