At present, it is possible to estimate binary NS merging rate in two ways: using the binary radiopulsar statistics observed and making various computations of binary stellar evolution.
toPhinney,1991,ApJ, 380, L17 
yr
toNarayan et al.,1991, ApJ, 397, L17 yr
toCurran and Lorimer, 1995, MNRAS, 347, 276 
yr
tovan den Heuvel and Lorimer, 1996, MNRAS, 282, L37 
yr
to``Bailes limit'', 1996 [15] 
yr
toClark et al., 1979, A&A, 72, p.120 
- yr
toLipunov et al., 1987, A&A, 176, L1 yr
toHils et al., 1991, ApJ, 369, p.271 yr
toTutukov and Yungelson, 1993, MNRAS, 260, p.675 
- yr
toLipunov et al., 1995, A&A, 298, p.667 
yr
toPortegies Zwart and Spreeuw, 1996, A&A, 312, p.670 
yr
toLipunov et al., 1997, MNRAS, 288, p.245 - yr
We emphasize that although theoretical merging rates are systematically higher than observational ones, both estimates do not contradict each other.
The main argument is that the first (observational) estimates of binary NS merging rate are based on the statistics of binary systems, in which only one of the components shines as radiopulsar, which is not at all the necessary condition for merging to occur. Thus the observational estimates are in fact only lower limits. There are also some selection effects that can change estimates from both groups.