Modeling central bank intervention as a threshold regression: evidence from Turkey∗

We conjecture that the Central Bank of Turkey intervenes differently depending on whether a single measure of market disorder exceeds a threshold value or not. As possi-ble such measures we consider deviation from the trend and excess volatility of the spot exchange rate. The questions we ask are what is the threshold value and whether the mea-sure of market disorder is consistent with the announced policy of the Bank. Interestingly, during the managed float period we find no support for a threshold in the Bank’s reaction function; during the free float period we find evidence of a threshold in volatility, which is in line with the Bank’s policy objectives.