Multinational Enterprises (MNEs) are crucial players in the world economy. They shape global production through Foreign Direct Investment (FDI), are responsible for about two-thirds of international trade flows, and are important transnational employers. Their activities are thus of primary interest both for academic economists and policymakers.
From the host countries’ standpoint, FDI is frequently praised for transferring technological assets to domestic firms. Indeed, multinationals often bring more innovative production technology and promote more efficient management procedures than those adopted by domestic producers. For this reason, affiliates of foreign firms tend to be larger and more productive. Local governments hope that this knowledge may also flow to other domestic firms not directly linked to a multinational. If this happens, the potential benefits of MNEs’ entry propagate through the domestic economy.
However, because technology is often factor-biased (i.e., it raises employment and wages of some production factors more than others), there are distributional effects to consider beyond aggregate welfare. Therefore, measuring the technological content of FDI is critical when thinking about the consequences of policies that attract or discourage foreign investment.
In my recent research paper entitled “Multinational Enterprises, Technology Transfers, and Robot Adoption,” I look into technology transfers from MNEs to their recently acquired affiliates abroad and document a new and controversial activity. Leveraging detailed firm-level and cross-country data, I show that MNEs transfer industrial robots to their newly acquired domestic affiliates.
Compared to traditional capital-intensive technology, robots are perfect substitutes for human labor in many low- and middle-skill jobs. Nowadays, they are essential inputs in most manufacturing sectors (IFR, 2019). The main concernabout their spread is that robots take over humans in an increasing share of jobs and fundamentally change the division of labor.
Data, Methodology, and Results
I establish my baseline results leveraging detailed firm-level data for the Spanish manufacturing sector. Data comes from the Survey on Business Strategy (ESEE) administered by Madrid’s SEPI Foundation. The survey is designed to be representative of the population of manufacturing firms with ten or more employees located in Spain. Other studiesusing these data praise them for their high quality.
Besides standard characteristics (e.g., sales, employment, investment, etc.), the ESEE data is among the few currently available to report information about robot adoption and ownership at the firm-level. In particular, I examine whether a firm in a given industry (NACE 2 classification) employs robots or not and whether foreign investors own more than 10% of its equity, which is the standard OECD definition of FDI.
When thinking about how ownership status affects robot adoption, one must deal with a fundamental identification problem. A simple inspection of robot adoption patterns between domestic- and foreign-owned firms is insufficient to gauge the causal effect of ownership on the choice of investing in robots. It is well-known that multinational firms perform better than domestic ones. Therefore, observing higher robot adoption rates among multinationals could be a cause or consequence of a pre-existing superior status.
To circumvent this issue, I build upon the methodology developed by Guadalupe, Kuzmina, and Thomas (2012). The thought experiment is the following. Suppose that, in a given year, you observe two equally performant domestic firms (e.g., with the same total sales, number of employees, wages, productivity, etc.). Imagine that one year later, a foreign investor acquires only one of them. How does the foreign investor decide which firm to acquire between two identical potential targets? If one assumes that it just happens randomly, it is possible to identify the causal effect of switching from domestic to foreign ownership on robot adoption. This is the basic intuition behind the method that I use and formally explain in the paper.
Using this technique, I find that formerly domestic-owned Spanish manufacturing firms that switch to foreign ownership become around 20% more likely to employ robots than similar firms that stay under domestic control. I also find that newly acquired firms are more likely to employ robots within the first four years past the acquisition date. The effect attenuates afterwards.
The data also reveals a few more interesting patterns. First, I find that the effect systematically varies across sectors. In particular, FDI boosts robot adoption in sectors with high automation potential which do not have high rates of robot adoption in practice. For instance, I find a small effect in the automotive industry, but a larger effect than average (about +25%) in the electricity sector. The reason is that the automotive sector already experiences high rates of robot adoption. Therefore, it is hard to boost it even more. By contrast, although its typical job could be easily performed by a robot, automation in the electricity sector is still low. Hence, the benefits from robot adoption are potentially very high. In this regard, I find that FDI kick-starts robotic-oriented structural transformation in some industries.
