Suppose you are building many of some subsystem, like a converter for example. There is likely to be variation in the gain stages and the designer decides some of these variations can be reduced by using attenuators. The highest gains are reduced to the lowest gains. Now suppose there is cabling involved. For the lowest gain devices where no attenuation is required, if it is removed, then the cable may be too short and might need to be changed to a different cable. Think of the 0dB attenuator as a place saver so that only one cable is ever needed.
That's the first thought that comes into my head. Maybe somebody else can come up with other reason (phase?)?
I use the SMT ones all the time. As mentioned above; phase. Recently used them to balance gain on a microstrip quadrature hybrid feeding two amps. The 0 dB attenuator will give a better phase match to it's companion attenuator than substituting a jumper or lead frame. It will also have similar parasitics and discontinuities, which is important at high frequencies. I wouldn't even buy a series of attenuators unless they offer a 0 dB or 1 dB value. I use these a lot:
I like the other answers to this question, but let me add....
When designing a transmit or receive chain, it is always best practice to put in a little more gain than needed, and configure chip attenuators of low value (2 dB for example) wherever they have the least hit on performance (in front of a mixer, not in front of an LNA). Then when you build a prototype you can tweak the performance by swapping attenuation values in and out (so-called select-at-test components). In an example case, your customer comes by and says that you need to reduce DC current to meet a new requirement. You reduce the drain current in a driver amplifier, and its gain drops. Time to reduce that attenuator... you will be glad that you have a zero dB value available.
Good engineering is all about leaving options in a design that prevent an entire design re-spin. Software people call these options "hooks", they stole that word from us. Hmm, its about time to add that to the slang dictionary...
Ok, so what exactly is in a 0dB pad? I'm trying to put one in front of an etched planar filter but I think I need some sort of actual path to ground. Is a 0 dB pad just a pi configuration with a short in series and opens for the shunts?