Many countries around the world are finally past the first peak of the COVID-19 pandemic. As restrictions are cautiously being lifted, it is a good time to assess what could be done better if a second wave were to materialise. Two types of policies in particular – social distancing and health care investment – can dramatically affect the human and economic toll of COVID-19. The former reduces contagion but carries high economic and social costs (Toda 2020 and Baldwin 2020). The latter is key to ensure that care is provided to everyone in need (Shiva 2020). In a recent VoxEU column (Ciminelli and Garcia-Mandicó 2020a), we analysed death registry data for a selected sample of about 1,000 Italian municipalities to shed some initial light on the COVID-19 epidemic in Italy. In this
Ciminelli, Garcia-Mandicó considers the following as important:
This could be interesting, too:
Makridis, Wang writes Social networks and consumption patterns during a pandemic
Bircan, De Haas, Schweiger, Stepanov writes Coronavirus credit support: Don’t let liquidity lifelines become a golden noose
Schwandt writes Pregnancy during the pandemic
S. S. writes Declining worker power versus rising monopoly power
Many countries around the world are finally past the first peak of the COVID-19 pandemic. As restrictions are cautiously being lifted, it is a good time to assess what could be done better if a second wave were to materialise. Two types of policies in particular – social distancing and health care investment – can dramatically affect the human and economic toll of COVID-19. The former reduces contagion but carries high economic and social costs (Toda 2020 and Baldwin 2020). The latter is key to ensure that care is provided to everyone in need (Shiva 2020).
In a recent VoxEU column (Ciminelli and Garcia-Mandicó 2020a), we analysed death registry data for a selected sample of about 1,000 Italian municipalities to shed some initial light on the COVID-19 epidemic in Italy. In this column, we use newly available data – covering almost all municipalities – to expand on our earlier analysis. We examine two main issues. The first is whether the closure of all non-essential businesses was effective in reducing mortality. The second concerns the management of the health care system: we quantify how many lives could have been saved through better preparedness. The focus is on the eight regions of Italy’s north – Emilia-Romagna, Friuli-Venezia Giulia, Liguria, Lombardia, Piemonte, Trentino-Alto Adige, Valle d’Aosta, and Veneto – together accounting for about 50% of Italy’s population and 85% of all official COVID-19 fatalities. Here, we only emphasise the main results of our analysis, while leaving the details in a supporting paper (Ciminelli and Garcia-Mandicó 2020b).
Unfortunately, the new data confirm that COVID-19 may have killed almost 0.1% of the local population in just over a month and that its mortality is vastly underreported in official statistics. The extent of underreporting is even greater than what we had initially calculated. A conservative estimate, which discounts indirect deaths due to other causes, puts it at 2.2 – meaning that, for each officially recorded COVID-19 fatality, an additional 1.2 deaths went undetected. We can also confirm that COVID-19 had much larger effects in Lombardia than all other regions. The most striking contrast is between Lombardia and Veneto. Again, this is masked in official statistics. According to those, COVID-19 mortality in Lombardia is five times larger than in Veneto. Instead, we estimate the difference between these two regions to be up to eight-fold.
Closing down services sensibly reduced COVID-19 mortality, but shutting down factories did not
To fight COVID-19, the Italian government adopted some of the most drastic measures among Western countries. On 8 March, it ordered the closure of all non-essential services, such as bars, restaurants, shops, and many other activities involving interactions between workers and consumers. About ten days later, it compounded those measures by ordering the closure of all factories producing non-essential goods.1 Were these decisions effective in reducing mortality?
In Table 1 below, we compare the mortality effect of COVID-19 in the average municipality against the effect in those with a ten percentage point higher share of employment in businesses that were shut down. If the closure of non-essential businesses was indeed effective in reducing mortality, we should see more muted effects of COVID-19 in municipalities where a larger share of people was working in businesses that the government ordered to close down.
Regardless of whether we consider the full sample of municipalities or we only focus on those within the epidemic epicentre, our results indicate that shutting down factories did not lead to any significant decrease in mortality – the mortality effect of COVID-19 is virtually identical in the average municipality and in those with a high employment share in factories that had to close down.
