In a 1970 New York Times article, Milton Friedman famously wrote that “the social responsibility of business is to increase its profits” (Friedman 1970, p. 1). However, he also noted that firms may still devote resources to the communities they operate in to increase firms’ reputation and long-term financial interests. They may make charitable donations to fight poverty, decrease product prices to prevent inflation or invest in pollution reduction to protect the environment. In return, they may attract highly qualified employees at lower salaries while reducing their taxable income.

The idea that firms use social responsibility to maximize shareholder value still holds today (e.g., Cho et al. 2015). However, in addition to traditional CSR activities, stakeholders also want firms to take stances on sociopolitical matters such as abortion, climate change, immigration, gun control, and LGBTQ+ equality (Argenti 2020). Prominent examples of firms that do so are Google, which introduced gender-neutral bathrooms in its London offices (Detrick 2018), and Target, which announced stricter gun control measures following a 2019 shooting in one of its El Paso stores (Painter 2020). However, because sociopolitical activities address partisan values, they can create costs beyond the immediate financial expense of the activity. Polarization, which describes the existence of two divided groups without a middle ground, has grown in recent years to the extent that individuals reject any view opposing their own (Iyengar et al. 2019). This also affects firms and their products (Monahan et al. 2023).

Because activism poses reputational risk, only a few managers speak out on sociopolitical issues (Larcker et al. 2018; Wettstein and Baur 2016). However, public pressure may still force firms to take stances. In March 2022, Florida’s governor, Ron DeSantis, signed a bill prohibiting classroom discussions about sexual orientation and gender identity (Mazzei 2022). While reluctant at first, criticism from its employees led Disney to publicly oppose the bill and apologize for not speaking out sooner (Barnes 2022; Chapek 2022). This in turn incentivized DeSantis to challenge Disney’s special tax status as a self-governing local entity (Caspani and Chmielewski 2022).

A growing number of academic papers study firm value consequences of CEO or firm sociopolitical activism. Benedo and Siming (2020), for instance, examine abnormal returns around CEO resignations from Donald Trump’s presidential advisory council and detect losses of -0.57% around the resignation announcements. Bhagwat et al. (2020) examine returns around 293 activism events and document losses of on average ‑0.40%. Mkrtchyan et al. (2023b) similarly study returns around 1,402 CEO activism events and find value gains of around 0.20%. Gangopadhyay and Homroy (2023) detect returns of 1.3% around 187 events, which they attribute to higher future sales. Several further studies examine the reactions of consumers and employees. Mkrtchyan et al. (2023a) find higher employee satisfaction ratings at firms with activist CEOs. Burbano (2021) documents that sociopolitical activism has no, or negative effects on worker productivity. Hou and Poliquin (2022) and Painter (2020) use mobile phone geolocation data to study consumer behavior around company announcements of stricter gun control measures. Both studies find that activism decreases consumer visits in conservative counties. Painter (2020) further documents an increase in consumer visits in liberal counties.

Even though some of the results seem conflicting (e.g., Bhagwat et al. [2020] detect negative returns around activism events while Mkrtchyan et al. [2023b] detect positive returns), they are generally consistent with the anecdotal examples presented in Section 2.1. Sociopolitical activities signal that firms dedicate resources to controversial issues with uncertain outcomes. Some firms may benefit from this, others may experience losses. Effect directions and magnitudes depend on many factors, such as the specific topic, the political ideologies of the firms’ stakeholders, or the product market in which the firms operate (Painter 2020; Burbano 2021; Bhagwat et al. 2020; Mkrtchyan et al. 2023a, b). It may, for instance, pay off to win over one-half of customers at the expense of the other half when firms target niche product markets with high prices. However, when firms target large customer bases with low prices, it may be dangerous to alienate a substantial share of potential customers (Melloni et al. 2023).

Also, the credibility of the stances matters. Investors may not react to activities that they do not believe to be true (Mkrtchyan et al. 2023b). They may even discount the value of firms with authenticity concerns. Some consumers may shun firms not directly because of the ideological message but because they perceive it as hypocritical (Argenti 2020; Cho et al. 2015; Melloni et al. 2023). Such risk may in turn affect investor perceptions of expected future cash flows and explain negative returns around activism events.

This study focuses on changes in consumer behavior around activism events supporting the LGBTQ+ community. This includes (announcements of) selling LGBTQ+ merchandise, celebrating Pride Month, providing an LGBTQ+ inclusive work environment, or opposing anti-LGBTQ+ legislation.

