We have conducted a VLA search for radio pulsars at the positions of 44 nearby OB runaway stars. The observations involved both searching images for point sources of continuum emission and a time series analysis. Our mean flux sensitivity to pulsars slower than 50 ms was 0.2 mJy. No new pulsars were found in the survey. The size of the survey, combined with the high sensitivity of the observations, sets a significant constraint on the probability, $f_p$, of a runaway OB star having an observable pulsar companion. We find $f_p \le 6.5$\% with 95\% confidence, if the general pulsar luminosity function is applicable to OB star pulsar companions. If a pulsar beaming fraction of \onethird\ is assumed, then we estimate that fewer than 20\% of runaway OB stars have neutron star companions, unless pulsed radio emission is frequently obscured by the OB stellar wind. Our result is consistent with the dynamical (or cluster) ejection model for the formation of OB runaways. The supernova ejection model is not ruled out, but is constrained by these observations to allow only a small binary survival fraction, which may be accommodated if neutron stars acquire significant natal kicks. According to Leonard, Hills and Dewey (1994), a 20\% survival fraction corresponds to a 3-d kick velocity of 420 km s$^{-1}$. This value is in close agreement with recent revisions of the pulsar velocity distribution.