social media strategies in product harm crises · social media strategies in product harm crises...

Social Media Strategies in Product Harm Crises

Thirty Sixth International Conference on Information Systems, Fort Worth 2015 1

Social Media Strategies in

Product Harm Crises Completed Research Paper

Shu He

The University of Texas at Austin Austin, TX 78712

[email protected]

Huaxia Rui University of Rochester

Rochester, NY 14627 [email protected]

Andrew B. Whinston The University of Texas at Austin

Austin, TX 78712 [email protected]

Abstract

The impacts of product-harm crises on other firms in the same category are complex considering the combined spillover and customer encroachment effects. With responding to both effects, non-focal companies may adjust their social media strategies to attenuate the negative influence from product-harm crises and to exploit the benefit from their competitor’s misfortune. Using daily social media activity data on Twitter for 56 major airline companies around the time of the Germanwings Flight 9525 crash, we find that non-focal airlines tend to increase their effort on customer relationship management after the crash with panel data fixed effects negative binomial models. At the same time, while airlines with no direct competition with the focal company appear to decrease their brand marketing effort due to the spillover effect, the tendency is reduced by the customer encroachment effect for the airlines that directly compete with the focal airline.

Keywords: Social media, Customer relationship management, social media marketing, Econometric analyses

Introduction

Product-harm crises happen if a product is defective, contaminated or harmful to consumers and the information is widely publicized. Unfortunately, product-harm crises are quite common nowadays. For instance, in the first 4 months of 2015, there have already been about 100 new recalls reported on Consumer Product Safety Commission website from various industries such as food, toy, and vehicles. Apparently, product-harm crises are nightmares for the focal companies which not only suffer from short-term financial cost of compensating affected consumers but also from lost sales, damaged reputation and quality perception as well as market shares after the crises. However, the effects of product-harm crisis on firms similar to the focal firms (e.g., firms in the same industry) are more complex, depending on factors such as product similarities and competition. For example, if consumers deem similar products of similar quality even though they are from non-focal firms, then, the perceived risk of consuming those similar products will also increase after the crises. As a result, the misfortune of the focal firm extends to non-focal firms that offer similar products, resulting in a negative spillover effect of the product-harm crisis. On the other hand, it is quite possible that firms offering similar products could benefit from the focal firm’s catastrophe if they directly compete with each other for consumers and there is lack of close



substitute of the type of affected products (e.g., when the product is a necessity). For example, customers may decide to patronize a competitor in the future to be free from the risk of consuming the focal firm’s products. We call this the customer encroachment effect of product-harm crisis. The potential existence of these opposite effects suggests the overall effect of product-harm crisis on non-focal firms is not obvious at all. Consequently, it is unclear how a non-focal firm would respond to a product-harm crisis. A non-focal firm not only has to strategically attenuate the spillover effect and exploit the customer encroachment effect, but also has to walk a fine line between revenue-seeking and compassion management so as not to backfire amid public anger.

We are interested in non-focal firms’ social media strategies during a product-harm crisis. Social media platform has become an increasingly important arena for companies to spend their marketing budgets. A recent survey of 4,943 marketing decision makers shows that the expected spending on social media marketing will grow from 8.4% of firms' total marketing budgets in 2013 to about 22% in the next five years.1 Today, many companies regularly post content on their Facebook pages and constantly monitor their Twitter accounts ready to interact with customers. For example, American Airlines has posted more than 879,000 tweets since it joined Twitter in 2009. Unlike traditional marketing mediums such as TV and newspapers which are typically scheduled at least several months in advance, social media offers an extremely flexible alternative for firms to adjust their marketing strategies almost in real time in today’s fast-paced world. Hence, we believe non-focal firms’ are most likely to adjust their social media strategies in response to a product-harm crisis. Another motivation of studying non-focal firms’ social media strategies during a product-harm crisis is to understand whether and how firms compete on social media, which has not been studied in the literature.

Our study subject is airlines’ social media strategy on Twitter in the aftermath of a major aviation accident. Airlines regularly post marketing content as well as responses to customers’ requests for conversation on Twitter. In this study, we measure the intensity of an airline’s marketing effort on Twitter by counting the number of originally posted tweets. To measure the intensity of an airline’s customer service effort (responding to customers’ request for conversation) on Twitter, we compute the number of tweets sent to the airline that received replies.

The main source of data for our study is the daily social media activity in terms of marketing effort and social media effort of 56 major airlines around the time of the crash of Germanwings Flight 9525 on March 24, 2015. Severe aviation crash can be a disaster for the whole airline industry. Empirical analysis in the previous literature (Chance et al., 1987, Bosch et al., 1998, Wong et al., 2003, and Ho et al., 2013) has shown that aviation accident may result in negative effect on sales and financial performance.

Comparing an airline’s marketing effort and customer service effort on Twitter before and after the aviation accident of the focal firm (i.e., Germanwings), we find that airlines tend to increase their effort on customer service on social media by increasing their communications with followers and customers after the crash. Interestingly, we find that airlines with low safety ratings and more similar to the focal airline tend to increase customer service effort even more, which might have been driven by the fact that airlines with low safety rating or similar airlines are more likely to be affected by the negative spillover effect of the product-harm crisis. Regarding airlines’ marketing effort, we find that companies which do not compete with the focal firm on any route tend to decrease their social media marketing effort after the crash. However, such tendency is reduced for the airline firms directly compete with the focal one.

The rest of the paper is organized as follows. We first review relevant literature and then develop the hypotheses for our research question. After that, we describe our data, followed by the description of the econometric models. Then we estimate the models and present the results. We conclude the paper by discussing the implications of the findings and pointing out future research directions.

1 2013 Chief Marketing Officer Survey: http://www.cmosurvey.org



Literature Review

Our paper is related both to the literature on product-harm crisis and the literature on social media marketing.