As outlined above, sociopolitical activities such LGBTQ+ activism can have both positive and negative consequences for firms. On the one hand, firms may benefit from speaking out on LGBTQ+ matters, especially when it comes to attracting younger individuals who are more likely to place themselves on the LGBTQ+ spectrum (Gaubert 2023; Shortall 2019). Consistently, Pichler et al. (2017) and Fatmy et al. (2021) document a positive relationship between the presence of LGBTQ+ friendly corporate policies and firm performance. Artiga Gonzáles et al. (2022) further find that firms serving retail customers are more likely to adopt LGBTQ+ friendly policies, supporting the idea that LGBTQ+ activism is successful in attracting (liberal) customers. On the other hand, LGBTQ+ matters are polarizing, and many conservative consumers reject brands that engage in LGBTQ+ activism (e.g., Aratani 2023). Hence, activist firms, especially those targeting end-users, may experience lower sales or store visits around activism events. In addition, some consumers may perceive activism as inauthentic (Melloni et al. 2023; Ciszek and Lim 2021; Zheng 2021). Corporate LGBTQ+ activism has a high ‘pinkwashing’ potential, which means that firms may take ‘cheap’ observable LGBTQ+ stances but do not support LGBTQ+ individuals in more meaningful ways. Merchandise, in particular, is considered a ‘money grab’, which benefits the income of the firm but not the community (Barker 2022). An example of pinkwashing criticism can be found in Figure 1. Panel A on the left shows an Instagram post by queeeerchameleon, a comic artist who focuses on LGBTQ+ content. The post suggests that firms use the rainbow logo instead of, for instance, investing in an inclusive work environment. Panel B on the right shows a tweet by Amazon stating that Amazon “stand[s] together with the LGBTQIA+ community” and a reply by @vexwerewolf asking Amazon to explain discrepancies in Amazon’s talk and actions. If LGBTQ+ allies boycott firms due to hypocrisy concerns, then activism can cause a decline in store visits due to backlash from both liberal and conservative consumers.

Figure 1.

Examples of rainbow-washing criticism. Source Panel A (left): https://www.instagram.com/p/CsmPtaTr37w/; Source Panel B (right): https://twitter.com/vexwerewolf/status/1400785796651700230.

That said, we also note that there is some uncertainty whether boycott calls are effective (e.g., Sykes 2023). Due to brand loyalty and switching costs, only a minority of consumers who disapprove of corporate actions ultimately boycott firms (Klein et al. 2004; Liaukonytė et al. 2023). Moreover, negative news can lead to greater brand awareness, in which case boycott calls may increase consumer visits (Dong et al. 2021). Whether and how LGBTQ+ activism affects consumer behavior is thus an open empirical question. Accordingly, we phrase our hypothesis as no relationship between LGBTQ+ activism and consumer reactions:

H1: Consumers do not react to corporate LGBTQ+ activism

SafeGraph Inc. provides daily store visits (foot traffic) based on anonymized smartphone geolocation data. These data come from smartphone users who agreed to sharing their locations with location-tracking applications such as mapping or weather services (Painter 2020). These users make up about 10% of the U.S. population (Hou and Poliquin 2022). SafeGraph provides data on store locations, which means that our sample is based on brands with physical establishments. At the moment of download, the dataset included 38,755,100 observations and 905,584 store locations of 400,030 brands between January 2018 and November 2021.⁴

We apply several sample selection steps summarized in Table 1 Panel A. First, we drop 30,174,200 observations of stores with missing stock symbols. 85% of these stores have only one store location. They are likely small neighborhood stores whose stances we cannot observe. Second, because we later control for county and time-specific shopping behavior, we drop 305,778 observations of brand locations with zip codes that are not included in simplemaps.com’s zip-code-to-county crosswalk. Third, we exclude observations of brand locations that are not covered in the full (47-months) sample timeframe. These are locations that opened or closed at some point during the time frame, and whose foot traffic may generally differ from the foot traffic of well-established locations. Combined, this leaves us with 7,688,488 observations (i.e., brand-location-months) of 94 listed firms operating 180 brands. Note that a parent firm can operate multiple different brands, which in turn can operate multiple stores. The brand with the most locations is McDonald’s, with 11,790 stores, and the brands with the fewest locations are BUILT Custom Burgers, Mucho Burrito, Family Fresh Market, and Ranch One, with two locations each.⁵

We define LGBTQ+ activism as stances taken by firms supporting the LGBTQ+ community.⁶ To identify such stances, we search for keyword combinations in press releases available on FactSet and in company tweets.⁷ We focus on the first time a firm speaks out on LGBTQ+ matters. If consumers permanently change their shopping behavior following LGBTQ+ activism, then later stances by the same firm may not affect consumer visits any longer.

