Adding a new year to the data results in changes of indices in earlier years. Is that correct?

Yes, inclusion of the data of a new year may affect the model estimations and imputations for all years of the time series. Usually, however, the indices in earlier years hardly change.