The literature on product-harm crisis typically focuses on effects of product-harm crisis on the focal company as well as the whole industry. In addition to the short-term financial cost of recalling defective products and compensating consumers who are affected by the defective products, the focal firms also suffer from lost sales, damaged reputation and quality perception as well as market shares after the crises, which could last for a prolonged period of time as in Dawar(1998), Van Heerde et al. (2007) and Chen et al. (2009). Furthermore, previous literature documents that product harm crises, or aviation crashes in our context, can induce huge negative influence not only on the focal company, but also on its competitors in the same category through this spillover effect. For example, Freedman et al. (2012) finds that there are industry wide spillovers of recalls with 276 recalls of toys and other children’s products in 2007. The negative spillover from one product to similar products of other firms might arise from increased uncertainty on products’ mean quality and the precision of the signals perceived by consumers in Zhao et al. (2011). In addition, customers may regard similar products in the same category as guilty by association. Hence, researchers have further investigated factors that may impact the likelihood of spillover effects in Roehm and Tybout (2006). Previous literature also investigated firms’ different strategies in product-harm crises. Cleeren et al., 2013 and Cleeren 2014 investigate competitors’ post-crisis strategies and their effectiveness, including advertising and price adjustments. Bala et al. (2015) theoretically analyze the competitors’ sales effort in crises considering an effort allocation among a portfolio of products across categories.

Aviation crashes, as the most severe product-harm crises for airline companies, are disasters for the whole industry. The existing literature focuses on the negative effect of aviation crash on passenger traffic and financial performance of the airline companies. In Wong and Yeh (2003), data from Taiwan shows that aviation accident decreases the whole market’s passenger traffic. Chance and Ferris (1987) regards aviation crash as an isolated incidence in stock market and it won’t influence other airline companies’ stock price significantly. However, in Bosch et al. (1998) and Ho et al. (2013), researchers study the stock prices’ changes of the whole industry after the aviation crash to show the evidence of both spillover and switching effect.

Although our paper is the first in the literature to study firms’ social media strategy in product-harm crisis, the literature on social media marketing in general is abundant. Most of the papers in this stream of literature either focus on exploring the link between firms' social media activities and consumers' social media engagements, or studying the link between consumers' engagement and their brand choices. For example, Lee Hosanagar and Nair (2014) study how persuasive content and informative content work differently in generating engagement on Facebook. Goh, Heng and Lin (2013) combine consumer transactions data with user-marketer interaction content data from a Facebook brand fan page to study the economic value of such engagement. Rishika et al. (2013) combine consumer transactions data with customers' social media participation data from a major social networking website to study the impact of such participation on the frequency of customer visits and customer profitability. Chung et al. (2014) estimates a system of simultaneous equations where the endogeneous variables include both consumer engagement measures such as liking or commenting on a firm's post, and firm’s financial performance measures, such as abnormal stock returns. They find that the richness and responsiveness of a firm's social media efforts are significantly associated with the firm's performance measured by abnormal returns and Tobin's q. In general, there is strong evidence of the business value of firm’s social media marketing efforts (Luo et al., 2013; Chung et al., 2014; Hitt et al., 2014). However, little is known about how firms leverage social media to compete with each other. Our paper contributes to this stream of literature by conceptualizing two aspects ─ brand marketing and customer relationship management ─ of firms’ social media strategies and studying how they are adjusted by firms during a product-harm crisis.

Hypotheses

Social media has been regarded as the “next big thing” for marketing (Yan 2011). Nowadays, firms regularly post marketing content and frequently engage with customers on social media sites. Based on



these two major types of activities, we conceptualize two primary functions of social media marketing, brand marketing and customer relationship management.

Social media platforms are ideal for brand marketing because firms can reach a large number of current or potential customers at relatively low cost, which offers a cost-effective way for a brand to engage with them with messages promoting its vision and building brand loyalty. The flexibility of social media and the wide support of its use on mobile devices allow a company to craft and deliver its brand message almost in real-time, which is vital in today’s fast-paced, technology-driven marketplace.

Twitter, as a prime example of social media platform, has also become a major platform for customer relationship management. This is particularly true for the airline industry. To this day, almost all commercial U.S. airlines and most of the major airlines in the world have official accounts on Twitter. Each day, thousands of conversations between customers and airlines occur on Twitter.

While these two functions of social media have distinct roles in a firm’s overall social media strategy, both can benefit a firm with more customer awareness and better customer relationship which may improve the firm’s real market and financial performance (Thomson et al. 2005, Woodcock et al. 2011, Hitt et al. 2014).

When there is product harm crisis in an industry, non-focal firms can adjust these two aspects of social media strategy to counter the negative spillover effect and to exploit the customer encroachment effect. We develop our hypotheses accordingly.

Brand Marketing

When an aviation accident happened, the negative spillover effect of the accident may render brand marketing by non-focal airlines less effective. Existing literature has found that advertisement by the focal firm’s competitors after the product harm crises may be a “double-edged sword” (Cleeren et al. 2013). On social media, consumers may regard non-focal airlines’ marketing messages at the height of media attention in the aftermath of a crash as opportunistic and insensitive. Furthermore, since aviation accident often results in heavy casualties, promotional messages immediately following the accident could easily backfire on social media. Hence, for airlines that do not directly compete with the focal airline, posting marketing messages on social media in the aftermath of a major aviation tragedy should be a suboptimal strategy. The above arguments lead to the following hypothesis.

Hypothesis 1: Non-focal firms that do not directly compete with the focal firm will decrease their brand marketing efforts on social media (Twitter) after the aviation accident.

