We have carried out a search for galaxies at z ~ 7-10 in ~14.4 sq. arcmin of new NICMOS parallel imaging taken in the Great Observatories Origins Deep Survey (GOODS, 5.9 sq. arcmin), the Cosmic Origins Survey (COSMOS, 7.2 sq. arcmin), and SSA22 (1.3 sq. arcmin). These images reach 5 sigma sensitivities of J110 = 26.0-27.5 (AB), and combined they increase the amount of deep near-infrared data by more than 60% in fields where the investment in deep optical data has already been made. We find no z>7 candidates in our survey area, consistent with the Bouwens et al. (2008) measurements at z~7 and 9 (over 23 sq. arcmin), which predict 0.7 galaxies at z~7 and <0.03 galaxies at z~9. We estimate that 10-20% of z>7 galaxies are missed by this survey, due to incompleteness from foreground contamination by faint sources. For the case of luminosity evolution, assuming a Schecter parameterization with a typical phi* = 10^-3 Mpc^-3, we find M* > -20.0 for z~7 and M* > -20.7 for z~9 (68% confidence). This suggests that the downward luminosity evolution of LBGs continues to z~7, although our result is marginally consistent with the z~6 LF of Bouwens et al.(2006, 2007). In addition we present newly-acquired deep MMT/Megacam imaging of the z~9 candidate JD2325+1433, first presented in Henry et al. (2008). The resulting weak but significant detection at i' indicates that this galaxy is most likely an interloper at z~2.7.