If people tend to disproportionately forget small price changes and remember large price changes when taking surveys, it could skew the results this way. Both

If people tend to disproportionately forget small price changes and remember large price changes when taking surveys, it could skew the results this way. Both the distribution of observed price changes, and the probability of forgetting/remembering, could be different between men and women. For example, suppose women buy equal amounts of two different goods X and Y. On even-numbered years, manufacturers raise the price of X by 6% and keep price steady on Y. On odd-numbered years, manufacturers raise price on Y by 6% and keep price steady on X. Actual inflation is 3% but women see a price jump of 6% per year in one of the products they buy. If they only notice/remember changes in prices and ignore/forget prices staying the same, then they'll report inflation as 6%. Men buy equal amounts of X, Y and Z, where Z increases by 2% per year. Thus, men report an average price increase of (2% + 6%)/2 = 4% per year, again ignoring the non-inflating good. All these numbers and effects are totally made-up speculation, but the idea seems sound and reasonable.