Do R&D expenditures really matter for TFP? VINCENZO ATELLA* and BENIAMINO QUINTIERI CEIS - University of Rome Tor Vergata - Via di Tor Vergata snc, 00133 Rome, Italy Recently, several studies have emphasized the role of R&D expenditure in determin- ing Total Factor Productivity (TFP). In this paper it is shown that the relationship between R&D variables and TFP is far from being established. In particular, by using data for the Italian economy, it is found that the estimated eŒects of R&D variables on TFP crucially depends on: (i) the way in which the production function, used to derive Solow residuals, is de®ned; (ii) the numbers of maintained hypotheses used to estimate Solow residuals; (iii) the level of aggregation of the data employed in the empirical analysis. I. INTRODUCTION Over the last decade several authors 1 have emphasized the role of R&D expenditure in determining the rate of growth of total factor productivity (TFP). In particular, Coe and Helpman (1995) conclude that there are `convincing empirical evidence that cumulative domestic R&D is an important determinant of productivity’ (p. 860). According to Atella and Quintieri, this `convincing empirical evidence’ can be criticized on at least two grounds. The ®rst refers to the problem of measurement and de®nition of TFP. In fact, most of these studies are based on growth accounting measures obtained from production functions in which only labour and capital are included as inputs. It is obvious that, in this case, Solow residuals incorporate the eŒects of other inputs such as energy and intermediate inputs, that over the years have represented an increasing fraction of productiv- ity change. An even more relevant criticism is based on the possible bias deriving from the use of the `growth accounting’ pro- cedure usually employed. This measure is based upon strong assumptions, such as constant returns to scale, com- petitive output market and no short-run ®xities, that very often are not representative of the real world. In these cases the use of growth accounting Solow residuals produces biases that can alter the relationship between productivity and its main determinants. The second criticism refers to the fact that in the Coe and Helpman analysis the relationship between R&D proxies and productivity measures is investigated at an aggregated level and this can determine a bias due to the adoption of a unique production function for diŒerent sectors of the economy. In this paper it is shown how taking into account these criticisms will weaken the `convincing’ empirical evidence relating productivity growth with R&D expenditure. In Section II the theoretical framework used to derive TFP measures and the result obtained for the Italian economy are presented, while in Section III the econometric results are discussed. II. TOTAL FACTOR PRODUCTIVITY: DEFINITION AND MEASUREMENT Most of the studies on TFP that have been carried out over recent years rely on nonparametric measures derived from the original approach suggested by Solow (1956). According to that methodology, TFP measurement is based on assumptions that are dicult to accept as main- tained hypotheses: perfect competition, absence of scale Applied Economics ISSN 0003±6846 print/ISSN 1466±4283 online # 2001 Taylor & Francis Ltd http://www.tandf.co.uk/journals DOI: 10.1080 /0003684001000793 9 Applied Economics, 2001, 33, 1385±1389 1385 * Corresponding author. E-mail: Atella@uniroma2.it 1 See in particular Coe and Helpman (1995), Coe and Maghadam (1993), Griliches (1988, 1991), Grossman and Helpman (1994), and Englander et al. (1988).