On the monotonicity of tail probabilities