This paper predicts recessions using the dispersion of deposit rates offered by banks on insured deposits. An increase in the dispersion of deposit rates can accurately predict recessions over long time horizons at the county, state, and national levels. We find that the growth of deposits, particularly uninsured deposits of riskier banks, slows down at the onset of a downturn, regardless of whether the downturn was preceded by a credit boom. In turn, riskier banks increase their deposit rates to attract more funding to support their balance sheet. The resulting increase in the dispersion of deposit rates predicts an impending economic downturn.

I show that the long-term decrease in the nominal short rate since the 1980s contributed to a decline in banks' supply of business loans, firm investment and new firm creation, and an increase in banks' real estate lending. The driving force behind these relationships was the shift in banks’ funding mix from time deposits (CDs) to savings deposits, which was caused by the decrease in the nominal rate. I show that banks finance business lending with time deposits because of their matching interest-rate sensitivity and liquidity. A lower nominal rate reduces the spread on liquid deposits (e.g., savings deposits), leading households to substitute towards them and away from illiquid time deposits. In response to an outflow of time deposits, banks cut the supply of business loans and increase their price. The decrease in business lending leads to reduced investment at bank-dependent firms and a lower entry rate of firms in industries that are highly reliant on external funding. I document these relationships both in the aggregate, and in the cross-section of banks, firms and geographic areas. For identification, I exploit cross-sectional variation in banks' market power and business credit data. I develop a general equilibrium model which captures these relationships and shows that the transmission mechanism I document is quantitatively important.