Exist a big study corpus abour reserve salary. But I don't have clear what is the best form of measure it. The reserve salary is the salary that somebody accept to start a work.
I don't meet to be "smart" or flip, but the best format to measure this (or anything) in is the format that works best for your research question or reporting purposes. If one of your goals is to match to someone else's measure (e.g., a federal or official statistic, then using their question format is a good idea so at least that part of the study is the same). But if you're not doing that, I recommend looking around to see what others have done, then writing your own question. Be sure to do some cognitive and field pretesting too . See Gordon Willis's book on cognitive testing for advice.
There is an increasing number of studies on self-reported reservation wages. A recent contribution is Krueger, Alan B. and Andreas I. Mueller, 2014.A Contribution to the Empirics of Reservation Wages. NBER Working Paper No. 1987. There is also a long tradition of indirect measures, i.e. computed on the basis of econometric estimates. For example, say that you observe a sample of workers. For each of them you know the wage rate (w), hours of work (h) and a vector of personal characteristic Z. Say you specify the labour supply function as h = Z'b + u and the reservation wage as r = Z'a + e, where a and b are parameters and u and e are random variables. Then one can use the conditional density
f(h - Z'b)/Prob(Z'a + e > w)
(where f(.) is the density of u)
to estimate a and b. The estimate of a will allow you to estimate r.