On the running electromagnetic coupling constant at $M_Z$

We present a discussion on how to define the running electromagnetic coupling constant at $M_Z$ or some other intermediate scale as e.g. $m_\Upsilon$. We argue that a natural definition consistent with general requirements of the renormalization group should be based on Euclidean values of the momentum of the photon propagator as the appropriate scale. We demonstrate in an explicit example of evaluating the running coupling constant at the scale of the $\Upsilon$ resonance mass that the usual definition of the hadronic contribution with a principal value prescription is inconsistent. In the determination of the value of $\alpha$ at $M_Z$ the numerical difference due to using a Euclidean definition rather than the principal value one is comparable in size to the errors caused by existing experimental and QCD inputs to the evaluation of $\alpha(M_Z)$.