Uncovered interest rate parity provides a crucial theoretical underpinning for many modelsin international finance and international monetary economics. Though theoretically sound,this concept has not been supported by the empirical evidence. Typically, econometrictests not only reject the null hypothesis, but also find significant slope coefficients with thewrong sign. Following the approach employed in Kool and Thornton (2004), we show thatthe empirical procedure conventionally used to test for UIP may produce biased slopecoefficients if the true data-generating process slightly differs from the theoreticallyexpected one. Using monthly data for ten industrial countries during the period W75-2004,we estimate the UIP relation for all possible bilateral country pairs for each of the six fiveyearsub-periods. The evidence supports the biasedness hypothesis: when the interest ratevolatility of the anchor country is very high (very low), this estimation procedure reportssignificantly higher (lower) slope coefficients.