The time plot of the series NPER exhibits an overall downward trend with a deep depression in late 2008. No regular seasonality is evident. A 12-month differencing yields a series SDNPER which has an overall slightly upward trend with no clear seasonality. A nonseasonal differencing of SDNPER yields a series DSDNPER with an overall horizontal trend. The visual inspection of its time plot hardly gives an impression of any regular seasonality. However its autocorrelation function shows a significant negative spike at lag 12, indicating a 12-month seasonality and a seasonal moving average component of order one. Moreover the partial autocorrelation plot has significant spikes at lags 12 and 24, suggesting the involvement of a seasonal autoregressive component of order two. Consequently, a (0, 1, 0)x(2, 1, 1)₁₂ SARIMA model is hereby proposed, fitted and shown to be adequate.