The impact of the MeToo Scandal on women's perceptions of safety

Abstract

8 This study assesses whether the negative exogenous informational shock of the MeToo 9 scandal has affected women’s perception of safety. The MeToo movement was first re10 ported in the media worldwide in October 2017, and has received enormous press coverage 11 since then. The exogenous and unanticipated nature of the scandal provides a natural 12 experiment that we can use to quantify how wider external information affects ‘ordinary’ 13 women’s perceptions of safety and their willingness to report feelings of dissatisfaction 14 with safety. To do so, we use a case study of women’s’ perception of safety in 25 metro 15 systems across the world, for which we have large-scale unique customer satisfaction data 16 over the years 2014 to 2018. We use interview dates to determine perceptions pre and post 17 scandal, and we take the MeToo scandal as a ‘treatment’ affecting women in the sense that 18 they were the primary target of the informational shock. Using pre and post treatment 19 data, for our defined treated (women) and control (men) units we apply a difference-in20 differences estimator to identify the impact of the scandal on perceptions of safety. Our 21 results show an 2.5% increase in the probability of women being dissatisfied with security 22 in stations and in metro carriages post-scandal. These results support the conclusion that 23 revealed preferences are only revealed up to a certain extent. A change in the context (e.g., 24 an informational shock), can alter perceptions and in turn, can encourage individuals to 25 disclose lower (or higher) satisfaction levels as perceptions of the norm change.

26 Keywords: Gender; Perceptions; Safety, Security; Metro; Revealed Preferences; Stated Pref27 erences.

28 1 Introduction

29 People’s perceptions and preferences depend on the nature and context of the information 30 available to them. Within the abundant literature on individual responses to additional in31 formation, many studies explore the importance of inferring from motivated reasoning (e.g. 32 Kunda 1990, Bénabou and Tirole 2002, 2011). This concept proposes that even in settings 33 in which individuals are constrained to be accurate, they remain unable to control their own 34 preconceptions when making a decision. This view of belief formation is consistent with the 35 renewal of interest in emotions and their influence on decision-making in psychology and neu36 roscience (e.g., Dolan and Sharot (2011)). Motivated reasoning is different than cognitive 37 dissonance and confirmation bias, representing situations in which people overweight their 38 prior beliefs when updating their views (Festinger 1957, Tetlock 1983, Rabin and Schrag

1 1999, Taber and Lodge 2006). Yet, confirmation bias can coexist with motivated reasoning 2 and there is ample experimental evidence of this form (e.g. Taber and Lodge 2006, Kahan 3 et al. 2012, Nyhan and Reifler 2010, Nyhan et al. 2013).

4 Motivated reasoning implies that individuals do not fully control their own preconceptions 5 when making decisions, which creates a tension between accuracy and the perseverance of 6 beliefs. Under the assumption of rationality, the collection and integration of new information 7 should be independent of one’s prior judgement (Evans and Over 2013). Depending on the 8 credibility of a piece of evidence, individuals’ perceptions should be adjusted upwards or 9 downwards (Anderson 1981). However, individuals are not purely rational.

10 Our starting premise is that all reasoning is at least in part motivated (Kunda 1987, 1990), and 11 that consequently we expect an informational shock to alter perceptions. Since its unravelling 12 on October 15, 2017, the MeToo movement has incentivised individuals to reveal experiences 13 of harassment, whether they experienced it themselves or have shared someone else’s expe14 rience(s). Although this event initially started in America, the press coverage around these 15 themes spread worldwide (see Fig. 1) and sparked wider debates around women’s security. 16 Since the scandal, individuals have been actively searching this topic on the internet, indi17 cating that they are aware of the movement and its content as well (see Fig. 2). As such, it 18 constitutes an excellent candidate for a quasi-experimental study of the impacts of information 19 on perceptions.

Figure 1: Detailed daily count of headlines containing the keywords “MeToo” (source: Factiva)

Figure 2: Web search intensity for the keywords “MeToo” (source: Google Trends)

1 We exploit the exogenous and unanticipated nature of the MeToo scandal, and the unique 2 setting surrounding this event (intensive press coverage, worldwide magnitude of the event), to 3 identify changes in women’s perceptions and propensity to reveal their preferences. Building 4 upon the principles of motivated reasoning, our analysis tests empirically the extent to which 5 behaviour changes post-scandal. Our specific research questions are as follows. Did the 6 informational shock associated with the MeToo scandal change women’s propensity to state 7 their level of contentment with safety? Do negative informational shocks leads to a significant 8 update in the stated preferences?

9 To answer these questions we conduct a case study of women’s’ perception of safety in 25 urban 10 metro systems around the world, for which we have large-scale unique customer satisfaction 11 data over the period 2014 to 2018. While the MeToo movement had no specific association with 12 public transport, women’s safety on public transport has been of long standing concern, and 13 in fact as Figure 3 shows, articles with headlines referring to MeToo and transport increased 14 massively after the MeToo scandal. Thus, we believe that our unique data provide a route to 15 examine the effects of MeToo on the perceptions of ’ordinary’ women that use public transport 16 systems to go about their daily business.