On the other hand, the closure of non-essential service activities was very effective in reducing mortality – COVID-19 killed about 15% less people in municipalities with a ten percentage points higher share of employment in close down services. These results are robust to considering a time sample spanning until 15 April and to controlling for a battery of factors potentially driving mortality, such as population density, air pollution, external commuting, and many others.
Table 1 The closure of non-essential businesses and the effect of COVID-19 on mortality
Notes: the table reports the estimated effects of COVID-19 on daily mortality rates per 100,000 inhabitants in the average municipality and in municipalities with a share of employment in shut down factories and service activities of ten percentage points above the mean. The mean is 20% for services and 30% for factories (factories also includes construction sites). The ‘full sample’ includes all municipalities in Italy’s north with more than 1,000 inhabitants. The ‘within epicentre’ sample includes municipalities within an area of 50km from the towns of Alzano Lombardo and Codogno, (both in the Lombardia region).
Why did closing down services reduce mortality while shutting down factories did not? Workers in the service sector interact with consumers everyday – the opposite of social distancing. Closing down service activities likely reduced contagion and thus mortality. But, for the most part, factory workers only interact with other workers in the same unit. In Italy’s north, about 35% of all employment in manufacturing and construction industries is concentrated in micro firms, with less than ten workers per unit. Another 40% is in firms with between ten and 50 workers per unit. Contact is thus fairly limited and, if infected workers are readily identified, contagion may be effectively contained without having to close them down.
COVID-19 mortality peaked at 12 deaths per 100,000 inhabitants per day in the outbreak epicentre
Next, we zoom in on the epicentre of Italy’s COVID-19 outbreak, which we define as the area including all the municipalities at a distance of 50km or less from either Codogno or Alzano Lombardo, in the Lombardia region. This is the most densely populated and wealthiest area of Italy. With almost 1,000 municipalities, it makes up for about 10% of Italy’s total population and more than 15% of its national income. It also has the best health care infrastructure in the country, which itself has one of the highest rates of intensive care units (ICUs) per capita in the all world (McCarthy 2020). However, the force with which the virus struck was so intense that it brought the system close to collapse. We estimate that, at its peak, around 20 March, COVID-19 was killing an average of 12 people per day per 100,000 inhabitants in Italy’s COVID-19 outbreak epicentre.
But mortality was up to 50% lower in municipalities with an ICU in town
Although Italy has many ICUs, they are not evenly distributed. Using health care data from the Lombardia region, which hosts 95% of all municipalities in the COVID-19 epicentre, we find that less than 4% of them have an ICU in town. For about 40% of municipalities, the closest is further than ten kilometres as the crow flies.2 As COVID-19 was making inroads among the population, calls to the emergency line soared, and, as the system became overwhelmed, waiting times for emergency transportation swelled. To make a trip that usually took only eight minutes, ambulances were taking an hour, and in some cases they were not getting in on time (Sorbi 2020).
We estimate that COVID-19 mortality was sensibly lower in municipalities with an ICU in town relative to those that did not have one. Figure 1 below reports our estimates of COVID-19 mortality by age bins, in the average municipality (dots) – where the closest ICU was ten kilometres away – and in municipalities with an ICU in town (crosses).3
Figure 1 The effect of COVID-19 with and without an ICU in town, by age groups
Notes: The Figure reports the effect of COVID-19 on mortality in the average municipality (with the closest ICU being ten kilometres away) and in municipalities with an ICU in town. Spikes denote 95% confidence intervals. The effects are for men and are estimated for the sample of municipalities in Lombardia that lie within an area of 50km from the towns of Alzano Lombardo and Codogno (the epidemic epicentres). For women the effects are qualitatively similar but lower.
Strikingly, we cannot reject that COVID-19 had a significant effect on mortality on men as young as 40 years old. Among men above 55, municipalities with an ICU in town consistently experienced 30% to 50% lower mortality across all age bins (Figure 1). Why was this the case? In Figure 2, we consider the overall population (that is, not differentiating by age) and compare the extra effect, over time, of being ten kilometres away from the ICU to the daily volume of incoming emergency calls for respiratory reasons or infectious diseases.
