Can radio waves be formed into a pencil beam?

Laser light is spatially and temporally coherent. The stimulated emission is mainly responsible for the temporal coherence.

So the answer is yes, you can create an electromagnetic beam that is spatially but not temporally coherent by placing a pinhole close to the source, and then another pinhole in the far field of the first pinhole. This beam will not spread out very much. (But also remember that laser light does spread out.)

Note that for RF frequencies, a "pinhole" is probably several meters in diameter. The far field distance is given by this inequality: $L \gg a^2/\lambda$, where L is the distance, a is the diameter of the hole, and $\lambda$ is the wavelength.

However, creating a RF pencil beam is probably not practical. The term "pencil beam" mentioned in the Wikipedia article is explained as being diffraction-limited. The size of a diffraction-limited beam gets larger with, I believe, the square root of the wavelength. It would be more like a gas-pipeline beam than a pencil beam.

It depends on how big a pencil you're thinking about. There's no fundamental reason why radio waves can't be collimated in the same sort of way that visible light beams are. In fact, some radar systems send out fairly collimated beams at radio frequencies.

If you want to make a radio-wave beam that is the same size as a typical laser beam, though, you're out of luck. You can't focus light of whatever wavelength down to a distance much smaller than a wavelength and expect it to stay there for very long. Making a reasonably collimated laser beam with a width of a millimeter or so isn't really a problem because the wavelengths of visible light are in the neighborhood of 500 nm, or about 2000 times smaller than the beam. Radio waves, however, have wavelengths that are measured in centimeters or even meters, and those aren't going to let you make a tight beam a millimeter across.

Well, you just need Maser :-) http://en.wikipedia.org/wiki/Maser