How Big Data Reveals Identity-Reaffirming Consumer Response to Covid-19: the Case of Political Identity

New big data findings show that online interest in covid-19 correlates with online interest in toilet paper in Democrat states, but with online interest in guns in Republican states. 

In uncertain times, it is not uncommon for individuals to react in surprising ways through consumption – and the covid-19 sanitary crisis is no exception. These situations demand rapid diagnosis and effective call for action. Governments and business actors thus increasingly need to quickly assess and analyze changing behaviors and effectively tackle them. Online search can greatly help – provided it sheds light on where and why people change their behavior.  

Let’s take a simple and recent example. As the Covid-19 progressively touched countries one after the other, vast media coverage revealed flocks of consumers rushing to buy toilet paper – but also guns (see here and here). In fact, “toilet paper” and “bulk ammunition” are among the top 5 keywords most searched in between March 2nd and April 1st 2020. And I write this piece, a rapid search on Google news revealed 124,000,000 entries for “covid-19 & toilet paper” and 357,000,000 entries for “covid-19 & guns”. 

What prompted consumers to stock toilet paper, guns and ammunitions – all seemingly unrelated to the disease? A simple but unsatisfactory explanation is that as consumers feel threatened by the virus and experience a loss of control, they turn back to their most basic human needs for survival and security. Back in 1962 during the Cuban missile crisis, when nuclear war seemed imminent, American consumers exhibited similar response by stockpiling canned goods, water… and toilet paper. A similar explanation could underlie the rush for guns given their use as a means to control.

But research shows consumers are not (always) irrational. In fact, when facing a threat, consumers selectively turn towards identity-rich products that can symbolically or effectively restore their sense of self. For instance, in one famous experiment, Gao et al. 2009 found that consumers who momentarily lost confidence in their intelligence (e.g., after failing a test) were more likely to buy a fountain pen (a symbol of intelligence) over a candy bar, because the former, but not the latter restored their self confidence. Contrary to a loss-of-control-based explanation, the identity-based account suggests that in our example, online searches and purchases of toilet paper and for guns may reveal two different responses an identity threat.

To unpack this phenomenon, my research assistant Yejin Park and I looked at post-covid-19 reactions in the US. Using Google search data in the US, we computed state-level correlations between queries for “covid-19,” “toilet paper,” and “gun” - two of the top words searched and covered in - for two months (14/01/2020 - 13/03/2020). To illustrate, a correlation of 1 indicates that if the number of searches for “covid-19” would increase by, say, 20%, the number of searches for “gun” would increase by the same percentage and at the same rate.

Consumer identity is strongly linked to one’s voting behavior – a strong marker reflecting one’s own world view based on one’s media consumption and exposure to different life narratives. We looked into whether state-level voting behavior would explain the search differences. In particular, we expected states which voted for Clinton would respond to covid-19 by preparing themselves for the unknown and for durable change (for which hoarding toilet paper or buying dumbbells makes sense). In contrast, we expected that states which voted for Trump should react to covid-19 by taking defensive action to preserve the status quo – a marker of political conservatism – and turn toward products that can help preserve it (for more on political identity and consumption, see Kim et al. 2018).

The results show that the states with the highest covid-19/gun correlations were states which voted for Trump over Clinton. In contrast, the states with the highest covid-19/toilet paper correlations tended to be states which voted for Clinton over Trump (see Figures 1 and 2). To illustrate, some of the states with the highest covid-19/toilet paper correlations include Oregon, New Hampshire or Rhodes Island; in contrast some of the states with the highest covid-19/ gun correlations include Alabama, Oklahoma, Florida or Tennessee. Consumers living in states which favored Trump in 2016 turn toward guns and bulk ammunition because the idea of preserving the status quo and defending themselves is an central part of their personality. In contrast former Clinton supporters tend to turn toward toilet paper because they are anticipating the change in lifestyle and prepare for a long lockdown. In fact, in a related analysis, online searches for covid-19 and toilet paper in Australia were both highly correlated with an increase in searches for survivalism – a doctrine preaching how to live during a world in crisis. 

Figure 1: intensity of correlations between searches for covid-19 and toilet paper

Figure 2: intensity of correlations between searches for covid-19 and gun

The findings show that consumers’ response to the sanitary crisis is not (entirely) irrational and offer a systematic methodology for organizations to quickly assess how a crisis affect consumer patterns and effectively react to consumers strategically turning toward certain products that reaffirm their identity: online search intensity mapping. Importantly, assessing the overall increase in search over a period is not enough. Search analytics should instead leverage geographical disparities in search levels to shed light on where and why changes in online searches (and offline behaviors) occur. This methodology rests on three simple steps: 1) recover the search volumes for one or several consumption-related products (e.g., toilet paper, canned food, yogurt etc..) and for the target event (e.g., Covid-19) for a specific period) 2) compute the correlation levels between each of the consumption keywords in the analysis and the target event and 3) producing geographical maps revealing variations in the strength of correlations across different geographies. We believe this methodology has a number of applications for consumer welfare and public health. It can significantly help uncover how consumers react locally – and limit potentially self-destructing consumption such as purchasing guns or drugs that governments will set the stage for a stronger and more stable post-crisis marketplace. 

Copyright ? 2020 David Dubois. All Rights Reserved.