In this article we propose to implement a covariance structure analysi
s to deal with the estimation of a stochastic frontier production func
tion on panel data and the measurement of a time-varying technical eff
iciency. First, this method solves the potential problem of correlatio
ns between input quantities and individual effects. Second, individual
effects and efficiency measures can be recovered as a byproduct of th
e analysis through the so-called factor scores. We implement this appr
oach by fitting to a balanced panel of French grain producers, a parsi
monious version of the Cornwell, Schmidt, and Sickles [1990]'s model w
here technical efficiencies are individual-specific linear functions o
f time. A specification search shows that this model is preferred to t
he traditional production function. Results shed light on the temporal
pattern of efficiency in the French grain production sector.