VoxEU Column Global crisis

Input specificity and the propagation of idiosyncratic shocks in production networks

Little is known about how adverse shocks spread through production networks. This column presents quantitative evidence on inter-firm contagion using natural disasters as exogenous instruments. Adverse shocks to upstream suppliers lower sales growth and valuation of a downstream firm.

The Great Recession shed light on the interdependence of firms and the risk of bankruptcy cascades, but there is currently little empirical evidence on the transmission of shocks within firm networks. Using the occurrence of natural disasters in the US, this column shows that idiosyncratic shocks to individual firms propagate significantly along the supply chain.

The transmission of micro shocks between individual firms

Frequent interactions between firms occur through supplier-customer relationships. Prior research examined the role of industry linkages as a source of amplification of microeconomic shocks (e.g. Long and Plosser 1987, Horvath 1998, 2000, Foerster et al. 2011, Acemoglu et al. 2012). While sector-level linkages have been extensively studied, little is known on the diffusion of micro shocks between individual firms.

At the firm level, shocks should be easily absorbed in production networks. It is plausible that organize their operations to avoid being affected by shocks to their supplies. Even when they face such disruptions, firms are supposedly flexible enough to recompose their production mix, or to switch to other suppliers. On the other hand, frictions may prevent firms from making adjustments when trade partners are adversely hit by shocks. Idiosyncratic shocks might then propagate from firm to firm, gradually amplifying.

Evidence from natural disasters

In recent research, we exploit the occurrence of natural disasters in the past thirty years in the US and supplier-customer links reported by publicly listed firms to shed light on the role of supply chain linkages in propagating firm-level shocks (Barrot and Sauvagnat 2014). First, we find that the sales growth of supplier firms directly hit by a natural disaster drops by around five percentage points. As the red line in Figure 1 shows, this has an effect on the customers of these disrupted suppliers. The sales growth of customers drops by around two percentage points when one of their suppliers is hit by a natural disaster. We also find evidence of horizontal propagation – other suppliers of affected customers also experience a drop in sales growth.

Figure 1. Natural disaster strikes and sales growth

We proceed to examine whether the drop in firms’ sales translates into value losses. When a natural disaster hits a county in which a supplier is located, the news of input disruption should be quickly reflected in the customer firm’s share price, allowing us to compute the associated drop in firm value. As shown in Figure 2, firms experience a drop in stock returns of around one percentage point when one of their suppliers is hit by a natural disaster. This is direct evidence that input disruptions have an effect on firm value.

Figure 2. Natural disaster strikes and stock returns

The role of input specificity

Finally, we show that input specificity is a key driver of the propagation of firm-level shocks. We rely on three different proxies to measure the specificity of any given supplier:

  • The classification of suppliers’ industries in either differentiated goods industries, standardized goods industries, or services;
  • Suppliers’ R&D expenses; and
  • The number of patents held by suppliers.

We find that the transmission of input shocks to customers and their other suppliers only occurs when a natural disaster initially hits one specific supplier.


How much can we learn from this research?  Our results can plausibly be extended to other forms of firm-specific idiosyncratic shocks, such as strikes or management turnover. They probably also extend to the specificity of inputs within the boundaries of the firm. While customer-supplier links reported by publicly listed firms allow us to pin down the nature of the input, we would expect similar results to be obtained within a firm, when the division producing a specific part of the final good is hit by a shock.


Acemoglu, D, V M Carvalho, A Ozdaglar, and A Tahbaz-Salehi (2012), “The Network Origins of Aggregate Fluctuations”, Econometrica, 80(5), 1977-2016.

Barrot, J-N and J Sauvagnat (2014), “Input Specificity and the Propagation of Idiosyncratic Shocks in Production Networks”, Working Paper.

Foerster, A T, P-D G Sarte, and M W Watson (2011), “Sectoral versus Aggregate Shocks: A Structural Factor Analysis of Industrial Production”, Journal of Political Economy, 119(1), 1-38.

Horvath, M (1998), “Cyclicality and Sectoral Linkages: Aggregate Fluctuations from Independent Sectoral Shocks”, Review of Economic Dynamics, 1(4), 781-808.

Horvath, M (2000), “Sectoral Shocks and Aggregate Fluctuations”, Journal of Monetary Economics, 45(1), 69-106.

Long, J B and C I Plosser (1993), “Real Business Cycles”, Journal of Political Economy, 91(1), 39-69.

945 Reads