It has been argued that estimating the spectral density function of a stationary stochastic process at the zero frequency (or the so-called long-run variance) is an illposed problem so that any estimate will have an infinite minimax risk (e.g., Pötscher 2002). Most often it is a nuisance...