On the other hand, non-focal firms that directly compete with the focal firm may take the crisis an opportunity to acquire the trouble company’s current and future customers (Chance et al., 1987, Bosch et al. 1998). In addition, with tremendous attentions around the world on the airlines industry after the accident, advertisement can even get more visibility for all companies in the same industry (Dawar 1998, Berger et al 2010). Indeed, Rubel et al. (2011) show that customers may pay more attentions to focal categorized brands, which may lead to an increase of advertising effectiveness.

In the airline industry, markets are naturally divided by routes (Ciliberto and Tambe 2009). Without any shared routes, airline companies face different groups of customers, resulting in little competition among them. Hence, companies who directly compete with the focal company may have more incentive for brand marketing through social media. Hence, we propose the following hypothesis.

Hypothesis 2: Non-focal firms that directly compete with the focal firm will decrease brand marketing efforts on social media less compared to non-focal firms that do not directly compete with the focal firm after the aviation accident.

Customer Relationship Management

Because consumers have higher uncertainty on overall airline quality due to the negative spillover effect (Zhao et al.2011), a non-focal airline may find it beneficial to be more attentive to customers’ requests for engagement on social media which could help reduce consumers’ anxiety and foster a positive customer relationship. On the other hand, from the perspective of resource allocation, non-focal firms may spend



more efforts engaging with customers on social media given the reduction in brand marketing effort on social media2. These two lines of arguments lead to the following hypothesis.

Hypothesis 3: Non-focal firms will increase their effort of customer relationship management on social media (Twitter) after the aviation accident.

Due to the customer encroachment effect, non-focal companies that directly compete with the focal airline will have more incentive than those not competing with the focal airline to increase their efforts to engage with customers on social media. On the other hand, considering the perspective of resource allocation, firms that directly compete with the focal firm may not be able to increase their effort on customer relationship management due to the relative fixed total available resources on social media. Hence, we propose the following hypothesis to test:

Hypothesis 4: Non-focal airlines that directly compete with the focal firm will increase their effort on customer relationship management on social media (Twitter) more after the aviation accident compared to firms that do not directly compete with the focal firm.

The following diagram summarizes the conceptual model for our hypotheses.

Figure 1: Conceptual model.

Data

To measure the companies’ brand marketing strategy, we use the number of original tweets posted by airline companies’ Twitter account each day. Companies frequently post tweets such as sales, attractive destinations, company’s activities and topics to attract attentions on social media. To measure an airline’s customer relationship management strategy, we use the number of tweets sent to the airline’s Twitter account that have been replied by the airline each day. Twitter users, including those who do not follow the focal company Twitter account, can send tweets to the company by “@” company’s Twitter account if they have questions, complains or information that they want to share with the airline company. In general, companies frequently communicate with their customers and actively reply to the incoming tweets.

2 The underlying assumption is that the total available resources invested on social media will not decrease within a short period after the aviation accident. This is reasonable given that the accident is completely exogenous and drastically reducing the labor force of the social media team within a short period is not practical.



To analyze the above related variables, we have generated a novel panel data set merged variables from various data sources. We collected Twitter information of 56 major airline firms around the world from September 2014 to June 2015. During this period, there are two major aviation crashes around the world: Indonesia AirAsia Flight 8501 on December 28th 2014 and Germanwings Flight 9525 on March 24th 2015. Since the first crash occurred during the holiday season, which was close to the beginning of a new year, there are potentially more unobserved confounding factors. Therefore, we use data from the Germanwings’ to test our hypotheses and report the analysis from AirAsia as a robustness check. The airline companies in our dataset are selected based on their size3 and based on whether they had verified Twitter account at the time when we started building the data collection system. More specifically, our data includes all the tweets posted by each airline during this period and all the tweets sent to each airline’s Twitter account.4 We also gather all airline companies’ route information5 and their safety ratings from airlineratings.com6. In particular, we define an airline firm as focal airline’s direct competitor if it shares at least one route with the focal firm.

Another variable we include is the social media similarity measure between a non-focal airline and the focal airline. In order to have a relatively comprehensive measure of the similarity, we take advantage of the followers for each firm’s Twitter account. In particular, we regard two airlines with more overlapped followers as more similar to each other. To evaluate the level of overlapping, we use the following measure:

The definitions and summary statistics for the main variables are listed in Table 1.

Empirical Analysis

Model Specification

As we mentioned above, our key dependent variables include the number of original tweets posted by airline companies’ Twitter account and the number of tweets sent to individual Twitter users (i.e., reply tweets) by the airline’s Twitter account. In order to compare companies’ strategies before and after the aviation accident, we apply panel data fixed effect models to control for potential unobserved airline companies’ fixed characteristics. Also, there may be a time trend on companies’ strategies. For instance, companies may gradually increase their effort on social media customer service over time, thus we add a time trend variable to control for it. In addition, we observe that some companies do not post any information during weekends on social network at all. So we also include dummy variables for each specific day of a week in our regression.

Given the fact that the numbers of original tweets or the number of reply tweets are integers with over-dispersion, we apply negative binomial model to analyze the data as in Haussman et al. (1984). More specifically, we use the following model to analyze the number of original tweets or reply tweets for airline at date t:

.

3 All these airlines had more than 10 million passengers annually in 2010.

4 For the main results, we exclude the Lufthansa airline which is the parent company of Germanwings.

5 See the link: http://ourairports.com/airlines/

6 See the link: http://www.airlineratings.com/airlines-ratings.php



Here, either represents the airline account ’s number of original posts or the number of reply tweets on day . For the independent variables, is the dummy variable that indicates whether it is before or after the airplane crash, is the categorical variable for each day of a week, is a control variable for time trend, and are the airline specific time invariant unobservable characteristics.

