Credit Constraints and the Cyclicality of R&D Investment: Evidence from France

GAICIL Insights

Recessions are supposed to be the cheapest time to invest in the future. For many firms, they are the time the future gets canceled. Credit Constraints and the Cyclicality of R&D Investment: Evidence from France, by Philippe Aghion, Philippe Askenazy, Nicolas Berman, Gilbert Cette and Laurent Eymard, uses a French firm-level dataset of 13,000 firms over 1994-2004 to analyze how credit constraints shape R&D behavior over the cycle, a question with consequences for long-run productivity.¹

The results are striking. R&D investment is countercyclical without credit constraints and becomes procyclical as firms face tighter constraints. Unconstrained firms raise their R&D share in downturns, when the opportunity cost of diverting resources from production is lower. Constrained firms cannot, because their access to finance falls during recessions. Procyclicality appears only for firms in sectors heavily dependent on external finance or characterized by low asset tangibility. The textbook logic works only for firms with a bank willing to say yes.

The asymmetry is the damage. In more credit-constrained firms, R&D plummets in recessions and does not rise proportionally in upturns. Downturns leave lasting damage on constrained firms' R&D. Results carry over from the R&D share of investment to R&D levels, so the share's procyclicality is not primarily driven by variation in physical investment. The authors also examine how productivity growth responds to sales volatility interacted with credit constraints. The lab closes in the slump and only partly reopens in the boom.

Credit constraints thus turn recessions into lasting innovation losses. Without constraints, firms would use downturns to invest in R&D, smoothing the cycle's effect on long-run growth. With constraints, R&D falls when it would be most valuable, and the losses are not recovered. Every recession leaves a hole in the innovation record.

The parallel with today's technology is close. AI firms carry every risk factor. AI investment shares the traits linked to procyclical R&D: heavy reliance on external finance and low asset tangibility. Software, data and skills are hard to pledge as collateral. AI-intensive firms, especially young ones, may see R&D cut sharply in downturns, and the asymmetry suggests cuts may not be fully reversed in recoveries. The next recession could set AI research back further than the headlines suggest.

Policy follows. Easing credit constraints in downturns could protect AI investment. Countercyclical R&D support, through tax credits, grants or financing programs that expand in recessions, could offset constraint-induced procyclicality, sustaining innovation through the cycle and limiting recessions' lasting damage. An R&D credit that grows in bad times is worth more than one that is flat.

Tangibility deserves emphasis. Financial systems built around tangible collateral may fit poorly with the intangible-intensive AI economy. Instruments able to value and finance intangible assets, such as intellectual property and data, could reduce credit constraints for AI firms. That structural reform would complement countercyclical policy. Lending against code is a frontier in finance that AI needs crossed.

Macro and innovation policy meet here. Recessions affect long-run growth through R&D, and credit constraints amplify the effects. Monetary and fiscal policies stabilizing credit conditions support innovation. Macroeconomic decisions carry innovation consequences that deserve explicit attention.

The data are deep. Observing thousands of firms over a decade relates R&D behavior to firm-specific credit constraints and sectoral traits, valuable evidence for targeted policy.

Firms can draw a lesson too. Maintaining finance access through downturns lets a firm sustain R&D while competitors cut back, potentially gaining an edge in the recovery. In fast-moving AI markets, the ability to keep investing through a downturn may be decisive. A cash buffer is also a research strategy.

Stepping back, credit constraints make R&D procyclical, so innovation collapses in recessions without full recovery in booms. As AI investment, intensive in intangibles and external finance, becomes central to growth, policies easing credit constraints and supporting R&D through the cycle will help protect the long-run productivity gains AI promises.

Governments and investors might start here: the right moment to double down on AI research support may be the moment budgets and markets are telling everyone to cut. Countercyclical courage is what turns a recession into an opportunity.