How can production network analysis inform policy on Covid-19?

The economy is characterised by intricate linkages between consumers, businesses and sectors across national and global supply chains. Production network analysis provides insights into the effects of imposing and lifting lockdowns – and policies to support recovery.

Covid-19 is a global shock but the implications for different sectors vary greatly in what is an increasingly interconnected economy. Government interventions – such as fiscal policies, furlough programmes, bailouts and lockdowns – should make use of economy-wide supply chain data and network analysis to ensure that they are as effective as possible.

Super spreaders of the economic system

All production creates input-output linkages among economic sectors along intricate national and global supply chains: one firm buys supplies and components from many others, and in turn sells on its own outputs to many business or final customers. Households are part of this system through their labour market participation and consumption of final goods.

The effects of an economic policy – such as subsidies to a specific economic sector (or to households with certain characteristics) or lockdown of a specific industry – should be assessed based on the way it diffuses beyond the target of the policy to the rest of the economy. An effective economic policy is one that is able to make the best use of the links and feedbacks within the economic network.

An analogy between epidemiology and economics may illustrate this. In the same way that epidemiologists recognise the importance of tracking and tracing ‘super spreaders’ to avoid new waves of Covid-19, so economists must track and trace the ‘super spreaders’ of the economic system to create strong and robust economic recovery with balanced fiscal policies.

Production networks and lockdowns

Locking down a community and closing down economic activity in certain sectors are costly economically, and those costs are not equally spread throughout the economy. Key production activities that are more central to economy-wide supply chains (such as construction) have a much wider impact on the economy when they are idle.

What’s more, even if these central production sectors are kept open in lockdown, their scale of operations is likely to fall as downstream activities serving UK households directly (such as retail) are kept shut, thus depressing their own demand for intermediate goods and services.

Finally, employment declines in sectors under lockdown ‘spill over’ to the rest of the economy via stagnating final demand, as households scale back consumption in the face of job loss and income uncertainty.

Over the last decade, research on production networks has provided a framework for organising these complex interactions into an economic model and, consequently, has determined how the centrality of a production process in its supply chain affects the propagation of economic shocks and aggregate outcomes (see Acemoglu et al, 2012, for an early contribution and Carvalho and Tahbaz-Salehi, 2019, for a recent overview of this body of research).

This framework has been used successfully to understand how a shock to part of the supply chain – for example, the 2011 Japanese earthquake – disrupted a specific set of firms concentrated in a given geographical area and then spread throughout the supply chain, creating large national and global economic effects (Carvalho et al, 2020a).

In the current crisis, this framework can help to quantify the economic costs of a particular lockdown strategy, either already implemented or proposed, taking account of the complex propagation patterns of the ensuing economic disruption.

For example, based on labour supply restrictions by sector as implied by lockdown legislation, Barrot et al (2020) estimate that six weeks of economic lockdown – roughly the time elapsed between the 23 March announcement of UK lockdown and the 10 May announcement of progressive easing of lockdown – imply an annualised GDP decline of 5.5%.

It is particularly notable that upstream sectors in the production network – which were not directly hit by administrative closures – are estimated to suffer large losses due to the decline in downstream demand (see Barrot et al, 2020; and Baqaee et al, 2020a).

The coupled contagion and production problem

Production network analysis also helps to understand the potential effects of policy interventions that re-open some sectors but not others. Hence, it allows policy-makers to think about optimal targeted lockdown strategies. This is an important tool as the economy is partially re-opened, or in anticipation of future local lockdowns that may be needed.

For example, the partial re-opening of some economic sectors induces more labour mobility, thereby increasing the likelihood of new local outbreaks of the infection. These effects depend on the geographical distribution of labour of the sectors for which lockdown has been relaxed, and on how this is interposed with the centrality of such communities in the process of diffusion of Covid-19. An epidemiological model of diffusion could quantify those effects (see, for example, Ferguson et al, 2020).

Integrating the economic model – which allows us to compute economic costs for any possible lockdown strategy – and the epidemiological model – giving the implied disease transmission dynamics for a given strategy – allows policy-makers to answer a key question: what is the optimal sector lockdown strategy such that, given supply chain transmission, aggregate GDP losses are minimised subject to constraints on the level of disease transmission that society is willing to tolerate?

Baqaee et al (2020b) and Pichler et al (2020) make headway on this class of problems by studying a simplified coupled contagion and production environment. They both report that smart lockdown strategies can indeed improve on the trade-off between economic and health costs.

Related question: Coronavirus and the economy: what are the trade-offs?

Should exit strategies be coordinated across countries?

Given the nature of supply chains, spanning the entire world, the resolution of the coupled contagion and production problem within a country is just one piece of a jigsaw puzzle of the entire world economy. The effectiveness of how a country like the UK handles the coupled contagion and production problem will depend on how other countries, tightly connected to UK via trade linkages, deal with their very same problem.

If exit strategies are not coordinated across countries, the very nature of the Covid-19 epidemiological contagion process will create asynchronous delays in the supply chains. When parts of the supply chains will be able to operate at full capacity (such the part operating in China or South Korea once they had contained the first wave of Covid-19), others will be disrupted by local lockdowns (including parts of Europe now and many emerging markets). This suppresses production complementarities that are so valuable for boosting global productivity.

As a concrete example, in the aftermath of the 2011 Japanese earthquake, Boehm et al (2019) document that US automakers, relying on Japanese automotive supply chains, were brought to a near standstill in production given supply chain failures.

For the Covid-19 crisis, Bonadio et al (2020) report that one-third of the total Covid-19-related GDP contractions can be attributed to foreign lockdown propagating via trade linkages. Asynchronous lockdowns and uncoordinated exit strategies across countries may thus impose significant costs through global supply chain linkages even if the receiving country is not itself imposing lockdown.

