The impact factor's Matthew Effect: A natural experiment in bibliometrics