As Code Monkey cited, it is implementation defined and implementation varies -- it isn't just a BigEndian/LittleEndian and charset difference.  I've tested four implementations (all using ASCII) with the program
#include <stdio.h>
int main()
{
    unsigned value = 'ABCD';
    char* ptr = (char*)&value;
    printf("'ABCD' = %02x%02x%02x%02x = %08x\n", ptr[0], ptr[1], ptr[2], ptr[3], value);
    value = 'ABC';
    printf("'ABC'  = %02x%02x%02x%02x = %08x\n", ptr[0], ptr[1], ptr[2], ptr[3], value);
    return 0;
}
and I got four different results
Big endian (AIX, POWER, IBM compiler)
'ABCD' = 41424344 = 41424344
'ABC'  = 00414243 = 00414243
Big endian (Solaris, Sparc, SUN compiler)
'ABCD' = 44434241 = 44434241
'ABC'  = 00434241 = 00434241
Little endian (Linux, x86_64, gcc)
'ABCD' = 44434241 = 41424344
'ABC'  = 43424100 = 00414243
Little endian (Solaris, x86_64, Sun compiler)
'ABCD' = 41424344 = 44434241
'ABC'  = 41424300 = 00434241