I was chatting to Justin, G4TSH last night. I told him about my experiments with WSPR on 7MHz and of course, couldn't resist telling him that my 1 watt WSPR signal had been heard in Tasmania. Justin wondered whether a 1W CW signal would have been heard in the same circumstances.
My feeling, at least based on the propagation at the time was that it probably wouldn't have been possible; it was a typical noisy evening on 7MHz, so my guess was that any QRP CW transmission would have been wiped out. Justin felt you could have probably made a rubber stamp '599/599' type QSO! Maybe, in the right conditions, but I think he might have been a little optimistic!
Thinking about it, I wondered if there's been any comparison about the advantage of WSPR over CW for reporting. A quick Google found me an interesting link on Roger, G3XBM's blog, who'd considered this very point.
Roger had asked Joe, K1JT, father of WSPR and the 'WS modes' for his comments and I've taken the liberty of reproducing Joe's reply to Roger;
"For an answer to your question about relative sensitivities of CW, WSPR, and some of the other modes implemented in WSJT, let me suggest going to the "References" link on the WSJT web site, http://physics.princeton.edu/pulsar/K1JT/refs.html ,
and select reference #11. Table 3 on page 9 gives the information you asked about. The difference given there is about 11 dB, in favor of WSPR over ear-and-brain CW. For most operators, the difference is more like 15 dB."
If we assume 6dB per S-point and that a 1 watt CW signal was S1 (or so), then a WSPR signal could be very weak and still detected. And of course, the repetitive nature of WSPR transmissions means that they may be heard where a 'random' CW transmission might not be.
Of course, in some ways, this is an unfair comparison. WSPR is purely a reporting mode whereas CW is a conversational tool, but it certainly suggests considerable advantages in using WSPR for propagation testing.