The baseline model above is used to test Hypothesis 1 and 3. To test Hypothesis 2 and Hypothesis 4, we include the interaction term of and the route-sharing dummy. This can also be regarded as a difference-in-difference model to further test Hypothesis 1 and 3 and alleviate the concern that there might be other events which happen to influence the companies’ social strategies along with the aviation accident. First of all, there is no additional major incident in the airline industry during our sample period that may potentially influence the majority or a significant portion of airline companies. Secondly, thanks to the heterogeneity in impacts that companies received from the same incident, we can take advantage of difference-in-differences method to alleviate potential endogeneity issue. The rational of the “difference-in-differences” estimation method is as follows: Airline companies with different background and from different markets may be affected differently by the same aviation accident. By testing whether the difference of companies’ strategic changes before and after the crash are consistent with the prediction based on the heterogeneous influence of the aviation accident, we can be more confident about the causality revealed by the empirical model. So we further propose the regressions as follows based on above specifications:

where represents airline company ’s characteristic that may define the impact magnitude of airline company from the accident and will be other time varying control variables. In our analysis, we use the dummy indicating whether the company directly competes with the focal company and the follower overlapping measure. For the control variables, there are daily total number of tweets sent to each company account, daily number of followers, daily number of original tweets, average growth rate of follower numbers for the last seven days of airline firm , as well as airline’s safety rating, whether it is a low

budget airline and its number of destinations. To exclude potential autocorrelation problem, we also include lagged dependent variables as control variables.

From the Table 1, we can see that the variance of original tweet number posted by airline daily is about 3.290^2/2.168=4.993 times its mean. Similarly, the variance of reply tweet number is much larger than its mean. Hence, we applied fixed effects negative binomial model to estimate regression (1) and (2) due to the over dispersion distribution of the dependent variables.

Results

The results from baseline regressions are reported in Columns 1-3 Table 2 and Columns 1-3 in Table 3. Clearly, after the aviation crash, airlines tend to shift their social media effort from brand marketing to customer relationship management. The results support Hypothesis 1 and Hypothesis 3.

Columns 4-9 of Table 2 presents the results of the DID model for brand marketing. We see that airline firms that directly compete with the focal Germanwings spent more effort on social media marketing compared to non-focal airlines not competing with Germanwings with the two-month and three-month data. This finding supports Hypothesis 2, indicating the customer encroachment effect. On the other hand, for the companies with more overlapped followers, they might experience more negative spillover from the crash. Hence we observe that those firms will decrease their marketing effort even more after the crash, further supporting Hypothesis 1.

Columns 4-9 of Table 3 presents the results of the DID model for customer relationship management. Overall, non-focal airline companies’ effort on social media has increased after the crash. Furthermore, firms more affected by the negative spillover effect from the aviation accident tend to increase their effort on social media more, which gives us more evidence of firms’ effort shifting from brand marketing to customer service. This may contradict to the results in previous literature, recording an increase of advertisement as in Dawar (1998) and Berger et al. (2010). This may be due to the fact that we are looking at a relatively short time period: one to three months after the crash. The flexibility of Twitter also enables airline firms to adjust their business strategies in time.



Robustness Check

AirAsia Crash

On December 28th 2014, Indonesia AirAsia Flight 8501 crashed into the Java Sea during bad weather. Since this tragedy happened during the holiday season at the end of 2014, there are too many confounding factors around the time period, which may introduce bias to our estimation results. Nevertheless, we repeated the main analysis on the airlines’ Twitter data from September 28th 2014 to March 24th 2015 as a robustness check. The results for brand marketing are reported in Columns 1-6 in Table 4. While Hypothesis 3 continues to hold, the support for Hypothesis 1 is much weaker. This might have been caused by the airlines’ tendency to increase their social media marketing effort during the holiday season.

Amtrak Derailment

Unlike an aviation accident, a rail incident will affect the airline industry differently: while the customer encroachment effect should still exist, the negative spillover effect should be minimal. On May 12, 2015, an Amtrak Northeast Regional train from Washington, D.C. bound for New York City derailed and crashed on the Northeast Corridor in the Port Richmond neighborhood of Philadelphia, Pennsylvania, killing 8 passengers and injuring more than 200 passengers. The 2015 crash was the deadliest on the Northeast Corridor since 1987. As one of the major alternative transportation in the east coast, this train derailment incident may undermine travelers’ perception of rail safety, which offers an opportunity to the airline industry to attract more demand. As a result, we expect to detect customer encroachment effect among U.S. airlines but not to find any negative spillover effect. To test this, we collected additional data for all airlines in our initial sample during the period from April 12th 2015 to July 12th 2015. We ran similar analysis and the results are presented in Columns 7-10 in Table 4. The results confirm our arguments above and further support our hypotheses.

Tweets with Hashtags

Generally, companies’ tweets include two categories: informational tweets and promotional tweets. For informational tweets, companies may post airline related or company related information on Twitter such as airline schedule and delay information, company’s major decisions and industry’s major affairs. The number of informational tweets may be less elastic to marketing strategies. On the other hand, companies also frequently post promotional tweets such as sales, attractive destinations, company’s activities and topics. Usually the promotional tweets may attract more attentions and interactions. Twitter hashtag, which is regarded as a powerful social media marketing tool for brand engagement as reported in Forbes, has been widely applied by companies to promote participation and attentions for their products and services. As a result, we use firms’ daily number of original tweets with Hashtags as an alternative measure of the firm’s brand marketing effort on social media. Since about 80% of the observations in our dataset have less than 3 hashtags on one day, we also generated the dummy variable for whether one company posted any hashtag tweets on a day. The results are presented in Table 5. Overall, the results are qualitatively similar to our main findings, thus further supporting all of our hypotheses.

Seemingly Unrelated Estimation

The dependent variables in customer relationship management and brand marketing effort may also be correlated. As a result, we further did a seemingly unrelated estimation for robustness check. The results are reported in Table 6. Again, the results support our hypotheses.

