Abstract
This paper investigates the likelihood of rounding as a source of bias in survey data. The analysis uses thirty years of March Current Population Survey data, the 1990 Census, and the 1996 Survey of Income and Program Participation matched to Social Security Administration data to measure the magnitude and direction of rounding by worker characteristic. The results show three distinct trends. First, more people round their earnings today than in the past. Second, groups that are more likely to round their earnings include men, people in their prime working years (35 to 64), and those with high levels of education or with high levels of earnings. And third, using the matched data, these respondents appear to round their earnings down.