One obvious question is why and how do foreign owners spur robot adoption by their affiliates. Consistently with Guadalupe, Kuzmina, and Thomas (2012), I find that newly acquired firms face higher domestic and international demand past the acquisition date. Robot adoption is a cost-effective way to scale up production and meet the increased demand. Besides, this process is accompanied by a higher reliance on foreign technology. Upon acquisition, new affiliates substitute in-house innovation with import of foreign technology. Overall, foreign parents want their new affiliates to perform well and expand. Transferring knowledge on innovative and efficient production technology such as robots is one way to achieve this goal.
Finally, I also investigate how likely my results are to hold beyond the Spanish manufacturing industry and, therefore, how general the phenomenon is. To answer this question, one would ideally have firm-level data about robot adoption and foreign ownership across many countries. Unfortunately, such information is not available. As a second-best approach, I construct a new dataset with information about the stock of deployed industrial robots and foreign-owned firms’ gross output at the country-industry-year level. The dataset contains information about 40 middle- and high-income countries and 19 industries from 2005 to 2014.
Using this data, I find that a 5% increase in foreign-owned firms’ gross output induces, on average, the adoption of one more robot per thousand workers. According to estimates for the US from Acemoglu and Restrepo (2020), this reduces the employment-to-population ratio by about 0.2 percentage points and wages by up to 0.5 percentage points. Overall, I conclude that FDI-induced robotization is a pervasive phenomenon in most advanced economies.
The Aggregate Implications of FDI-induced Robotization
In the last part of my research project, I try to understand the aggregate implications of FDI-induced robotization for productivity and inequality. On the one hand, one expects significant productivity gains coming from the adoption of innovative technology and better management practices. On the other, robot adoption calls for a skill-biased change in firm supply chains.
I address this question in two stages. First, building upon the methodology developed by Doraszelski and Jaumandreu (2018), I estimate the contribution of foreign-ownership and robot adoption on firm-level total factor and capital-biased productivity.[1] Second, I ask myself how would aggregate productivity and the labor share look like absent robot adoption and/or multinational presence in the Spanish manufacturing industry. More precisely, I simulate counterfactual aggregate productivity and labor share (my proxy for inequality) in the Spanish manufacturing industry by shutting down the firm-level robot adoption and foreign ownership premium one at a time.
This exercise delivers two major insights. First, robot adoption explains about 10% of aggregate productivity in the average sample year. FDI-induced robotization accounts for about half of it. Second, absent robots, the aggregate labor share would be about 30% higher in the average sample year. In a nutshell, my results suggest that robot adoption and foreign ownership increase the size of the pie but also slice it more unevenly.
Conclusions and Food for Thoughts
The use of industrial robots in manufacturing has surged over the last 25 years thanks to rapid technological developments. This process is not free from concerns. Although robots may free up human workers from routine, sometimes risky jobs, policymakers confront robots-induced displacement of many low- and middle-skilled workers. In addition, scholars have shown that robots’ diffusion produces societal effects even beyond the labor market and influences public finance through taxation, electoral outcomes, health, etc.
In this paper, I show that multinational enterprises employ robots more intensively than domestic producers and transfer robotic technology to their network of affiliates. Although the focus of my analysis is primarily on the Spanish manufacturing industry, I show that the effect also holds beyond the Spanish context. These findings suggest that policymakers face an efficiency-equity trade-off when designing policies to attract foreign investment. Knowing the technological content of FDI is the first necessary step to design suitable and correctly predict the impact on the host economy.
Fabrizio Leone is a PhD Candidate in Economics at ECARES, Université Libre de Bruxelles. This post is adapted from his paper “Multinational Enterprises, Technology Transfers and Robot Adoption,” available on SSRN.
[1] The difference between total factor and capital-biased productivity can be easily understood using the following example. Suppose that firms employ workers and machines to produce their output. Saying that robot adoption and foreign ownership increase total factor productivity means that workers and machines suddenly become more productive in equal proportions. By contrast, if robot adoption and foreign ownership increase capital-biased productivity, they increase the relative productivity of machines versus workers.