Figure 3: Cumulated number of headlines with keywords related to Me Too and transport (metro, transport, public transport) – Source: Factiva

1 The paper makes two main contributions. The first is empirical. To our knowledge we are 2 the first statistical study to explore the impact of the MeToo scandal on the perception of 3 safety in public transport worldwide. Our data on the stated preferences of over 137,500 4 individuals from 25 cities across the world provides an excellent opportunity to explore this 5 issue in a detailed and comprehensive fashion. Our second contribution lies in the nature 6 of our statistical approach we use to identify the impact of MeToo on women’s perceptions 7 of safety. We define the MeToo scandal as ’treatment’ creating pre- and post- treatment 8 periods, and we define women as treated ’units’ and men as controls. However, we also 9 acknowledge that men could be affected by this informational shock and incentivised into 10 revealing different preferences. To deal with this seeming inconsistency, we use a difference11 in-differences methodology which provides us with clear-cut causal estimates and eliminates 12 trend biases, while also controlling for men’s potential reactions to the informational shock. 13 Our DiD analysis can also explore the existence and extent of differentiated responses to test 14 how informational shocks affect subgroups based on socio-demographics.

15 Using pre and post scandal (treatment) data, for our defined treated (women) and control 16 (men) units, we apply a difference-in-differences (DiD) estimator to identify the impact of the 17 scandal on perceptions of safety. Our results show evidence that the MeToo scandal did lead 18 to an increase in the propensity of women to reveal and state concerns over safety. We find this 19 effect essentially for respondents in European countries for which we find an average increase 20 of 2.5% in women’s propensity to agree that metro stations are not secure, following the 21 MeToo scandal. Our DiD results are stable across specifications and to location controls and 22 time fixed effects. Testing for heterogeneity in responses show that there are no heterogeneous 23 effects by age levels, and that women’s reaction is of a larger magnitude if they are daily metro 24 users.

25 Our results support the conclusion that revealed preferences are only revealed up to a certain

1 extent, and that informational shocks encourage individuals to disclose different satisfaction 2 levels as their perception of the norm change. The paper is structured as follows. Section 3 3 presents the data and Section 4 the empirical model. Section 5 presents and interpret results 4 along with robustness checks and sensitivity analyses. Section 6 concludes.

5 2 The MeToo Scandal

6 The MeToo movement was first reported in the media worldwide in October 2017, and has 7 been subject to a substantial press coverage since, as presented in Figures 2 and 1 that 8 indicate respectively an increased web search intensity and the number of headlines. While the 9 MeToo movement had no specific association with public transport, women’s safety on public 10 transport has been of long standing concern. One might object that the MeToo movement 11 was not associated with public transport in articles, yet Figure 3 shows that articles with 12 headlines referring to MeToo and transport started and increased massively after the MeToo 13 scandal1 .

14 In this study, the MeToo scandal is analysed from a perspective that is moved away from 15 testimonies within the Hollywood industry and is closer to the public debate. Our rationale 16 revolves around Europe as public transport is widely used and spread in these countries, and 17 metros are used by general population and across all walks of life. Therefore, it is relevant to 18 focus our study on public transport patronage in Europe as it differs from that in Asia and 19 the Americas, with customer data in European transport systems being more representative 20 of the population. Thus, we believe that this unique event provides a route to examine the 21 perceptions of women that use public transport systems to go about their daily business.

22 To the best of our knowledge, there has been no quantitative analyses of the impact of the 23 MeToo movement. Analyses led so far are quite scarce, of a qualitative nature and do not 24 analyse individual responses to the scandal at a large scale. More precisely, one study present 25 the interest of an application aimed at helping harassed women (Sheikh and Fayyaz 2019) 26 in the context of MeToo but does not analyse the phenomenon in itself. Another study uses 27 the MeToo context to analyse qualitatively how anger is mediated after the scandal but uses 28 the MeToo scandal from a Hollywood viewpoint to analyse individual responses (Orgad and 29 Gill 2019). Gu et al. (2019) refers to the MeToo scandal to mention gender differences and 30 discrimination but does not study the scandal per se.

31 Prior research has tested the impact of informational shocks on individuals (Sides 2011, Cruces 32 et al. 2013, Kuziemko et al. 2015, De Neve et al. 2017). However, individual responses to 33 informational shocks are often studied in the field of public economics and more often it is its 34 impact on voting behaviours that is explored (Ait Bihi Ouali 2019). According to Alesina and 35 Giuliano (2011) and Margalit (2013), negative experiences can change individuals’ preferences 36 and make them less optimistic or more risk averse. Fiscal scandals are considered a negative 37 experience, triggering a revision of beliefs.

38 Our previous research shows that women have different priors than men on the perception of

Shortly after the scandal first broke in the media, a survey suggested that individuals are more willing to speak out against harassment post-MeToo scandal, which already hints at the fact that this informational shock affects individual perceptions, see: https : //www.f awcettsociety.org.uk/metoo − one − year.