You may not need a balun, but you may still need a transformer (or isolator)to get a better match.  It depends on the S11 you can live with.  You need to determine the RF link requirements and concider how much reflected power the Tx can handle.  As a guide a 2:1 VSWR (-9.6 dB return loss) means approx 11.5% of the power from the Tx will be "reflected" back to the Tx.  A return loss of 7 dB is approx 2.6:1 VSWR and means approx 20% of the power will bw reflected back to where it came from.
Hope this helps.

Expanding on the previous post, it is worth clarifying that any power that happens to be reflected by the mismatched antenna back towards the transmitter is not necessarily lost.  As I'm sure almost everyone knows, once the steady state is achieved, the reflection simply changes the apparent loading* on the transmitter.

(*In the general case the transmitter load is not likely to be the same as the antenna impedance due to the effect of the length of the transmission line - per the Smith Chart.  In this particular case, where the antenna is reportedly connected 'directly', then the transmission line length is probably essentially zero - as compared to the wavelength of 915 MHz - and you don't need to refer to the Smith Chart.  This assumption should be confirmed...)

So long as the transmitter is kept within its performance specifications, then it will cope with the load and emit power in accordance with it performance characteristics.  There may be a power foldback circuit to deal with mismatched loads - the performance specification for the transmitter chip should inform on this point.  I believe that is the point that 'Photistor' was leading towards and it is perfectly correct.

Also, if the specified 'matched loss' of the transmission line is significant (probably not in this particular case), then the mismatch within the tranmission line can increase the line loss above its 'matched loss' specification.  This increased line loss is due to increased currents and voltages within the line resulting from the SWR.

In summary, even with reflections, the actual transmitter power output will either be radiated by the antenna, or it will be lost heating the transmission line.  This is obviously in keeping with the principle of the conservation of energy.

In this case, the only thing to check is the performance curves for the transmitter chip as applies to mismatched loads.

The point of this posting is just to clarify that the reflected power ('11.5%' or 'approx 20%' in the examples above) is not necessarily lost power.

With respect to baluns, if you don't think you need it - then you probably don't need it.

Suggestion: Sites addressing balun (impedance transformer) applications:
http://n-lemma.com/calcs/dipole/
http://www.educatorscorner.com/media/Exp101i.doc
http://www.educatorscorner.com/media/Exp101i.pdf
http://www.hard-core-dx.com/nordicdx/antenna/wire/t2fd.html
http://www.cebik.com/wbfd.html
etc.
The balun is often aligned with coax line.

"...balun (impedance transformer)..."

The word 'balun' is a contraction of the words 'balanced' & 'unbalanced'.  A balun isn't necessarily an 'impedance transformer'; in fact, 1:1 baluns are quite common.

Of course, baluns can also be 'impedance tranformers' at the same time, but this is not necessary.

I wonder at the wisdom of accepting unbalanced feed to a dipole.  The effect of reflected power on the transmitter power is not the issue, although the effect of the reactive load on spurious emissions is.  Also with a dipole, you will have unbalanced currents in the two arms and while this is generally not a problem in a simple antenna, it is in an array or a reflector based system where you'll end up with a skewed beam pattern.

In the case of a coaxial cable (unbalanced) feeding a dipole antenna (unbalanced) - if you can force the currents to remain inside the coaxial cable (as opposed to some current running along the outside), then the currents will also be forced to be balanced.  Thus the current into each leg of the dipole will be equal and opposite and everything will be hunky-dory.

Generally, if you can keep your coax cable perfectly normal (90 degrees) to the dipole, and everything else if perfectly symmetrical, then the currents should be equal and opposite on the two legs of the dipole.

The unbalanced feedline - the coaxial cable (by itself) - does not necessarily cause an unbalanced antenna.  The currents are still likely to be balanced.

If they're not, then balance can be accomplished by simply coiling the cable to provide a common mode choke a few feet (at HF) down from the feed point.  This coil might be termed a 'balun' if you're feeling generous.

The coil balun helps to ensure the currents are balanced when things are not perfectly symmtrical.  It does this by making sure that the currents are forced into the cable where nature ensures that they're balanced.