This study investigated the effects of mathematics test format on item characteristics in high school. The test format varied across classes. In Study 1, high school students in three out of five classes completed the fill-in-the-blank and multiple-choice test, while the students in the remaining two classes completed the descriptive test. The results revealed that fill-in-the-blank and multiple-choice tests had a higher rate of correct answers and a lower nonresponse rate compared with descriptive tests. However, there was little difference in item discrimination. There were many items in which inter-group differences were confirmed, particularly in multiple-choice items and items for which the meaning of the symbols in the question could be deduced from the number of digits in the answer column; therefore, these items were excluded from Study 2. In Study 2, students in three of the five classes completed the fill-in-the-blank test, while the students in remaining two classes attempted the descriptive test. The results revealed that there was little difference in the item discrimination, and there were no differences in the non-response rates. There was a group difference in the correct answer rates when the number of digits in the correct answer was different from the typical mistaken answer. These findings suggest that if multiple-choice items that promote correct answers by guessing and items that have different number of digits in correct answer from the typical mistaken answer are eliminated, there may be no difference in item characteristics depending on the test format.
