Recent pulsar surveys have increased the number of observed double neutron stars (DNS) in our galaxy enough so that observable trends in their properties are starting to emerge. In particular, it has been noted that the majority of DNS have eccentricities less than 0.3, which are surprisingly low for binaries that survive a supernova explosion that we believe imparts a significant kick to the neutron star. To investigate this trend, we generate many different theoretical distributions of DNS eccentricities using Monte Carlo population synthesis methods. We determine which eccentricity distributions are most consistent with the observed sample of DNS binaries. In agreement with Chaurasia & Bailes (2005), assuming all double neutron stars are equally as probable to be discovered as binary pulsars, we find that highly eccentric, coalescing DNS are less likely to be observed because of their accelerated orbital evolution due to gravitational wave emission and possible early mergers. Based on our results for coalescing DNS, we also find that models with vanishingly or moderately small kicks (sigma < about 50 km/s) are inconsistent with the current observed sample of such DNS. We discuss the implications of our conclusions for DNS merger rate estimates of interest to ground-based gravitational-wave interferometers. We find that, although orbital evolution due to gravitational radiation affects the eccentricity distribution of the observed sample, the associated upwards correction factor to merger rate estimates is rather small (typically 10-40%).