Studies find price increases for additions to the S&P 500 index but no decreases for deletions. Additions come with good earnings news, suggesting these studies are not just measuring an indexing effect. We develop a regression discontinuity design using Russell Indices for cleaner identification. Stocks are assigned to indices based on their end-of-May market capitalizations. Stocks ranked just below 1000 are in the Russell 2000. The indices are value-weighted so these stocks receive index buying whereas those just above 1000 have close to none. Using this random assignment, we find price effects for both additions and deletions.