Safety Ratings

Roehm and Tybout (2006) found that a competing company is more likely to be influenced by the crisis if it is similar to the troubled company on the key attribute. Similarly, Janakiraman et al. (2009) found that spillover only happens among similar brands. With the aviation crash, consumers would naturally regard



the focal airline as not safe and would be more reluctant to purchase from similar airlines (i.e., airlines with dented safety record). So the worse a non-focal airline’s safety rating is, the more severe the negative spillover effect the airline will experience after the disaster, which implies an even lower effectiveness of brand marketing on social media. In line with this reasoning, we expect that non-focal firms with poorer safety rating will decrease their brand marketing efforts even more on social media (Twitter) after the aviation accident.

Similarly, airline firms with dented safety record will be more negatively influenced by the aviation accident, which would strengthen the arguments leading to Hypothesis 3. Hence, we expect non-focal firms with poorer safety record will increase their customer relationship management efforts more on social media (Twitter) after the aviation accident.

We test the moderating role of safety rating by introducing the interaction term of the crash dummy and the safety rating. The results presented in Table 7 suggest that non-focal airlines with poorer safety rating (i.e., lower value of safety rating) decrease brand marketing effort more and increase customer service effort more. In other words, the social media responses by those non-focal airlines that might have experienced more negative spillover effects are further amplified by their low safety ratings.

Conclusions

In this paper, we study airline companies' social media strategies after a major aviation disaster using Twitter activity data from 56 main airline companies within a six-month period. The two main aspects of companies' social media strategies include brand marketing and customer relationship management. From the first aspect, social media helps companies to reach large number of new customers and build up brand loyalty for the existing customers with relatively low cost. From the second aspect, social media provides companies a convenient channel to give prompt feedbacks to customers and to build up close relationship with customers. When a product-harm crisis happened to one focal company, the whole industry might be influenced through both the negative spillover effect and the customer encroachment effect. Although it is difficult to adjust their marketing strategies for traditional media, companies can adjust their social media strategies almost in real time. With panel data analysis, our empirical results indicate that companies would increase their social media effort on customer relationship management but decrease their social media efforts on brand marketing. Moreover, we find that companies that directly compete with the focal company will decrease their social media brand marketing effort compared with those that do not directly compete with the focal company, which indicates the customer encroachment effect. Furthermore, our results show that companies that are similar to the focal company will receive more negative spillover effect.

Our present paper can have multiple potential extensions. First of all, we plan to see how the strategies will influence companies’ Twitter account performance. Existing literature shows contradictive evidences on the effectiveness of advertisement after the product-harm crises (Zhao et al. 2011, Cleeren et al. 2013). With social media data, though it is difficult to see how the strategies will change the real world market share and sales in short time period, we can evaluate the effectiveness by company account's social media performance such as the growth rate of follower number, customers' retweet and like behaviors. In addition, we can further investigate companies' strategic changes on other aspects. For instance, companies may be more patient or humble when they are replying to their customers after the aviation accident. And it is also interesting to see if there are any systematic changes in tweets' content posted by non-focal companies in the category.



Table 1. Variable Definition and Summary statistics for Germanwing airline from December 2014 to June 2015

Variable Definition Observations Mean Std. Dev.

num_post Daily number of original tweets posted by

company Twitter account 9702 2.1678 3.290317

num_hashtag Daily number of original tweets with hashtag

posted by company Twitter account 9702 1.235415 2.627715

num_replies Daily number of tweets that sent to company

Twitter account are being replied 9702 53.31139 106.436

crash Dummy variable indicates before or after the

airline crash 9702 .5231911 .4994876

route_share Dummy variable indicates direct competitor of

focal company or not 9702 .1112142 .3144134

follower_overlap Standardized percentage of followers that are

overlapped with the focal company 9702 -4.81e-09 1

safety_std Standardized safety rating for each airline

company from airlineratings.com 9342 1.25e-08 1

low_cost Dummy variable indicates whether the airline

company is a low –cost carrier 9702 .3150897 .4645755

destination_std Standardized number of destinations for each

airline company Twitter account 9702 3.84e-09 1

follower_log Log of number of followers each day 9702 11.88939 1.442477

lag_growth Average growth rate of follower numbers for

the last seven days 9270 .0011938 .0013021

num_at Total number of tweets sent to the airline

company Twitter account each day 9702 205.0855 425.1841

weekday Categorical variable for weekdays 9702 2.975469 1.991689

Table 1. Variable Definition and Summary statistics7

7 The time period starts from December 29th 2014 to June 24th. We started from December 29th 2014 since Indonesia AirAsia Flight 8501 crashed on December 28th 2014.



Table 2. Companies’ Marketing Strategies around the Germanwings Crash

Baseline Route share interaction Route share + Follower overlapping interaction

1-month 2-month 3-month 1-month 2-month 3-month 1-month 2-month 3-month

crash -0.188*** -0.181*** -0.115*** -0.192*** -0.193*** -0.131*** -0.212*** -0.206*** -0.148***

(0.0432) (0.0421) (0.0389) (0.0438) (0.0426) (0.0393) (0.0463) (0.0440) (0.0405)

route_share 0.108 0.0993 0.132 0.106 0.0746 0.0757 0.0980 0.0621 0.0539

(0.224) (0.190) (0.168) (0.226) (0.196) (0.176) (0.228) (0.198) (0.179)

crash *

route_share

0.0461 0.164** 0.209*** 0.112 0.210** 0.263***

(0.109) (0.0838) (0.0746) (0.117) (0.0914) (0.0813)

follower_overlap -0.240*** -0.232*** -0.220*** -0.242*** -0.239*** -0.230*** -0.217*** -0.217*** -0.192***

