Provided are a mask blank substrate which has effectively and extremely high principal surface flatness while a reduction in the manufacturing throughput of the mask blank substrate is suppressed, a mask blank, and a transfer mask. Also provided are manufacturing methods therefor. virtual reference surface that becomes an optically effective flat reference surface defined by a Zernike polynomial which is composed of only terms in which the order of a variable related to a radius is the second or lower order, and includes one or more terms in which the order of the variable related to the radius is the second order is set, and a mask blank substrate satisfying the condition that data (PV value) relating to the difference between the maximum value and the minimum value of the difference data between the reference surface and the measured shape of the mask blank substrate is one-eighth or less of an exposure wavelength (λ) is selected.