The additional effect on mortality of being far from the ICU became larger as the number of incoming emergency calls for respiratory or infectious diseases swelled – a sign that the congestion of the emergency care system may have prevented critical patients from being transported to the ICU on time. At the peak, municipalities with the closest ICU being ten kilometres away experienced six more deaths per day per 100,000 inhabitants than municipalities with an ICU in town – almost twice as much.
Figure 2 Effect of being ten kilometres from the closest ICU at time of hospitals congestion
Notes: The Figure reports the additional effect of COVID-19 on mortality in municipalities with the closest ICU being ten kilometres away (blue solid line). The blue shaded area denotes 95% confidence interval. The effects are for men and are estimated for the sample of municipalities in Lombardia, within an area of 50km from the towns of Alzano Lombardo and Codogno (the epidemic epicentres). For women the effects are qualitatively similar but lower. The dashed line reports the daily number of emergency calls for respiratory and infectious disease reasons (right axis).
Our results suggest that, given resource and time constraints, medical staff may have had to prioritise serving more patients at the expense of reducing geographical coverage. Going forward, Italy should invest to strengthen its emergency care response, by improving pre-hospital emergency services – such as clarifying first point of contact for possible COVID-19 cases, expanding capacity to manage large volumes of calls, improving phone triage to better prioritise care delivery – building up ambulance capacity, and ideally mobilising ICUs more evenly across the territory. All these factors would be key to help reduce mortality if a new outbreak were to materialise.
The scale of underreporting increases to a factor of three among women in their 80s
We started this column by noting that COVID-19 deaths are vastly underreported in official statistics. Underreporting is particularly high among the elderly, and especially so among older women. Figure 3 below compares our estimate of COVID-19 deaths against official statistics, by age and gender. For each officially recorded death of a woman in her 80s, two others went undetected. The ratio of official to undetected deaths increases to a striking seven-to-one when considering women above the age of 90. After properly accounting for all these undetected deaths, the gender gap in COVID-19 mortality is sensibly reduced. According to official statistics, COVID-19 killed more than two men for each woman, while our estimates indicate that deaths among men were only 30% higher than those among women.
Figure 3 COVID-19 deaths from registry data and in official data, by age and gender
Notes: The figures compare our estimate of COVID-19 deaths from registry data to official COVID-19 fatalities, by age and gender. Source: ISTAT (2020) and SARS-CoV-2 Surveillance Group (2020)
Why do we observe so many undetected deaths among the elderly, and particularly so among women? All available anecdotical evidence points to nursing homes. As emphasised by The Economist (2020a), care home residents are acutely vulnerable to COVID-19 and little has been done to protect them. Among European countries that count deaths in nursing homes in official statistics, these make up for 30% of all official fatalities, on average. This ratio increases up to 50% in Belgium (The Economist 2020b).
Living in a nursing home may have significantly increased the chances of dying for the very old
Since Italy does not include nursing homes deaths in its COVID-19 statistics, it is likely that they account for a big chunk of the undetected deaths. However, our registry data do not provide information on the circumstances of death, and therefore we cannot precisely say how much is the share of nursing home deaths in the underreporting. But there are other ways in which we can shed more light on COVID-19 deaths in nursing homes.
Using data on the exact location and number of beds of each nursing home in the Lombardia region, we find that nursing homes are much more evenly distributed than ICUs. About 50% of all municipalities have a nursing home in town. Overall, the population of nursing home residents tops 65,000.4 Was COVID-19 mortality higher in nursing homes than outside them, across the same demographic group? In other words, does the mere fact of living in a nursing home increase the chance of dying of COVID-19?