We search for activism events at the parent firm level. 26 of the 94 SafeGraph sample firms operate more than one brand. For instance, Yum! Brands Inc. operates KFC, Pizza Hut, and Taco Bell. Amazon.com Inc. operates Amazon Fresh, Amazon Go, and Whole Foods Market. Darden Restaurants Inc. and Kroger Co. operate eight restaurants and 21 grocery store brands, respectively. The choice to search for activism events at the parent level assumes that parent-firm stances are reciprocated by the individual brands. However, even if individual brands do not take stances themselves, association with the parent firms’ stances may still affect them.

After manually collecting firms’ Twitter handles, we add two more sample selection steps, summarized in Table 1 Panel B. First, we drop 17 firms (1,536,806 observations) without Twitter accounts as we cannot observe their stances. In fact, most of our sample firms take LGBTQ+ stances in tweets rather than in press releases. Also, LGBTQ+ related tweets precede most LGBTQ+ related press releases. Second, we drop seven firms that had their first activism event prior to the start of our dataset (1,955,952 observations) as these firms would have missing values on our treatment variable in all months (see Section 3.3 for more details). Consequently, our final sample consists of 70 firms, of which 14 have activism events. Table 2 Panel A and B present the summary statistics for stores with and without activism events.⁸ Tables A.1 and A.2 in the Appendix present an overview of all LGBTQ+ activism events, and a list of brands per activist parent firm.

Most activism events fall in June, which is to be expected as June is dedicated to celebrating LGBTQ+ pride in the U.S. With respect to the summary statistics, we observe slightly lower foot traffic at activist firms. Treatment (control) group stores receive on average 408 (468) visits and 267 (322) visitors per month, respectively (values in Table 2 are log transformed). At the same time, activist firms have notably more store locations, suggesting that larger brands with more stores are more likely to take LGBTQ+ stances. The treatment group has 2,190,717 observations and consists of 14 firms, 41 brands, and 46,611 locations (3,329 locations per firm). The control group has 2,004,973 observations and consists of 56 firms, 72 brands, and 42,659 locations (762 locations per firm).

Note that we do not expect a causal link between firm disclosures and consumer visits. Instead, tweets and press releases may just be a signal of general LGBTQ+ activism that can be observed more directly by consumers. Shake Shack, for instance, tweeted about serving a ‘pride shake’. The shake, which was sold in the restaurants, was likely more visible than the tweet. In other words, a difference in store visits at Shake Shack locations may be caused by consumers noticing the shake, rather than by consumers reading the tweet.

We use a difference-in-differences design to test the effect of LGBTQ+ activism on consumer store visits. We estimate the following OLS regression model, where i reflects parent firms, j reflects brands, k reflects store locations, and t reflects year-months:

Ln (store visits)_ijkt = α + β₁ Treatment_it + ϵ_ijkt

The dependent variable, Ln (store visits) is the natural logarithm of the number of store visits at a brand’s local store in a given year-month. The independent variable Treatment is equal to one in the three months following a parent firm’s LGBTQ+ activism event (the first month is the event month), zero in the three months prior to the event, and missing otherwise. The control group, for which Treatment is always zero, consists of firms that did not engage in LGBTQ+ activism events before or during our sample time frame.⁹ Treatment thus estimates the change in store visits at locations belonging to activist parent firms relative to the change in store visits at locations belonging to the non-activist control group. Figure 2 presents a visualization of the design, and Table 3 presents an overview of all variables used. Note that the regression model by default only uses observations in year-months for which Treatment is at least once equal to one or zero.¹⁰ All other months, i.e., year-months with missing values for Treatment, are omitted from the regression. In a three-month event window, this reduces the number of observations by 1,911,051 observations, leaving 2,284,639 observations to estimate β₁ (see also Table 1 Panel C).

Figure 2.

