In an influential paper from 1995, "International R&D spillovers", David Coe and Elhanan Helpman use data on trade and R&D expenditures for 22 countries to estimate the effects of a country's own R&D effort and the R&D effort of its trade partners on the country's TFP. They use a model of endogenous, innovation-driven growth as basis for the empirical equations. The main idea is that a country benefits from the cumulative R&D stock of its trade partners as well as its own R&D stock, through spillover effects from trade in intermediate inputs.
The intermediate inputs can be horizontally differentiated, which means that a new input is equally good as an old, and investment in R&D increases the number of available varieties. The inputs can also be vertically differentiated, so that the development of a new input will replace an old one because of its increased quality. In the endogenous growth framework, R&D investments are made by firms seeking monopoly profits in imperfectly competetive markets. This R&D effort generates a product or process than can be patented and give rise to a profit for the firm, as well as a non-appropriable product that will increase the country's stock of knowledge. This non-appropriable part is what generates the spillover effect. In the case of horizontally differentiated inputs, the investments in R&D increase the stock of knowledge through an increase in the number of available inputs, which in turn decreases future R&D costs. With vertically differentiated inputs, the development of a higher quality input will enable future entrepreneurs to build on a higher quality foundation. In both cases, the stock of knowledge increases and there are spillovers from current to future R&D activities.
Coe and Helpman (1995) argue that since intermediate inputs are traded internationally, a country's productivity depends on its own R&D stock as well as the R&D stock of its trade partners. The domestic R&D stock is constructed as the sum of the country's cumulative R&D expenditures, multiplied by a constant depreciation factor. In the basic specification used in Coe and Helpman (1995), the foreign R&D stock variable is constructed as a weighted sum of the R&D stocks of the country's trade partners, where the weight is the imports from one country as share of the total imports. They find a positive and significant effect of both R&D stocks on the country's TFP, and interpret this as evidence that there are significant international R&D spillovers from trade.
Norway is one of the countries used in the paper. I use panel data on Norwegian firms to see whether this effect of imported R&D spillovers can be found on the firm level in Norway. The data set consists of data from the accounting and manufacturing statistics, as well as data on trade and R&D expenditures for Norwegian firms. This is combined with data on R&D expenditures on the sector level for 22 member countries of the OECD. The variables are constructed on the basis of the variables used in Coe and Helpman (1995) with some adjustments to fit the data set at hand.
I estimate the basic specification from Coe and Helpman (1995) using the fixed effects estimator. The software used for the estimations is Stata 11. I find a positive and significant effect of the firm's own R&D stock. However, the foreign R&D stock appears to have no significant impact on the productivity of the firms. This indicates that any spillover effects that Norway might get from foreign R&D, are not transmitted through each firm's imported intermediate inputs.
I also estimate several extensions of the original specification. In the first extension, I substitute the firm's own R&D stock with the lagged R&D expenditure, because of the nonstationarity of the R&D stock variable. In the second, the weights are changed to reflect the fact that a firm can purchase intermediate inputs domestically as well as internationally. The foreign R&D stock variable is then constructed using the imports from one country as share of total purchase of intermediate inputs as weights. The third extension is aggregating the foreign R&D stock variable to the sector level, to see if I can detect any intra-industry spillovers that are transmitted through trade in intermediate inputs.
The positive and significant effect of the firm's own R&D stock is stable in the various specifications. The foreign R&D stock variable is however not statistically significant at any of the conventional levels. This confirms the indication that any spillover effects do not emanate from each firm's imports of intermediate inputs, and it also indicates that any spillover effects have to appear at a more aggregated level.