To some extent, selfishness. Remember that natural selection acts within species.
Combat is dangerous. Very dangerous. Even if you are not immediately killed and successfully fend off a predator, wounds can weaken you, damage your senses, and/or kill over time with infection.
If you are a zebra, or a sheep, and you are in a group of other conspecifics, the safest action for you may be to not be the individual that is attacked. Sometimes, you might achieve this by simply advertising being a tough target. Other times, your best bet is to run away. This concept is sometimes humorously discussed in the context of humans and bears: you don't have to be faster than the bear, you have to be faster than the slowest person with you. This is called selfish herd behavior.
Fleeing predation is likely to be the default behavior
If you imagine an allele that changes an animal's behavior to combat an attacking predator (when other conspecifics predominantly run from predators), it is difficult for this allele to grow in the population. Individuals with the allele will get in fights, and although they may reduce predation of the herd as a whole, they will individually be selected against.
Similarly, imagine an allele that changes an animal's behavior to run from an attacking predator (when other conspecifics predominantly attack predators). Individuals with this allele will avoid fights, stay healthier, and breed better than their conspecifics that are getting beaten up, and eventually the herd will be mostly escapers.
Exceptions to the default
However, as pointed out in a comment by gardenhead, not all prey try to escape.
To avoid the 'default' escape behavior, there must be some other motivation: fighting back has to be special some how. In the context of group selection, fighting back may be selected for if it means improved survival of related conspecifics. This behavior is often seen in defense of offspring, but can expand to larger related groups. However, even defense of offspring can be risky, and it is not uncommon in nature for mothers to abandon offspring they cannot defend. It may be safer and more productive in the long run to survive and produce new offspring next year.
Another alternative would be some sort of social/sexual selection, where strong individuals who defend the group (likely males) have increased likelihood to mate.
Fischhoff, I. R., Dushoff, J., Sundaresan, S. R., Cordingley, J. E., & Rubenstein, D. I. (2009). Reproductive status influences group size and persistence of bonds in male plains zebra (Equus burchelli). Behavioral Ecology and Sociobiology, 63(7), 1035-1043.
Hamilton, W. D. (1971). Geometry for the selfish herd. Journal of theoretical Biology, 31(2), 295-311.
King, A. J., Wilson, A. M., Wilshin, S. D., Lowe, J., Haddadi, H., Hailes, S., & Morton, A. J. (2012). Selfish-herd behaviour of sheep under threat. Current Biology, 22(14), R561-R562.
Sjare, B., & Stirling, I. (1996). The breeding behavior of Atlantic walruses, Odobenus rosmarus rosmarus, in the Canadian High Arctic. Canadian Journal of Zoology, 74(5), 897-911.
Smith, W. P. (1987). Maternal defense in Columbian white-tailed deer: when is it worth it?. The American Naturalist, 130(2), 310-316.
Wood, A. J., & Ackland, G. J. (2007). Evolving the selfish herd: emergence of distinct aggregating strategies in an individual-based model. Proceedings of the Royal Society of London B: Biological Sciences, 274(1618), 1637-1642.