The Relationship between Investment in R&D and Productivity. The Econometric Evidence

Citation

Vecchi, M. (2001), "The Relationship between Investment in R&D and Productivity. The Econometric Evidence", Journal of Economic Studies, Vol. 28 No. 1, pp. 65-68. https://doi.org/10.1108/jes.2001.28.1.65.2

Investment in high‐tech products is considered one of the main factors in fostering a country’s economic growth and economic performance. Both governments and private firms have directed large amounts of resources towards increasing their endowment in the knowledge‐intensive sectors. The economic analysis could not remain indifferent to this phenomenon and it has in fact for a long time followed its development, trying to quantify its actual impact on the economic activity. One measure of the relationship between technological growth and economic activity is the Solow’s productivity residual or total factor productivity (TFP). However, this methodology is not fully satisfying as it implies a purely exogenous technological progress. As a consequence increasing effort has been devoted to trying to measure the relationship between technology and productivity in an alternative way. Professor Zvi Griliches has played a central role in this research programme and this volume presents a collection of his most influential papers, summarising the main methodological issues that have arisen in almost two decades of work on the topic. Rather than relying on a residual measure of technology, the research led by Griliches has focused on the firm/industry’s investments on R&D and how this investment affects productivity. The standard Cobb‐Douglas production function framework has been enriched with R&D capital and the contribution of technological change on productivity can be directly estimated.

From a first reading of the papers in the volume it is evident that the economic analysis of the relationship between R&D and productivity has to face several problematic issues. Two main difficulties are met at the very beginning of the empirical investigation, that is in the data collection. First, the data are not easily available, especially at the micro level. This fact has constrained the analysis to a few countries, mainly the USA, France and Japan (chapters 7 and 8). The investigation also has to deal with missing observations. There are many firms that do perform R&D investments but they do not report them in their account or the figures reported are very inaccurate (chapter 8).

Secondly, it is very difficult to measure output in some sections of the economy, for example in the Government sector or in the private sector for those goods that are subject to fast technological changes (chapter 2). In the latter case, the data do not always account for the changes and their effect on the quality and the price of the products. A typical example is the computer industry where the lack of an appropriate price index can impair the correct evaluation of the relationship between R&D and productivity (chapter 14). Other problems specifically affect the measurement of R&D capital. What is usually available to the researcher is R&D expenditure. From this variable a measure of capital has to be derived. This implies the evaluation of the lags between the implementation of the research activity and the finalisation into a new product or a new process, the obsolescence of the R&D capital and the spillover effects (chapter 11).

The analysis of the relationship between R&D and productivity also has to deal with econometric problems. Most empirical studies are based on the production function framework, which is extended in order to include R&D capital as a right‐hand side variable. Its contribution to production growth is assumed to be independent of the contribution of the other inputs. This is a very strong assumption but it is not specific to the introduction of R&D capital. Multicollineary always characterises the estimation of production functions. More serious is the simultaneity problem. Future output depends on past investments in R&D and, at the same time, R&D depends on past and expected output (chapter 2). One of the methods used to address this problem is instrumental variable (IV) estimation. More recent studies have presented generalised method of moments (GMM) estimates next to the more standard sets of OLS results. However, the lack of suitable instruments in this context means that only lagged values of the regressors can be used to carry on IV estimation (chapter 12). The results, at this stage, appear to be quite sensitive to the set of instruments used and, as a consequence, it is difficult to asses the effect of R&D on productivity. Possibly, the IV method should be extended and further model specifications should be tested in order to obtain results that are more robust.

The attention of researchers has focused not only on the private returns to R&D but also on the relationship between productivity and R&D spillovers. The contribution of R&D activities can easily spread across different firms and industries as it is very difficult to impose boundaries on knowledge. Innovations can spread using different channels. One of them is via the exchange of products, incorporating the results from R&D. But the most interesting issue, and the more difficult to measure, is “the impact of the discovered ideas or compounds on the productivity of the research endeavours of others” (chapter 11). Griliches reviews the literature on R&D spillovers and discusses how these are measured and how the research effort has shifted from the agricultural to the manufacturing sector. Results seem to suggest the presence of significant extemalities, although more research is needed in order to account for the numerous interactions across firms/industries.

The understanding of the relationship between R&D and productivity has come a long way in 20 years of study. The various difficulties that can arise in the empirical investigation have not discouraged the research effort. On the contrary, the literature on the topic is very rich and the existing evidence suggests that the results are quite consistent over time (chapter 4) and “great advantages have been made in theory and in econometric techniques” (chapter 14). In this volume, Griliches presents a very detailed discussion of the bad and the good news about investigating the impact of R&D on productivity, as well as posing several questions for further research. The volume is certainly a very good reference for anyone new to the topic and an important tool for those who are already working in this area.