Harmful responding effects as moderated by response types (Dummy variable measure)
| DV | Proportion of negative comments | Number of negative comments | Number of total comments | Sentiment of comments |
|---|---|---|---|---|
| Data | 2 Month Before/After | 2 Month Before/After | 2 Month Before/After | 2 Month Before/After |
| Compensation × Post-Policy | 0.541336* | 4.241997** | 3.780426 | −0.801662*** |
| (0.264968) | (1.404860) | (2.550712) | (0.233554) | |
| Apologize × Post Policy | 0.082477 | −0.626327 | 0.510043 | 0.133563 |
| (0.122803) | (1.055628) | (2.521199) | (0.248621) | |
| Treatment × Post-Policy | 0.056596 | 3.123971** | −0.102404 | 0.094806 |
| (0.105324) | (1.030993) | (2.239799) | (0.270575) | |
| Compensation | 0.104709 | −2.003517 | −1.153003 | 1.045446*** |
| (0.099070) | (1.240022) | (1.143140) | (0.240972) | |
| Apologize | 0.012413 | −0.109119 | 0.059888 | 0.144883 |
| (0.087526) | (0.767834) | (0.902874) | (0.199510) | |
| Treatment | −0.176562 | 0.394014 | −1.011061 | 0.185072 |
| (0.172763) | (1.459082) | (2.318928) | (0.455418) | |
| Post-Policy Period | 0.118885+ | −0.235301 | −0.246638 | −0.025345 |
| (0.068070) | (0.665190) | (0.630641) | (0.210049) | |
| Number of Followers | 0.000001 | 0.000018* | 0.000020* | −0.000000 |
| (0.000001) | (0.000008) | (0.000009) | (0.000000) | |
| Lag Sentiment | −0.011738 | 0.196729 | 0.032441 | |
| (0.024595) | (0.294891) | (0.361427) | ||
| USDOT Complaints | −0.000838 | 0.011593 | 0.008869 | −0.005062** |
| (0.001086) | (0.013020) | (0.016024) | (0.001680) | |
| Number of Likes | −0.000007 | −0.000031 | −0.000036 | −0.000002 |
| (0.000006) | (0.000023) | (0.000024) | (0.000002) | |
| Number of Positive | 0.040209+ | 1.024298*** | ||
| Comments | (0.021291) | (0.021683) | ||
| Constant | 0.220945* | 1.295785 | 2.151815 | −2.958843*** |
| (0.105262) | (1.385871) | (2.069326) | (0.353000) | |
| Chi-Sq | 69.60 | 168.55 | 8990.68 | 40.20 |
| N | 268 | 268 | 268 | 268 |
| DV | Proportion of negative comments | Number of negative comments | Number of total comments | Sentiment of comments |
|---|---|---|---|---|
| Data | 2 Month Before/After | 2 Month Before/After | 2 Month Before/After | 2 Month Before/After |
| Compensation × Post-Policy | 0.541336* | 4.241997** | 3.780426 | −0.801662*** |
| (0.264968) | (1.404860) | (2.550712) | (0.233554) | |
| Apologize × Post Policy | 0.082477 | −0.626327 | 0.510043 | 0.133563 |
| (0.122803) | (1.055628) | (2.521199) | (0.248621) | |
| Treatment × Post-Policy | 0.056596 | 3.123971** | −0.102404 | 0.094806 |
| (0.105324) | (1.030993) | (2.239799) | (0.270575) | |
| Compensation | 0.104709 | −2.003517 | −1.153003 | 1.045446*** |
| (0.099070) | (1.240022) | (1.143140) | (0.240972) | |
| Apologize | 0.012413 | −0.109119 | 0.059888 | 0.144883 |
| (0.087526) | (0.767834) | (0.902874) | (0.199510) | |
| Treatment | −0.176562 | 0.394014 | −1.011061 | 0.185072 |
| (0.172763) | (1.459082) | (2.318928) | (0.455418) | |
| Post-Policy Period | 0.118885+ | −0.235301 | −0.246638 | −0.025345 |
| (0.068070) | (0.665190) | (0.630641) | (0.210049) | |
| Number of Followers | 0.000001 | 0.000018* | 0.000020* | −0.000000 |
| (0.000001) | (0.000008) | (0.000009) | (0.000000) | |
| Lag Sentiment | −0.011738 | 0.196729 | 0.032441 | |
| (0.024595) | (0.294891) | (0.361427) | ||
| USDOT Complaints | −0.000838 | 0.011593 | 0.008869 | −0.005062** |
| (0.001086) | (0.013020) | (0.016024) | (0.001680) | |
| Number of Likes | −0.000007 | −0.000031 | −0.000036 | −0.000002 |
| (0.000006) | (0.000023) | (0.000024) | (0.000002) | |
| Number of Positive | 0.040209+ | 1.024298*** | ||
| Comments | (0.021291) | (0.021683) | ||
| Constant | 0.220945* | 1.295785 | 2.151815 | −2.958843*** |
| (0.105262) | (1.385871) | (2.069326) | (0.353000) | |
| Chi-Sq | 69.60 | 168.55 | 8990.68 | 40.20 |
| 268 | 268 | 268 | 268 |
Note(s): **p < 0.01, *p < 0.05, + p < 0.10. Treatment is an indicator variable that is equal to 1 if the firm responds to negative comments post policy change. Compensation (or Apologize) is an indicator variable that is equal to 1 if the firm uses compensation (or apologize) type response. Lag Sentiment measures the sentiment of all negative comments for the firm in the previous month. We estimate a regression with endogenous treatment effects using the number of employees and the days since a brand page has been on Facebook as instruments for the treatment variable
Source(s): Authors’ own work