With a territory of over 3 million square miles and a population of more than 210 million, Brazil is geographically the fifth-largest country in the world and the sixth-largest in population. It encompasses a wide range of cultural, political, demographic, economic, and behavioural variations that originate from Europe to Africa.
Charles Wyplosz questioned the extent to which policy reactions to COVID-19 are driven by political factors, and which are driven by other influences like history, culture, ethnic divisions, political regimes, and election laws on one hand, and the price that societies attribute of life on the other (Baldwin and Welder di Mauro 2020).
Wyplosz’s question is particularly pertinent to Brazil. Given this country’s social complexity, it would not be difficult to assume that Brazil would face great challenges in controlling the COVID-19 epidemic. Indeed, Brazil saw major standoffs in designing policies to combat the epidemic, with incoordination between federal and state powers over containment policies.
On the one hand, the state governments defended their autonomy in implementing these policies, aiming at more severe measures, while on the other hand, the federal power, in the words of Ajzenman et al. (2020), “publicly flaunted social distancing measures and downplayed the seriousness of the disease in at least two well-publicized instances”.
The impact of this lack of coordination between federal and state actions was noted on 22 May 2020 when, amid these political disputes, the WHO declared Brazil the new epicentre of the coronavirus pandemic. Wyplosz’s words that “...governments’ reactions will reveal the nature of their leaders and, more widely, that of societies” (Baldwin and Welder di Mauro 2020) could not be more prophetic.
But is it justifiable for states to have the independence to implement their own containment policies? Are the characteristics of Brazilian individual states different to the point that epidemic dynamics differ across the country, thus justifying this state-level independence?
Remarkable epidemic heterogeneities in Brazil
A recent study by Hallal et al. (2020) provides insights into these questions. The study examines first-wave seroprevalence surveys of household probabilistic samples of 133 large sentinel cities in Brazil, including 25,025 participants from all 26 states and the Federal District.
It finds that the seroprevalence of antibodies to SARS-CoV-2, assessed using a lateral flow rapid test, varied markedly across the cities and regions, from below 1% in most cities in the South and Center-West regions to up to 25% in the city of Breves in the Amazon (North) region.
From these findings by Hallal et al. (2020), it is possible to observe that the regions with the greatest differences in seroprevalence of SARS-CoV-2 antibodies were precisely those regions that are most heterogeneous culturally, economically, socially, and demographically. While urban metropolises in the Southeast had, on average, low prevalence of antibodies, as in the case of São Paulo (3.3%), smaller cities in the interior of the North, mainly coastal regions of the Amazon River, presented levels on average above 10%, as in the case of Tefé (19.8%).
Figure 1 São Paulo, a Brazilian urban megalopolis
Figure 2 Tefé, a city in the interior of the Amazonas state with a predominantly agricultural culture
Such results suggest that regional differences in epidemic consequences may be deeply related to differing economic, cultural, and behavioural characteristics of these states. Thus, we explore the possible influence on epidemic dynamics of the different characteristics intrinsic to each state or region.
A theoretical investigation for Brazil
In our study (Borelli and Góes, 2020), we use the SIR-macro model (Eichenbaum et al. 2020) to investigate how differences in local characteristics may have affected epidemic dynamics and their economic consequences in five Brazilian states with the most critical epidemic situations to date, namely, São Paulo (SP), Amazonas (AM), Ceará (CE), Rio de Janeiro (RJ) and Pernambuco (PE).
The SIR-macro model (Eichenbaum et al. 2020) extends the canonical SIR model (Kermack and McKendrick 1927), allowing the behaviour of economic agents to influence epidemic dynamics. In this model, economic agents can voluntarily choose to reduce their consumption and work activities to lower their chances of being infected. They can also be induced to reduce these activities through containment measures adopted by the authorities.
The dynamics of the model follow a logic similar to that by Gourinchas (2020), that flattening the infection curve inevitably accentuates the macroeconomic recession curve. Reducing the economic activities of agents reduces the severity of the epidemic on the one hand, but deepens the severity of economic recessions on the other. The model also incorporates endogenous mortality rates and probabilities of discovering effective treatments and vaccines over time. This extended framework helps the results to better mimic reality and allows us to assess the macroeconomic impact of epidemics.
To incorporate into this model the differences in state characteristics, we selected nine demographic, economic, and behavioural variables that show the greatest diversity (Table 1). These data were used to capture the different forms and intensities in which infections in each state are distributed among consumption, work, and other activities.
Table 1 Main data used for model calibration for each state
Note: Imperial College’s estimated infection fatality rates (IFR) were used to calibrate parameters related to the endogenous mortality rates to simulate the capacity of the health systems of each state.
