Persistence in the long-run expected rate of return for corporate pension plans

Firms that sponsor a defined benefit pension plan are suspected of managing earnings through the choice of the expected long-run rate of return to pension assets. However, data on this rate show it to be quite persistent with more than 50% of firms leaving their ERR unchanged from one year to the next. To capture this persistence, I model the rate using a first-order autoregression. Asset allocation information is included in the model. Endogeneity bias is addressed by estimating the dynamic panel data model using a system GMM estimator. No evidence of earnings management, measured by the relative size of the pension plan to net income and by acquisition activity, was found.