Speech Can Kill: Bolsonaro And The Social Distancing Experiment

When President Trump suggests that injecting disinfectant could cure COVID-19 both health experts and the disinfectant manufacturers quickly came out and refuted the claim. Nevertheless, there was a massive spike in Google searches (Figure 1), a rise in calls to poison control centres, and several people even who consumed it. This highlights how recommendations from political leaders can have a substantial impact on their followers. We delved into the latest study by researchers from the University of Cambridge and Sao Paolo School of Economics, who decrypt how politicians can prompt their followers into dangerous behaviour using Brazil as a case in point.

The Setup

Why Brazil?

First, President Bolsonaro was dismissive of the virus and, in fact, encouraged people to not engage in social distancing. His controversial remarks were that is was ‘just a little dose of flu’, a ‘media trick’, and that Italy (one of the worst hit) was different. It provides an excellent opportunity to explore the impact of a leader’s actions and public speeches on the citizen behaviour. In particular two dates were important, the 15 March demonstration and the 24 March official presidential pronouncement. They prove to be outlier events where he boldly and comically dismissed the virus.

Second, Brazil is politically divided: the president has an equal level of approval and disapproval ratings. That acts as an effective counterfactual strategy to study how non-followers would react to his action.

The Data

Using Geo Localised data of 60 million (random and anonymous) mobile devices in Brazil (Dataset provider), researchers could track individual device locations and movement. That, in effect, helped them construct and map out social distancing index (Figure 2).

Figure 2: Social Distancing Index: Before and After

Source: Page 17 of “More Than Words: Leaders’ Speech and Risky Behavior during a Pandemic“

Unsurprisingly, social distancing rose as the pandemic gripped the nation (index increases). However, you can see that these changes were not homogeneous; social distancing increased in some regions more than the others. The researchers had a hunch that these differences were not random, so they collected electoral data to examine whether the geographical unit of adverse social distancing outcomes matched the areas where Bolsonaro was popular.

Empirical Model

To estimate the causal effect of whether Bolsonaro’s anti-social distancing message impacted his follower’s behavior, the researchers specified a model with ten days leads and lags. The model also allowed the authors to determine placebo and dynamic effects. For instance, to track how Bolsonaro’s message led to an increase/decrease in social distancing over time. The level of poverty, rurality, type of non-pharmaceutical intervention already in place and political preferences were also controlled for given these factors could impact the result of social distancing.

The Results

As shown in Figure 3, all leads, except one (t = -9), were statistically significant. The negative values of the leads imply Bolsonaro’s words and actions have had a more substantial effect on easing in social distancing in the areas where he had more support compared with where he did not. Roughly, one million people failed to practice adequate social distancing on each of the ten days following the key events (15 and 24 March).

Figure 3: Average Effect on Social Distancing*

Source: Page 18 of “More Than Words: Leaders’ Speech and Risky Behavior during a Pandemic“

*Note: Municipalities in which Boldonao had at least 50% of the votes in the first round of the 2018 election

One potential problem with the above estimation is that lockdown measures are implemented at the state level. There are some states in which he might have < 50% of the votes in every single municipality and vice versa. To combat this problem and for robustness, the researchers ran the same regression but changed the definition of ‘pro-Bolsonaro’ (Figure 4). They found, if any, the effect becomes more significant.

Figure 4: Average Effect on Social Distancing**

Source: Page 18 of “More Than Words: Leaders’ Speech and Risky Behavior during a Pandemic“

**Note: Municipalities in which Boldonao had more than the state median percentage of votes in the first round of the 2018 election

Bottom Line

This study highlights that, in some cases, medical advice is not what drives peoples behavious but rather the words and actions of their leader. Through the use of alternate dates, researchers have confirmed that a leader’s words can pose a negative externality on society (Brazil has the world second highest coronavirus cases). That’s why you see analysts tracking Trump’s speeches word by word and conducting sentiment analysis of political tweets religiously.

Mehdi is a research analyst at Macro Hive. He’s currently pursuing an MSc in Finance & Investment at Nottingham University Business School and he is a CFA level 3 candidate. Mehdi has previously pursued roles as an Equity Research Analyst, Junior Economist & in Proprietary Trading.