(0.0659) (0.0548) (0.0509) (0.0661) (0.0553) (0.0512) (0.0691) (0.0578) (0.0553)

crash * -0.0818 -0.0547 -0.0677*

follower_overlap (0.0613) (0.0448) (0.0399)

Other control yes yes yes yes yes yes yes yes yes

weekday yes yes yes yes yes yes yes yes yes

time trend yes yes yes yes yes yes yes yes yes

airline fixed effect

yes yes yes yes yes yes yes yes yes

observations 5,759 7,349 9,098 5,759 7,349 9,098 5,759 7,349 9,098

number of airlines 53 53 53 53 53 53 53 53 53

Table 2. Companies’ Marketing Strategies around the Germanwings Crash8

8 This table represents the empirical results comparing companies’ social strategies before and after the Germanwings aviation crash. For 1-month specification, the observations include daily Twitter information from December 28th 2014 to April 24th 2015. For 2-month specification, the observations include daily Twitter information from December 28th 2014 to May 24th 2015. For 3-month specification, the observations include daily Twitter information from December 28th 2014 to June 24th 2015. The control variables include log of follower number, lagged growth rate of follower number, whether it is a low budget carrier and the standardized number of destinations. We use panel data fixed effect negative binomial model to estimate the coefficients. *** p<0.01, ** p<0.05, * p<0.1



Table 3. Companies’ Customer Relationship Management Strategies around the Germanwings Crash

Baseline Route share interaction Route share + Follower overlapping interaction

1-month 2-month 3-month 1-month 2-month 3-month 1-month 2-month 3-month

crash 0.0791*** 0.0773*** 0.0889*** 0.0631*** 0.0690*** 0.0759*** 0.104*** 0.0892*** 0.100***

(0.0185) (0.0183) (0.0172) (0.0197) (0.0189) (0.0177) (0.0206) (0.0193) (0.0180)

route_share 0.674*** 1.092*** 0.873*** 0.648*** 1.070*** 0.829*** 1.115*** 1.070*** 0.830***

(0.128) (0.108) (0.0941) (0.129) (0.110) (0.0963) (0.126) (0.109) (0.0945)

crash *

route_share

0.0707** 0.0447* 0.0780*** 0.000809 -0.0137 -0.00312

(0.0288) (0.0259) (0.0246) (0.0328) (0.0279) (0.0265)

follower_overlap 1.184*** -0.312*** -0.304*** 1.180*** -0.309*** -0.297*** -0.383*** -0.319*** -0.306***

(0.136) (0.0327) (0.0287) (0.135) (0.0331) (0.0294) (0.0389) (0.0351) (0.0330)

crash * 0.101*** 0.0953*** 0.117***


Other control yes yes yes yes yes yes yes yes yes

weekday yes yes yes yes yes yes yes yes yes

time trend yes yes yes yes yes yes yes yes yes


yes yes yes yes yes yes yes yes yes

observations 5,857 7,476 9,092 5,857 7,476 9,092 5,857 7,476 9,092

number of airlines 54 54 54 54 54 54 54 54 54

Table 3. Companies’ Customer Relationship Management Strategies around the Germanwings Crash9

9 This table represents the empirical results comparing companies’ social strategies before and after the Germanwings aviation crash. For 1-month specification, the observations include daily Twitter information from December 28th 2014 to April 24th 2015. For 2-month specification, the observations include daily Twitter information from December 28th 2014 to May 24th 2015. For 3-month specification, the observations include daily Twitter information from December 28th 2014 to June 24th 2015. The control variables include log of follower number, lagged growth rate of follower number, whether it is a low budget carrier and the standardized number of destinations. We use panel data fixed effect negative binomial model to estimate the coefficients. *** p<0.01, ** p<0.05, * p<0.1



Table 4. Companies’ Brand marketing Strategies around the AirAsia Crash and Amtrak Derailment

AirAsia Amtrak

Baseline interaction Baseline U.S. interaction

1-month 2-month 3-month 1-month 2-month 3-month 1-month 2-month 1-month 2-month

crash -0.0298 -0.00502 -0.0302 -0.0893* -0.0436 -0.0619 -0.0164 -0.0165 -0.0630 -0.0709

(0.0427) (0.0410) (0.0394) (0.0460) (0.0428) (0.0410) (0.0566) (0.0423) (0.0611) (0.0474)

route_share -0.451*** -0.303*** -0.210** -0.484*** -0.364*** -0.270***

(0.109) (0.0980) (0.0910) (0.111) (0.101) (0.0942)

crash *

route_share

0.268*** 0.186*** 0.149***

(0.0724) (0.0563) (0.0516)

follower_overlap -0.0574 -0.184*** -0.208*** -0.0563 -0.175*** -0.198***

(0.0814) (0.0417) (0.0379) (0.0752) (0.0422) (0.0392)

crash * -0.0785 -0.0279 -0.0237


if_U.S. 6.326*** 4.258*** 6.307*** 4.247***

(0.544) (0.468) (0.546) (0.467)

crash * 0.117** 0.141**

if_U.S. (0.0596) (0.0567)

Other control yes yes yes yes yes yes yes yes yes yes

weekday yes yes yes yes yes yes yes yes yes yes

time trend yes yes yes yes yes yes yes yes yes yes


yes yes yes yes yes yes yes yes yes yes

observations 6,413 8,056 9,328 6,413 8,056 9,328 2,853 4,473 2,853 4,473

number of airlines 53 53 53 53 53 53 54 54 54 54

Table 4. Companies’ Brand Marketing Strategies around the AirAsia Crash and Amtrak Derailment



Table 5. Companies’ Brand Marketing Strategies with Hashtags around Germanwings Crash

