What can AI tell us about COVID-19’s impact on infrastructure?

|

This page in:

Image
What can AI tell us about COVID-19’s impact on infrastructure? Makhtar Diop Finance
Image depicting the effects of COVID-19 on the market | © m.mphoto, Shutterstock

It goes without saying that COVID-19 is affecting infrastructure—for both projects in place and in planning. Instinctively, we can imagine that demand disruptions yield revenue losses for PPPs; construction is affected; and future planning while sands are still shifting is extremely challenging. At the World Bank we’ve noted that, in the medium term, we should anticipate a lasting downward trend in revenues, adverse impacts on access to financing for projects yet to reach financial close, and potentially continued disruption of construction schedules of projects already underway as the virus’ trajectory changes.

What are the implications of the hit infrastructure is taking on crisis recovery, once we get there? On the one hand, a downturn in infrastructure balance sheets, creation, and maintenance does not bode well—given the sector’s role in getting people to jobs, goods to market, electricity to hospitals, and digital access to students. On the other hand, we know from previous experience in other crises that many governments will use infrastructure spending to stimulate their economies, so there may be a ray of hope for the sector that allows it to continue to improve quality of life for people across the globe.

It’s too early to speculate how much economic stimulus will come from infrastructure spending in the (we hope near) future, especially as fiscal space continues to shrink. What is helpful, however, is an attempt to measure the pandemic’s current impact on infrastructure projects.

We trained an artificial intelligence (AI) model to collect data from the internet and report back with cases of project delays and cancellations in developing countries. We’ve been collecting this data since February and find it provides much more of a backstory than what we achieve with other data-gathering methods. The algorithm allows us, yes, to see the impact of COVID-19 in real-time, but also the major reasons behind cancellations and delays reported in different languages across the developing world—at minimum cost.

What did we find? In terms of developing-country infrastructure projects, 256 have been reported as cancelled or delayed. Project disruptions in pipeline peaked in April, stabilizing since then. For projects already under construction, the number of projects facing disruptions peaked in May and also have been decreasing. Read short report here

Most disruptions were due to travel limitations and disrupted supply chains. Projects sponsored by foreign entities were severely affected—as they often need foreign engineers and technicians at construction sites. With respect to supply-chain delays, the construction industry is heavily reliant on manufacturers in China, where operations were strongly affected by COVID-19 early on. The global solar PV value chain was particularly hard-hit because manufacturing capacity is concentrated in a few major markets, including China.

The second most reported reason for project disruption is the non-availability of laborers due to lockdown measures. Countries like the Philippines, India, and Colombia enforced enhanced community quarantines, resulting in labor shortages at construction sites.

Other reasons include delayed or cancelled tender processes, lower demand projections, and government budget funds reallocation to tackle COVID-19 containment.

It’s noteworthy that the major reasons for delays or cancellations—travel limitations, supply changes, and labor availability—are temporary issues. This indicates that the infrastructure business may improve once the pandemic is better controlled. Many projects delayed in the Philippines, for example, resumed construction as soon as lockdown measures lifted in May. More generally, data shows that nearly 20 percent of disrupted projects have (for now) resumed activity.

Unfortunately, we must pay strict attention to fresh spikes in COVID-19 cases. We must also consider that infrastructure projects are often sequential and interdependent. For example, we’ve already witnessed delays transmission line construction that put six development-stage wind projects under financial stress. A series of open tenders—once delayed in April—are now cancelled due to the government's non-indebtedness policy amid the pandemic-induced fiscal crisis.

What we know for sure is that development finance institutions (DFIs) face a test of whether they can live up to the responsibility of ensuring recent development gains are not lost. In order to pass, DFIs—among many other responsibilities—should help countries address both immediate and long-term infrastructure needs.

DFIs can, for example, assist governments in diagnosing risks early to mitigate the virus’ impacts on project portfolios through expert support, training, and tools. Projects in pipeline should be re-reviewed often in this changing environment and projects under construction or in operation should be reexamined for any contractual remedies available to mitigate the effect of potential delay and funding needs. Where governments lack the expertise and resources to take this on, DFIs can step in to help.

Since our AI data shows that travel restrictions hinder many aspects across the project cycle, we need to pay particular attention here—as this can affect the strong due diligence that ensures projects are well-executed, necessary, and sustainable. We can remind ourselves that services provided by DFIs have evolved to cope with risks unique to different moments throughout history. This is an opportunity to be agile and creative: can DFIs share due diligence to move good projects forward? We know that many governments already were struggling with the extra expertise, time, and cost that effective due diligence requires; this may be a moment to streamline onerous processes.

Returning to the importance of AI, we’re seeing it play a critical role in many aspects of the global COVID-19 crisis response. It’s accelerating research progress by processing large volumes of data, enabling policymakers and the medical community to understand the virus faster and more cheaply.

We should follow this lead, actively exploring how we can apply AI to support the infrastructure sector more. Our work using AI for monitoring the virus’ impact on projects is novel, and we strongly believe there are more ways to support policymakers and society to manage different stages of the crisis and its aftermath. 

There’s significant energy around this topic. It can be one of the many ways that countries recover from the crisis stronger, more resilient, and better prepared for future disruptive events like pandemics.

 

Related Posts

2019 brought private investment in infrastructure to more countries: Is there a message for policymakers amid COVID-19?

How the World Bank is looking at COVID-19 and public-private partnerships, right now and post-crisis

COVID-19 and infrastructure: A very tricky opportunity

What gets measured gets done: 50 countries reform their PPP regulations

Bouncing back is not enough. Let’s bounce forward to infrastructure resilience

 


 

This blog is managed by the Infrastructure Finance, PPPs & Guarantees Group of the World Bank. Learn more about our work here.

Peter Gatheca
November 30, 2020

Makhtar,
This is a very interesting read on the application of AI in economic performance and development monitoring. I would definitely like to read more case studies on such since it's rather obvious AI will play an even bigger role in the near and distant future. Developing countries such as Kenya could certainly benefit from this proactive use of emerging technologies.

Peter

David Boroto
October 12, 2021

Very insightful data on project disruptions, and an interesting application of AI to collect it. What were the primary data sources for project information when webscraping with the AI algorithm? Public government websites, news articles, any centralized data sources maybe?