Research design. This figure illustrates our regression model using activism events from Kroger Co and Yum Brands! Inc. Our variable Treatment is equal to one at store locations in the three months following a firm’s LGBTQ+ activism event, zero in the three months before the event, and missing otherwise. Note that September 2019 has missing values on Treatment, which means it will be omitted from the regressions. Our model and this figure are based on Heese et al. (2022).

In addition, we add store and county-year-month fixed effects. Store fixed effects estimate coefficients at the store-location-level and account for time-invariant store characteristics such as stores’ local popularity. Year-month fixed effects account for time-specific differences in consumer visits. Consumers may, for instance, visit stores less in colder months. By using county-year-month fixed effects, we additionally account for differences in consumer visits in the same year-months in the same county. Such differences include county-specific weather events and local economic wealth (Painter 2020). Finally, we cluster standard errors by county in all tests.

Our main results are presented in Table 4, column (1). The coefficient on Treatment is negative and significant, which means that consumers reduce their store visits when parent firms take pro-LGBTQ+ stances. The coefficient of -0.025 means that visits at these locations decreased by 2.5% relative to the stores included in the control group. This effect size compares to Painter (2020), who documents a 3.9% decrease in store visits following Walmart’s announcement of stricter gun control measures.

Our base results in column (1) present the effects of comparing LGBTQ+ activist parent firms to all other non-activist control group firms. In columns (2) to (4), we re-estimate effects within the specific store types (Hou and Polquin 2023; Painter 2020). We use the top categories given in SafeGraph, which leaves three store types: Gasoline stations, grocery stores, and restaurants. The results are presented in Table 4, columns (2) to (4). The coefficient on Treatment remains significantly negative for grocery stores and restaurants but is insignificant at gasoline stations, suggesting that LGBTQ+ activism does not affect consumer behavior at all store types. The effect also seems to differ across grocery stores and restaurants: The coefficient on Treatment for grocery stores suggests a decrease of 10.9%, while the coefficient on Treatment for restaurants suggests a substantially smaller decrease of 1.1%.

An explanation for the insignificant coefficient at gasoline stations may be that activism is less observable at those stores. In addition to announcing their support, grocery stores and restaurants can offer LGBTQ+ themed products such as Shake Shak’s ‘Pride Shake’ mentioned in Section 3.2. Besides, boycotts may be less effective at gasoline stations, where price and location are the main drivers of consumption choices (Fernandes 2023). With respect to the difference in effects between restaurants and grocery stores, one reason may be that consumers may visit restaurants less frequently than grocery stores, in which case the effects of boycotts would not be as obvious. Alternatively, consumers may be less likely to boycott their favorite restaurant compared to their favorite grocery store.

We run several robustness tests. First, we change the event window for Treatment to one and six months, respectively. Second, we use the natural logarithm of the number of individual visitors, Ln(visitors), as an alternative dependent variable. Third, to rule out that our results are driven by one specific activist parent firm, we rerun the test in Table 4 columns (1) to (4) and drop one individual firm at a time. The results support prior inferences. Fourth, for each activist firm, we search for negative events that may occur at the same time as the LGBTQ+ activism events. Except for Winn-Dixie, which had to issue a product call back due to possibly contaminated kale, we do not find bad news that could explain our results.¹¹ Finally, we plot treatment coefficients by event window in Figure 3. To do so, we create an events dataset, including for each treatment firm the event month (month zero) and the five months before and after the activism event.¹² The control group again consists of firms that did not engage in LGBTQ+ activism during the sample time frame. The coefficients are again based on regressions with Ln(visits) as dependent variable and Treatment as independent variable, including county-year-month fixed effects.¹³ Panel A to D present the treatment effects for all stores, gasoline stations, grocery stores, and restaurants, respectively. Consistent with Table 4, we observe a decrease in store visits following LGBTQ+ activism events (month zero) at grocery stores and restaurants, but not at gasoline stations. Again, the effects are more pronounced at grocery stores. Interestingly, the coefficients on Treatment are generally positive for grocery stores, but generally negative for restaurants. We do not have an explanation for this phenomenon, but it is possible that the net benefits of engaging in activism differ across store types, making large grocery store chains (restaurants) with high consumer turnover more (less) likely to engage in LGBTQ+ activism.

Figure 3.

Stacked events. This figure presents the coefficient estimates from regressions of ln(visits) on Treatment and county-year-month fixed effects in a stacked events dataset. Treatment is equal to one if the firm has an LGBTQ+ activism event during our sample time frame, and zero otherwise. Month zero is the event month. Standard errors are clustered by county.

