How shocks propagate through the economy and contribute to fluctuations has been one of the central questions of macroeconomics. We argue that a major mechanism for such propagation is input-output linkages. Through input-output chains, shocks to one industry can influence ‘downstream’ industries that buy inputs from the affected industry, as well as ‘upstream’ industries that produce inputs for the affected industry. These interlinkages can propagate and potentially amplify the initial shock to further firms and industries not directly affected, influencing the macro economy to a much greater extent than the original shock could do on its own.
The significance of the idea that a shock to one firm or disaggregated industry could be a major contributor to economic fluctuations was downplayed in Lucas’ (1977) famous essay on business cycles. Lucas suggested that due to the law of large numbers, idiosyncratic shocks to individual firms should cancel each other out when considering the economy in the aggregate, and therefore the broader impact should not be substantial. Recent research, however, has questioned this perspective. For example, Gabaix (2011) shows that when the firm size distribution has very fat tails, the power of the law of large numbers is diminished and shocks to large firms can overwhelm parallel shocks to small firms, allowing such shocks to have a substantial impact on the economy at large. Acemoglu et al. (2012) show how microeconomic shocks can be at the root of macroeconomic fluctuations when the input-output structure of an economy exhibits sufficient asymmetry in the role of some disaggregated industries as (major) suppliers to others.
In Acemoglu et al. (2016), we empirically document the role of input-output linkages as a mechanism for the transmission of industry-level shocks to the rest of the economy. Our approach differs from previous research in two primary ways.
- First, whereas much prior work has focused on the medium-term implications of such network effects (e.g. over more than a decade), we emphasise the influence of these networks on short-term business cycles (e.g. over 1-3 years).
- Second, we begin to separate types of shocks to the economy and the differences in how they propagate.
We build a model that predicts that supply-side (e.g. productivity, innovation) shocks primarily propagate downstream, whereas demand-side shocks (e.g. trade, government spending) propagate upstream. For example, a productivity shock to the tire industry will tend to strongly affect the downstream automobile industry, while a shock to government spending in the car industry will reverberate upstream to the tire industry. We then demonstrate these findings empirically using four historical examples of industry-level shocks, two on the demand side and two on the supply side, and confirm the predictions of the model.
Model and prediction
We model an economy building on Long and Plosser (1983) and Acemoglu et al. (2012), in which each firm produces goods that are either consumed by other firms as inputs or sold in the final goods sector. The model predicts that supply-side (productivity) shocks impact the industry itself and those consuming its goods, while a demand-side shock affects the industry and its suppliers. The total impact of these shocks – taking into account that customers of customers will be also affected in response to supply-side shocks, and suppliers of suppliers will also be affected in response to demand-side shocks – is conveniently summarised by the Leontief inverse that played a central role in traditional input-output analysis.
The intuition behind the asymmetry in propagation for supply versus demand shocks relates to the Cobb-Douglas form of the production function and preferences. If productivity in a given industry is lowered by a shock, firms in that industry will produce fewer goods and the price of their goods will rise. Due to the Cobb-Douglas structure, these effects cancel each other out for upstream firms, leaving them unaffected, while downstream firms feel the increase in prices and consequently lower their overall production. On the other hand, if demand in a certain industry increases, firms in that industry increase production, necessitating a corresponding increase in input production by upstream firms. Because of constant returns to scale, however, the increased demand does not affect prices, and so downstream firms are not changed.
We also incorporate into the model geographic spillovers, showing that shocks in a particular industry will also influence industries that tend to be concentrated in the same area, as shown empirically by Autor et al. (2013) and Mian and Sufi (2014). The idea is that a shock to the first industry will influence local demand generally, and therefore will change demand, output, and employment for other local producers.
We test the model’s prediction by examining the implications of four shocks: changes in imports from China; changes in federal government spending; total factor productivity (TFP) shocks; and productivity shocks coming from foreign industry patents. The first two are demand-side shocks; the latter two affect the supply side. For each of these shocks, we show the effects on directly impacted industries as well as upstream and downstream effects. Our core industry-level data is taken from the NBER-CES Manufacturing Industry Database for the years 1991-2009, while input-output linkages were drawn from the Bureau of Economic Analysis’ 1992 Input-Output Matrix and the 1991 County Business Patterns Database.
For brevity we focus here on the first example, where changes in imports from China influence the demand in affected industries. Of course, rising import penetration in the US for a given industry could be endogenous and connected to other factors, such as sagging US productivity growth. We therefore instrument import penetration from China to the US with rising trade from China to eight non-US countries relative to the industry’s market size in the US, following Autor et al. (2013) and Acemoglu et al. (2015). Chinese imports to other countries can be taken as exogenous metrics of the rise of China in trade over the last two decades.
The empirics confirm the predictions of our model. A one standard-deviation increase in imports from China reduces the affected industry’s value added growth by 3.4%, while a similar shock to consumers of that industry’s products leads to a 7.6% decline.
- In other words, the upstream effect is nearly twice as large as the effect on the directly hit industry in a basic regression.
- Downstream effects, on the other hand, are of opposite sign and do not change in a statistically significant manner, confirming the model’s prediction.
Figure 1 shows the impulse response function when our framework is adjusted to allow for lags and measure multipliers. Again, a one standard-deviation shock to value added through trade produces network effects that are much greater than the own effects on the industry.
- We calculate that the effect of a shock to one industry on the entire economy is over six times as large as the effect on the industry itself, due to input-output linkages.
Similar effects are found for employment, and the findings are shown to be robust under many different specification checks.
Figure 1. Response to one SD value-add shock from Chinese imports
The other three shocks – changes in government spending, TFP shocks and foreign patenting shocks – also broadly support the model’s predictions, with the first leading to upstream effects and the latter two leading to downstream effects. Similarly, extensions quantify that geographical proximity facilitates the propagation of the shocks, particularly those on the demand side.
Shocks to particular industries can reverberate throughout the economy through networks of firms or industries that supply each other with inputs. Our work shows that these shocks are indeed powerfully transmitted through the input-output chain of the economy, and their initial impact can be substantially amplified. These findings open the way to a systematic investigation of the role of input-output linkages in underpinning rapid expansions and deep recessions, especially once we move away from simple, fully competitive models of the macro economy.
Acemoglu, D, U Akcigit, and W Kerr (2016), “Networks and the Macroeconomy: An Empirical Exploration”, NBER Macroeconomics Annual, forthcoming. NBER Working Paper 21344.
Acemoglu, D, V Carvalho, A Ozdaglar, and Al Tahbaz-Salehi (2012), “The Network Origins of Aggregate Fluctuations”, Econometrica, 80:5, 1977-2016.
Acemoglu, D, D Autor, D Dorn, G Hanson, and B Price (2015), “Import Competition and the Great U.S. Employment Sag of the 2000s”, Journal of Labor Economics, 34(S1), S141-S198.
Autor, D, D Dorn, and G Hanson (2013), “The China Syndrome: Local Labor Market Effects of Import Competition in the United States”, American Economic Review, 103:6, 2121-2168.
Gabaix, X (2011), “The Granular Origins of Aggregate Fluctuations”, Econometrica, 79, 733-772.
Long, J and C Plosser (1983), “Real Business Cycles”, Journal of Political Economy, 91:1, 39-69.
Lucas, R (1977), “Understanding Business Cycles”, Carnegie Rochester Conference Series on Public Policy, 5, 7-29.
Mian, A and A Sufi (2014), “What Explains the 2007-2009 Drop in Employment”, Econometrica, 82:6, 2197-2223.