Variable capital utilization is an important element in recent DSGE models. To allow capital utilization to respond to shocks, the cost of varying utilization is modelled in terms of either accelerated depreciation of capital (Greenwood et al., 1988) or foregone consumption (Christiano et al., 2005). We perform a Bayesian estimation of the standard medium-scale DSGE model augmented with news shocks to examine which specification is supported by the data. The former exhibits a superior fit with the data relative to the latter. We show that it is attributable to the fact that the former explains better the properties of nominal variables, such as inflation and nominal interest rates.