For decades, economists have relied on a standard tool based on average effects to understand the workings of markets in which individuals and organisations match with each other. In the world of banking, for example, this tool assumes that a bank’s behaviour is consistent across all of its customers, and that a firm’s appetite for borrowing is the same for each bank with which it interacts. The tool also assumes that if a bank decides to scale back lending, it does so for everyone equally.
But we know that reality is more complex. Banks specialise in certain regions or industries, and firms often have deep ties to specific lenders.
In a new study with Olivier De Jonghe (National Bank of Belgium), we reveal that the ‘average’ behaviour of a bank or firm is only half the story. The other half – the distinct effect specific to each relationship – is just as important, but the traditional mathematical analysis has been ignoring it (De Jonghe and Lewis, 2026).
Why traditional models fail
When researchers use traditional models to measure ‘credit demand’ (how much firms want), they may get results that defy common sense. For example, when applied to our study’s real-world data, the leading model suggests that when demand for loans goes up, interest rates go down. This violates all economic theory in a standard market, where higher demand usually drives prices up.
We explain why this happens: banks’ desire to lend to a firm is determined by both the banks’ own considerations and their assessment of that firm’s characteristics. Similarly, firms’ desire to borrow from a particular bank is determined both by their credit needs and that bank’s identity.
In simple terms, banks do not view all firms equally, and vice versa. Assuming that they do conflates banks’ supply with firms’ demand and can have serious unintended consequences when a model is applied to data.
A smarter way to measure the economy
We propose a new solution that uses covariance restrictions. Instead of looking at just loan volume, as in previous studies, we look at the relationship between volume and price (interest rates). By analysing how these two variables move together across a vast network of thousands of firms and banks, we can mathematically unscramble the forces of demand and supply, without making the strong assumptions of previous approaches.
Crucially, our method doesn’t require complex assumptions about why a bank likes a firm. It simply looks at the mathematics of the relationships: if a bank behaves differently towards two similar firms, the model uses that difference to identify a relationship-level ‘shock’ to credit supply.
Europe’s credit markets: a large-scale test case
We apply our method to the AnaCredit dataset, which covers almost every significant corporate loan between 2019 and 2023 in nine Eurozone countries, including France, Germany, Italy and Spain. This period included the Covid-19 pandemic, the subsequent inflation spike and the aggressive interest rate hikes that followed.
Our analysis produces several key insights.
Massive variation
We find that the differences within a single bank’s relationships were often just as large as the differences between two different banks.
National differences
Even though the Eurozone uses one currency, the slopes of the supply and demand curves were very different across countries. For example, during the pandemic, German firms were desperate for cash regardless of the cost (inelastic demand), while in other countries, borrowing was more sensitive to price.
A hidden contraction
Our most significant finding concerns the 2022 monetary policy tightening. Traditional methods suggest that the credit crunch wasn’t that bad, even for the firms that were most affected. But our new method reveals a sharp and persistent drop in credit supply for firms exposed to interest rate risks.
Winners and losers from interest rate hikes
By looking at the relationship level, our research shows how the European Central Bank’s rate hikes filtered through the economy. Firms with floating-rate contracts (loans where the interest rate jumps when market rates rise) were hit by a double penalty. Not only did their debt become more expensive, but banks also became less willing to lend to them, which is likely to have been because they were seen as riskier.
Conversely, banks increased the relative supply of credit to firms with fixed-rate contracts. Why? Because those firms were insulated from rising costs and were seen as safer bets. This kind of detail – who gets the money and why – is completely lost in traditional models based on average effects.
A new toolkit for policy
Our study concludes that making use of relationship-level data is essential for understanding the effects of policy decisions. If policy-makers rely on models that ignore these specific links, they might underestimate the pain being felt by certain businesses or fail to spot how financial risks are becoming concentrated in specific parts of the market.
The new method provides a clearer way to track the pulse of the economy. It proves that in the world of finance, who you borrow from – and your relationship with them – can be just as important as how much you are borrowing.
