There are some extended ascii characters in the csv download. It seems jupyter doesn't mind, but if you're doing this on IDLE or some other thing with CPython, it balks on reading the file, starting at character 535. This can be avoided by adding """encoding='utf-8'""" to the open statement.