THE PERFORMANCE OF NON-SURVEY TECHNIQUES FOR CONSTRUCTING SUB-TERRITORIAL INPUT-OUTPUT TABLES

This study is a contribution to the ongoing debate on the performance of various non-survey techniques for constructing sub-territorial Input-Output tables. Three aspects of the behaviour of the methods are analysed: performances in reproducing ‘true’ input coefficients, variability of error, and direction of bias. The analysis uses real data and in particular the World Input-Output Table. The most important aspect that emerges from the analysis is that even though SLQ has been identified as one of the most robust methods, its performance drops when confronted with Input Output Tables characterized by a high percentage of technical coefficients close to zero. Also the CILQ, SCILQ and RLQ methods behave in a similar way. On the contrary, the performance of the methods FLQ for δ=0.2 and AFLQ for 0.2≤δ≤0.3 are not affected by this situation.