Related question: What happens if trade and mobility are permanently reduced?

Can the network approach inform policies beyond lockdown and exit strategies?

Governments have already injected an unprecedented amount of money into the economy to support households and cushion unemployment. That will help to avoid a prolonged loss of revenue leading to cascading business failures or seizures in production, as well as supporting the health of the financial system.

Related question: Will government measures protect the most vulnerable in society?

To be effective, these injections must be targeted to generate positive economic feedbacks. In addition, the government’s limited fiscal capacity implies that subsidies to certain economic sectors should be financed by taxing other sectors. Which sectors should be subsidised and which sectors should finance such subsidies?

A sector that is taxed will increase the price of that sectoral output and so the tax will, partly, be passed through to consumers. A sector that receives a subsidy can decrease the price and may become more competitive, benefiting consumers. The optimal tax and subsidy scheme will then tax sectors that do not pass on much of the tax to consumers in higher prices and use the revenue to subsidise other sectors.

Related question: Why should the government provide income protection in a recession?

The practical problem in designing those fiscal policies is to identify the characteristics of those sectors with high and low pass-through to prices. A network analysis capturing production linkages across sectors, the level of competition within sectors and household consumption patterns can identify and categorise sectors for tax and subsidy schemes.

As an illustration, King et al (2019) adopt a production network analysis to analyse carbon tax reforms. They show that to reduce aggregate emissions effectively, a carbon tax should be targeted at sectors based on their location in the production network.

Galeotti et al (2020) determine optimal fiscal policies when the market is not competitive; Grassi and Sauvagnat (2019) provide an overview of how production network analysis informs a range of economic policies; and Galeotti et al (2020) provide a general analysis of targeted network interventions.

Related question: What is the size of the fiscal multiplier?

What data are needed to improve existing analysis of production networks?

There are three sets of data that are needed to develop targeted network interventions like the ones described above.

Reconstructing the production network

Data on global supply chains at the sector-country level are available from the World Input Output Database. The Office of National Statistics (ONS) provides input-output data for UK. The ONS data allow researchers to reconstruct the production network at an intermediate level of disaggregation and can be coupled with publicly available national accounts data (on final consumption and employment per sector).

Policy design would benefit from firm-to-firm transaction data, so that the UK production network can be constructed in more detail. This is important to develop a better understanding of the supply chain disruptions observed throughout the Covid-19 crisis, and the role of bottleneck firms in meeting demand for particular goods and services, be it flour, toilet paper or ventilators (on this see, for example, Carvalho et al, 2020b).

Related question: What should we do about price gouging?

In addition, it would allow policy-makers to understand how rapidly certain sectors can adjust their upstream and downstream supply chains relationships. Unfortunately, unlike in many other countries, these data are not available in the UK.

Mapping lockdown policies

Another data input to research on production networks during the Covid-19 crisis concerns the restrictions implied by a given lockdown plan. Thus far, this body of analysis has mapped lockdown restrictions as sector-specific labour supply restrictions. In particular, the work typically considers three different aspects of lockdown policies:

First, mandated closures of certain sectors (such as retail) are coded as direct restrictions to labour supply in those sectors.
Second, general confinement measures imply that a sector’s labour supply is limited by the amount of work that is possible to conduct remotely, by telecommuting. It is possible to obtain data that are proxies for the amount of telecommuting per sector by taking into account the mix of occupations in each sector. These are available from the ONS.
Third, school closures may again affect sectors differently. Barrot et al (2020) show how to leverage from census data to obtain the fraction of workers in each sector whose labour supply is most affected by school closures, that is, parents of school age children. On the other hand, detailed data on furloughed workers per sector are not yet available. This information is necessary to understand fully how final demand is affected during the crisis.

Data for integrated contagion and production models

Designing targeted network interventions requires accurate modelling of the health consequences of opening up or locking down a given sector. Furthermore, the spatial concentration of both economic activity and disease means that geographically fine-grained data are needed.

The ONS holds microdata on the spatial distribution of production (sector-by-sector) and the associated commuting patterns across the UK. But high quality data on the contact rate and disease transmission per sector (either on the worker side, when engaged in production, or on the household side, when engaged in consumption) are currently not accessible.

Where can I found out more?

From micro to macro via production networks: an accessible overview of research on production networks by Vasco Carvalho in the Journal of Economic Perspectives.

In and out of lockdowns: identifying the centrality of economic activities: Giorgio Barba Navaretti and colleagues describe targeted network exit strategies for Italy at VoxEU.

The role of global supply chains in the Covid-19 pandemic and beyond: Barthélémy Bonadio and colleagues discuss how the Covid-19 pandemic shock propagates along global supply chains at VoxEU.

Production networks and epidemic spreading: re-opening the UK economy: Anton Pichler and colleagues discuss how the coupling of production networks and epidemic modelling can guide lockdown relaxation scenarios at VoxEU.

Re-opening scenarios: James Stock presents his work (Baqaee et al, 2020b) on a epi-macro model featuring production networks, at NBER’s Youtube Channel.

Social distancing: what will the economic fallout be? Jean-Noel Barrot summarises his work on production networks and the economic costs of social distancing in France, at Knowledge@HEC.

Supply chain bottlenecks in a pandemic: Vasco Carvalho and colleagues discuss how firm-to-firm data can help policy-makers locate and shore up firms that are essential for the provision of key goods and services during a pandemic, at Cambridge-INET Covid-19 Analysis.

Trade and global value chains in the age of Covid-19: Richard Baldwin, Vasco Carvalho and Simon Evenett present at a World Bank webinar on trade, global value chains and supply chain disruptions, at the World Bank.