We present z > 9 candidates identified in Hubble Frontier Fields imaging of the first four clusters and blank parallel fields (two-thirds of the complete program). Based on the deeper Hubble imaging, we revise the redshift estimate of the CLASH z ~ 9.6 candidate MACS1149-JD to z = 9.2+0.2-0.5 (95% C.L.). We identify a new possible fainter z ~ 9 companion 3” away (~1 kpc in the source plane). And we discover a new z ~ 9.2 candidate in the MACS1149 parallel field. Combined with previously published candidates at z ~ 9.8 and 9.1 (in A2744 and its parallel field, respectively), these five z > 9 candidates fall below our published expectation of 8 – 47 at this stage in the program. We attribute some of this shortfall to incompleteness, which we have yet to quantify. At z ~ 8 (7.5 – 8.5), we detect 26 candidates down to F160W H < 28.7 AB, roughly one-third the ~82 we expect. If our z > 9 incompleteness is similar (~68%), our results would support the sharp drop in z > 9 number counts claimed by some (but not all) previous works and supported by several (but not all) theoretical models. Properly quantifying our incompleteness will require adding simulated high-redshift galaxies into the images and testing our recovery rate. Additionally, incorporating the deep Spitzer imaging into our analysis could potentially significantly improve our identification of z > 9 candidates by rejecting low-redshift (z ~ 2) interlopers. Data from the full Frontier Fields program will provide strong evidence for or against accelerated evolution and a sharp drop in the cosmic star formation rate density at z > 9.