Interest rate parity and the behavior of the bid-ask spread

We use a stochastic frontier regression model to test interest rate parity (IRP) with bid- ask spreads for the Belgian franc, the Deutschmark, and the Swiss franc. The forward markets tested have become efficient in the sense that IRP holds well. The bounds provided by IRP do not appear to be binding, however. We provide evidence that in spite of the overall goodness of fit of the model, the arbitrage margins are sometimes violated, implying possible arbitrage opportunities. The percentage bid-ask spread is consistently higher for the Belgian franc than for the Deutschmark and the Swiss franc. Spreads are increasing functions of the time to maturity and volatility. Spread-size clustering is more severe than price-level clustering and appears to be inversely related to volatility and positively related to the trading volume. We find no evidence of significant calendar-day effects on spread size.