What caused the decline of US manufacturing employment in recent decades?

One of the most contentious aspects of globalisation is its impact on national labour markets. This is particularly true for advanced economies facing the emergence and integration of large, low-wage, and export-driven countries into the global trading system. Contributing to this controversy, between 1990 and 2011 the US manufacturing sector lost one out of every three jobs. A body of research, including recent work by Bloom et al. (2019), Fort et al. (2018) and Autor et al. (2013), has attempted to understand this decline in manufacturing employment. The focus of this research has been on two broad explanations. First, this period could have coincided with intensive investments in labour-saving technology by US firms, thereby resulting in reduced demand for domestic manufacturing labour. Second, the production of manufacturing goods may have increasingly occurred abroad, also leading to less demand for domestic labour.

New facts on manufacturing employment, trade, and multinational activity

On the surface, the second explanation appears particularly promising. Manufacturing employment declined from nearly 16 million workers in 1993 to just over 10 million in 2011, shown by the black line in Figure 1. This large decline in manufacturing employment coincided with a surge in outward foreign direct investment (FDI) by US firms (the blue line in Figure 1). Nevertheless, existing theories of trade and multinational production make ambiguous predictions regarding the link between foreign production and US employment. Further, due to a lack of suitable firm-level data on US multinationals, there has been limited research on their role in the manufacturing employment decline (see Kovak et al. 2018 for a recent exception).

Figure 1 US manufacturing employment and US outward FDI

In a recent paper, we address the question of whether foreign input sourcing of US multinationals has contributed to a decline in US manufacturing employment (Boehm et al. 2019). We construct a novel dataset, which we combine with a structural model to show that US multinationals played a leading role in the decline in US manufacturing employment. Our data from the US Census Bureau cover the universe of manufacturing establishments linked to transaction-level trade data for the period 1993-2011. Using two directories of international corporate structure, we augment the Census data to include, for the first time, longitudinal information on the direction and extent of firms' multinational operations. To the best of our knowledge, our dataset is the first to permit a comprehensive analysis of the role of US multinationals in the aggregate manufacturing decline in the US. With these data, we establish three new stylised facts.

Fact 1: US-owned multinationals were responsible for a large share of the aggregate manufacturing employment decline

Our first finding is that US multinational firms, defined as those US-headquartered firms with foreign-owned plants, contributed disproportionally to the decline in US manufacturing employment. While 33.3% of 1993 employment was in multinational-owned establishments, this group directly accounted for 41% of the subsequent decline.

Fact 2: US-owned multinationals had lower employment growth rates than similar non-multinationals

In Figure 2, we show that multinationals exhibited consistently lower net job creation rates in the manufacturing sector, relative to other types of firms. Compared to purely domestic firms and non-multinational exporting firms, multinationals created fewer jobs or shed more jobs in almost every year in our sample. Of course, these patterns may not be causal, and other characteristics of multinationals could be driving the low job creation rates. To address this concern, we control for all observable plant characteristics, and find that multinational plants experienced lower employment growth than non-multinational owned plants in the same industry, even when the size and age of the plants are held constant.

Figure 2 Net US manufacturing job creation rates by type of US firm

Fact 3: Newly multinational establishments experienced job losses, while the parent multinational firm expanded imports of intermediate inputs

An alternative way to assess the role of multinational activity on US employment with our data is to use an ‘event study’ framework. We compare the employment growth trajectories of newly multinational-owned plants to otherwise similar plants in terms of industry, firm age, and plant size. As can be seen in Figure 3a, prior to the plants becoming part of a multinational, their growth patterns are not different from the control group. However, in the years following the multinational expansion, there is a brief positive but then sustained negative trajectory of employment at these manufacturing plants. Ten years after the transition, these newly multinational-owned plants have manufacturing employment that is about 20% smaller than an otherwise similar plant.

Figure 3 US employment and import dynamics at new multinational plants

b) Cumulative relative employment (Index)

a) Relative imports

Further, these newly multinational firms increase imports following the expansion abroad. As Figure 3b demonstrates, these firms substantially increase imports both from related parties and other firms (at arms-length), relative to their control group. Taken together, Figures 3a and 3b suggest that offshoring might explain the observed negative relationship between trade and employment.

Structural analysis: Did the offshoring of intermediate input production result in a net employment decline in the US at the firm level?

While the patterns we identify above are suggesting that increased foreign input sourcing by multinational firms led to a decrease in US manufacturing employment, they are not necessarily causal. Standard models of importing, such as Halpern et al. (2015), Antras et al. (2017) or Blaum et al. (2018), make ambiguous predictions as to whether foreign sourcing is associated with increases or decreases in domestic employment. At the heart of this ambiguity are two competing forces. First, a reduction in the costs of foreign sourcing leads firms to have access to cheaper intermediate inputs. As a result, their unit costs fall and their optimal scale increases. This ‘scale effect’ raises their US employment. On the other hand, firms respond by optimally reallocating some intermediate input production towards the location with lower costs. This ‘reallocation effect’ reduces US employment. Theoretically, the scale effect could dominate the reallocation effect and lead to positive employment effects of offshoring, or vice versa.

We use our microdata to estimate the relative strengths of these two competing forces. We show that in a conventional class of models and in partial equilibrium, the value of a single structural constant – the elasticity of firm size with respect to firm production efficiency – completely determines which of the two forces dominates. Our estimation approach is to develop a method to structurally estimate an upper bound on this constant using our data on the universe of US manufacturing firms. While a high value of the upper bound leaves open the possibility that foreign sourcing and domestic employment are complements, a low value of the bound unambiguously implies that the two are substitutes.

Our estimates of the bound are small, indicating that during the period 1993-2011, the reallocation effect was much larger than the scale effect. In other words, during this period of aggregate manufacturing employment decline, multinationals’ foreign input sourcing was leading to a net decline of manufacturing employment within these firms.

Aggregate implications for US manufacturing employment

It is important to point out that the model we use only speaks to employment changes within existing firms and does not take into account general equilibrium forces that can also affect employment. Since such general equilibrium effects are inherently difficult to assess, estimates of how much of the observed aggregate decline can be attributed to offshoring of multinational firms are uncertain and often require strong assumptions. We thus proceed under two alternative sets of assumptions. In the first, we conduct a simple partial equilibrium aggregation exercise, which uses observed changes in firm cost shares of domestic inputs together with our estimated parameter bounds to obtain model-implied predictions of the employment loss due to foreign sourcing. This approach captures both the direct impact of foreign sourcing by existing firms as well as the first-order impact on domestic suppliers, holding all else equal. Under the second, we model these indirect, general equilibrium effects, such as firm entry and exit, explicitly. In both of these scenarios, we find that the offshoring activities of multinationals explains about one-fifth to one-third of the aggregate US manufacturing employment decline.

Policy implications

Our research shows that the global sourcing behaviour of US multinational firms was an important component of the manufacturing decline observed in the past few decades. These firms set up production facilities abroad and imported intermediate goods back to the US, with the consequence of reduced demand for domestic manufacturing workers. While our research suggests that offshoring had a negative impact on employment, we caution that it does not support the view that offshoring and trade should be contained with tariffs or other policy interventions. Previous research has shown that both trade and offshoring are critical for consumers’ access to affordable goods in the US. Instead, our research implies that government assistance for displaced manufacturing workers could facilitate their transition to new jobs in other sectors.