Measurement of total factor productivity: Evidence from French construction firms

This paper reviews eight (8) methods of calculating total factor productivity (TFP) in the "construction of residential and non-residential buildings" sector in France. These include fixed-effects estimators; instrumental variables and the generalized method of moments (Blundell and Bond, 1999); Olley and Pakes, 1996; Levinsohn and Petrin, 2003; Wooldridge, 2009; Ackerberg, Caves, and Frazer, 2015; the calibration method; and the data envelopment analysis (DEA) method. Then, using firm-level data from 2009 to 2018, we show that the market structure can be likened to an oligopoly situation and that capital intensity is also very low in this sector. Furthermore, the fixed-effects estimator provides the lowest capital coefficient and overestimates both the absolute value of the scale effect and the intermediate inputs coefficient. The highest capital coefficient is provided by the Wooldridge (2009) estimator. But there is little difference between the TFP measures, especially when semi-parametric methods are used. While the calibration of elasticities shows that the construction sector is labor intensive, the DEA method shows that on average only large firms are fully efficient. To our knowledge, the Ackerberg, Caves and Frazer method be a good estimator of TFP in the French construction sector. Finally, when comparing TFP levels, all estimation methods (fixed effects; Wooldridge, 2009; Olley and Pakes, 1996; Levinsohn and Petrin, 2003 and Ackerberg, Caves, and Frazer, 2015) are strongly positively correlated with each other (over 92%). However, the correlations between these methods and the non-parametric methods (DEA and calibration methods) are very low, even negative with the calibration method.