To assess the results, we considered two scenarios. In the first, no containment measures are adopted by the authorities. In the second, optimal containment policies are adopted. Our results indicate that in both scenarios, the states present relevant differences in their epidemic dynamics due to their heterogeneous characteristics. From a qualitative point of view, key differences are verified in:
- Peak sizes of the infected population curves;
- Moments of the epidemic progress when the peaks of infected population curves are reached;
- Depths of economic recession curves;
- Moments of the epidemic progress when economic recovery begins;
- Shares of the total pre-epidemic population that is infected by the end of the epidemic;
- Shares of the total pre-epidemic population that die by the end of the epidemic.
An overview of these differences is in Figure 3, which presents all the results for the main variables both in the scenario in which containment policies are not adopted and in the scenario in which optimal policies are adopted.
Figure 3 Epidemic progression with adoption of optimal containment policies (solid lines) and without adoption of containment policies (dotted lines)
Notes: São Paulo (SP), Amazonas (AM), Ceará (CE), Rio de Janeiro (RJ) and Pernambuco (PE).
We found that states differ not only in the intensity of their local epidemics but also in their timing, a factor of great importance for the design of public containment policies. Table 2 summarises the moments in the progress of the epidemic when the main epidemic phenomena occur, showing how epidemics in states can experience different temporal dynamics.
Table 2 Moments of epidemic progress when the main peaks and valleys of the model variables occur, for each state.
Note: The moments are reported as progress percentage of the total time interval of the epidemic in each state. The beginning of the epidemic in each state is considered the moment when each state reaches the 100th infection (not necessarily the same date). This means that it would be incorrect to conclude, for example, that ‘the peak in the infected population would occur first in Amazonas and then in São Paulo’. A more correct interpretation would be: ‘in relation to the moment when the 100th infection occurs, the peak would occur faster in Amazonas (AM) than in São Paulo (SP)’.
The observed heterogeneities between states in epidemic dynamics imply the need for heterogeneous containment policies. When optimal trajectories of containment rates are achieved, we see significant differences in the measures that should be adopted by each state.
We find that trajectories vary between states in the initial severity required, the moment they must be accelerated, the extent they should be elevated, and when they should finally start to be reduced.
Table 3 Optimal containment rates for each state
To illustrate the differences, note, for example, time dynamics. In Rio de Janeiro the peak of the containment rate occurs at 48.67% of the epidemic progress, while in Amazonas it is at 40.67%, a difference of 8 percentage points. Considering an average epidemic duration of one year, this difference would mean that Rio de Janeiro, relative to the moment of the 100th infection, would take approximately one month longer than Amazonas to start relaxing its containment measures.
On the intensity of optimal containment policies, while Pernambuco requires the containment rate to reach 53.40%, São Paulo needs an increase only up to 38.76%, that is, 14.64 percentage points less than in Pernambuco.
Because these containment policies are heterogeneous, the effects are also quite heterogeneous. Table 4 shows how the implications of adopting optimal policies differ across states.
Table 4 Effects of adopting optimal containment policies, by state
Note: These numbers are obtained by calculating the difference between the results of the two scenarios.
Conclusion and lessons for other countries
Qualitatively corroborating our findings with empirical evidence in the COVID-19 literature for Brazil, we conclude that the characteristics of Brazilian states may significantly affect state epidemic dynamics, making them differ substantially.
From a policymaking perspective, this interstate heterogeneity implies the need for optimal containment policies that are also heterogeneous and varying in extent and duration for different states. Disregarding the importance of such heterogeneities and not taking them into account to coordinate containment policies may amplify both the severity of the economic recession and the number of infected and deaths resulting from the epidemic.
This, in part, can explain Brazil’s situation today and should alert any other large countries with similarly heterogeneous characteristics to the need to emphatically consider interregional heterogeneities in the fight against COVID-19.
Ajzenman, N, T Cavalcanti and D Da Mata (2020), “Leaders’ speech and risky behaviour during a pandemic”, VoxEU.org, 2 May.
Baldwin, R, and B Weder di Mauro (2020), Economics in the time of COVID-19, Voxeu.org eBook, a VoxEU.org eBook, CEPR Press.
Borelli and Góes (2020), “Macroeconomics of epidemics: Interstate heterogeneity in Brazil”, Covid Economics: Vetted and Real-Time Papers 30.
Eichenbaum, Martin S, S Rebelo and M Trabandt (2020), “The macroeconomics of epidemics”, NBER Working Paper 26882.
Gourinchas, P-O (2020), “Flattening the pandemic and recession curves”, in R Baldwin and B Weder di Mauro (eds.), Mitigating the COVID economic crisis: Act fast and do whatever it takes, a VoxEU.org eBook, CEPR Press.
Hallal, P, et al. (2020, May) “Remarkable variability in SARS-CoV-2 antibodies across Brazilian regions: Nationwide serological household survey in 27 states”, medRxiv, preprint.
Kermack, W, and A McKendrick (1927), “A contribution to the mathematical theory of epidemics”, Proceedings of the Royal Society of London, series A 115, no. 772: 700-721.