The long-term temporal stability of a spectrograph is one of the most important characteristics affecting the spectrograph's radiometric performance. For many applications, from monitoring ocean color and lunar irradiance to laboratory irradiance measurement standards, the stability of a spectrograph is a primary factor in the overall measurement uncertainty and therefore is the major criterion for the suitability of the spectrograph as an optical-scale transfer standard. Here we report a facility built for testing the long-term radiometric stability of commercial, fiber-coupled spectrographs. The facility uses tungsten quartz-halogen irradiance standard lamps, type "FEL," of the National Institute of Standards and Technology (NIST) as light sources. To ensure the highest stability of these lamps during spectrograph tests, parameters such as lamp current, lamp voltage, and signals from an independent filter radiometer were continuously recorded to monitor any possible instability caused by such effects as lamp aging. Using this facility, we report the stability study of four spectrographs with spectral coverage from the UV to short-wave infrared over an interval of two months during which the lamp irradiance was stable to better than 0.02%. The tested spectrographs show good stability in general, ranging from 0.02% to 0.1% in the visible over a span of 11 days. For a longer two-month test, the variation in spectrograph responses increases by less than 0.1% with no discernable long-term drifts. In addition, we measured the response variation of two of the test spectrographs before and after they were sent to remote field locations and subjected to adverse environmental conditions. In this case, a larger response variation of up to 1.0% dependence on the wavelength was observed. We discuss the performance of the facility and the implications for using these spectrographs for several of NIST's remote sensing projects as radiometric transfer standards based on these stability measurements.