Number of Tweets with Hashtags If Hashtag Tweets

Baseline Interaction Baseline Interaction

1-month 2-month 3-month 1-month 2-month 3-month 1-month 2-month 3-month 1-month 2-month 3-month

crash -0.315*** -0.298*** -0.222*** -0.373*** -0.348*** -0.290***

-0.711*** -0.671*** -0.432*** -0.786*** -0.788*** -0.557***

(0.0571) (0.0565) (0.0524) (0.0621) (0.0587) (0.0545) (0.123) (0.120) (0.107) (0.127) (0.123) (0.110)

route_share

10.91*** 0.559 0.485* 10.94*** 0.544 0.320

(1.912) (0.340) (0.258) (1.975) (0.402) (0.297)

crash *

route_share

0.270* 0.428*** 0.528*** 0.666** 1.066*** 1.152***

(0.153) (0.128) (0.116) (0.291) (0.234) (0.211)

follower_overlap -1.649*** -0.284*** -0.237*** -1.594*** -0.261*** -0.202***

(0.253) (0.0727) (0.0637) (0.255) (0.0805) (0.0709)

crash * -0.204** -0.121** -0.101** -0.155 -0.138** -0.115*

follower_overlap (0.0904) (0.0576) (0.0483) (0.0957) (0.0699) (0.0618)

Other control yes yes yes yes yes yes yes yes yes yes yes yes

weekday yes yes yes yes yes yes yes yes yes yes yes yes

time trend yes yes yes yes yes yes yes yes yes yes yes yes


yes yes yes yes yes yes yes yes yes yes yes yes

observations 5,542 7,072 8,755 5,542 7,072 8,755 4,997 6,377 7,895 4,997 6,377 7,895

Number of airlines

51 51 51 51 51 51 46 46 46 46 46 46

Table 5. Companies’ Brand Marketing Strategies with Hashtags around Germanwings Crash



Table 6. Seemingly Unrelated Estimation on Companies’ Social Media Strategies around Germanwings Crash

1-month 2-month 3-month

Baseline Interaction Baseline Interaction Baseline Interaction

Marketing Customer Marketing Customer Marketing Customer Marketing Customer Marketing Customer Marketing Customer

crash

-0.196*** 0.0424** -0.228*** 0.0403** -0.185*** 0.0196 -0.210*** 0.0215 -0.0983** 0.0267* -0.133*** 0.0316*

(0.0522) (0.0181) (0.0555) (0.0189) (0.0516) (0.0179) (0.0537) (0.0184) (0.0464) (0.0162) (0.0483) (0.0167)

route_share

-12.61*** 0.242 -12.10*** 0.110 -7.329** -1.476 -6.876** -1.793 -4.178 -1.537 -3.642 -1.871*

(4.336) (1.586) (4.348) (1.589) (3.389) (1.322) (3.402) (1.323) (2.720) (1.128) (2.729) (1.124)

crash *

route_share

0.157 0.0294 0.223** 0.00460 0.280*** -0.0134

(0.127) (0.0335) (0.0933) (0.0300) (0.0826) (0.0276)

42.14*** 9.919* 40.20*** 10.36* 23.40* 16.53*** 21.42* 17.65*** 12.20 16.64*** 9.752 17.84***

follower_overlap (15.46) (5.635) (15.50) (5.646) (12.09) (4.705) (12.14) (4.709) (9.702) (4.015) (9.738) (4.000)

crash * -0.122** 0.0150 -0.0475 0.0382** -0.0735* 0.0510***

follower_overlap (0.0582) (0.0195) (0.0477) (0.0178) (0.0434) (0.0161)



time trend


airline fixed effect yes yes yes yes yes yes yes yes yes yes yes yes

observations 5,868 5,868 7,488 7,488 9,270 9,270 5,868 5,868 7,488 7,488 9,270 9,270

number of airlines 53 53 53 53 53 53 53 53 53 53 53 53

Table 6. Seemingly Unrelated Estimation on Companies’ Social Media Strategies around Germanwings Crash



Table 7. Companies’ Social Media Strategies with Safety Measures around Germanwings Crash

Brand Marketing Customer relationship

Baseline Interaction Baseline Interaction

1-month 2-month 3-month 1-month 2-month 3-month 1-month 2-month 3-month 1-month 2-month 3-month

crash -0.166*** -0.161*** -0.0973** -0.166*** -0.171*** -0.111*** 0.145*** 0.135*** 0.144*** 0.139*** 0.129*** 0.134***

(0.0443) (0.0432) (0.0398) (0.0448) (0.0435) (0.0401) (0.0207) (0.0203) (0.0193) (0.0211) (0.0205) (0.0195)

route_share

-0.0681 0.105 0.0121 -0.0647 0.0748 -0.0561 1.677*** 1.691*** 1.210*** 1.630*** 1.621*** 1.130***

(0.254) (0.229) (0.195) (0.257) (0.238) (0.206) (0.261) (0.233) (0.160) (0.260) (0.235) (0.167)

crash *

route_share

0.0296 0.176* 0.242*** 0.0919* 0.109*** 0.178***

(0.125) (0.0946) (0.0847) (0.0500) (0.0409) (0.0394)

safety_std 0.245*** 0.233*** 0.235*** 0.232*** 0.219*** 0.221*** -0.174*** -0.274*** -0.239*** -0.171*** -0.255*** -0.229***

(0.0376) (0.0338) (0.0305) (0.0377) (0.0346) (0.0319) (0.0328) (0.0295) (0.0267) (0.0334) (0.0300) (0.0271)

crash * 0.0580** 0.0353* 0.0274 -0.00581 -0.0309*** -

0.0208**

safety_std (0.0272) (0.0210) (0.0190) (0.0124) (0.00982) (0.00941)



time trend yes yes yes yes yes yes yes yes yes yes yes yes



observations 5,541 7,071 8,754 5,541 7,071 8,754 5,419 6,918 8,416 5,419 6,918 8,416

Number of airlines

51 51 51 51 51 51 50 50 50 50 50 50

Table 7. Companies’ Social Media Strategies with Safety Measures around Germanwings Crash



References

Bala, R., Bhardwaj, P., and Chintagunta, P. K. 2015, “Product Recalls, Category Effects and Competitor Response.” (Online at http://www. fin. gc. ca/consultresp/Airtravel/airtravStdy_e. html).

