From public labs to private firms: magnitude and channels of RD spillovers

Evidence from France suggests public research spending spills over to the private sector. Firms in the top 25% of exposure to academic funding increase their R&D effort by 20% relative to firms with the lowest exposure. "...Firms in the top quartile of exposure to increase their R&D effort by 20% compared to the bottom quartile. We exploit reports produced by funded clusters, complemented by data on labor mobility and R&D public-private partnerships, to provide evidence on the channels for these spillovers. We show that spillovers are driven by outsourcing of R&D activities by the private to the public sectors and, to a lesser extent, by labor mobility from one to the other and by informal contacts. A few years after the treatment, industry-location pairs exposed more to the local public research funding shock witnessed increased spending on R&D labor. Five years after treatment, we observe an increase of around 1.5% in R&D wage bill when the exposure of an industry in a commuting zone is doubled...."
“..We first propose a new measure of scientific proximity between public research groups and industrial sectors, allowing us to measure the exposure of private firms to the program. Based on this measure, we use the funding shock to estimate the causal impact of public research on private sector outcomes. We find that firms spatially and scientifically “close” to funded research groups increase their spending on R&D inputs and achieve higher levels of R&D outputs compared to less exposed firms…”

Three Primary Findings

“…Our first contribution is to construct a new measure of proximity. The key idea is to use the distance between the science used by firms and the science produced by research groups, as a proxy for the share of ideas produced by the research group which are relevant for a firm in a given industry.2 Importantly, this measure assigns a scientific position to firms rather than using the typical approach in the literature that attempts to assign a technological position to an academic group. An academic group and an industry can be close according to our measure, even if the papers published by the group are not yet cited in patents, and would thus be categorized as distant with the traditional approach. We then test and exploit this measure in the context of the LabEx program that selected 170 academic clusters in 2010 and 2011. These clusters bring together public researchers from different research units, not necessarily from the same institution, planning to work together on a common theme. The funding was run as a competition, with an international jury evaluating 436 proposals. We obtained a number of key pieces of data from the agency organizing the competition, for both accepted and rejected projects, in particular the full bibliography of the proposals and the grades awarded by the jury. The articles listed in the bibliography of the proposal are used to construct the proximity between LabEx projects and industrial sectors through the above procedure, while the identity and grades of rejected projects are the basis for various robustness checks.

“…Our second contribution is to use this natural experiment to estimate the causal effect of a positive shock in funding of public sector research on private sector R&D. For a given pair of commuting zone (henceforth “CZ”) and industry, we calculate exposure as the sum of funding obtained by the LabEx proposals in the CZ, weighted by the proximity of the LabEx to this industry. We then implement a difference-in-differences estimation that exploits the fact that a given industry in a given commuting zone will be more exposed to the shock if it is close technologically and geographically to a funded LabEx. We show that firms in high exposure pairs of CZ and sector significantly increase their employment in R&D after the start of the program, controlling for CZ specific time effects. The magnitude is large: a firm in the top quartile of exposure increases total spending on wages of R&D workers by more than 20% compared to the bottom quartile. We also find significant impacts on outputs of the R&D process in the more exposed sectors, in particular on the creation of new plants and on the production of new patents….”

“…Our third main contribution is to shed light on the channels through which the important spillovers from public to private research occur. There are three main candidates. First, spillovers could be due to direct collaboration between researchers in the academic cluster and the exposed firms (see e.g. Fernandes and Ferreira, 2013). Second, they could result from mobility of researchers from the public to the private sector or creation by these researchers of startups (see e.g. Agrawal and Henderson, 2002). Third, spillovers could simply be due to informal contacts between researchers from the public and the private sectors, for instance during events organized by the LabEx…”

The Evidence

R&D efforts

“..We first study how the financing of public research affects private R&D efforts. Figure 1 presents the results of the estimation of the dynamic model (equation (4)) using as outcome variable the total R&D wage bill (in log

"… We present the results using the continuous measure of exposure in Figure 1(a) and we show the effect by quartile of exposure in Figure 1(b). The results show that a few years after the treatment, industry-location pairs that were more exposed to the local public research funding shock witnessed an increase in spending on R&D labor. 5 years after treatment, we observe an increase of around 1.5% in R&D wage bill when the exposure of an industry in an commuting zone is doubled. This Figure also shows the absence of differential pre-trends in the years building up to the financing, which is evidence in support of our identification strategy. Figure 1(b) shows that the effect is mostly driven by pairs of industry-location that are in the top quartile of exposure. On average, these pairs increase their spending on employment in R&D by 20% compared to the least exposed units (which received no funding)...."

R&D outputs

“..The impact on patenting is explored in the second part of Table 3….To measure the effect of the LabEx program on the creation of new plants, we use the registry of plants administrated by the Insee and calculate the number of new establishments in each CZ and each sector. Table 3 shows a positive effect of LabEx exposure on the probability of creation of a new plant in an industry-CZ pair, corresponding to an increase of 0.2 percentage points when exposure doubles. Decomposing this effect between plant creation from new and incumbent firms shows that it stems almost entirely from new firms. The impact on patenting is explored in the second part of Table 3. As is standard in the literature, we use a Poisson regression to estimate the coefficient, which takes into account the very large number of observations with no patent. The baseline model yields an insignificant static coefficient along with non zero pre-trend. However, when sector-year fixed effects are added to the model, the coefficient becomes positive and significant and the pre-trends are no longer significant…”

Spillovers By Sector

“…. While an empirical comparison of the relative impact of these two instruments on private sector innovation is beyond the scope of this paper, we can shed light on the question of what sectors are more affected by these two instruments. To perform this comparison, we allocate the total funding of the two instruments by industry: the LabEx funding is allocated using the sum of exposures over all labs, while the research tax credit is using the claims made by firms. The results are presented in Figure 3. We see that the distribution of exposure to the LabEx program is much more skewed towards a few sectors. The two sectors that benefit the most are the scientific R&D (almost 30%) and the pharmaceutical sector (more than 20%).On the contrary the benefits of the CIR are much more evenly distributed across sectors, including in sectors where innovation should not be central, such as computer consultancy. This evidence is coherent with the idea that financing public research to spur private sector innovation might be a better instrument to target truly innovative firms than more direct tools such as tax credit programs. Comparing these instruments and understanding their complementarities is an important avenue for future research….”

Antonin Bergeaud, Arthur Guillouzouic, Emiric Henry and Clément Malgouyres, "From public labs to private firms: magnitude and channels of R&D spillovers,"London School Of Economics, October 2022, https://poid.lse.ac.uk/textonly/publications/downloads/poidwp041.pdf