We find that corporate LGBTQ+ activism is associated with reduced store visits. As described in Section 3.2, this negative effect may be caused by two distinctly different consumer groups: conservative consumers who may boycott firms because they disapprove of LGBTQ+ activism, or liberal consumers who may perceive the stances as hypocritical. To shed more light on the source of the effect, we examine differences in consumers’ political ideologies. If the decrease in store visits is predominantly caused by fewer conservative consumers, we would expect a stronger effect in conservative counties.

To proxy for ideology, we use data on presidential election results by county provided by the MIT Election Lab.¹⁴ Following Painter (2020), we split vote shares into quintiles: Q1 represents counties where the Republican presidential candidate received less than 20% of votes, Q2 represents counties where the Republican presidential candidate received at least 20% but less than 40% of votes, Q3 represents counties where the Republican presidential candidate received at least 40% of votes but less than 60%, and so on. We then regress ln(visits) on Treatment and on the interactions of Treatment with the Q2, Q3, Q4, and Q5 (Qi) vote share quintiles, denoted as Treatment × Qi. We continue to use county-year-month fixed effects, which subsume the main effects (i.e., the coefficients on Qi). The results are presented in Table 5. Columns (1) to (3) present the results for gasoline stations, grocery stores, and restaurants, respectively.

Note that because we now use interactions of Treatment with Qi, the interpretation of Treatment differs: Treatment presents the effect of LGBTQ+ activism in the highly liberal Q1 counties. The interaction terms Treatment × Qi present the differences in post-activism store visits in Qi relative to the Q1 base group.

Except for one coefficient on Treatment × Q4 in the grocery stores group, none of the interaction terms is significant at conventional levels (p < 0.05). This means that there are no significant differences between highly liberal and more conservative counties. The effects are similar when we use sample splits instead of interactions (untabulated). In these subsamples, we observe a significantly negative coefficient on Treatment for grocery stores and restaurants in both liberal and conservative counties.

These effects are surprising given that liberals generally support LGBTQ+ activism (Gaubert 2023; Shortall 2019), in which case we would expect an increase in foot traffic or no effect for this group. We can think of at least three explanations for our findings. First, firms may take stances because they anticipate decreases in store visits, and hope to attract new (liberal) consumers. We consider this explanation less likely given the risk that comes with sociopolitical activism (Wettstein and Baur 2016). After all, by September 2021, only about a quarter of the SafeGraph sample firms took (observable) LGBTQ+ stances. Second, LGBTQ+ activism may decrease visits by conservative consumers but may not increase visits from liberal consumers. In that case, we may observe a significantly negative effect even in liberal counties, but the decrease in liberal counties would likely be smaller than the decrease in conservative counties. We observe this to some extent at grocery stores, for which the coefficient on Treatment × Q4 is significantly negative. However, we do not observe significant differences between Q1 counties and the less liberal Q3 and Q5 counties. Also, we do not observe significant differences at restaurants. Third, liberal consumers may reduce their store visits due to authenticity concerns or perceived hypocrisy. To find some anecdotal support for this hypothesis, we search NexisUni and Google News for media coverage of the activism events. This search reveals some negative sentiment about Exxon Mobil’s and Amazon’s activism.¹⁵ However, we also find newspaper articles that praise the activist sample firms for their commitment (see e.g., Moussa 2018).

Overall, we find that LGBTQ+ activism is associated with decreased store visits. These findings are consistent with prior research documenting firm value losses around activism events (e.g., Bhagwat et al. 2020), but inconsistent with studies detecting greater financial performance at firms with LGBTQ+ friendly corporate policies (Fatmy et al. 2021; Pichler et al. 2017). In addition, we document similar effects in liberal and conservative counties. These results are surprising, as they suggest that conservative consumers are not the only consumer group that responds negatively to LGBTQ+ activism. While the current data does not allow us to test this assumption empirically, one possible explanation is that liberal consumers reduce their store visits when they perceive a firm as dishonest. For instance, as illustrated in Figure 1, some firms may take (rhetorical) stances but fail to provide more direct support, such as implementing policies for an inclusive workspace. Other firms may provide such support, but may also donate to politicians voting against pro-LGBTQ+ legislation (see e.g., Place 2021). Overall, we conclude that sociopolitical activism is a complex issue, requiring firms to carefully consider how a stance may be perceived by their respective target group.