With a large number of residents sharing the same common spaces and having close contacts with multiple staff members, nursing homes may have acted as hotbeds of contagion. Moreover, as nursing homes do not qualify as medical centres in Italy, they were heavily understaffed and unprepared to deal with the crisis, lacking protective equipment for staff, and emergency care equipment for infected patients (ATS Insubria 2020). In Lombardia, these inherent characteristics of nursing homes may have been particularly aggravating, as the regional authority decided to relocate COVID-19 positive patients with mild symptoms from hospitals to nursing homes (La Stampa 2020).
Figure 4 below shows the average mortality effect of COVID-19 (dots) and the effect that would have prevailed if nobody lived in nursing homes (crosses). Among ages between 65 and 84, we do not find a significant difference neither among men nor among women. For those aged 85 and over, however, our results suggest that living in a nursing home may have significantly increased the chance of dying during the COVID-19 epidemic – possibly by one-third among men and about one-half among women.
Figure 4 The mortality effect of COVID-19 on average and had there been no nursing homes
Notes: the figure reports unconditional (average) effects on mortality, and effects conditional on there being no nursing homes. This conditional effect is estimated after constructing a measure of the proportion of the elderly residing in a nursing home, which uses information on the composition by age and sex of nursing home residents and on the number of nursing homes beds in each municipality. Spikes denote 95% confidence bands. Estimates are for a sample of municipalities in Lombarida, within an area of 50km from the towns of Alzano Lombardo and Codogno (the epidemic epicentres). The effects would be qualitatively similar but lower if we considered the entire Lombardia region
Authors' note: The views expressed in this column are those of the authors and do not represent those of the OECD or its member countries.
ATS Insubria (2020), “Liste di Attesa nelle RSA”, 1 May.
Baldwin, R (2020), “Remobilising the workforce: A two imperatives approach”, VoxEU.org, 13 April.
Ciminelli, G and S Garcia-Mandicó (2020a), “COVID-19 in Italy: An analysis of death registry data”, VoxEU.org, 22 April.
Ciminelli, G and S Garcia-Mandicó (2020b), "Mitigation Policies and Emergency Care Management in Europe's Ground Zero for COVID-19”, available at SSRN.
Finizio, M (2020), “Chiamate al 118, in lombardia oltre il 30 percent per motivi respiratori e infettivi. picco il 16 marzo”, Il Sole 24 Ore, 7 April.
ISTAT (2020), “Decessi del 2020”, database, Accessed on: 12 April.
La Stampa (2020), “Fontana su anziani morti nelle Rsa lombarde: ‘L'utilizzo di quelle case di riposo per i malati Covid è responsabilità dell'Ats’”, 17 April.
McCarthy, N (2020), “The countries with the most critical care beds per capita”, Statista.com, 12 March.
SARS-CoV-2 Surveillance Group (2020), “Characteristics of sars-cov-2 patients dying in italy. report based on available data on April 2nd, 2020”, Istituto Superiore Sanità.
Shiva, M (2020), “We need a better head start for the next pandemic”, VoxEU.org, 26 April.
Sorbi, M (2020), “Un'ora di attesa per un'ambulanza. Ora anche il 118 rischia il collasso”, Il Giornale, 21 March.
The Economist (2020a), “The impact of covid-19 on care homes”, 16 April 2020 edition.
The Economist (2020b), “Many covid deaths in care homes are unrecorded”, 9 May 2020 edition.
Toda, A A (2020), “Early draconian social distancing may be suboptimal for fighting the COVID-19 epidemic,” VoxEU.org, 21 April.
1 Among large western countries, only the government of Spain ordered the closure of all non-essential factories.
2 This distance should be seen as an underestimate of the true distance to reach an ICU. It measures distance between two municipality centres as the crow flies. Therefore, it does not account for roads, nor within-city travelling to reach the ICU.
3 Ten kilometres is the average distance to ICU across municipalities in our sample. Therefore, the estimates of ICU at ten kilometres should be interpreted as the average effect of COVID-19 on mortality.
4 Almost every nursing home in Lombardia has hundreds of people in their waiting lists. We have calculated that there are, on average, 157 people waiting for every bed that is still available (ATS Insubria 2020). For our calculations, we assumed full occupancy.