International Spillovers and Local Credit Cycles

Using the universe of Turkish corporate credit transactions matched to bank balance sheets over 2003-13, this paper shows that an easing of global financial conditions lowers domestic borrowing costs and raises lending mainly through domestic banks funded in international wholesale markets, that local currency borrowing cheapens more than foreign currency borrowing because the UIP premium moves with the cycle, and that collateral constraints do not relax during the boom. The data are the Turkish credit register collected by the Banking Regulation and Supervision Agency – roughly 53 million loan records, aggregated to 18.3 million firm-bank-currency-quarter observations plus 10.0 million individual new loan issuances – carrying interest rates, maturity, posted collateral, currency and bank-assigned risk measures, merged with quarterly bank balance sheets. Four facts follow, corresponding to four sets of regressions. First, in firm-bank panel regressions with firm-by-bank fixed effects, macro controls and bank characteristics, the elasticity of the loan interest rate to log(VIX) is 0.019 – implying a one percentage point fall in the average firm’s borrowing rate over the interquartile range of log(VIX) – and the loan-volume elasticity is minus 0.067, which the paper’s aggregation exercise translates into 43 percent of observed cyclical aggregate corporate loan growth. Second, the transmission runs through domestic banks reliant on non-core wholesale funding rather than through foreign banks: the interaction of a high-non-core dummy with log(VIX) is 0.013 with firm-by-quarter fixed effects absorbing credit demand, almost as large as the whole macro elasticity, and without those effects the implied interest-rate elasticity for high-non-core banks is double that for low-non-core banks (0.03 against 0.015). Strikingly, the loan-VIX elasticity is negative and strongly significant for domestic banks but “slightly positive and statistically insignificant for foreign banks” – “a result that differs drastically from the literature that focuses on the role of foreign banks.” Third, foreign currency loans carry a 7 percentage point average price advantage that widens to 8 points in high-VIX episodes and narrows to 6 in low-VIX ones, so lira borrowing cheapens relatively during booms, consistent with the aggregate UIP premium co-moving with the VIX; but with firm-by-bank-by-quarter effects the differential becomes insignificant, so “for the same firm borrowing from the same bank over time in different currencies, there is no UIP deviation.” Fourth, using posted collateral at origination with firm-by-bank-by-month effects, there is “no significant relationship between the collateral-to-loan ratio and loan volumes in all columns,” and none in high or low VIX episodes, while a horse race shows the interest rate strongly significant and collateral never significant – so “credit growth during low VIX episodes is driven by low interest rates, regardless of collateral values.” The paper’s scope is one country over one sample, and identification rests on cross-bank heterogeneity in funding structure rather than on an exogenous shock to individual banks; the authors also report a battery of null results – no effect of exchange rate fluctuations interacted with firm currency mismatch, no role for bank leverage once non-core funding is included – and frame the collateral result as a demand on theory rather than a settled mechanism.

Şebnem Kalemli-Özcan | Macro Paper Warehouse

International Spillovers and Local Credit Cycles