The New Approach to Modeling Foreign Exchange Intervention Rationales

Many papers which treat about foreign exchange intervention approach the problem by explaining the intervention with economic variables such as exchange rate volatility or exchange rate misalignment. Intervention is considered as binary variable. In a binary choice model  takes on value 1 if intervention is conducted ( ) and 0 otherwise. Monetary authorities can sell or purchase foreign currency, therefore two types of intervention are distinguished. It is obvious that the decision of the Central Bank on purchase is determined  by different values of economic categories than the decision on sale. Binary choice model is inappropriate because sale ( ) shall not be treated the same as purchase ( ). In order to avoid this mismatch, ordered data model can be applied. In such a model  takes value 1 if foreign currency is sold, -1 if foreign currency is purchased and 0 otherwise. This solution, although it captures different types of intervention, is not optimal because it excludes the differentiation between purchased and sold amounts of foreign currency. Furthermore,  classical regression model can-not be applied, because 0 value appears in many countries for many periods. Tobin (1958) considered censored data model, in which for many observations limit value 0 is observed. But in our model value 0 is observed for many periods , however this value is not a limit.                                                                                                                 This paper proposes a new type of tobit model, where the 0 value is observed for many periods, but it is not a limit value. Empirical investigation has been done for foreign exchange intervention in Slovakia before entering the ERM II and after joining this mechanism. On that basis, parameters of the model have been estimated by maximum likelihood. Asymptotic distribution of the ML-estimator in the case of stationary and integrated covariates is derived.           Empirical investigation concerns foreign exchange intervention in Slovakia before entering the ERM II and after joining this mechanism. The results of a different paper [Grabowski (2009)], in which equilibrium exchange rate for Slovakia was calculated using CHEER approach, are used here. We expect that an impact of exchange rate volatility and exchange rate misalignment on the probability of intervention increased after joining the ERMII.  Next important economic category, which has very significant meaning in explaining the probability and the size of intervention is so called “distance from band”. If the exchange rate was close to margin, then the National Bank of Slovakia conducted foreign exchange intervention. First trials conducted by authors justify these hypotheses.                                           Poland, Czech Republic and Hungary are still prior to joining the ERM II. Before the situation on the foreign exchange market in these countries will be analyzed and the risk of intervention will be calculated, the case of Slovakia should be analyzed. Slovakia seems to be a good study to refer to. Since foreign exchange intervention is a specific variable and differs from classical binary variable and censored variable, new type of tobit model has to be proposed.