The Center for Data Innovation spoke with Fabio Pammolli, professor of Economics and Management at the Polytechnic University of Milan in Italy. Pammolli discussed how he is using mobility data from Facebook’s Data for Good program to inform policies that can bolster Italy’s post-pandemic economic future.
Hodan Omaar: What is the goal of your work, and how are you using Facebook data to achieve this goal?
Fabio Pammolli: Our goal is to analyze the impact that the COVID-19 lockdown has had on the socio-economic conditions of Italian citizens. Italy was the first European country to experience a major outbreak of COVID-19, with very high mortality rates in some of the northern regions. It was also the first European country to apply a national lockdown in response to the pandemic, despite already having a weak economic outlook. Now, there is growing concern that the economic consequences of lockdown are disproportionately impacting the poor.
To understand the relationship between the lockdown and the economy, we performed a study that explored how variations in mobility relate to some fundamental economic variables. First, we used data from Facebook’s Data for Good program to understand how mobility changed in different municipalities. Next, we compared this data against official statistics on the economic conditions of municipalities from the Italian Ministry of Economy and Finance, including data on individual indicators such as average income and aggregate indicators at the level of municipalities, to investigate the features of the territories that were most affected.
We found that the lockdown did not produce homogenous results across regions. Instead, two seemingly opposite patterns emerge. Individual indicators showed that poor people are more exposed to the economic consequences of the lockdown. Conversely, aggregate indicators such as fiscal capacity, which show a municipality’s ability to collect tax revenues, revealed that more efficient municipalities are those more severely hit by restrictions in mobility.
Omaar: So poor people in rich regions are the worst off. Why is that the case?
Pammolli: To understand these apparently contrasting results, you first have to understand the relationship between inequality and mobility contraction. We measure inequality as the ratio between mean and median income; this means that if two municipalities have the same average income, but one has a greater inequality, the latter has a greater share of poor people.
Poorer people tend to live and work in regions that have industries that are less economically productive and are strongly dependent on mobility flows and direct contact with customers. Industries like tourism, retail, and services. Mobility trends associated with these industries experienced a sudden contraction of more than 90 percent during the lockdown in Italy. This explains why poor people are more exposed to the economic consequences of COVID-19. But that is only half of the story.
Wealthy municipalities (i.e., those with high fiscal capacity) need to be able to sustain poor individuals. But if inequality in a wealthy municipality is high, then there are more poor people to sustain—more so than in other municipalities with relatively less inequality. This creates an additional strain on territories. Furthermore, the pandemic is causing all areas to experience a sharp reduction in the fiscal revenues their tax bases can generate. This is the reason why poor people in richer, more unequal municipalities are worse off.
Omaar: What can policymakers do to balance efforts to stop the spread of the contagion and bolster economic recovery?
Pammolli: The impact COVID-19 will have on a territory depends on the socio-economic structure of that territory. Therefore, policymakers should not embrace a one-rule-fits-all approach for their mitigation efforts. Rather, they should adopt policy actions tailored to the socio-economic contours of their territories.
I co-authored a study in October that simulated the economic impact of implementing different mobility restriction measures to contain COVID-19 in Italy. When trying to stop the spread of COVID-19 and support economic growth, policymakers need to focus on the two sources of economic losses: the number of infected people and the restrictions in mobility that prevent individuals from working. Obviously, mobility restrictions are useful for containing high losses in the number of available workers as they mitigate the epidemic spreading. But restrictive policies also have a detrimental effect on the economy as they force non-infected workers to vacate workplaces.
We used Facebook mobility data to model the daily movements of over four million individuals in the country, then modeled the economic outcomes derived from changing the timing and modality of government mobility restrictions. Our study shows that the total reduction of disposable income can vary anywhere from 10 percent in the best case to 40 percent in the worst case, and that the outcome is due to non-linear interactions between mobility policies and infection transmission rates. We believe that our findings can contribute to inform data-driven policy design.
Omaar: Has using Facebook data improved the models you are creating?
Pammolli: Yes. While there is a well-established literature incorporating mobility patterns in epidemic modeling, economic models of epidemics have mostly focused on one type of macroeconomic model that assumes individuals uniformly distributed in space and randomly moving. But this is not a realistic assumption. By using near real-time mobility data from Facebook, we have been able to create simulations within population models that are better tailored to address the role of mobility in the contagion.
In general, I think having such data available for research is very useful and the Facebook model is one that other providers should follow.
Omaar: Does it matter that the Facebook data doesn’t include users under the age of 18?
Pammolli: Not in our case. Our work is focused on using mobility flows that proxy economic activities, so we’re focusing mostly on the mobility of the active fraction of the population. Even when validating the data, we compare it against mobility matrices that refer to the active population. So, for us it is not a problem. Of course, if you want to study other phenomena, excluding such users may become an issue.