The financial accelerator grew out of the observation that standard DSGE models could not explain the severity of financial crises. Bernanke, Gertler, and Gilchrist (1999) - hereafter BGG - embedded a costly state verification (CSV) debt contract into a New Keynesian framework, building on Townsend (1979), Gale and Hellwig (1985) and Bernanke and Gertler (1989); Carlstrom and Fuerst (1997) had already put the same contract into a real business cycle model. The CSV setup gives rise to an external finance premium that is inversely related to borrower net worth, creating an endogenous wedge between the risk-free rate and the cost of external funds. After 2008 it became the standard way to put credit spreads and borrower balance sheets into a DSGE model, because it gives falling asset prices and widening spreads a route back to investment and output.
The core mechanism works through three interlocking channels. First, entrepreneurs borrow to fund capital purchases, facing an optimal debt contract derived from costly state verification - the lender pays a monitoring cost μ when the borrower defaults, so the contract charges a spread above the risk-free rate to cover expected monitoring costs. Second, this external finance premium is a decreasing function of the entrepreneur's net worth relative to the capital stock: higher leverage means higher default probability, higher expected monitoring costs, and a wider spread. Third, the balance-sheet amplification loop closes the circle - when asset prices fall after an adverse shock, entrepreneurial net worth drops, the external finance premium rises, investment contracts, which further depresses asset prices and net worth. This positive feedback loop is the financial accelerator proper.
Financial-accelerator blocks spread through policy institutions after 2008. The New York Fed's DSGE model (Del Negro, Giannoni and Schorfheide 2015) adds a BGG entrepreneurial sector to a Smets-Wouters core. Christiano, Motto and Rostagno - two of them ECB economists - built the contract into an estimated medium-scale model (2010) and added risk shocks (2014); versions of that model are used in ECB policy analysis. The Bank of England's COMPASS model, by contrast, contains no financial-frictions block - a much-discussed criticism - and financial frictions are instead handled by other models in its forecasting suite. Iacoviello (2005) took the parallel collateral-constraint route into housing, tying household borrowing limits to property values; that model, not BGG's contract, became the standard tool for mortgage and loan-to-value questions. Models from both traditions are used for macroprudential stress testing, countercyclical capital buffer analysis, and the evaluation of credit policy.
The literature branched in several directions after BGG. Kiyotaki and Moore (1997) developed a parallel collateral-constraint tradition in which borrowing is capped by the value of collateral - a limit that comes from lenders' inability to force repayment - rather than priced through a premium. Gertler and Kiyotaki (2010) and Gertler and Karadi (2011) shifted the friction from the entrepreneur to the banking sector, creating models where bank leverage constraints drive credit supply. Jermann and Quadrini (2012) introduced financial shocks - disturbances to the borrowing constraint itself - and showed they account for a substantial share of business cycle variation. The debate between CSV-based and collateral-constraint-based approaches, and whether the friction belongs on the borrower or the intermediary side, remains active.