During the subprime crisis some sectors had a productivity decline of about 15-20% (according to sna/bls data). Did the firms reinstall old machines from the attic for a few quarters? No. Will a firm immediately fire employees if turnover/value added goes down? No. Can the statistical office diffentiate between employees hard working at some times and picking in their noses at other times? No (for the most part). Labour hording can explain a great part of the productivity movement during the cycle (take for example a simple multiplicator-accelerator-model, assume some degree of labour hording and you will get tfp fluctuations with the "right" standard deviation and correlations). You don't need references for burning antic libraries...