Berger, J., Sorensen, A. T., and Rasmussen, S. J. 2010. “Positive effects of negative publicity: When negative reviews increase sales,” Marketing Science (29:5), pp. 815-827.

Bosch, J. C., Eckard, E. W., and Singal, V. 1998. “Competitive Impact of Air Crashes: Stock Market Evidence,” The. JL and Econ (41), pp. 503.

Chance, D. M., and Ferris, S. P. 1987. “The effect of aviation disasters on the air transport industry: a financial market perspective,” Journal of Transport Economics and Policy (21:2), pp.151-165.

Chen, Y., Ganesan, S., and Liu, Y. 2009. “Does a firm's product-recall strategy affect its financial value? An examination of strategic alternatives during product-harm crises,” Journal of Marketing (73:6), pp. 214-226.

Chung, S., Animesh, A., Han, K., and Pinsonneault, A. 2014. “Firms' Social Media Efforts, Consumer Behavior, and Firm Performance: Evidence from Facebook. Consumer Behavior, and Firm Performance: Evidence from Facebook,” (Online at (Online at http://www. fin. gc. ca/consultresp/Airtravel/airtravStdy_e. html).).

Ciliberto, F., and Tamer, E. 2009. “Market structure and multiple equilibria in airline markets,” Econometrica ( 77:6), pp. 1791-1828.

Cleeren, K. 2014. “Using advertising and price to mitigate losses in a product-harm crisis,” Business Horizons (58:2), pp. 157-162.

Cleeren, K., Van Heerde, H. J., and Dekimpe, M. G. 2013. “Rising from the ashes: How brands and categories can overcome product-harm crises,” Journal of Marketing (77:2), pp. 58-77.

Dawar, N. 1998. “Product-harm crises and the signaling ability of brands,” International Studies of Management & Organization (28:3), pp. 109-119.

Freedman, S., Kearney, M., and Lederman, M. 2012. “Product recalls, imperfect information, and spillover effects: Lessons from the consumer response to the 2007 toy recalls,” Review of Economics and Statistics (94:2), pp. 499-516.

Goh, K., C. Heng, Z. Lin. 2013, “Social Media Brand Community and Consumer Behavior: Quantifying the Relative Impact of User- and Marketer-Generated Content.” Information Systems Research, 24(1), 88-107

Haussman, J., Hall, B., and Griliches, Z. 1984. “Economic models for count data with an application to the patents-R&D relathionship,” Econometrica (52), pp. 909-938.

Hitt, L., Jin, F., and Wu, L. 2014. “Who Benefits More from Social Media: Evidence from Large-Sample Firm Value Analysis,” (Online at http://misrc.umn.edu/wise/2014_Papers/127.pdf).

Ho, J. C., Qiu, M., and Tang, X. 2013. “Do airlines always suffer from crashes?,” Economics Letters (118:1), pp. 113-117.

Janakiraman, R., Sismeiro, C., and Dutta, S. 2009. “Perception spillovers across competing brands: a disaggregate model of how and when,” Journal of Marketing Research (46:4), pp. 467-481.

Lee, D., Hosanagar, K., and Nair, H. 2014. “The Effect of Social Media Marketing Content on Consumer Engagement: Evidence from Facebook,” (Online at http://papers.ssrn.com/sol3/papers.cfm? abstract_id=2290802).

Luo, X., Zhang, J., and Duan, W. 2013. “Social media and firm equity value,” Information Systems Research (24:1), pp. 146-163.

Rishika, R., A. Kumar, R. Janakiraman, and R. Bezawada. 2013. “The Effect of Customers' Social Media Participation on Customer Visit Frequency and Protability: An Empirical Investigation.” Information Systems Research, 24(1),108-127.

Roehm, M. L., and Tybout, A. M. 2006. “When will a brand scandal spill over, and how should competitors respond?,” Journal of Marketing Research, (43:3), pp. 366-373.

Rubel, O., Naik, P. A., and Srinivasan, S. 2011. “Optimal Advertising When Envisioning a Product-Harm Crisis,” Marketing Science (30:6), pp. 1048-1065.

Thomson, M., MacInnis, D. J., and Park, C. W. 2005. “The ties that bind: Measuring the strength of consumers’ emotional attachments to brands,” Journal of consumer psychology (15:1), pp. 77-91.

Van Heerde, H., Helsen, K., and Dekimpe, M. G. 2007. “The impact of a product-harm crisis on marketing effectiveness,” Marketing Science (26:2), pp. 230-245.



Wong, J. T., and Yeh, W. C. 2003. “Impact of flight accident on passenger traffic volume of the airlines in Taiwan,” Journal of the Eastern Asia Society for Transportation Studies (5), pp. 471-483.

Woodcock, N., Green, A., and Starkey, M. 2011. “Social CRM as a business strategy,” Journal of Database Marketing & Customer Strategy Management (18:1), pp. 50-64.

Yan, J. 2011. “Social media in branding: Fulfilling a need,” Journal of brand management (18:9), pp. 688-696.

Zhao, Y., Zhao, Y., and Helsen, K. 2011. “Consumer learning in a turbulent market environment: Modeling consumer choice dynamics after a product-harm crisis,” Journal of Marketing Research (48:2), pp. 255-267.

social media strategies in product harm crises · social media strategies in product harm crises...

Documents